130 Fulltext

The Syntax of Compound Tenses in Slavic
Published by
LOT phone: +31 30 253 6006
Trans 10 fax: +31 30 253 6000
3512 JK Utrecht e-mail: lot@let.uu.nl
The Netherlands http://wwwlot.let.uu.nl/
Cover illustration: Cyril and Methodius, the first Slavic linguists. A picture by
Barbara Tomaszewicz inspired by G. Čapkŭnov’s work.
ISBN-10: 90-76864-99-3
ISBN-13: 978-90-76864-99-0
NUR 632
Copyright © 2006: Krzysztof Migdalski. All rights reserved.

The Syntax of Compound Tenses in Slavic
Proefschrift
ter verkrijging van de graad van doctor

aan de Universiteit van Tilburg,
op gezag van de rector magnificus,
prof. dr. F.A. van der Duyn Schouten,
in het openbaar te verdedigen ten overstaan van
een door het college voor promoties aangewezen commissie
in de aula van de universiteit
op dinsdag 30 mei 2006 om 14.15 uur
door
Krzysztof Marek Migdalski
geboren op 4 maart 1975 te Bielsko-Biała, Polen

Promotor: Prof. dr. H.C. van Riemsdijk
Copromotor: Dr. H.J.W.M Broekhuis
Contents
ACKNOWLEDGMENTS ................................................................................... XI
ABBREVIATIONS........................................................................................... XIII
INTRODUCTION................................................................................................. 1
CHAPTER 1 THE DIACHRONY OF COMPOUND TENSES IN SLAVIC7
1.1 Introduction.................................................................................................7
1.2 The division of the Slavic languages ..........................................................7

1.2.1 The common ancestor........................................................................................8
1.2.2 Sources of the division........................................................................................9
1.3 The Tense and Aspect system ...................................................................11

1.3.1 The Tense System in Proto-Indo-European .................................................11
1.3.2 Modifications of tense and aspect systems in Proto-Slavic .........................13
1.3.3 Old Church Slavonic.........................................................................................16
1.3.3.1 Simple past tenses in Old Church Slavonic...............................................16
1.3.3.1.1 The aorist.................................................................................................16
1.3.3.1.2 The imperfectum ....................................................................................17
1.3.3.2 Aspect in Old Church Slavonic...................................................................18
1.3.3.3 Some interactions between tense and aspect ............................................20
1.3.3.3.1 Imperfectum + Perfective Aspect........................................................21
1.3.3.3.2 Aorist + Imperfective Aspect ...............................................................22
1.3.3.4 The Compound Tenses in Old Church Slavonic .....................................22
1.3.3.4.1 Future tenses ...........................................................................................22
1.3.3.4.1.1 Future I.............................................................................................23
1.3.3.4.1.2 Future II (Futurum Exactum/Future Perfect)............................23
1.3.3.4.2 The pluperfect.........................................................................................24
1.3.3.4.3 Present Perfect ........................................................................................25
1.3.3.4.3.1 The form of the Present Perfect ...................................................25
1.3.3.4.3.2 The meaning of the present perfect .............................................26
1.3.3.4.4 The conditional mood............................................................................27
1.3.3.5 Participial forms in Old Church Slavonic..................................................28
1.3.3.5.1 The l-participle ........................................................................................30
1.3.3.5.1.1 The meaning of the l-participle.....................................................30
1.3.3.5.1.2 Development of the l-participle....................................................31
1.3.3.5.1.3 L-participles versus l-adjectives.....................................................32
1.3.3.5.1.4 Grammatical properties of the l-participle ..................................32
1.3.3.5.2 The auxiliary / copula ‘to be’ ................................................................33
1.3.3.5.2.1 Tense.................................................................................................33
1.3.3.5.2.2 Aspect...............................................................................................34
1.3.3.5.2.3 Auxiliary vs copula..........................................................................34
1.3.4 Development of the tenses in Modern Slavic languages..............................35
1.3.4.1 Simple past tenses .........................................................................................37
1.3.4.1.1 East and West Slavic ..............................................................................37
1.3.4.1.2 South Slavic .............................................................................................37
1.3.4.2 Present perfect...............................................................................................39
1.3.4.2.1 East Slavic................................................................................................39
1.3.4.2.2 West Slavic...............................................................................................40
1.3.4.2.2.1 Changes to the auxiliary in Polish.................................................40
1.3.4.2.2.2 Changes to the present perfect in Polish .....................................42
1.3.4.2.2.3 The auxiliary versus copula distinction in Czech........................45
1.3.4.2.3 A reanalysis of the present perfect in Serbo-Croatian .......................47
1.3.4.3 The pluperfect ...............................................................................................48
1.3.4.3.1 South Slavic .............................................................................................48
1.3.4.3.2 West and East Slavic ..............................................................................49
1.3.4.4 Ways of expressing the future .....................................................................49
1.3.4.4.1 South Slavic .............................................................................................50
1.3.4.4.2 West and East Slavic ..............................................................................52
1.3.4.4.2.1 Compound future tense forms......................................................52
1.3.4.4.2.2 Expressing the future with perfective verbs................................53
1.3.4.5 New types of compound tenses in Slavic ..................................................54
1.3.4.5.1 The renarrated mood .............................................................................54
1.3.4.5.1.1 Bulgarian and Macedonian.............................................................54
1.3.4.5.1.2 Beyond Bulgarian and Macedonian ..............................................55
1.3.4.5.2 ‘Have’-perfects ........................................................................................56
1.3.4.5.2.1 Macedonian......................................................................................56
1.3.4.5.2.2 Beyond Macedonian .......................................................................58
1.4 Conclusions ............................................................................................... 59
CHAPTER 2 THE SYNTAX OF 'BE'-PERFECTS AND THE L-

PARTICIPLE .................................................................................................... 61
2.1 Introduction............................................................................................... 61
2.2 Head-movement accounts of participle fronting in Bulgarian and Serbo-

Croatian ................................................................................................................. 62
2.2.1 Participle fronting as long head movement ...................................................63
2.2.2 Participle fronting as head adjunction ............................................................65
2.3 Towards an alternative analysis ................................................................ 67

2.3.1 Properties of the l-participle.............................................................................67
2.3.2 Participle fronting as locative inversion..........................................................68
2.3.3 Details of the present analysis..........................................................................71
2.3.3.1 Movement of adjuncts..................................................................................71
2.3.3.2 Movement of internal arguments................................................................72
2.3.3.2.1.1 Object shift requirement................................................................73
2.3.3.2.1.2 Direct object shift ...........................................................................73
2.3.3.2.1.3 Indirect object shift.........................................................................75
2.3.3.2.1.4 Movement of the indirect object PPs in Bulgarian ....................76
2.3.3.2.1.5 Movement of PP- and small clause complements......................79
2.3.4 Word order in the Slavic left periphery ..........................................................81
2.3.4.1 Two types of l-participle fronting ...............................................................81
2.3.4.2 Constituents in the left periphery ...............................................................84
2.3.4.3 Is it a case of stylistic fronting? ...................................................................86
2.3.4.4 Focus and word order ..................................................................................88
2.3.5 L-participle fronting as XP movement - elaboration....................................92
2.3.5.1 Auxiliary-participle adjacency ......................................................................92
2.3.5.1.1 Short verb/participle movement in Serbo-Croatian..........................94
2.3.5.1.2 Short verb/participle movement in Bulgarian....................................95
2.3.5.2 Double participle constructions..................................................................98
2.3.5.3 The future auxiliary šte in Bulgarian..........................................................102
2.3.6 Interaction between participle movement and negation............................103
2.3.6.1 Introduction.................................................................................................105
2.3.6.2 Serbo-Croatian ............................................................................................107
2.3.6.2.1 Properties of negation..........................................................................107
2.3.6.2.2 Negation and participle movement....................................................109
2.3.6.3 Bulgarian ......................................................................................................114
2.3.6.3.1 Properties of negation..........................................................................114
2.3.6.3.2 Negation and participle movement....................................................116
2.3.6.3.3 Interaction between the interrogative complementizer li and
negation in Bulgarian ...............................................................................................118
2.4 Conclusions ..............................................................................................123
CHAPTER 3 THE SYNTAX OF ‘HAVE’-PERFECTS AND PASSIVE

PARTICIPLES ...................................................................................................125
3.1 Introduction..............................................................................................125
3.2 A typology of passive participles..............................................................125
3.3 Generalizations concerning ‘have’-perfects.............................................128
3.4 Properties of ‘have’-perfects in Slavic ......................................................129

3.4.1 Kashubian.........................................................................................................129
3.4.2 Macedonian ......................................................................................................133
3.4.2.1 Properties of the construction...................................................................134
3.4.2.2 Participle fronting across the auxiliary ‘have’ ..........................................136
3.5 Impersonal participles in Polish ..............................................................142

3.5.1 Properties of the impersonal participles.......................................................142
3.5.2 Historical development of the impersonal participles................................146
3.6 Grammaticalization of ‘have’-perfects .....................................................147

3.6.1 Romance ...........................................................................................................148
3.6.2 Germanic ..........................................................................................................151
3.6.3 Slavic .................................................................................................................153
3.6.3.1 Properties of the stative perfect ................................................................154
3.6.3.2 Towards an analysis ....................................................................................157
3.7 Conclusions ..............................................................................................160
CHAPTER 4 CLITICS IN SOUTH SLAVIC ................................................163
4.1 Introduction..............................................................................................163
4.2 What are clitics?........................................................................................163
4.3 Diachrony of clitics in Slavic....................................................................164

4.3.1 Reduction of the copula/auxiliary ‘to be’.....................................................165
4.3.2 Reduction of the pronominal clitics..............................................................165
4.3.3 Changes in the clitic positions .......................................................................166
4.3.3.1 Wackernagel’s law .......................................................................................166
4.3.3.2 Violations of Wackernagel’s law in Old Church Slavonic.....................167
4.4 Clitic positions in contemporary South Slavic languages .......................170

4.4.1 Serbo-Croatian .................................................................................................171
4.4.1.1 The clitic paradigms....................................................................................171
4.4.1.2 Properties and positions of the Serbo-Croatian clitics...........................173
4.4.1.3 The second position effect in Serbo-Croatian: syntax or phonology?.178
4.4.1.3.1 Apparent splitting of syntactic constituents by clitics in Serbo-
Croatian .................................................................................................................178
4.4.1.3.2 Motivation for the second position requirement .............................180
4.4.2 Macedonian ......................................................................................................182
4.4.2.2 Clitic doubling .............................................................................................184
4.4.2.3 Direction of cliticization in Macedonian .................................................185
4.4.2.3.1 Proclisis ..................................................................................................185
4.4.2.3.2 Enclisis ...................................................................................................186
4.4.2.3.3 Postposition...........................................................................................186
4.4.2.3.4 The relation between types of cliticization and stress assignment.188
4.4.2.4 Towards an analysis of cliticization in Macedonian ...............................190
4.4.2.4.1 A note on previous scholarship..........................................................191
4.4.2.4.2 An alternative analysis of cliticization in Macedonian .....................193
4.4.2.4.2.1 Proclisis with case-assigning verbs..............................................193
4.4.2.4.2.2 Explaining the Person Case Constraint .....................................197
4.4.2.4.2.3 Enclisis with imperatives and gerunds .......................................203
4.4.2.4.2.4 Patterns with non-case-assigning categories..............................206
4.4.2.4.2.4.1 Cardinaletti and Starke’s (1999) division of deficient
elements .................................................................................................206
4.4.2.4.2.4.2 Weak and clitic forms in Macedonian ................................206
4.4.3 Bulgarian ...........................................................................................................208
4.4.3.2 Properties and positions of the Bulgarian clitics.....................................209
4.4.3.3 Clitic doubling in Bulgarian .......................................................................211
4.4.3.4 Towards an analysis of cliticization in Bulgarian ....................................212
4.4.3.4.1 Clitics are in head positions.................................................................213
4.4.3.4.2 Clitics are in specifier positions ..........................................................214
4.4.3.4.3 An alternative analysis..........................................................................215
4.4.4 Concluding remarks ........................................................................................218
4.4.4.1 Ordering of the clitics.................................................................................218
4.4.4.1.1 Bošković (2001, in press).....................................................................218
4.4.4.1.2 Tomić (1996a) .......................................................................................220
4.4.4.1.3 Some alternative suggestions...............................................................220
4.4.4.2 Why do clitics climb?..................................................................................221
CHAPTER 5 THE STATUS OF CLITICS AND COMPOUND TENSES IN

POLISH .................................................................................................. 223
5.1 Introduction............................................................................................. 223
5.2 The pronominal forms ............................................................................ 223

5.2.1 The paradigm of pronouns ............................................................................223
5.2.2 The positions occupied by the pronominal forms......................................224
5.3 The auxiliary forms and the syntax of compound tenses....................... 228
5.3.1 The paradigm of the auxiliaries......................................................................228
5.3.2 Positions of the auxiliaries..............................................................................230
5.3.2.1 Auxiliaries as clitics .....................................................................................232
5.3.2.1.1 Placement of the auxiliary clitic – syntax or phonology? ................232
5.3.2.1.2 Prosodic and morphological interactions between the auxiliary and
its host .................................................................................................................234
5.3.2.2 Auxiliaries as affixes....................................................................................235
5.3.3 Some previous accounts of the syntax of compound tenses in Polish ....238
5.3.3.1 Borsley & Rivero (1994).............................................................................239
5.3.3.2 Bošković (1997)...........................................................................................243
5.3.3.3 Szczegielniak (1997)....................................................................................244
5.3.3.4 Witkoś (1998) ..............................................................................................246
5.3.4 An alternative account ....................................................................................250
5.3.4.1 Auxiliary encliticization ..............................................................................253
5.3.4.1.1 The conditional / subjunctive auxiliary by.........................................253
5.3.4.1.1.1 The position of by.........................................................................253
5.3.4.1.1.2 The meanings of by ......................................................................254
5.3.4.1.2 Topicalization........................................................................................259
5.3.4.1.3 Że-support..............................................................................................262
5.3.4.2 Auxiliary affixation on the l-participle......................................................266
5.3.4.2.1 Participle movement in Polish and in South Slavic .........................266
5.3.4.2.2 Position and types of negation in Polish ...........................................267
5.3.4.2.2.1 Position of sentential and constituent negation........................268
5.3.4.2.2.2 Prosodic properties of negation..................................................270
5.3.4.2.2.3 Genitive of negation .....................................................................270
5.3.4.2.3 Negation and types of verbs ...............................................................272
5.3.4.2.4 Position and types of the auxiliaries...................................................274
5.3.4.2.4.1 Syntactic properties of functional and lexical auxiliaries .........274
5.3.4.2.4.2 Semantic properties of functional and lexical auxiliaries .........279
5.3.4.2.5 Why is XP-movement of the l-participle impossible in Polish?.....281
5.3.4.2.5.1 An excursus on participle fronting in Czech.............................281
5.3.4.2.5.2 The solution...................................................................................283
5.4 Conclusions ............................................................................................. 285
SUMMARY ......................................................................................................... 287
REFERENCES .................................................................................................. 289
SAMENVATTING IN HET NEDERLANDS................................................. 305

Acknowledgments
Writing this dissertation would not be possible without the presence and assistance of a
number of people. I am happy to acknowledge my gratitude to them here.
First of all, I would like to thank my supervisor Henk van Riemsdijk for introducing me
to the Models of Grammar Group, his support, fairness, and wise advice. I have
benefited a lot from his inspiring comments and guidance on how high quality linguistic
research should be done. I am also grateful for his valuable suggestions on non-
linguistic pleasures of life, such as wine tasting and travelling.
I would also like to express my appreciation to my daily supervisor Hans Broekhuis for
the courage to supervise a thesis on Slavic languages, enthusiasm, and always having the
time to discuss my ideas. His eagerness to provide me with meticulous comments on
earlier versions of the chapters (and then on comments on the comments) has been
admirable, and certainly contributed to a clearer presentation of my ideas.
Moreover, I thank Ad Backus, Bożena Rozwadowska, Maaike Schoorlemmer, Craig

Thiersch, and Olga Tomić for agreeing to join my ph.d. assessment committee. Special
thanks are due to Maaike Schoorlemmer for calling me with very detailed remarks on
the manuscript a few days before its submission.
Going back a few years in time, I would like to stress the importance of Professor
Bożena Rozwadowska, who was my first syntax teacher at Wrocław University. I thank
her for introducing me to the world of modern linguistics, enthusiasm, and firm
support over time. I am glad that she is a member of my ph.d. committee, and that she
will be in Tilburg during my defence.
Between the years of 2000 and 2001 I had the privilege to study at the University of
Tromsø, and I thank the linguists I met there, in particular Tarald Taraldsen and
Anders Holmberg. Life in the subarctic was fun thanks to my friends Patrycja
Jabłońska, Svetoslav Marinov, Terje Wagener, and Barbara Gąsior-Chrzan.
It is difficult to write a dissertation that deals with languages that one speaks little or not
at all. I thank the following people, who agreed to act as language informants: Svetoslav
Marinov, Polja Vitkova, Iliyana Krapova, Mariana Lambova, Yovka Tisheva, Mila
Tasseva-Kurktchieva (Bulgarian), Ljupka Grujoska, Olga Tomić, Svetlana Spasovska
(Macedonian), Patrycja Jabłońska, Dorota Klimek (Polish), Nataša Milićević, Boban
Arsenijević, Tanja Milićev, Olga Tomić, Monika Bašić, Amela Čamdžić (Serbo-
Croatian), Jakub Dotlačil (Czech), Nino Grillo (Italian), and the participants in the
Nasze Kaszuby internet forum, who provided Kashubian data.
I owe special thanks to fellow linguists, who gave me much needed feedback. Željko
Bošković kindly commented on parts of this thesis, and I am very happy that he will be
my postdoc supervisor at the University of Connecticut in the near future. Alexandra
Cornilescu was of great help with sorting out intriguing patterns of Macedonian
cliticization. Olga Tomić served as an authority on data from different Balkan
languages; Ben Hermans was the phonology guru, while Ewa Willim and Dorota
Klimek discussed certain aspects of Polish syntax with me. I also acknowledge the input
xii
I received from Bożena Cetnarowska, Steven Franks, Gilbert C. Rappaport, Giuliana

Giusti, Tarald Taraldsen, participants in numerous conferences, LOT and EGG-
schools, and many other people I may have forgotten to mention here. I am also
grateful to Basia Tomaszewicz for making the cover illustration, and to Anne
Breitbarth, Hans Broekhuis, and Henk van Riemsdijk for translating the summary into
Dutch.
On a personal note, I thank the current models of grammar: Anne Breitbarth, Jutta
Hartmann, Vera Hegedűs, and Nataša Milićević for the fun we had together in different
offices, buildings, and outside. Tilburg is not the most charming city in the
Netherlands, but a number of people have made it a pleasant place to live by their
presence here or away. For this I would like to thank Bertjan Busser, Pim Derks,
Mihaela Dogaru, Magda Golędzinowska, Yann Girard, Ljupka Grujoska, Riny
Huybregts, Patrycja Jabłońska, Małgorzata Jedynak, Anna Kijak, Joke Klene, Dorota
Klimek, Maja Lubańska, José Magalhães, Svetoslav Marinov, Gertjan Meijering, Tanja
Milićev, Anna Młynarczyk, Piroska Lendvai, László Molnárfi, Grzegorz Pawlina, Meli
Poenaru, Stefan Schevers, Craig Thiersch, Grzegorz Trojanowski, Basia Tomaszewicz,
Luis Vicente, Raoul Wijnands, and Ewa Willim.
Most of all, I would like to thank my family in Poland: my sister Ela and her husband
Piotrek, my brother Tomek, and my parents Alina and Ryszard Migdalscy, to whom I
dedicate this dissertation. I would have never achieved anything without their love and
support.
Abbreviations
ACC accusative
AOR aorist
AUX auxiliary
Bg Bulgarian
COND conditional
CL clitic
COM comparative
Csb Kashubian
EMPH emphatic
F feminine
FOC focus particle
FUT future auxiliary
GEN genitive
GER gerundive/gerund
IMP imperfectum (tense)
IMPF imperfective (aspect)
IMPV imperative
INF infinitive
INST instrumental
IPT impersonal participle (in Polish)
M masculine
Mac Macedonian
N neuter
NOM nominative
NV non-virile
OCS Old Church Slavonic
PAP present active participle
PARTIT partitive
PASS present passive participle
PAST.PAP past active participle
PAST.PASS past passive participle
PAST past tense
PART the l-participle
PF synthetic perfect (in Latin)
Pl Polish
PL plural
PLQPF pluperfect
POSS possessive
PRAP present active participle
PRES present tense
PRF perfective (aspect)
PRON pronoun
PROX proximate (demonstrative, in Macedonian)
PTC particle
PTP past participle
xiv Abbreviations
Q the question/focus particle li

REFL reflexive
Rus Russian
S-C Serbo-Croatian
SUBJ subjunctive
VIR virile
VOC vocative
Introduction
It has often been noted that Slavic languages show considerable resemblance in their
lexicon and morphology. However, their syntactic structures differ to a large degree.
This dissertation is concerned with one facet of the syntactic variation. It examines the
syntax of compound tenses in Slavic, taking into account their diachronic development
from Proto-Indo-European via Old Church Slavonic to the current stages. The two
research questions it seeks to explore are what structure compound tenses have in
general, and how and why they differ in Slavic languages.
1. Theoretical embedding
The analysis is couched in the framework of generative linguistics; in particular it

follows its recent version that is currently being developed as the Minimalist Program
(cf. Chomsky 1995 and subsequent work). One of the cornerstones of this framework
is the observation that children attain grammatical competence very rapidly and to the
utmost perfection even though the linguistic data they are exposed to are fragmentary
and seem insufficient to acquire the level of complexity represented by natural
language. This paradox is explained through the existence of Universal Grammar, which
is understood as a set of basic, biologically inherited principles that condition the way
language-specific Grammars are constructed. These principles are parameterized, so
that each language may set their values differently. However, given the speed at which
language is attained, these rules must be very simple and general, while the parameter
values must be deducible from very limited language input. Thus, the task of the
generative linguist is to study language data and to try to capture them into simple and
insightful generalizations.
The data investigated in this dissertation cover compound tenses in a selection of
Slavic languages. Compound tenses are structures formed with at least two verbs: the
auxiliary and the main verb. Auxiliaries differ from other types of verbs in several
respects.1 They form a closed class of functional categories, which in many languages
show distinct morphological and phonological properties. For example, the auxiliary be
is the only verb in English with a suppletive paradigm. The peculiar morphosyntactic
properties are especially evident in the case of the auxiliaries which have lexical
counterparts. For instance, in some varieties of English only the auxiliary have, but not
the main verb have, can be negated or undergo subject inversion.
The distribution of auxiliaries in the clause structure is also quite restricted. They
are always very selective about the category of their complements, as normally they
accept only verbs of a specific category. For instance, the auxiliary have in English is
always complemented by the past participle. Moreover, the type of auxiliary may
influence the morphological form of its complement. Thus, past participles in many
Romance and Germanic languages agree in φ-features with the subject when they are
selected by the auxiliary ‘be’, but not ‘have’.
The semantic function of auxiliaries is rather limited, as they merely express
grammatical properties of predicates, such as tense, aspect or mood. In this way they
perform a role akin to that of inflectional morphology on the verb. In some more
radical accounts, auxiliaries are claimed to be devoid of any semantic value. For
1See Barbiers and Sybesma (2004) for a recent overview of the properties of auxiliaries, which
was also used in this introduction.
2 Introduction
instance, Chomsky (1993) argues that all auxiliaries are uninterpretable at LF. Likewise,
Emonds (2000) states that they are lexicalized post-syntactically (that is, at PF), because
they encode only formal features, which do not play any role at LF. This might indeed
be true of some of them, such as the spurious do in English or the copula be, but in
general this view seems too strong, given that some auxiliaries, such as modal verbs,
clearly contribute to the clause interpretation (cf. Barbiers and Sybesma 2004).
According to Pollock (1989: 385ff), auxiliaries do not assign theta roles to the
constituents that they are subcategorized for. On the basis of data from English and
French, he relates this property to syntactic movement, which in his opinion is only
available to verbs that are not theta role assigners, but it is not really clear why this
should be so. This is not a widely-accepted view, though, and this dissertation follows
the proposals due to Hoekstra (1984, 1986), Roberts (1987), and Broekhuis and Van
Dijk (1995), who argue that have is a transitive auxiliary, whereas be is an unaccusative
auxiliary. The former introduces an agent and assigns accusative case to the object,
while the latter is unable to perform these functions, on a par with other unaccusative
verbs.
The relation of auxiliaries with respect to other verbs has been a matter of a long
debate in generative grammar. On the one hand, Chomsky (1965) proposed that
auxiliaries are categorially different from the main verb, therefore they do not project a
separate VP, but rather they are extensions of the VP template. On the other hand,
Ross (1969) claimed that auxiliaries have the same functions as main verbs;
consequently, they project their own VP and their own maximal clausal projection. In
the current framework these two lines of reasoning correspond, respectively, to a
monoclausal versus bi-clausal interpretation of the compound tense structure. This
dissertation will not tackle this issue (see Erb 2001, Julien 2001, Van Riemsdijk 2002,
and Breitbarth 2005 for some discussion), and I will assume that in Slavic compound
tense constructions are uniformly monoclausal, with the participle projecting the lexical
head Part, which is dominated by several functional projections up to TP.
Finally, it is important to point out that auxiliaries differ from main verbs also in
their phonological properties. For instance, in English their onset and the nucleus may
be eliminated, as in (1).
(1) If I’d known you’re coming, I’d have baked a cake
Correspondingly, the perfect auxiliaries in Slavic are often clitics, and their phonological
requirements are reflected in the syntactic patterns of compound tenses. Moreover, it
will be shown that the phonological reduction of certain verbs indicates the reanalysis
of a lexical verb as an auxiliary.
The status of the past participle, which constitutes the main verb in compound
tenses, is subject to equal controversy. In contrast to auxiliaries, participles form an
open class of lexical items to which new members may be added, and are as rich in their
semantic content as other lexical verbs. However, their categorial status is far from
clear. In languages with overt agreement morphology, participles are specified for
number and gender, on a par with adjectives, but not for person, in contrast to finite
verbs. They can also appear in the contexts which are inaccessible for verbs; for
instance, they may modify nouns (e.g. the forgotten story vs. *the forget story), or appear with
linking verbs, such as remain (e.g. the door remained closed vs *the door remained close).
Moreover, participles occur in many types of constructions. That is, unlike
auxiliaries, whose distribution is limited to compound tenses, participles perform a
variety of functions. For instance, in many Indo-European languages the same type of
Introduction 3
participle that is used in compound tenses as an “active” past participle appears in

passive structures, as in I have forgotten this story and a forgotten story. This fact has given rise
to the idea, most recently explicitly verbalized by Emonds (2000 ch. 5), of a uniform
categorial treatment of past and passive participles. Even though this proposal is not
unproblematic, because unaccusative verbs such as arrive may occur as perfect
participles but never as passive participles, I will demonstrate that it is basically correct
and receives more support from Slavic data. I will argue that just as passive participles,
past participles are unable to project an external theta role and assign structural case,
but through the process of grammaticalization of the compound tense formed with the
auxiliary ‘have’, the participle increases its verbiness and may be formed from a wider
spectrum of verbs.
2. The compound tenses in Slavic
Most of the generalizations concerning the nature of participles and auxiliaries have
been made in the generative literature on the basis of Romance and Germanic data. The
Slavic languages have not received much attention, even though they use compound
tense constructions that are not found in many other Indo-European language groups.
They also exhibit a wide range of morphologically and functionally diversified
participles. Thus, their examination may lead to a verification of some assumptions
concerning the properties of compound tenses.
The Slavic languages have developed a compound tense which is formed with the
verb ‘be’ as the exclusive auxiliary in all contexts, irrespectively of the transitivity of the
main verb. This is a rare pattern outside Slavic. In Germanic and Romance languages it
is found only in the dialect of Terracina (Italo-Romance) and Shetlandic (a variety of
Scots English, cf. Bentley and Eythórsson 2004). In other Germanic and Romance
languages the verb ‘be’ is selected as the auxiliary only in unaccusative and passive
structures; that is, when the subject is an internal argument of the verb.
In Slavic, the auxiliary is accompanied by the so-called “l-participle”, which is used
as the main verb (cf. 2a). In contrast to the Germanic and Romance languages, the
participle in the compound tense is morphologically different than in the passive
construction. As (2b) shows, the l-participle may never be used as the passive participle.
(2) a. Ivan e čel knigata

Ivan bePRES.3SG readPART.M.SG book-the
“Ivan has read/been reading the book”
b. Knigata e četana/*čela ot Ivan
book-theF bePRES.3SG readPASS.F.SG/readPART.F.SG by Ivan
“The book is being read by Ivan” (Bg)
The l-participle is also not a past participle, because in some Slavic languages it is used
to express future meanings, as shown in (3a) for Polish and in (3b) for Serbo-Croatian.
Example (3b) represents the so-called Future II construction.
(3) a. Jan będzie pisał list

Jan bePRF.1SG writePART.M.SG letterACC
“Jan will be writing a letter” (Pl)
4 Introduction
b. Kad budemo govorili s Marijom…

when bePRF.1PL speakPART.PL with Marija
“When/if we speak with Marija…” (S-C)
The l-participle renders the aspectual meaning of the predicate. Thus, the form čel in
(2a) is specified for the imperfective aspect. It can also appear with an aspectual prefix,
such as pro- in (4), which characterizes perfective meaning.
(4) a. Ivan e pročel knigata

“Ivan has read the book” (Bg)
However, the auxiliary ‘be’ shows aspectual distinctions as well. For instance, when it is
used in the imperfective aspect in Old Church Slavonic (cf. běaxõ in 5a), the complex
tense is interpreted as the pluperfect. When the verb ‘be’ occurs in the perfective (cf.
bõdemъ in 5b), it gives rise to the future perfect interpretation. Note that both of the
examples in (5) contain the l-participle as the main verb. This indicates that the
temporal interpretation of these sentences depends on the form of the auxiliary, rather
than the participle. This fact casts a serious doubt on the idea that auxiliaries do not
have any semantic value.
(5) a. Mъnoзi že ot ijudei běaxõ prišьlo kъ Martě

many FOC from Jews beIMP.3PL comePART.SG.N to Martha
i Marii da utěšętъ i
and Mary to comfort them
“And many of the Jews had come to Martha and Mary to comfort them”
(OCS, Lunt 1974: 98)
b. … vъskõjõ sę i rodili bõdemъ
why even bearPART.PL bePRF.1PL
“Why will we have been born?” (OCS, Schmalstieg 1983: 159)
Diachronically, the l-participle is a Slavic innovation. It derives from a class of Proto-

Indo-European adjectives ending in *-lo, which were completely verbalized and
reanalyzed as participles. It still has adjectival morphology, and agrees with the subject
of a clause in gender and number, but is virtually not found outside the compound
tenses. In this respect the l-participle differs from the corresponding categories in many
other Indo-European languages, which can be used as adjectives outside the compound
tense paradigm.
Previous accounts of compound tenses in Slavic did not pay attention to the
special character of the l-participle described above or to the fact that they are always
constructed with the verb be as the auxiliary. Hence, there are a number of interesting
issues that have not been addressed. For instance, how is the l-participle different from
the past participle in the Germanic and Romance languages? Why does it always agree
with the subject? What grammatical roles do the l-participle and the auxiliary ‘be’
perform in the compound tense structure? How do its properties influence the syntax
of compound tenses in Slavic?
The examination of the l-participle structures and their relation to compound
tenses found in other Indo-European languages will be facilitated by the fact that two
Slavic languages, Kashubian and Macedonian, have alongside developed a compound
tense which is structurally the same as the corresponding constructions in Germanic
Introduction 5
and Romance. As shown in (6a) for Macedonian, the auxiliary ‘have’ is accompanied by
the past participle skinato, which is morphologically the same as the passive participle
(cf. 6b). However, the past participle always occurs in an invariant, singular neuter
form, and never agrees with the subject or its complement.
(6) a. Ja imam skinato mojata kosula

herCL.ACC have1SG tearPTP.N my-the shirtF.SG
“I have torn my shirt”
b. Novoto palto mu e skinato
new-theN coatN himCL.DAT be3SG tearPASS.N
“His/her new coat is torn” (Mac)
Some other Slavic languages use a related construction exemplified in (7) for Polish.
(7) Mam już wszystkie ciasta upieczone

have already all cakesACC.NV.PL bakeACC.PASS.NV.PL
“I have already baked all the cakes” (Pl)
The principal difference between (6) and (7) is agreement in case, gender, and number
between the direct object and the participle in the latter example. Moreover, the
structure in (7) displays various lexical and aspectual restrictions on the participle.
Diachronic studies show that the construction corresponding to the one in (7) was the
source of the compound tense formed with the auxiliary ‘have’ in the Germanic and
Romance languages (the so-called ‘have’-perfect). The fact that it is found in a number
of Slavic languages at different stages of its grammatical development permits an
investigation of this diachronic process from a synchronic point of view. The questions
that will be posed in relation to the ‘have’-perfect will include the way its syntactic
configuration becomes modified with its grammaticalization. The most evident
indication of the process is the loss of the object agreement on the participle. How and
why is it lost? Does the grammaticalization of the ‘have’-perfect imply a categorial shift
of the auxiliary or the participle? Does it reverse the inability of the participle to assign
structural case and a theta role? How does it influence the status of the auxiliary ‘have’?
And, on a more general level, what is the structure of the ‘have’-perfect in comparison
to the ‘be’-perfect? What is the function of the auxiliaries and the participles in the
respective constructions?
Apart from looking at the structural properties of compound tenses, the
dissertation also investigates typological differences in a number of Slavic languages. All
of them originate from a common ancestor, but the inventory and the structure of their
tense systems have been considerably diversified throughout history. It seems that most
studies of language change focus on external sources of linguistic variation, which arise
due to contacts among speakers of different dialects. An issue that is investigated less
often is how internal properties of a language may trigger a modification of its
grammar. In the case at hand, I will examine the tense and aspect system in Old Church
Slavonic (that is, the oldest written variant of Slavic), which according to some studies
(cf. Hewson & Bubenik 1997) was imbalanced because of an overlap in marking
aspectual distinctions by both aspectual morphology and aspectual past tenses. Due to
this overlap, the whole system was uneconomical, unstable and hence prone to
simplification. The simplification occurred either via a semantic reanalysis, which gave
rise to new meanings of semantically superfluous constructions, or through a
6 Introduction
morphological reduction of certain compound tense structures. It will be demonstrated

that the morphological reduction is directly reflected in syntax.
3. Organization of the thesis
The outline of the thesis is as follows. Chapter 1 is a detailed introduction to the tense
and aspect system that Old Church Slavonic inherited from Proto-Indo-European. It
discusses the simple and compound tense forms found in this language, and shows how
they are represented in the language groups that subsequently evolved. The
presentation is very detailed, because I believe that it is crucial for any linguist to
ground his/her analysis on solid, crosslinguistic data in order to be able to make valid
generalizations. The analysis will help the reader understand the complexity of the tense
and aspect distinctions in Slavic as well as the sources of structural differences among
these languages.
Chapter 2 analyzes the compound tenses formed with auxiliary ‘be’ and the l-
participle in Bulgarian and Serbo-Croatian. In particular, it examines properties of the l-
participle and argues that in contrast to the past participle in Germanic and Romance, it
is able to assign structural case and project an external theta role. These assumptions
are used to make specific claims about the structure of the VP in Slavic and are applied
in the analysis of the widely-discussed l-participle fronting across the auxiliary to the
clause-initial position. Contrary to all previous accounts, which advocate a head
movement approach via “Long Head Movement” (cf. Lema and Rivero 1989; Rivero
1994a) or head adjunction (Wilder and Ćavar 1997; Bošković 1997), it is argued that the
l-participle undergoes remnant XP-movement and lands in Spec, TP to check the φ-
features of T.
Chapter 3 is devoted to an analysis of the compound tense formed with the
auxiliary ‘have’ and the past participle. As was mentioned above, this is the default
compound tense in Germanic and Romance, but among the Slavic languages, it has
been fully developed only in Kashubian and Macedonian. Some other Slavic languages,
such as Polish and Czech, use non-grammaticalized variants of the construction, which
permits studying its development in detail. The chapter will also investigate past
participle movement across the auxiliary ‘have’ in Macedonian and in this way verify the
claims made about the l-participle and its movement in chapter 2.
Chapter 4 explores phonological and syntactic properties of the auxiliary ‘be’,
which in Old Church Slavonic and South Slavic is a clitic. The auxiliary always clusters
with pronominal clitics, so it is necessary to examine them together. In contrast to
other elements in the clause in Slavic, clitics must rigidly appear in designated positions.
The South Slavic languages exhibit a remarkable diversity in the distribution of clitics. It
is shown that this reflects a process of language change, which involves a shift from the
second position clitics in Old Church Slavonic and Serbo-Croatian to verb-adjacent
clitics in Bulgarian and Macedonian. The change is argued to have a syntactic reflex:
second position clitics target specifier positions, whereas verb-adjacent clitics are
adjoined to T.
Chapter 5 discusses the syntax of compound tenses in Polish. In comparison with
South Slavic languages, such as Bulgarian and Serbo-Croatian, its structure has been
simplified. For instance, the auxiliary ‘be’ has been largely reanalyzed as an affix on the
l-participle. It will be demonstrated that the impoverishment of the auxiliary form has
an effect on the syntax of its compound tenses. For instance, it will be claimed that
unlike in Bulgarian and Serbo-Croatian, the l-participle is not able to undergo XP-
movement. It moves as a head and incorporates into the auxiliary.
Chapter 1 The diachrony of compound
tenses in Slavic
1.1 Introduction
The aim of this chapter is to set the scene for the analysis carried out in the subsequent
parts of the thesis and to discuss the diachronic development of compound tenses in
the Slavic languages. Even though the main part of this work is synchronic in nature,
the analysis will become more insightful by paying attention to the historical changes
that have taken place. It will be shown that the present variation in the syntax of
compound tenses in Slavic is a direct result of the different solutions adopted in each of
the languages in order to eliminate certain inconsistencies in the tense and aspect
system in Proto-Slavic and Old Church Slavonic.
Furthermore, the chapter is meant as an extensive overview of the tense and aspect
systems in the Slavic languages. Consequently, some of the topics that are discussed
here may receive scarce attention in the subsequent sections of the thesis. However,
they are included here because it is hoped that at least some of the readers will treat the
chapter as a thorough descriptive introduction to the system of compound tenses in
Slavic.
The chapter is organized as follows. After a brief typological overview of the Slavic
languages in section 1.2, it presents the emergence of tense and aspect specification in
Proto-Indo-European in section 1.3. Next, it demonstrates that Proto-Slavic inherited a
rather conservative model of expressing temporal relations, which was further modified
and elaborated over time. One of the features of the model was an excess of aspectual
marking, which was subsequently reduced in distinct ways in particular Slavic groups.
The syntactic effects of the implemented solutions will be investigated in the later parts
of the thesis. The present chapter concludes by describing the current state of affairs in
the tense and aspect system of the modern Slavic languages in section 1.3.4.
1.2 The division of the Slavic languages

The present section describes typological divisions among the Slavic languages. The
languages comprise three major groups: South, West, and East. The thesis is concerned
mainly with three South Slavic languages: Bulgarian, Macedonian, Serbo-Croatian, and
one West Slavic representative, Polish. These languages were selected on the basis of
distinct features of their compound tense structures. Since the contemporary East
Slavic languages have virtually no compound tenses any more, they will receive little
attention. For ease of reference, the map in (1) presents the distribution of the Slavic
languages. 2
2The map comes from Comrie and Corbett (2002: 2; map 1.1). I thank Routledge for granting
permission for the reproduction of the map.
8 The diachrony of compound tenses in Slavic
(1)
1.2.1 The common ancestor

All Slavic languages derive from a common ancestor, which is referred to as Proto-
Slavic. Since there are no written records of this language, it is a reconstruction based
on a comparison of grammatical forms of other Slavic and Indo-European languages.
Most likely, Proto-Slavic started to differentiate into dialects around the 6th century,
when Slavs spread into south-eastern and central Europe. However, it is usually
assumed that the unity of Proto-Slavic was finally split around the 9th-10th century, with
The division of the Slavic languages 9
the attainment of statehood by Bulgaria, Carantania, Croatia, Serbia, Bohemia, Moravia,

Pannonia, Poland, and Kievan Rus’ (cf. Schenker 2002: 114).
Old Church Slavonic was the first literary and liturgical Slavic language. Its
grammar was reconstructed on the basis of the earliest Slavic relics. The manuscripts
were never dated, but the oldest of them are assumed to originate from approximately
863 (Huntley 2002). The earliest texts were not preserved, and the few manuscripts
available come from the end of the 10th century. These are translations of Greek
ecclesiastical texts made by two monks from Salonika, Constantine (Cyril) and
Methodius. They were delegated by the Byzantine Emperor Michael the 3rd to go to
Moravia (currently part of the Czech Republic) and to translate the most important
liturgical books into the local dialect. Gradually, Methodius’s followers were moving
southwards, and established two cultural centers: in Macedonia and at the court of the
Bulgarian Tsar Simeon (893-927). After a period of growth, the Bulgarian state was
destroyed by the Byzantine armies at the end of the 10th century. The state of
Macedonia lost its independence some time later, and as a result, the language and
culture were preserved only in Croatia and some Bulgarian, Macedonian, and Serbian
monasteries. After the baptism of the ruling prince Vladimir in 988, a variant of the
Church Slavonic language evolved in Russia (Lunt 1974: 2).
Constantine and Methodius devised an alphabet for the Slavic language.
Presumably, the authors’ native dialect was South-Eeastern Macedonian, but the texts
may have been influenced by local Moravian varieties. At any rate, the manuscripts
demonstrate that the differences between the Slavic dialects, which were all used over a
very large geographical area, were still insignificantly small in the 9th century. It was only
after the year 1100 that independent descendants of Old Church Slavonic started to
differentiate into Macedonian-Church Slavonic, Serbian-Church Slavonic, and Russian
(Rus’ian)-Church Slavonic (Lunt 1974). Notably, Rus’ian texts had some distinct
features already before 1100, and that is why the language is also referred to as Old East
Slavonic (Whaley 2000b). However, as far as the tense system is concerned, Old
Russian did not differ in any fundamental way from Old Church Slavonic (cf. Van
Schooneveld 1959: 142).
1.2.2 Sources of the division

The contemporary division of Slavic languages has both historical and linguistic
motivations. Historically, it is related to the influence of the Byzantium culture and the
Orthodox religion on Bulgaria, Macedonia, Russia, Ukraine, and Byelorussia, and the
Latin/Roman culture. The current states of Croatia, the Czech Republic, Poland,
Slovakia, Slovenia, and the area of Lusatia in the eastern parts of Germany were
influenced by the Catholic or the Protestant creed. In theological or culture studies the
split is referred to as Slavia Orthodoxa versus Slavia Romana (cf. Dalewska-Greń 1997:
560), and in fact it cuts through the group of the South Slavic languages. For example,
speakers of Slovene and Croatian belong to Slavia Romana, while speakers of Serbian,
Bulgarian, and Macedonian belong to Slavia Orthodoxa. The division is also evident in
the alphabetical systems that are used. The languages of Slavia Orthodoxa are codified in
the Cyrillic script, whereas the languages of Slavia Romana use the Latin alphabet.
Let me investigate the criteria of the linguistic divisions. As far as tense marking is
concerned, all the Slavic languages use or have used the present perfect tense composed
of the resultative l-participle (cf. section 1.3.3.5.1) and the auxiliary ‘be’ (cf. section
1.3.3.5.2). In South Slavic, the auxiliary (cf. 2a) has largely the same distribution and the
same morphological forms as the copula (cf. 2b).
(2) a. Čel sŭm knigata

readPART.M.SG beAUX.PRES.1SG book-the
“I have read the book”
b. Dovolen sŭm
gladM.SG bePRES.1SG
“I am glad” (Bg)
In West Slavic, the forms of the copula and of the auxiliary diverge (cf. the discussion
of Czech in section 1.3.4.2.2); and the auxiliary has been morphologically impoverished,
most severely in Polish, where it was reanalyzed as a suffix on the l-participle (cf. 3a).
An example of a construction with a copula is given in (3b).
(3) a. Czytał-em książkę

readPART.M.SG+AUX.PRES.1SG bookACC
“I (have) read a book”
b. Jestem zadowolony
bePRES.1.SG gladM.SG
“I am glad” (Pl)
In the East Slavic languages the auxiliary has disappeared completely, and the l-
participle functions as the past tense preterite (cf. section 1.3.4.2.1).
(4) Ja čital knigu

I readPART.M.SG bookACC
“I (have) read a book” (Rus)
Apart from the compound tense based on the auxiliary and the l-participle, some of the
South Slavic languages use aspectual past tenses, the aorist and the imperfectum. They
were inherited from Old Church Slavonic, and I will characterize them in sections
1.3.3.1 and 1.3.4.1.2. Outside South Slavic, the aspectual tenses were preserved only in
Sorbian. Elsewhere, the present perfect tense formed with the l-participle took the role
of the default past tense. However, the meaning of the present perfect is undergoing
changes in Bulgarian and Macedonian, too; and it is more and more often used to
characterize events that have not been witnessed by the speaker (cf. section 1.3.4.5.1).
The Slavic languages also vary with respect to ways of expressing the future. In
South Slavic the future tense is constructed with a variant of the verb ‘want’ as the
auxiliary, which is followed by a subjunctive form of the main verb. In East and West
Slavic, the future meanings are rendered through finite verbs marked for the perfective
aspect, or with a perfective form of the auxiliary ‘to be’, followed by the infinitive or the
l-participle. More details will be given in section 1.3.4.4.
The East and West Slavic languages share some characteristics; therefore they will
sometimes be jointly referred to as the North Slavic group. Most of the South Slavic
languages are members of the Balkan Sprachbund, and as such, they share a number of
features with non-Slavic languages of the region. As far as the system of verbal
categories is concerned, the striking property of the Sprachbund is the absence of the
infinitive. Another characteristic feature is the analytic future tense marked with the
auxiliary that is a descendant of the verb ‘to want’. Lindstedt (2000a) argues that the
The Tense and Aspect system 11
Sprachbund features should also include the opposition between the aorist and the
imperfectum, because even though the tenses are of a Proto-Slavic origin, they most
presumably have been retained due to the presence of related tenses in the non-Slavic
languages of the area. Non-Slavic innovations, which arose exclusively because of
contacts with genetically unrelated languages, include have-perfects in Macedonian (cf.
section 1.3.4.5.2) as well as the renarrated mood in Bulgarian and Macedonian (cf.
section 1.3.4.5.1). Certainly, none of these features can be considered exclusively
Balkan. For example, we find some not fully developed forms of have-perfects in Czech
and Polish. Still, it is the combination of all these grammatical properties that
characterizes the Balkan Sprachbund.
Outside the tense-aspect system, the division of the Slavic languages is related to
the availability of morphological case. Most Slavic languages have seven case
morphological distinctions including vocative. The only exceptions are Bulgarian and
Macedonian, which have only preserved some case distinctions on pronouns (cf.
chapter 4). These are also the only Slavic languages that have developed the definite
article. The article is postnominal and emerged as a reduced form of the demonstrative.
1.3 The Tense and Aspect system

The following sections will investigate the tense and aspect systems in more detail. I will
begin with a brief diachronic discussion of the ways in which tense and aspect were
expressed in Proto-Indo-European in section 1.3.1. Subsequently, I will show how the
systems developed in Old Church Slavonic and contemporary Slavic languages in
sections 1.3.3 and 1.3.4, respectively.
1.3.1 The Tense System in Proto-Indo-

European
The notions of tense and aspect are both related to the concept of time. The difference
is that while tense locates the eventuality described in a clause in relation to speech time
or other points in time, such as past, present, or future, aspect is independent of other
points in time: it expresses the internal temporal organization of an event, and the
degree of its completeness. In other words, tense is a deictic category, but aspect is not.
For instance, in order to determine whether the proposition expressed by the sentence
in (5) is true or not, it is necessary to establish who the speaker is, as well as where and
when the sentence was uttered.
(5) I was in Longyearbyen last Monday
Conversely, the truth of the propositions represented by the sentences in (6) is the
same, even though they differ in aspect (Smith 1989: 108; cf. also Osawa 1999).
(6) a. Thatcher treats her Cabinet colleagues like children

b. Thatcher is treating her Cabinet colleagues like children
It is commonly assumed (cf. e.g. Lehmann 1974: 139-141, 186) that tense did not exist
as a separate grammatical category in Early Proto-Indo-European. There were no
independent verbal affixes for marking tense, and verbs were specified only for aspect.
The primary distinction was between imperfective (non-terminative, stative) versus
perfective (terminative). The distinction was marked by different morphological
endings, represented by the m-inflection for perfective forms and the h-inflection for
imperfective forms, with the paradigm given in (7).
(7) The m-inflection and the h-inflection in Early Proto-Indo-European3
m-inflection h-inflection
1SG *-m *-x
2SG *-s *-th
3SG *-t *-Ø
(Lehmann 1974: 141)
Another type of aspectual contrast was made between momentary and durative, which
was signaled by changes to the form of the root. For instance, it is hypothesized that
the durative (continuous) action was signified by an accented e vowel in the root,
whereas the roots of the verbal forms expressing momentary actions had no principal
accent and hence a Ø (null) vowel (Lehmann 1974: 186). This pattern is found in
Sanskrit and Greek, for instance in the Greek forms éleipon ’I was leaving’ and élipon ’I
left’ (Lehmann 1993: 179).
Implicitly, the aspectual marking on verbs rendered the time of an event. For
example, the events characterized by verbs in the perfective aspect were understood as
occurring in the past. As a result, time reference in Proto-Indo-European was
computed from aspectual distinctions.
The temporal relations could also be expressed by adverbs or adverbial particles.
This strategy can be observed in Sanskrit or Greek texts, where particles define the time
of action denoted by the verb. As an illustration, consider the Greek example from the
Iliad 1.70, in which the past reference of the event is indicated by the particle pró
‘before’.
(8) Hòs ēidē tá t’ eónta tá t’ essómena pró t’ eónta

who knew those PTC being those PTC will-be before PTC being
“Who knew the things happening now, those that will happen and those that
have happened?” (Ancient Greek, Lehmann 1974: 139)
Explicit expression of tense was a later innovation. The first real tense marker was the
present tense suffix -i. Presumably, it originated from an enclitic deictic particle, which
had the meaning of ‘here and now’, and was related to a corresponding locative case
ending in noun declension (Watkins 1962: 102-103). Thus, it is argued (cf. Shields 1992)
that the earliest method of expressing temporal relations was based on the opposition
between “now-here” and “not-now-here”, which was in fact deictic in nature, because it
could also refer to spatial relations.
The tense marking was slowly spreading through the entire verb system, but the
division of the concept of “not-present” into the notions of the past and the future
took place much later, in dialectal Indo-European. Consequently, three types of tenses
evolved: present, aorist, and perfect. The aorist expressed a past event, because it
referred to an action that was completed at the moment of speech. The present
3 As is customary in the literature, the asterisk marks a reconstructed form.

rendered uncompleted actions, whereas the perfect emphasized the result of an event,
and thus linked the past to the moment of speech.
The semantic interactions between tense and aspect were rather complex and gave
rise to modification of the system in particular Indo-European dialects. For example,
Albanian, Old Armenian, Baltic, and Slavic largely retained the aspectual distinctions.
Some other varieties, such as Celtic, Germanic, Latin, Late Sanskrit, and Biblical Greek
strengthened the tense markings. However, none of them developed a purely aspectual
or a purely tense system for establishing temporal relations. They always used a
combination of the two (Lehmann 1993: 181).
Proto-Indo-European did not have a future tense; the future was initially indicated
through modalities of subjective or optative forms (Schenker 2002: 94). Separate forms
for expressing the future developed only in Late-Proto-Indo-European and were found
only in some of the dialects (for example, in Sanskrit, Greek, Italic, and Baltic;
Lehmann 1974: 190).
The pluperfect tense was also a later invention. It was used to refer to actions that
took place before a narrated event. The imperfect tense (the imperfectum) emerged
rather late, too, and it had a clear aspectual flavour in the sense that it was used to
describe long-lasting or repetitive actions that were not completed. Once the
imperfectum came into existence, it contrasted with the aorist, which referred to
punctual, completed events. However, some Indo-European languages never developed
the imperfectum. For example, Germanic lacks it completely.4
1.3.2 Modifications of tense and aspect systems

in Proto-Slavic
This section discusses modifications of the Proto-Indo-European tense system within
the Slavic family. The languages of the family inherited the simple past tenses of Proto-
Indo-European, but in addition to them, they radically extended the system of aspect
marking. A major development includes the emergence of the perfect tense, which was
formed with verbal adjectives ending in *-lo reanalyzed as participles and the auxiliary
‘to be’. Furthermore, they introduced their own forms of the imperfective future and
the pluperfect.
As is well known, the Slavic languages robustly mark aspectual oppositions. The
opposition between the perfective and imperfective encompasses virtually all verbs,
both finite and nonfinite ones. Almost all verbs in the Slavic languages form aspectual
pairs. Each member of a pair describes the same kind of event, but one of them
appears in the non-perfective aspect (such as czytać ’to read’; kupować ‘to buy’ in Polish),
whereas the other member occurs in perfective aspect (such as przeczytać ‘to have read’;
kupić ‘to have bought’ in Polish). Aspectual distinctions are found even on verbal
nouns, as in the Polish examples of kupienie ‘an event of one purchase’ and kupowanie
‘an event of buying something taking place in time’ (cf. Rozwadowska 1997 ch. 3).
More examples will be provided in section 1.3.3.2; for the time being I will explain the
source of the aspect morphology in Slavic.
4 Note that the contemporary German tense called ‘imperfect’ does not express an imperfective
meaning. Germanic languages have never developed a real ‘imperfect’ tense, comparable to
‘imparfait’ in French, so the term used in German grammars may be inappropriate.
Inflected verbs in Proto-Indo-European had a three-element structure: the stem

was formed by a root followed by a thematic suffix and an inflectional ending. The
thematic suffix assigned a stem to a particular inflectional paradigm, but it could also
render aspectual information. For example, in Ancient Greek the aspectual specification
of the verb depended on the thematic suffix, underlined below. The form in (9a) is
imperfective, while the form in (9b) is perfective.
(9) a. deîk-ny-nai
“to be showing”
b. deîk-s-ai
“to have shown”
(Ancient Greek, Długosz-Kurczabowa & Dubisz 2001: 283)
The inflectional endings spelt out the inflectional categories, such as φ-features, supine
or infinitive morphology (Schenker 2002: 83). As an illustration, consider the
reconstructed paradigm of the Proto-Slavic verb *nesti ‘to carry’ (Długosz-Kurczabowa
& Dubisz 2001: 265). The first element of the verb is the root; the second one is the
thematic suffix, whereas the final element carries inflectional morphology.
(10) The paradigm of *nesti ‘to carry’ in the present tense
SINGULAR DUAL PLURAL

1 nes-ō-mь nes-e-vě nes-e-mъ
2 nes-e-šь nes-e-ta nes-e-te
3 nes-e-tь nes-e-te nes-o-nti
(Proto-Slavic, Długosz-Kurczabowa & Dubisz 2001: 265)
In Late-Proto-Indo-European, the verbal morphology became simplified, and the

majority of thematic suffixes blended with the inflectional endings. As a result, verbs
acquired a two-element structure. The modification can be observed in a later version
of the Proto-Slavic paradigm of the verb *nesti ‘to carry’ presented in (11), in which the
forms of the 1st person singular and the 3rd person plural has a two-element structure.
(11) The paradigm of *nesti ‘to carry’ in the present tense (later version)
SINGULAR DUAL PLURAL

1 nes-õ nes-e-vě nes-e-mъ
2 nes-e-šь nes-e-ta nes-e-te
3 nes-e-tь nes-e-te nes- õtъ
(Proto-Slavic, Długosz-Kurczabowa & Dubisz 2001: 265)
In Proto-Slavic the change was triggered by phonological readjustments, and it involved

the nasalization of the vowel ō when it was followed by nasal consonants, as in nes-o-nti
→ nes- õtъ in (10) and (11).
The fusion of the two verb-final elements in Late-Proto-Indo-European had also
semantic consequences. Due to the weakening of the distinction between the aspect-
marking thematic suffix and the inflectional endings, it was becoming more and more
difficult to mark the opposition between completed and ongoing events
morphologically. The change was taking place slowly, but the aspectual system of Late-
Proto-Indo-European started to show gaps.
In most of the Indo-European languages the inconsistencies were remedied

through the development of new aspectual tenses, such as the Imperfait and Passé
Simple in French. However, Proto-Slavic was in this respect the most conservative
language in the Proto-Indo-European family, because it retained the original ways of
marking aspect.
Still, the aspectual system it had inherited from Proto-Indo-European was
irregular, because sometimes there were no systematic pairs of verbs marked for
perfective and imperfective aspect throughout the whole paradigm (cf. the forms in 11).
Moreover, not all Proto-Indo-European verbs were available in both aspectual forms.
Therefore, Proto-Slavic had to reconstruct and regularize the whole verbal system.
The regularization was done by using old word-formation morphemes for the missing
member of an aspectual twin. For example, if there was a verb characterizing an on-
going event, a related verb with perfective meaning was derived from it by adding an
already existing perfectivizing suffix. Still, the formation of a consistent system of
aspectual pairs was a huge task, and the already existing perfectivizing or
imperfectivizing morphemes were not sufficient. Therefore, new morphemes had to be
coined by extracting suffixes from existing verbs. For example, there was a suffix nõ,
which was typical of a conjugation class which in the majority of cases listed verbs
characterizing completed events. The verbs from this class had imperfective equivalents
in the so-called -a- conjugation class. At some point the morpheme nõ left its own
conjugation class, and started to be used throughout the verbal system as a
perfectivizing morpheme. As a result of this spread, all the verbs containing the nõ
formant were reinterpreted as derived perfective verbs, whereas the corresponding
verbs from the -a- conjugation class were reanalyzed as basic imperfective forms (cf.
Młynarczyk 2004: 15-17; Klemensiewicz et al 1964: 242-253). The pattern can be seen
in the Old Church Slavonic examples in (12).5
(12) a. krič-a-ti
“to shoutIMP”
b. krik-nõ-ti
“to shoutPRF”
(13) a. dvig-a-ti
“to lift/carryIMP”
b. dvig-nõ-ti
“to lift/carryPRF”
Ultimately, a highly uniform system of aspectual pairs of verbs was created.

Apart from taking recourse to aspectual marking, Proto-Slavic further developed
the aspectual tenses, the aorist and the imperfectum, which had been inherited from
Proto-Indo-European. In this way aspect was doubly marked in Slavic: through the
tense forms of the aorist and the imperfectum, as well as through the
perfective/imperfective aspectual morphemes. Consequently, temporal relations in
(Proto-) Slavic were expressed via an intricate interplay of temporal and aspectual
markers. As an illustration, consider an example of an aspectual pair of the verb nesti ‘to
carry’ in Old Church Slavonic, presented in four different tenses. The forms indicate
5 Old Church Slavonic is transcribed in different ways in the literature. For consistency the
spelling of all the examples used in this dissertation follows the convention of Lysaght’s (1982)
dictionary.
that the distinction between the aspects is morphologically independent of the tense
distinctions, and that each of the tenses could appear in the perfective or the
imperfective aspect.
(14) Tense and aspect distinctions in Old Church Slavonic
TENSE/ASPECT imperfective perfective

3SG present nesetъ ponesetъ
3SG aorist nese ponese
3SG imperfect nesĕaše ponesĕaše
3SG perfect neslъ jestъ poneslъ jestъ
(OCS, cf. Van Schooneveld 1951: 97)
As far as the modifications of the Proto-Indo-European tense system are concerned,

the biggest innovation in Proto-Slavic was the replacement of the late Proto-Indo-
European synthetic form of the prefect tense with an analytic construction, which
consisted of the copula ‘to be’ and the resultative l-participle (cf. section 1.3.3.4.3 for
details).
Minor additions included the introduction of the imperfective future and the
pluperfect. The Proto-Slavic future tense was formed with an infinitive plus a finite
form of some phase verbs, such as ‘to begin’, ‘to have’, ‘to be’, or ‘to want’ (Schenker
2002: 95; cf. section 1.3.3.4.1). The pluperfect was formed with an imperfective form of
the verb ‘to be’ functioning as the auxiliary and the l-participle (cf. section 1.3.3.4.2).
1.3.3 Old Church Slavonic

This section will analyze the tense and aspect system of Old Church Slavonic. It will
describe the simple past tenses: aorist (cf. section 1.3.3.1.1) and imperfectum (cf.
section 1.3.3.1.2), and examine the ways the tenses interacted with perfective and
imperfective aspect (cf. section 1.3.3.2). Special attention will be paid to Slavic
innovations, such as the compound tense formed with the so-called l-participle and the
auxiliary ‘be’ (cf. section 1.3.3.4). The section will conclude with an overview of other
types of participles that were used in this language (cf. section 1.3.3.5).
1.3.3.1 Simple past tenses in Old Church Slavonic

I have just demonstrated that aspect is obligatorily specified on all verbs in Slavic. At
the same time, Slavic languages inherited two aspectual tenses from Proto-Indo-
European: aorist and imperfectum. Let me investigate the meanings of the past tenses
and aspects, as well as the semantic interpretations triggered by combinations of
particular types of tenses and aspects in detail.
1.3.3.1.1 The aorist

Throughout its history, Proto-Slavic had three different types of aorist formations. Two
of them, termed the ‘root’ (or ‘simple’, with suffixes added directly to the verb stem)
aorist and the ‘sigmatic’ aorist (with the tense marker s between the stem and the
suffixes) were inherited from Proto-Indo-European. The third type emerged alongside
and eventually replaced the two older variants (Lunt 1974: 90, Schenker 2002: 98) as the
only productive type, with the paradigm given in (15).
(15) The paradigm of the aorist in Old Church Slavonic
singular dual plural

1 -(o) xъ -(o)xově -(o)xomъ
2 -(e) Ø -(o)sta -(o)ste
3 -(e) Ø -(o)ste -(o)šę
(OCS, Lunt 1974: 87)
The aorist was the default, simple past tense in Proto-Slavic, which was used to refer to
actions regarded as basic in narrative texts. It related to concrete, temporally
independent events, without referring to the results of their occurrence. The actions
described could happen either once or in a series, but successive usage of the aorist did
not necessarily indicate distinct successive events. This was determined by the context.
Successiveness could be semantically marked by perfectivity.
The aorist normally denoted bounded eventualities, that is the ones with a specific
beginning and an end-point. However, it did not imply specific reference to the
duration of an event or to the stretch of time between the event and the moment of
speaking. Neither did it specify the time when an event took place or the consequences
of an action. I will contrast the meanings expressed by the aorist with the meanings
denoted by the present perfect in section 1.3.3.4.3.
Długosz-Kurczabowa and Dubisz (2001: 276-277) point out that the semantics of
the aorist underwent some changes in the history of Slavic. In Proto-Indo-European
and Proto-Slavic, the tense denoted past events, without referring to any aspectual
distinctions. With the development of the imperfectum, the present perfect, and the
pluperfect in Late Proto-Slavic, the aorist started to be used for denoting
momentaneous, completed events. An example of usage of the aorist is provided in
(17) below, where it is contrasted with the imperfectum.
1.3.3.1.2 The imperfectum

The imperfectum was inherited from Proto-Indo-European, but its morphological
realization is a Slavic innovation. It consisted of a verbal root, followed by the suffix –
ĕax, plus the suffixes of the root aorist.
(16) Paradigm of the imperfectum in Old Church Slavonic

1 -ĕaxъ -ĕaxově -ĕaxomъ
2 -ĕaše -ĕašeta -ĕašete
3 -ĕaše -ĕašete -ĕaxõ
(OCS, Lunt 1974: 86)
The imperfectum characterized actions as non-completed and emphasized the duration

or the repetition of an action (Schenker 2002: 101). Most of the time, it referred to a
background action that was happening simultaneously with some other occurrence in
the past. The other occurrence could be explicitly specified, and usually expressed by
the aorist, or understood from the context.
As an example, consider the following fragment from Savvina Kniga.
(17) oni že poslušavъše cara idošę i sĕ zvĕzda

they FOC hearPAP.NOM.PL kingACC goAOR.3PL and behold star
jõže vidĕšę na vustoče idĕaše prĕdu nimi
whichF.ACC seeAOR.3PL in eastLOC goIMP.3SG before theyINSTR
“They, having heard the king, went, and behold, the star which they saw (had
seen) in the east went before them”
(OCS, Savvina Kniga, Matthew 2.9, cf. Schmalstieg 1983: 154)
The sentence contains the verb ‘to go’ in two different tenses. The aorist form idošę
‘went’ narrates the main event, and merely denotes that the Magi left. The same verb in
the imperfectum idĕaše concentrates on the background of the action. It indicates that
the star moved before the Magi during the entire time of their journey.
1.3.3.2 Aspect in Old Church Slavonic

Old Church Slavonic distinguished three basic aspectual forms: the imperfective, the
perfective, and the retrospective.6 Their morphological forms are exemplified in the
chart in (18).
(18) The types of aspect in Old Church Slavonic
Imperfective Perfective Retrospective

Non-Past nes-õ (Pres) pri-nes-õ nes-lъ jesmь
“I carry” “I will carry-in “I have carried”
(=bring)”7
Past nes-ĕaxъ (Imp) nes-oxъ (Aor) nes-lъ bĕaxъ (Plqpf)
“I was carrying” “I carried” “I had carried”
(OCS, cf. Hewson & Bubenik 1997: 86)
The perfective aspect indicates that the event characterized by the verb is limited by
some boundary. The imperfective aspect does not denote any absolute boundary or
culmination of the action described by the verb. As was remarked earlier, virtually all
verbs in Old Church Slavonic formed aspectual pairs, one of them was perfective, and
the other one imperfective. The distinction between two members of a pair could be
morphologically expressed in the following ways (Lunt 1974: 74):
a) the use of aspectual prefixes (cf. also the form pri-nesõ in 18)
(19) a. tvoriti b. sъ-tvoriti

“to do” “to have done”
b) the difference in thematic (derivational) suffix
6 The traditional term is the “perfect” aspect. Hewson & Bubenik (1997: 13) decide to dub it
“retrospective” in order to avoid confusions due to the similarity of the terms perfect and perfective.
7 For some reason, Hewson & Bubenik (1987: 89) leave the slot for the non-past perfective form
empty, so I filled it in myself. In Old Church Slavonic, non-past perfective forms of verbs convey
future-time reference, the way they also do in the contemporary North Slavic languages (cf.
Whaley 2000a: 95 and section 1.3.4.4).
(20) a. stõp-aj-õtъ b. stõp-i-ti

“to tread” “to have trodden”
c) the difference in thematic suffix plus a modification of the root
(21) a. prašt-aj-õtъ b. prost-i-ti

“to forgive” “to have forgiven”
d) suppletion, that is the use of completely different stems (an exceptional strategy)
(22) a. glagola-ti b. rek-õtъ

“to say” “to have said”
Since imperfective aspect marks the incompleteness of the event denoted by the verb, it
is closely related to the imperfectum, which is an aspectual past tense. Importantly, the
perfective aspect (found in Old Church Slavonic, but also in Greek and expressed with
the aorist) is significantly different from the tense dubbed “perfect” in the Germanic
languages. The tense denotes the anteriority of the action with respect to the speech
time or some other past time. The aspect is a feature of an event which may be
described with a present, past, or future tense. It never refers to the moment of
speaking. In German the distinction is conveniently expressed by the pair of terms
perfektisch, which refers to tense, and perfektivisch, which refers to aspect (cf. Kuryłowicz
1964: 90ff).
The retrospective aspect did not have a separate verbal form, and was expressed
through the use of compound formations that consisted of the l-participle and the
auxiliary ‘to be’. It could relate to the past (in the present perfect, cf. section 1.3.3.4.3)
or to the future (in Future II, cf. section 1.3.3.4.1.2).
What is the difference between the perfective and retrospective aspects? It is
somewhat delicate. The perfective aspect always views an event as bounded or
completed, whereas the retrospective aspect does not necessarily do so. The latter
merely gives a retrospective view of an event, which need not be completed.
According to Hewson & Bubenik (1997), the three types of aspect also differ as to
the way the subject is positioned with respect to the event. In the imperfective aspect,
the subject is positioned “in the middle” of an event. In the perfective aspect, the agent
is presented at the very last moment of the action, in the position just before the
completion of the event. In the retrospective aspect, the agent is presented right after
the event, as immediately “external” to it, and occurring in a “resultant” state.
(23) Event time

|--------x- - - - - - - | (imperfective aspect)
|-------------------x- | (perfective aspect)
|----------------------|x (retrospective aspect)
(cf. Hewson & Bubenik 1997: 32)
In Old Church Slavonic retrospective aspect was represented analytically with finite
forms of the auxiliary to be and the l-participle. The perfective aspect was expressed
synthetically through aorist forms, which contained aspectual prefixes (cf. the chart in
18).
The important thing is that in Old Church Slavonic the two kinds of aspect could
be expressed on one verbal form. Retrospective aspect was marked via the compound
tense formed with the l-participle. However, the l-participle could at the same time
represent perfective or imperfective aspect, signified by a prefix or a derivational suffix.
As an example, compare the two Old Church Slavonic compound tense forms.
(24) a. Neslъ jesmь imperfective + retrospective

carryIMPF.PART.M.SG be1SG.PRES
“I have been carrying”
b. Poneslъ jesmь perfective + retrospective
carryPRF.PART.M.SG be1SG.PRES
“I have carried”
The variant in (24a) represents retrospective aspect, because it occurs in the compound
tense, with a finite form of the auxiliary ‘to be’ and the l-participle. However, the l-
participle is also specified for imperfective aspect. The l-participle in (24b), which
contains the aspectual prefix po, represents perfective aspect. Since the sentence is
constructed with the l-participle, at the same time it expresses retrospective aspect. A
similar contrast obtains in English, even though English does not have uniform
markers of perfective aspect. A sentence like I have been writing a letter is imperfective and
retrospective, whereas I have written down a letter is perfective and retrospective.
1.3.3.3 Some interactions between tense and aspect

Chart (14) in section 1.3.2 demonstrated that in Old Church Slavonic the aspectual
distinctions were morphologically independent of the tense distinctions. On the one
hand, almost all verbs occurred in aspectual pairs; with one member of the pair being
perfective, and the other one imperfective. On the other hand, each of the aspectually-
marked verbs could be found in either of the aspectual tenses, the aorist or the
imperfectum. How did aspect interact with tense, then?
Since the imperfectum expressed durative or uncompleted past actions which were
simultaneous with some other actions, it was used almost exclusively with verbs
specified for imperfective aspect. As far as the aorist is concerned, the situation was
slightly more complex. I pointed out in section 1.3.3.1.1 that in Late Proto-Slavic the
aorist acquired some aspectual meaning. Most of the time it rendered instantaneous
completed events, so as a rule it was used with perfective forms of verbs. However, the
aorist could also characterize a series of short-term events, or a long-lasting past event,
when it was viewed as occurring in its entirety (cf. Hewson & Bubenik 1997: 89-90).
Let me examine the interactions in more detail, starting with the “default” cases of
imperfective verbs in the imperfectum, and the perfective verbs in the aorist. In the
following translation of Luke 24.14 into Old Church Slavonic, the short-term event
described with the perfective aorist occurs in its entirety and is contrasted with an
ongoing process, which is expressed with the imperfective imperfectum.
(25) I ta besĕdovaašete kъ sebĕ o vsĕxъ sixъ… i

and converseIMPF.IMP.3PL to REFL about all this and
samъ Isusъ približi sę i idĕaše sъ ńima
himself Jesus approachPRF.AOR.3SG REFL and goIMPF.IMP.3SG with them
“And they were conversingIMPF.IMP with one another about all these things…
and Jesus approachedPRF.AOR and was going alongIMPF.IMP with them”
(OCS, Zographensis, Luke 24.14, Huntley 2002: 151)
It may seem semantically contradictory to find perfective forms of the imperfectum

tense or (somewhat less so) imperfective forms in the aorist. Yet, such cases are readily
observed in Old Church Slavonic texts. Dostál (1954: 599-600) provides ample statistics
concerning the occurrence of the imperfectum and the aorist tenses with imperfective
and perfective aspectual markings in Old Church Slavonic texts. The results of his
calculations are presented in chart (26).
(26)
Verbal Aktionsart/ Imperfective Perfective

Aspecto-Temporal Category
Imperfectum 99% 1%
Aorist 40% 60%
(Dostál 1954: 599-600)
The fact that the “unexpected” combinations of the imperfective aorist and perfective
imperfectum are found in the corpus proves the intricate nature of the Old Church
Slavonic tense system. Because of this complexity, the language could express very
intricate temporal relations, but admittedly, the system also had some weak points,
which will be discussed in section 1.3.4. For the time being, let me discuss the meanings
of the semantically unexpected combinations of tense and aspect values.
1.3.3.3.1 Imperfectum + Perfective Aspect

The combination of the imperfectum with a perfective aspect could be used to express
repeated actions of single, short-term, momentary events, such as a repeated action of
giving a single kiss in the translation of Luke 7.38.
(27) Načętъ močiti nodzĕ ego slъzami i

beginPRF.AOR.3SG batheINF footNOM/ACC.DUAL heGEN tearINST.PL and
vlasy glavy svoeję otiraaše i oblobyzaaše
hairINSTR.PL headINSTR her-ownGEN wipeIMPF.IMP.3SG and kissPRF.IMP.3SG
nodzĕ ego i mazaaše mũromь
footNOM/ACC.DUAL heGEN and anointIMPF.IMP.3SG myrrhINSTR
“She began PRF.AOR to bathe his feet with her tears and wipedIMPF.IMP [them]
with the hair of her head, and kissedPRF.IMP his feet, and anointedIMPF.IMP
[them] with myrrh” (OCS, Luke 7.38, cf. Hewson & Bubenik 1997: 90)
In this example there is one verb in the perfective aorist: načętъ ‘began’ and two
imperfective verbs in the imperfectum: otiraaše ‘was wiping’ and mazaaše ‘was anointing’.
The verb oblobyzaaše ‘kissed’ is also marked for the imperfectum, but has a perfective
aspectual form. It expresses repeated events of imprinting a single kiss. In the original
version of the text in Greek the verb is in the imperfectum.
The example in (28) contains two verbs in the perfective aspect. The perfective-
marked imperfectum characterizes the event as an ongoing process, while the perfective
aorist indicates the immediacy of the reaction.
(28) I ĕviše sę prĕdъ ńimi ĕko bledi glagoli ixъ

and appearPRF.AOR.3PL REFL before them as pale words themACC
i ne imĕaxõ imъ vĕry
and not havePRF.IMP.3PL themDAT faith
“And their words appearedPRF.AOR to them to be nonsense and they would
not believePRF.IMP them”
(OCS, Zographensis, Marianus; Luke 24:11, OCS, Huntley 2002: 151)
1.3.3.3.2 Aorist + Imperfective Aspect

The imperfective marking of aorist verbs was considerably more common, as the aorist
could be used to describe long-lasting events in their entirety. As an example, consider
the following sentence taken from Suprasliensis.
(29) Aky kъ človĕku bo besĕdova i

as to man beCOND.3.PL converseIMPF.AOR.3SG and
vъzira na ńъ
lookIMPF.AOR.3SG to him
“For he conversedIMPF.AOR with him and lookedIMPF.AOR at him as if he were a
man” (OCS, Suprasliensis 122.30, Huntley 2002: 151)
The use of the imperfective form of the aorist signified that that the event took place,
but it did not imply whether the event was completed or not.
Summarizing, it has been shown that Old Church Slavonic combined all types of
aspect and tense distinctions with each other. In the past tense the imperfective aspect
normally coincided with the imperfectum, whereas the perfective aspect with the aorist.
The combinations with contradictory aspectual values, that is the perfective
imperfectum or the imperfective aorist, are statistically less frequent (cf. the chart in
26). However, Lunt (1974: 137) points out that that this is because the situations
requiring them are not common. They occur in narrations of complex past events, but
these type of passages are scarcely attested in Old Church Slavonic. However, the fact
that they do appear indicates that tense and aspect were two independent systems.
I will present the current state of relations between tense and aspect in
contemporary Slavic languages in section 1.3.4.1.2. The subsequent section will analyze
the compound tense forms in Old Church Slavonic.
1.3.3.4 The Compound Tenses in Old Church Slavonic

The Proto-Slavic compound tenses were constructed with the (resultative) l-participle
(cf. section 1.3.3.5.1) and a present tense form of the auxiliary verb byti ‘be’. Both the l-
participle and the auxiliary could occur in either perfective or imperfective aspect. I will
investigate the tenses in turn according to the aspectual form of the auxiliary ‘be’.
1.3.3.4.1 Future tenses

Old Church Slavonic lacked a uniform future marker, so a number of different
strategies were used to render the future time. Most frequently future events were
expressed by perfective verbs in the present tense; cf. chart (18). However, the future
could also be expressed periphrastically with two types of tenses: Future I and Future
II.
1.3.3.4.1.1 Future I
Future I was rendered by combinations of the infinitival form of the main verb with a
finite perfective variant of certain verbs, usually bõdõtъ ‘to be’ (cf. 30) and imamъ ‘to
have, be destined to’, xotętъ ‘to want’ and less frequently načьnõtъ and vъčьnõtъ ‘to begin’
(Lunt 1974: 136-137). The infinitival forms occurring with these verbs could be in
either perfective or imperfective aspect (cf. stradati in 30).
(30) I mъně bõdõtъ stradati podružija i volę svojeję

and meDAT bePRF.1SG go-withoutINF.IMPF marriage and freedom REFL
“And I will have to be deprived of marriage and my [own] freedom”
(OCS, Suprasliensis 237 1-2, Whaley 2000a: 24)
Eventually, one type of the verb prevailed as the auxiliary: a descendant of bõdõtъ in
North Slavic and a descendant of xotętъ in South Slavic.
1.3.3.4.1.2 Future II (Futurum Exactum/Future Perfect)

Apart from the Future I, there was another construction for characterizing future
events in Old Church Slavonic, termed Future II. The Future II was formed with the
perfective form of the auxiliary verb ‘to be’ followed by the l-participle. The l-participle
could be either in the perfective or, more commonly, in the imperfective aspect. The
tense was used to denote future events preceding some other future event, as in the
following example from Suprasliensis 379.10.
(31) Ašte na to sъtvorimъ vladyky podražali bõdem

if for it accomplishPRES.1PL Lord imitatePART.IMPF.PL bePRF.1PL
“We will accomplish that if we [will] have beenPRF imitatingIMPF the Lord”
(OCS, Huntley 2002: 152)
Here, the l-participle in the imperfective form podražali ‘imitate’ depicts an ongoing
process, which is a condition for occurrence of another event that will take place in
more distant future.
Future II was used very rarely in Old Church Slavonic, as only seven examples
have been found. This might be accidental, because, as Whaley (2000b) points out, the
scarcity might be due to contextual properties of Old Church Slavonic relics. Since the
texts were typically past-tense narrations, they seldom required future perfect forms.
Let us investigate the semantics of Future II. In a nutshell, it renders the meaning
of “past in the future”. According to Comrie’s (1985: 69ff.) definition, the future
perfect is a relative tense which characterizes an event (E) happening before or after the
moment of speech (S) which is described from the perspective of a future reference
point (R). In Old Church Slavonic, the future orientation of the reference point (R) was
expressed by the perfective form of the verb bõdõtъ ‘to be’, and the event (E) was
represented by the l-participle (Whaley 2000b).
Importantly, Whaley (2000a: 110) argues that the Future II in Slavic did not have
to express the future reference per se, as it could also characterize irrealis or even past-
tense meanings. Rather, Whaley claims that the tense rendered “a displaced perception
of state”. What she means by this is that the future reference point (R) is not just the
“vantage point” from which the event (E) is perceived. It also marks the point at which
there occurs a change in the speaker’s knowledge about the event.
As an illustration, consider the usage of the Future II in an example taken from

Codex Suprasliensis. The fragment is about Mary, who has been recently informed by an
angel that she will give birth to Jesus, and is not certain whether she should tell Joseph
about it.
(32) I Marija vь sebĕ si razmyšljaše provĕdĕ li se Iosifu

and Mary in herself REFL ponderPRAP speak-out Q REFL JosephDAT
ili pače sьkryjõ tainoje se jeda bõdetъ sъlъgalъ prixodivyi
or hide secret REFL in case bePERF.3SG liePART.M.SG comePAP
“And Mary pondered to herself, ‘Shall I tell this to Joseph, or hide this secret,
in case the one who had come [i.e., the angel] will have lied’”
(OCS, Suprasliensis fol. 239, II. 23-25, cf. Whaley 2000a: 110)
Given the context, the future event characterized by Future II bõdetъ sъlъgalъ ‘will have
lied’ may only refer to the moment before Maria finds out whether the angel was telling
the truth or lying. That is, the event of lying takes place before Mary’s thought, while
her realization of the lie will occur in the future. Furthermore, the use of jeda ‘in case’
signifies an irrealis meaning of the clause, and it refers to Mary’s speculations.
Whaley points out that the future reference of the future perfect was of a
secondary importance. In fact, a number of its occurrences in the oldest Slavic texts
were even completely incompatible with a pure future-tense reference. Most of the time
the interpretation of Future Perfect coincided with past-tense or irrealis meanings. It
did not imply the actual reference time with respect to the moment of speech. It only
established the chronology of a described state with the perception of this state by the
speaker. Section 1.3.3.5.1 will relate these types of meaning to the general semantic
properties of the l-participle.
1.3.3.4.2 The pluperfect

The pluperfect described past events that took place before some other past events. It
was formed with the l-participle and the auxiliary ‘to be’, which appeared either in the
imperfectum or in the imperfective form of the aorist. According to Lunt (1974: 98),
the use of the auxiliary in the imperfectum (cf. 33a) implied that the past action co-
occurred with some other event, which was explicitly mentioned or just implied by the
context, whereas the aorist form of the auxiliary (cf. 33b) simply stated an independent,
past action.
(33) a. Mъnoзi že ot ijudei běaxõ prišьlo kъ Martě

many FOC from Jews beIMP.3PL comePART.SG.N to Martha
i Marii da utěšętъ i
and Mary to comfort them
“And many of the Jews had come to Martha and Mary to comfort them”
b. Ne bĕ že ne u Isusъ prišьlъ vъ vьsь
not beIMPF.AOR3SG FOC not at Jesus comePART.M.SG in town
“Now Jesus had not yet come into the town”
(OCS, J 11.19; J 11.30, Lunt 1974: 98)
It appears, however, that the meaning distinctions related to the selection of

imperfectum form of the auxiliary versus the imperfective aorist variant were not
always so clear. For example, Huntley (2002: 153) states that there were no discernable
semantic differences in the uses of the auxiliary. The imperfectum form was possibly
used less often because it is attested only in third-person forms.
1.3.3.4.3 Present Perfect

The origin of the perfect tense deserves considerably more attention, because in most
of the contemporary Slavic languages the perfect replaced the aspectual past tenses (the
aorist and the imperfectum) and became the default way of characterizing past events
(cf. section 1.3.4.2). Moreover, it was subject to the most radical modifications in the
history of Slavic languages, both with respect to its meaning and its morphological
form. The details of these modifications will be presented in section 1.3.4.
1.3.3.4.3.1 The form of the Present Perfect

In Old Church Slavonic the present perfect was formed with the imperfective form of
the auxiliary ‘to be’ in the present tense and the l-participle. The participle usually
appeared in the perfective form, but imperfective variants were also frequently found.
The participle agreed with the subject and was specified for gender and number. The
chart in (34) presents the paradigm with the imperfective form of the l-participle neslъ
‘to have been carrying’ in all potential number, gender (masculine, feminine, and neuter)
and person variations.
(34) The form of the Present Perfect in Old Church Slavonic
singular m/f/n dual m/f/n plural m/f/n

1 neslъ/a/o jesmь nesla/ě/ě jesvě nesli/y/a jesmъ
2 neslъ/a/o jesi nesla/ě/ě jesta nesli/y/a jeste
3 neslъ/a/o jestъ nesla/ě/ě jeste nesli/y/a sõtъ
(OCS, Bartula 1981: 100)
Let us investigate a few examples of its usage. In (35), the l-participle is in the perfective
form, whereas the auxiliary ‘to be’ is in the imperfective form. The event characterized
by the perfect tense is completed and precedes the speech time.
(35) Prišelъ jestъ vasilisikъ

comePART.PRF.M.SG beIMPF.PRES.3SG Basiliscus
“Basiliscus has come” (OCS, Suprasliensis 20.2, Huntley 2002: 152)
When the participle appears in the imperfective form, the completion of the described
event is undetermined. In (36), the event of lying at home may still be taking place at
the speech time.
(36) Radujõtъ sę zĕlo jako u svoixъ sõtъ

rejoiceIMPF.PRES.3PL REFL completely as at own beIMPF.PRES.3PL
si doma ležali
this home liePART.IMPF.M.PL
“They rejoiceIMPF.PRES greatly, because they have (beIMPF.PRES) been lyingIMPF at
home with their people” (OCS, Suprasliensis 267.17, Huntley 2002: 152)
1.3.3.4.3.2 The meaning of the present perfect

In Old Church Slavonic the present perfect was used to refer to an action that takes
place in the past, but whose results are significant for the time of speaking. Since the
tense expressed the resultative aspect, the action did not necessarily have to be viewed
as completed.
(37) Otrokovica nĕstъ umrьla nъ sъpitъ

damsel NEG+be3SG diePART.F.SG but sleepPRES.3SG
“The damsel is not dead, but sleepeth” (OCS, Mk 5.39, Lunt 1974: 98)
Van Schooneveld (1959: 87) argues that the perfect tense relates the result of an action
to the subject. In this way it differs from the aorist, which concentrates on the action
itself, without any reference to the result or the influence of an event on the subject.
Moreover, the perfect tense presents events with no relation to their development in
time, as this type of meaning is rendered by the imperfectum.
Schmalstieg (1983: 156) quotes an example of a biblical passage that elucidates the
semantic difference between the aorist and the present perfect. The fragment describes
the story of Jairus’ daughter (Mark 5.22-43). Jesus is approaching the president’s house,
when some messengers come and say to Jairus, using the aorist.
(38) Dъšti tvoĕ umrĕtъ

daughter your dieAOR.3SG
“Your daughter just died”
Subsequently Jesus enters the president’s house and says, using the present perfect.
(39) Nĕstъ umrъla nъ sъpitъ

not+bePRES.3SG diePART.F.SG but sleepPRES.3SG
“(The child) is not dead but asleep”
The Old Church Slavonic translator skillfully presented the event from two
perspectives. The messengers report a simple past occurrence and use the aorist. Jesus
presents the result of the past event for the present moment, and therefore uses the
present perfect form. Hewson & Bubenik (1997: 91) remark that the semantic
difference was not expressed in the original Greek text, in which the aorist was used in
both sentences.
Słoński (1926: 8) carefully investigates Old Church Slavonic translations of Greek
biblical texts and notices that as a rule the Greek perfect tense was rendered as the
aorist in the Slavic versions. Moreover, Słoński’s study indicates that the perfect tense
in Old Church Slavonic was found very rarely. He examined Codex Marianus,
Suprasliensis, and Glagolita Clozianus, compared them with the original Greek texts and
found 190 occurrences of the perfect in the original Greek relics, which were matched
by only 17 perfects in the Slavic translation. By contrast, the present perfect forms in
the Old Church Slavonic texts were most of the time translations of the aorist. It is very
difficult to pinpoint any semantic differences in the two translation strategies. Possibly,
the Old Church Slavonic perfect might have had a different meaning from the Greek
perfect. One of the differences could relate to the availability of forms marked for the
imperfective aspect in the Slavic perfect tense, which were not possible in Greek.
However, the distribution of the perfect tense in the paradigm is more revealing.
Słoński (1926: 21) observes that the perfect tense was found almost exclusively in the
2nd and the 3rd person singular. According to him, this is due to the fact that the 2nd and
the 3rd person singular forms of the aorist were phonetically the same. Therefore, the
translator would replace the aorist with appropriate forms of the perfect for clarity.
What this suggests again is that in Old Church Slavonic the meaning of perfect tense
was very close to the aorist. As will be demonstrated in section 1.3.4, this was one of
the weaknesses of the system, which gave rise to radical modifications of the tense
system in the Slavic languages.
Dostál’s (1954: 599ff.) calculations confirm that in Old Church Slavonic the
perfect tense was used rather rarely, and usually in embedded clauses. In his study of
tense usages in Old Church Slavonic relics, Dostál finds 10 thousand uses of the aorist,
2300 of the imperfectum, and approximately 600 of perfect tenses (i.e., around 5%).
The striking thing is that the tense which was the least common in Old Church
Slavonic has become the one that is most widely used in modern Slavic languages.
Lunt (1974: 98) points out that the reason for the rare occurrence of the present
perfect could be the fact that it was not necessary to render the type of temporal
relation it characterizes in the written texts from that time. However, this does not
seem a valid explanation, since there are a lot of events in the Bible that took place in
the past, but the results of which are still significant for the present time.
Another reason for the scarce use of the present perfect could be a late emergence
of this tense in Slavic, which possibly arose later than the simple past tenses. For
instance, Damborský (1967) argues that in the earliest stages of Slavic, the l-participle
was unknown. Bartula (1981: 100) observes that there are few examples of the present
perfect in the earliest Old Church Slavonic manuscripts, whereas in the more recent
ones they are found more often, especially in Codex Suprasliensis and Savvina kniga (both
from the 11th century). Therefore, the translators of biblical texts may have been
reluctant to use a novel form that did not match the archaic nature of biblical texts.
Finally, Lindstedt (1994: 33-34) observes that the present perfect could never occur
in “plot-advancing” sentences, because it was not a narrative tense in Old Church
Slavonic. This is the reason why it was typical of dialogues as well as monologues
represented by psalms and prayers.
Section 1.3.4.2 will show that the meaning of the perfect tense has undergone
radical modifications in all the contemporary Slavic languages.
1.3.3.4.4 The conditional mood

The conditional mood was constructed with the l-participle and a special conditional
variant of the verb ‘to be’ (Schmalstieg 1983: 157). Alternatively, the aorist variants of
the verb ‘to be’ could be used as the conditional auxiliary. The chart in (40) shows only
the singular and the plural forms, because the dual was not attested in the conditional
mood.
(40) The conditional auxiliaries in Old Church Slavonic
singular plural
1 bimь bimъ
2 bi biste
3 bi bišę/bõ
(OCS, Schmalstieg 1983: 157)
The conditional mood expresses irrealis meaning, as indicated in the following example.
(41) G(ospod)i ašte bi bylъ sьde nebi

Lord if beCOND.2SG bePART.M.SG here NEG+beCOND.3SG
bratrъ moi umrъlъ
brother my diePART.M.SG
“Lord, if you had been here, my brother would not have died”
(OCS, J.11.21, Schmalstieg 1983: 157)
1.3.3.5 Participial forms in Old Church Slavonic

Old Church Slavonic relics exhibit a rich participial system. The participles combine the
functions of verbs and adjectives. Along with adjectives, they are inflected for gender,
number, case, and definiteness.8 They are also specified for the verbal categories of
tense, aspect, and genus.9 However, the temporal distinctions expressed by the
participles were relative, rather than absolute: events occurring simultaneously with the
tense of the main verb were expressed by present participles, whereas the events which
are anterior to the tense of the main verb were expressed by past participles. There
were five morphologically distinct forms of the participles: present active, present
passive, past active, past passive, and the (resultative) l- participle (Lunt 1974: 139-141;
Schenker 2002: 104-106; Dostál 1954: 614ff.) The resultative participle deserves a wider
mention, therefore it will be analysed in a separate section (1.3.3.5.1). The
morphological forms of the participles are described in the chart in (42).10
8 The definiteness marking is related to two declensions of adjectives and passive participles in
Old Church Slavonic: the nominal declension (which produced the so-called “short forms”) and
the pronominal declension (which had “long forms”). The pronominal declension contained the
demonstrative pronoun j, which functioned like a postpositional definite article (cf.
Klemensiewicz et al 1964: 323-326 and Lavine 2000 ch. 3).
9 The genus distinction is concerned with the opposition between active and middle
constructions (as in, for example, He opened the door versus The door opened). It had been
morphologically distinguished through inflection in Proto-Indo-European, but Proto-Slavic lost
these distinctions. This opposition was taken over by a newly developed contrast of reflexive and
non-reflexive forms. The middle construction was rendered with the reflexive particle sę (cf.
Schenker 2002: 94).
10 The suffixes listed in the chart sometimes do not correspond to the ones exemplified in the
data below. This is because they often undergo morpho-phonological alternations in the presence
of inflectional morphemes.
(42) Types of participles in Old Church Slavonic
Participles Present Past

Active present tense stem + -õšt- or infinitival stem + -ŭš- or -vŭš-
-ęšt
Passive present-tense stem + -m- infinitival stem + -tŭ, -nŭ, or
-enŭ
Resultative (the l-participle) infinitival stem + -l suffix
The Present Active Participle denoted actions that co-occur with the events expressed
by the main verb.11
(43) Reče že pritъčõ k nimъ glagolę

speakAOR.3SG FOC parableACC.F.SG to them sayPAP.NOM.M.SG
“And he spoke a parabole to them saying...” (Luke 12:16-21)
The Present Passive Participle often rendered the meaning of a possibility, as in vidimъ
‘being seen, visible’, nerazorimъ ‘indestructible’. It could be formed only from transitive
verbs.12
(44) Nitčьsomuže bõdetъ kъ tomu da isypana bõdetъ

nothingDAT bePRF.3SG to thatDAT to castPASS.F bePRF.3SG
vъnъ i popiraema tčlovjĕky
out and treadPASS.NOM.F.SG menINST
“(It is thenceforth good for) nothing, but to be cast out, and to be trodden
under foot of men” (OCS, Matthew 5:1-13)
The Past Active Participle characterized events that started before the action denoted
by the main verb.
(45) Uzьrjęvъ že narody vьzide na gorõ

seeingPAST.PAP.NOM.M.SG FOC multitude go-upAOR.3SG to mountain
i jěko šěde pristõpišę kъ nemu utčenitsi ego
and when sitAOR.3SG comeAOR.3PL to himDAT disciples his
“And seeing the multitudes, he went up into a mountain, and when he was
set, his disciples came unto him” (OCS, Matthew 5:1-13)
The Past Passive Participle was used to describe a state which was caused by an external
agent and which begun before the reference time. It may also be employed predicatively
in passive constructions. A number of passive participles were used as adjectives, for
instance sъměrenъ ‘humble’ or učenъ ‘learned’ (Lunt 1974: 141).
11Example (43) comes from: http://www.utexas.edu/cola/depts/lrc/eieol/ocsol-1-X.html

12Examples (44) through (46) come from: http://www.utexas.edu/cola/depts/lrc/eieol/ocsol-6-
R.html
(46) Blaženi milostivii jěko ti

blessedNOM.M.PL mercifulNOM.M.PL as they
pomilovani bõdõtъ
obtain- mercyPAST.PASS.NOM.M.PL bePRF.3PL
“Blessed are the merciful, for they shall obtain mercy”
(OCS, Matthew 5:1-13)
The inventory of the participles was subsequently modified in particular Slavic

languages. All of them were inflected for case and agreed in φ-features with the subject.
The only participle that has never been marked for morphological case (that is, which
only appeared in nominative) was the resultative l-participle.
1.3.3.5.1 The l-participle

Of particular relevance for the topic of this thesis is the l-participle, which is also
known as the “resultative participle” or the “participium praeteriti (perfecti) activi II”
(Bartula 1981: 94, Schenker 2002: 106). It is formed with the suffix -l attached to the
infinitival aorist stem.13
(47) Infinitive: da-ti

giveINF
l-participle: da-lъ/-la/-lo
givePART.MSG/F.SG/N.SG
In Old Church Slavonic the l-participle was used in all compound tenses and always
occurred with a finite form of the verb ‘to be’ as the auxiliary.
1.3.3.5.1.1 The meaning of the l-participle

The semantic role of the l-participle is to denote the result of a completed or
uncompleted action. In other words, it focuses on a past, future or present state that
results from a previous action.
According to Kowalska (1976: 20) the l-participle has always been used in relative
tenses. It conveys the meaning of a ‘distance’ from the speaker’s perspective. The
distance could be related to the chronology of events in the case of future or past tense
constructions, or to probability or possibility in the case of modal/conditional
constructions.
13 In contemporary Bulgarian and Macedonian the l-participle can be derived from both aorist
and imperfectum infinitival stems, as shown in (i) and (ii), respectively.
i. Xodil sŭm
goPART.AOR.M.SG bePRES.1SG
“I have gone”/“I went, they say”
ii. Xodel sŭm
goPART.IMP.M.SG bePRES.1SG
“I was going”/“I used to go, they say” (Bg, Lindstedt 1994: 44)
The meaning of the aspectual distinctions is reflected in the translations. The second part of the
translations represents ‘the renarrated mood’, which is described in section 1.3.4.5.1.
Lunt (1952: 91) claims that the meaning of an event which is “distanced”14 can be
realized in two ways. First, the use of the l-participle may specify that the speaker did
not witness the described event; therefore responsibility for accuracy of the statement is
withheld. This is the case in “renarrated mood” constructions (cf. section 1.3.4.5.1).
Second, the l-participle may denote an eventuality that has begun or took place in the
past, and which is still relevant at the moment of the utterance. This is one of the
original meanings of Slavic perfect tenses (cf. section 1.3.3.4.3).
The temporal meaning of the l-participle is of a secondary importance. This can be
readily observed in Future II constructions (cf. section 1.3.3.4.1.2), which make use of
the l-participle. Recall that the Future II does not necessarily characterize future events,
and may frequently express irrealis or even past tense meanings.
1.3.3.5.1.2 Development of the l-participle

It is generally assumed that the l-participle derives diachronically from a group of
Proto-Indo-European verbal adjectives ending in *-lo (cf. Damborský 1967). The
adjectives signified likelihood to perform a certain action or referred to a characteristic
feature of the person involved. The *-lo forms also served as a stem of nomina agentis
(agent participles) and proper names in many Indo-European languages. Examples of
such forms include discipulus ‘student’ or legulus ‘gatherer of fallen olives’ in Latin; dêlos
‘apparent, evident’ in Ancient Greek; bitil ‘applicant’, tribal ‘cattle driver’, or Unterläufel
‘delegate’ in Old German. In Slavic they were often the base of pejorative agent
participles, such krzykała ‘a shouter’ jąkała ‘stutter’ or guzdrała ‘dawdler’ in Polish (cf.
Wojtyła-Świerzowska 1974: 103ff). Many of them became sources of surnames, either
in the adjectival or nominal form, such as Szukała ‘searcher’ (agent) / Szukalski
‘searching’ (adjectival), Pękała ‘bursting’ / Pękalski ‘someone who bursts’ or Czekała
‘awaiting’ / Czekalski ‘someone who waits’ (Damborský 1967: 126ff).
At some point, some of the *-lo adjectives were reanalyzed as participles in
compound tenses. The process occurred in three Indo-European subgroups: Armenian,
Slavic, and Tocharian, and to a lesser extent in Umbrian (only in future perfect forms)
and Indic (Middle Indo-Aryan in active perfective participles; cf. Hewson & Bubenik
1997: 74). In Slavic, they emerged as l-participles. It is remarkable that the forms found
in Armenian and Tocharian are morphologically similar to the Slavic l-participle and
that they always occur with the copula ‘be’ as well.
In East Tocharian, l is a gerundive suffix attached to the present tense or to the
subjunctive stem. The resulting forms express necessity or possibility, respectively. The
gerundives are found with a copula, and can be used attributively or predicatively. The
copula may be dropped with predicative gerundives (cf. 48).
(48) Nervvamn-oko kälale

nirvana-fruit obtainGER
“(By praising God) the fruit of nirvana is obtainable”
(East Tocharian, Krause and Werner 1960: 186-187)
The only participle that is available in Classical Armenian is formed by adding the suffix
eal to the aorist stem. The participle is indifferent to voice distinctions, because Classical
14 Friedman (1977: 36) argues against the meaning of ‘distance’. For him the l-participle renders
the opposite meaning, because if an event is relevant for the present moment, it cannot be
distant.
Armenian had no separate passive participle. As in Tocharian and Slavic, it is

accompanied by the copula.
(49) Žamanek haseal ê

time comePART is
“Time has come” (Classical Armenian, Hewson & Bubenik 1997: 75)
1.3.3.5.1.3 L-participles versus l-adjectives

Even though the l-participle originated from a class of *-lo adjectives, the process did
not involve a reanalysis of all *-lo adjectives as participles. Both l-adjectives and l-
participles exist in Slavic, although their distribution varies across the languages. They
can be distinguished from each other by means of a few criteria (cf. Damborský 1967).
As an illustration, chart (50) presents the derivations of the form wytrwał (+
inflection), ‘persist’, which in Polish is a potential stem for both an l-participle and an l-
adjective with the meaning ‘persistent’. The form wytrwał is compared with the form
odeszł (+ inflection), ‘leave’, which can be a stem for the l-participle, but does not have a
corresponding adjectival variant.
(50)
l-adjectives l-participles
gradation wytrwal -szy “more persistent” impossible: *odeszł-szy
najwytrwal-szy “the most persistent” *najodeszł-szy
modification possible, e.g. bardzo wytrwali żołnierze
impossible
with ‘very’ “very persisten soldiers”
nominalization wytrwał-ość “persistence” impossible: *odeszłość
adverb formation wytrwał-o “persistently” impossible: *odeszł-szo
case morphology available, e.g. wytrwalych żołnierzach unavailable/only in nominative
“persistentLOC soldiersLOC” case
As the chart indicates, l-adjectives are potential stems for derivation of adverbs, nouns
and they are gradable. They also decline for case and can be premodified by the degree
adverb bardzo ‘very’. By contrast, even though l-participles are morphologically
adjectival, their distributional properties are different. This is related to the fact that the
l-participle may never appear independently of the auxiliary verb. For instance, the Old
Church Slavonic lexeme pisalь ‘written’ can only occur together with a finite form of the
copula as a part of a paradigm, as in pisalь jesm ‘I have written’. Occurences of the l-
participle outside compound tenses are unattested.
1.3.3.5.1.4 Grammatical properties of the l-participle

In contrast to the other participles found in Old Church Slavonic (cf. section 1.3.3.5)
and l-adjectives, the l-participle does not decline for case. Moreover, just as the related
formations in Classical Armenian, both the l-participles and the l-adjectives are
indifferent to voice distinctions, so they can have either active or passive meanings. For
instance, the Old Polish adjectives ukradły ‘stolen´, dowiodły ‘proven’ had a passive
meaning, whereas the contemporary Polish adjectives przybyły ‘arrived’ and powstały
‘arisen’ are active (Damborský 1967: 145).15 This property supports the claim made in
chapter 2 and 3 that unlike passive participles, l-participles are able to assign accusative
case to the object.
Finally, since l-participles are non-finite, they are not specified for tense. In fact,
this follows from their semantics. Given that they express the result following from a
past or future action, they are void of a temporal meaning.
1.3.3.5.2 The auxiliary / copula ‘to be’

The compound tenses in Slavic are constructed with various forms of the verb ‘to be’
functioning as the auxiliary. The following sections will provide the full paradigms of
byti in the present, aorist, and imperfectum tenses in Old Church Slavonic.
1.3.3.5.2.1 Tense
Chart (51) gives the paradigm of the verb ‘to be’ in the present tense.
(51) The paradigm of byti in the present tense

1 jesmь jesvě jesmъ
2 jesi jesta jeste
3 jestъ (je) jeste sõtъ (sõ)
(OCS, cf. Schmalstieg 1983: 138)
As marked in the chart, the 3rd singular form in the present tense jestъ has a reduced
variant je, whereas the 3rd plural form sõtъ can be reduced into sõ. These variants were
enclitic, which paved the way for further impoverishment of the forms in contemporary
Slavic languages (Vaillant: 1966: 441-442). The reduction was particularly important for
the construction of the paradigm of the copula in Polish and Serbo-Croatian, which will
be described in section 1.3.4.2.2, and in chapters 2 and 5.
The present forms have negated counterparts, which are constructed by addition
of the particle ne. The 3rd singular reduced variant je has the negated form ně.
(52) The paradigm of the negated form of byti in the present tense

1 něsmь něsvě nesmъ
2 něsi něsta neste
3 něstъ (ně) něste nesõtъ
The paradigms below represent the simple past tenses: aorist and imperfectum.
According to some grammarians (e.g. Lunt 1974: 121), the bracketed forms in the
paradigm of the imperfectum represent the imperfective aorist variants, but some other
researchers (e.g. Vaillant 1948: 298) argue that these are imperfectum forms.
15It seems that in contemporary Slavic languages the ambiguity of the l-participle with respect to
voice specification tends to be resolved, and that l-adjectives usually acquire an active meaning.
The contemporary Polish variants of the l-adjectives with the passive meaning mentioned above
have passive participle morphology: ukradziony ‘stolen’ and dowiedziony ‘proven’.
(53) The paradigm of the aorist form of ‘byti’

1 byxъ byxově byxomъ
2 bystъ (by) bysta byste
3 bystъ (by) byste byšę
(54) The paradigm of the imperfectum form of ‘byti’

1 běxъ běxově byxomъ
2 běaše (bě) běašeta (běsta) běašete (běste)
3 běaše (bě) běašete (běste) běaxę (běšę)
1.3.3.5.2.2 Aspect
The preceding sections have demonstrated that the forms of the verb ‘to be’ differ in
their temporal and aspectual specification in all the compound tenses. In the present
perfect the verb ‘to be’ occurs in the imperfective form, in the future tense it is
perfective, whereas in the pluperfect it appears either in the imperfectum or in the
aorist. The survey of the forms in (55) indicates that the verb ‘to be’ forms aspectual
pairs like most other verbs in Slavic.
(55) The aspectual forms of the verb ‘to be’ in the simple tenses
tense/aspect imperfective perfective

3sg present jestъ bõdetъ
3sg imperfectum bĕaše bõdĕaše
3sg aorist bĕ bystъ
(OCS, cf. Dostál 1954: 146; Van Schooneveld 1951: 103)
Whaley (2000a: 21) investigates the diachrony of the perfective form of the verb byti
used in the constructions with a future meaning. She points out that bõdetъ contains the
nasal vowel õ, which is a descendant of the nasal consonant found in the Late Proto-
Indo-European stem *bhū-n-d ‘to bePERF’. The nasal consonant was a perfectivizing
infix. In Old Church Slavonic the nasal vowel õ occurs in a small group of verbs
expressing ingressivity or inchoativity (i.e. the beginning of a process), such as sędõ ‘sit
(down)’, lęgõ ’lie (down)’, stanõ ‘stand (up)’; ‘become’ (cf. Meillet 1958: 169). The
presence of the nasal vowel proves that bõdetъ is perfective and implies a change of
state. Possibly, it could have the meaning of ‘become’ (cf. section 1.3.4.4 for more
details). As can be expected, the perfectivizing nasal vowel is absent in the imperfective
form, jestъ.
1.3.3.5.2.3 Auxiliary vs copula

Morphologically, the paradigm of the verb ‘to be’ in Old Church Slavonic is the same
regardless of whether it is used as a copula or as an auxiliary. However, its distribution
may differ. For example, Van Schooneveld (1959: 142) points out that whereas the
auxiliary in Old Russian was an enclitic and had to occur in the second position, the
copula was not. Therefore, the copula could occur at the beginning of a clause and
precede adjectives and present active participles.16 17
(56) I by obladją Olegъ Poljany

and beAOR.3SG rulePAP3SG Oleg Poljane
“And Oleg was ruling over the Poljane”
(Old Russian, Povest’, Van Schooneveld 1959: 143)
Moreover, the copula and the auxiliary behaved differently with respect to ellipsis. For
example, Růžička (1963: 202) argues that the finite forms of the verb byti ‘to be’ were
optional in copula constructions (‘be + NP’) in Old Slavic texts, but the perfect
auxiliary could never be dropped when it occurred with l-participles. Therefore,
Růžička concludes that there were two homophonous forms of byti, one for the perfect
tense, the other for copulas. However, there seems to be some disagreement among
researchers, because Van Schooneveld (1959: 107) claims that Old Russian could either
drop or retain both the auxiliary and the copula.
At any rate, it is evident that the copula and the auxiliary had the same
morphological forms in Old Church Slavonic. The next section will show that in the
languages that have lost the aspectual past tenses, the auxiliary has lost its tense
specification as well, becomes phonologically and morphologically weaker and is
eventually reanalyzed as a person and number marker. Conversely, in the languages that
retained the aspectual past tenses, there is no morphological distinction between the
auxiliary and the copula.
This concludes the presentation of the ways of expressing tense and aspect
distinctions in Old Church Slavonic. In the next section I will demonstrate how the
markings of tense and aspect developed in contemporary Slavic languages.
1.3.4 Development of the tenses in Modern

Slavic languages
Section 1.3.3 showed that Old Church Slavonic expressed temporal relations through
intricate tense and aspect markings. As can be expected, the tense system developed in
various directions in particular Slavic languages. Before I illustrate the modifications in
more detail, let me point out the triggers of the changes.
According to Hewson & Bubenik (1997: 285), the tense/aspect systems of late
Common Slavic and Old Church Slavonic were imbalanced in two respects. First of all,
as was already discussed in section 1.3.3.2, verbs could express three types of aspect:
retrospective, perfective, and imperfective. The retrospective aspect was marked by the
present perfect tense formed with the l-participle, whereas the (im)perfectivity was
signified by aspectual morphemes. The two types of aspect could be represented
simultaneously on one verbal form. For instance, both of the examples in (57) represent
retrospective aspect. In addition, (57a) renders imperfective aspect, whereas (57b)
marks perfective aspect by the prefix pri- on the l-participle.
16 The same holds for contemporary Czech, as will be shown in section 1.3.4.2.2.
17 Conjunctions were not potential clitic hosts in Old Russian or in Old Church Slavonic.
(57) a. Neslъ jesmь retrospective+imperfective

carryPART.IMPF.M.SG beAUX.1SG
“I have been carryingIMPF”
b. Pri-neslъ jesmь retrospective+perfective
carryPART.PRF.M.SG beAUX.1SG
“I have carriedPRF” (OCS)
The problem is that the difference between retrospective and perfective aspect is rather
minute. Figure (23), repeated below as (58), shows that perfective aspect situates the
agent at the very last moment of the action, but internally to the event. Retrospective
aspect locates the agent immediately after the end of the event.
(58) Event time

|--------x- - - - - - - | (imperfective aspect)
|-------------------x- | (perfective aspect)
|----------------------|x (retrospective aspect)
Given the semantic proximity of the two types of aspect, the natural expectation is that
the retrospective meaning might become too insignificant to maintain, so that the major
distinction will be drawn between the perfective and the imperfective.
Another weak point of the Old Church Slavonic tense/aspect system was the
coexistence of the aspectual tenses, the aorist and the imperfectum, with the perfective
and imperfective aspectual forms. Either of the aspectual tenses could be used with
both imperfective and perfective verbs, so the tense and aspect markings were in
principle independent of each other. However, in most cases the aspectual tenses
semantically coincided with the specifications of aspect. Since the most common
variants were the imperfective imperfectum and the perfective aorist, it was unnecessary
to mark the same aspectual distinction twice.
As an illustration, the Old Church Slavonic tense/aspect system is presented in
chart (59). The functionally overlapping aspectual forms are marked in bold. The
retrospective aspect is semantically redundant, because it expresses very similar
meanings to the perfective and imperfective aspects. Correspondingly, (im)perfectivity
is rendered in a morphologically uneconomical way, because it is marked both via
perfective prefixes (e.g. na-) and the aspectual tenses.
(59) The overlapping aspectual forms in Old Church Slavonic
Retrospective
Future - na-pьsa-lъ bõd-õ
Perfective Present na-pьš-õ (Fut) na-pьsa-lъ jesmь
Aorist na-pьš-axъ na-pьsa-lъ bĕ-xъ
Future xošt-õ pis-ati pisa-lъ bõd-õ
Imperfective Present pis-aj-õ pisa-lъ jesmь
Imperfectum piš-āxъ pisa-lъ bĕ-xъ
(Hewson & Bubenik 1997: 286)
The most plausible solution to enhance the tense/aspect system was to remove or
reanalyze the isofunctional categories. Thus, some of the South Slavic languages
retained the aspectual tenses, but at the same time have altered the semantics of the
compound tense formed with the auxiliary ‘be’ and the l-participle. In these languages
the construction usually does not express retrospective aspect any more, but has
acquired a new meaning and is used to characterize events that have not been witnessed
by the speaker (cf. the discussion of the renarrated mood in section 1.3.4.5.1). By
contrast, most of the West and East Slavic languages have lost the aspectual tenses,18
and have reanalyzed the compound formation constructed with the l-participle and the
auxiliary ‘be’ as the default past tense. The distinction between the perfective and
imperfective meanings is now rendered only by aspectual affixes.
The details of the modifications of the tense/aspect system in contemporary Slavic
languages will be fleshed out in the subsequent sections.
1.3.4.1 Simple past tenses

This section will describe the development of aorist and imperfectum in contemporary
Slavic languages. A division will be made between East and West Slavic, where the
tenses have been lost, and South Slavic, where the tenses have been preserved to
various degrees.
1.3.4.1.1 East and West Slavic

With the exception of Upper Sorbian, the aspectual past tenses have completely
disappeared from West and East Slavic languages. The former contrast between the
aorist and the imperfectum is currently marked only by aspectual morphemes. In Old
Russian the decline of the imperfectum in the 13th century preceded the decline of the
aorist in the 15th-16th century. In Old Czech the simple past tenses disappeared during
the 14th century, while in Polabian (the western-most Slavic language spoken in the area
of the Elbe River in Germany) they existed until the death of the language at the
beginning of the 18th century (cf. Stieber 1973: 45-46; 53).
The earliest written Polish texts contain examples of the aorist and the
imperfectum, but they are very rare. For example, Klemensiewicz et al (1964: 369) claim
that there were 8 examples of the aspectual tenses in Kazania Świętokrzyskie, which is the
oldest Polish literary relic from the 14th century and 13 instances in Psałterz Floriański
from the 14th-15th century. Later texts include infrequent usages of the aorist forms, but
only of the verb ‘to be’. Because of some phonological processes, the person endings of
the aorist and the imperfectum blurred in Old Polish and became difficult to
distinguish. Eventually, the past tenses vanished in the 14th century.
1.3.4.1.2 South Slavic

In contrast to East and West Slavic, the South Slavic languages have retained the
aspectual simple past tenses, albeit to different degrees. A possible reason for their
preservation could be the fact that the South Slavic languages are members of the
Balkan Sprachbund and thus stay in intensive cross-cultural contacts with ‘alien’, non-
Slavic languages, which have retained similar tense distinctions (cf. Tomić 1984).
The simple past tenses are most widely used in Bulgarian, where just as in Old
Church Slavonic, tense and aspect form two independent systems. Hence, the values of
the aspectual tenses need not correspond to the aspect marking. For instance, the
example in (60) shows that imperfective verbs may occur in the aorist.
18 Sorbian (a Western Slavic language spoken in eastern parts of Germany) is an exception,
because it has retained them.

(60) Toj caruva trijset godini

he reignIMPF.AOR.3SG thirty years
“He reigned for thirty years” Imperfective Aorist
(Bg, Comrie 1988: 73)
According to Scatton (1984: 321-322) the imperfective aorist is used when there is no
concrete, definite end-result of an event, or when the occurrence of a major event is the
main issue in the narration, rather than its termination. In the case at hand, the verb
caruva ‘reignIMPF.AOR.3SG’ inherently specifies an unbounded activity, which in Bulgarian
is expressed by means of imperfective aspect. However, the sentence contains the
adverbial ‘for thirty years’, which adds a terminal point to the situation characterized by
the aorist. The combination of the imperfective aspect with the aorist renders the
meaning of an atelic event that becomes terminated after a long period of time.
It is also possible for perfective verbs to carry the morphology of the imperfectum
tense. This is demonstrated in (61).
(61) Vseki pŭt, kogato izlezexme na poljana

every time when come-outPRF.IM.1PL to meadow
“Every time that we came out onto the meadow…”
Perfective Imperfectum
(Bg, Comrie 1988: 73)
Here, the combination of the perfective aspect with an imperfectum describes an

unbounded repetitive or habitual situation. Each of the individual occurrences of the
situation is regarded as bounded, which can be attributed to perfective aspect, but the
whole situation does not have to be bounded, which might be attributed to the
imperfectum form of the verb.
Undoubtedly, the semantically “contradictory” combinations of tense and aspect
forms require rather unusual contexts, which do not occur very often.19 Therefore, the
system that allows them is prone to simplifications.
In comparison to Bulgarian, the tense/aspect system of Macedonian is quite
reduced. The imperfectum is used as the default past tense, whereas the aorist is
becoming obsolete and can be found only in a limited set of expressions (Tomić 1989:
366). One reason for the loss of the aorist is morphological. The majority of verbs in
Macedonian have distinct forms of the aorist and the imperfectum endings only in the
2nd and the 3rd person singular. The forms are identical in all the other persons (Lunt
1952: 90).
(62) The morphological endings of the Macedonian aspectual tenses
aorist imperfectum
singular plural singular plural
1 -v -vme -v -vme
2 - -vte -še -vte
3 - -a -še -a
(Mac, Friedman 1977: 8)
19See Lindstedt (1985) for an extensive analysis of interactions between tense and aspect in
Bulgarian.
Hewson & Bubenik (1997: 288) attribute the decline of the aorist in Macedonian to the
fact that the imperfectum can be formed only from imperfective verbs, while the aorist
can be constructed only from the verbs marked for perfective aspect.20 The lack of the
distinction between the perfective and the imperfective aorist makes it redundant as a
separate tense. It is enough to mark perfectivity via aspect.
Other South Slavic languages have only remnants of the aspectual tenses. The
aorist and the imperfectum have completely vanished from Slovene, where the
compound tense formed with the l-participle has been adopted as the general past
tense. In Serbo-Croatian, the usage of the past tenses is restricted to certain dialectal
areas, even though they are still “taught” at schools. It seems that the aorist and the
perfect are more common in Serbian than in Croatian. Gradually, the present perfect is
taking over the old role of the aorist as the “narrative” tense in Serbo-Croatian. The
only exceptions are some Montenegrin dialects, where the aorist is still used as the main
narrative system by some modern fiction writers from this area (cf. Lindstedt 1994: 39).
Summarizing, it has been shown that the Old Church Slavonic aspectual tenses
have been preserved in Bulgarian, and (to a lesser extent) in Macedonian. All the other
Slavic languages have lost them, and make a distinction between perfective and
imperfective events by using aspectual morphology.
1.3.4.2 Present perfect

The present perfect has undergone the most profound changes in all Slavic languages,
which are related both to its semantics and grammatical form. In East and West Slavic,
as well as in Serbo-Croatian it was reanalyzed as the default tense for describing past
events. In some of these languages, especially in the East Slavic group, the process was
accompanied by morphological reduction of the auxiliary (cf. section 1.3.4.2.1 for East
Slavic and 1.3.4.2.2 for West Slavic). In Bulgarian and Macedonian the morphological
form of the present perfect remained largely the same, but the tense has acquired a new
modal meaning and is used for characterizing events that have not been witnessed by
the speaker. This type of usage is referred to as the “renarrated mood” and will be
discussed in section 1.3.4.5.1.
The modifications of the present perfect cross-cut the aerial boundaries between
the East, West, and South Slavic subgroups, which is evident in the discussion of
Serbo-Croatian (cf. 1.3.4.2.3.) versus Bulgarian and Macedonian (cf. 1.3.4.5.1). Likewise,
a number of geographically unrelated Slavic languages have developed new forms of
present perfect, with the verb ‘have’ acting as the auxiliary, which will be described in
section 1.3.4.5.2.
1.3.4.2.1 East Slavic

With the loss of the aorist and the imperfectum, the present perfect became the only
construction available to characterize past events in East Slavic. It is not clear why the
aspectual tenses disappeared from these languages, but it is certain that the usage of the
present perfect increased in Old Church Russian earlier than in Old Church Slavonic
(cf. Stieber 1973: 53), which indicates that the present perfect started to assume the role
of the default past events before Old Russian split into dialects.
20 Friedman (2002: 267) remarks that imperfective aorist had been found in the literature till the
middle of the 20th century, but now they are completely obsolete.
The decline of the aspectual tenses coincided with the loss of the auxiliary ‘be’ in
the present perfect. Unlike in South and West Slavic (cf. section 1.3.4.2.2.1 for a
discussion of Polish), it was lost without any prior morpho-phonological reduction into
a clitic or an affix. The auxiliary started to be left out already in the 11th century, while
in the 12th-13th century dropping the auxiliary became the norm. This development was
peculiar to the East Slavic group and Old Church Russian, because the auxiliary drop
was uncommon in Old Church Slavonic. The only Old Church Slavonic relic in which
the auxiliary was occasionally missing was Codex Suprasliensis, but the omission was
limited to the 3rd person singular form (Van Schooneveld 1959 ch. 4). In the 16th and
the 17th century the present perfect auxiliaries were still occasionally found in the 1st
and the 2nd person in the singular and in the plural, but afterwards, they fell out of use
completely. As a result, in the contemporary East Slavic compound tenses the l-
participle occurs without a perfect auxiliary, as shown for Russian in (63).
(63) Ana napisala pis’mo

Ana writePART.F.SG letter
“Ana wrote a/the letter” (Rus)
The disappearance of the auxiliaries had a clear syntactic effect: the East Slavonic
languages became non-pro-drop, and pronouns can be omitted only when they are
topics (cf. Franks 1995 ch. 7). Otherwise, it is impossible to mark person-number
distinctions.
As a result of the loss of the present perfect, the only compound forms available in
contemporary East Slavic are the future tense (cf. section 1.3.4.4.2) and the conditional
form (cf. 64), which consists of the invariant auxiliary by and the l-participle. Since the
auxiliary does not show person-number distinctions, the subject is always present.
(64) Ja čital by
I readPART.M.SG+COND
“I would read” (Rus)
With the decline of the perfect auxiliaries, the contemporary East Slavic group has lost
most of its compound tenses. Therefore, little attention will be given to these languages
in this thesis.
1.3.4.2.2 West Slavic

Just as the East Slavic languages, the West Slavic languages have lost the aorist and the
imperfectum as well. Semantically, they have also lost the present perfect, but unlike the
East Slavic languages, they have retained the auxiliaries in an impoverished form. I will
present the change using Polish data, with some examples of the auxiliary/copula
distinction in Czech added in section 1.3.4.2.2.3 for comparison.
1.3.4.2.2.1 Changes to the auxiliary in Polish

As was noted in section 1.3.3.5.2.1, Old Church Slavonic had two forms of the verb ‘to
be’ in the 3rd person: the strong (orthotonic) form jestъ in the singular, and sõtъ in the
plural, as well as the reduced variants, je and sõ, respectively. The placement of the
orthotonic forms was relatively free, whereas the reduced ones were enclitic and had to
occur in the second position. In the languages that emerged from Old Church Slavonic,
the reduced forms were extended to the whole paradigm. The chart in (65) presents the
development of the verb ‘to be’ in the history of Polish (Decaux 1955: 126ff; Andersen
1987: 24); the orthotonic variants are taken from Old Polish, whereas the reduced ones
in the last two columns are taken from 16th century and Modern Polish. The dual forms
were lost in the 16th century (Długosz-Kurczabowa & Dubisz 2001: 308), but can still
be found in some dialects.
(65) Diachronic development of the Polish verb ‘to be’
ORTHOTONIC FORM REDUCED FORMS

(Old Polish) 16th century Polish Modern Polish
1SG jeśm -(e)śm/-(e)m -(e)m
2SG jeś -(e)ś -(e)ś
3SG jest/jeść/je - -
1PL jesm(y) -(e)smy -(e)śmy
2PL jeśće -(e)śće -(e)śće
3PL są - -
1DUAL jeswa -(e)swa -
2DUAL jesta -(e)sta -
3DUAL jesta -(e)sta/-0 -
(cf. Andersen 1987:24)
Deceaux (1955: 127-128) observes that in Old Polish the orthotonic 3rd person
auxiliaries jest, je, and są were found only in emphatic predication structures. This
restriction may account for the fact that these forms already disappeared between the
15th and the 17th century. The singular variant jest fell out of use first, and the plural
form są was lost later.
The sentences in (66) exemplify the use of Old Polish orthotonic auxiliary forms.
They may co-occur with the enclitic variants, and their position in the clause is largely
unrestricted.
(66) a. Wiem ze stworzyciela wszego luda porodziła jeś

know1SG that creator allGEN mankindGEN bearPART.F.SG 2SG
“I know you bore the creator of all mankind”
b. Tom jest oglądała
That+1SG 3SG.EMPH seePART.F.SG
“That I did see”
c. Jest ja ciebie zepchnął albo uczynił-em
3SG.EMPH I youACC repulsePART.M.SG or doPART.M.SG+1SG
tobie co złego?
you any harm
“Did I repulse thee or do thee any harm?” (Old Polish, Andersen 1987:28)
The examples in (67) illustrate the use of the reduced auxiliaries, which must encliticize
on the first element in the clause, and thus appear in the second position.
(67) a. Ani-ś mię zepchnął, ani rzucił, ani-ś

not+2SG me repulsePART.M.SG nor desertPART.M.SG nor+2SG
niektóre złości uczynił
any harm doPART.M.SG
b. Bo-cie-m się cała darowała
for-you+1SG REFL entire givePART.F.SG
“For I gave myself wholly to thee” (Old Polish, Andersen 1987:28)
A Modern Polish example is given in (68). It will be shown in chapter 5 that the
auxiliary following the l-participle has been largely reanalyzed as an affix.
(68) Wczoraj pojechał-em do Szczyrku na narty

yesterday goPART.M.SG+AUX.1SG to Szczyrk on skis
“Yesterday I went skiing in Szczyrk” (Modern Polish)
With the decline of the orthotonic forms, the emphatic distinction had to be rendered
by other means, such as word order or sentence stress. However, the orthotonic forms
did not disappear from the language entirely. They were reanalysed as copula stems, to
which the reduced auxiliaries were added as person-number markers. The paradigm of
the copula in contemporary Polish is given in (69), with the person-number affixes
marked in italics.
(69) Paradigm of the copula in Modern Polish
Modern dialects Standard Polish

1SG jest-em jest-em
2SG jest-eś jest-eś
3SG jest jest
1PL są-śmy jest-eśmy
2PL są-śće jest-ście
3PL są są
(cf. Andersen 1987)
1.3.4.2.2.2 Changes to the present perfect in Polish

The previous section has shown that the modification of the present perfect in Polish
consisted in a gradual reanalysis of the auxiliary ‘to be’ as a person-number affix
attached to the l-participle. In other words, it involved a change from analytic to
synthetic verbal structures.
The loss of the aspectual tenses in the East and West Slavic languages increased
usage of the present perfect, which became the default past tense. Kowalska (1976: 42)
argues that because of this increase, the forms of the auxiliary became morphologically
and phonologically weaker. As a result, two different forms arose.
a) The analytic form: the auxiliary occurs in the Wackernagel position, that is, after the
first stress-bearing word in the clause. This is usually a conjunct, pronoun, wh-word or a
particle. This strategy is especially common in subordinate clauses.
(70) a. A teraz-eś mi tę robotę náznaczył

and now+AUX.2SG meDAT this work assignPART.M.SG
“And now you have assigned this work to me”
(Zwieciadło duchowej łaski 1645)
b. Ju-że-ście śie go dośytz namęcżyli
already+FOC+AUX.2PL REFL himACC enough tirePART.PL
“You have tired him enough already”
(Pl, Żywot Pana Jezu Krysta 1522; Kowalska 1976: 43)
b) The synthetic form: the auxiliary does not appear in the Wackernagel position but
immediately follows the l-participle, and is on the way to become an affix.
(71) a. Egiptowi podáli-śmy ręce

EgyptDAT givePART.PL+1PL hands
“We gave our hands to Egypt”
(Wereszczyński 1592 Excitarz do podniesienia wojny…)
b. Y zdrowie śwoie położyli-ście dla mnie
and health your givePART.PL+2PL for me
“And you gave your health for me”
(Pl, Grzegorz z Żarnowca 1582 Postylle część wtora; Kowalska 1976: 43)
Tableau (72) shows that the synthetic form steadily prevails over time: it is found most
often in clauses that contain verbs, nouns, adjectives or numerals in the initial position.
(72) The position of auxiliary verbs in Polish
Century analytic forms synthetic forms

14th-1520 (religious texts) 1153 (53%) 1007 (47%)
14-1520 (judical texts) 3651 (91%) 341 (9%)
1st half 16th 746 (71%) 301 (29%)
2nd half 16th 293 (67%) 145 (33%)
1st half 17th 296 (67 %) 146 (33%)
2nd half 17th 487 (55%) 394 (45%)
1st half 18th 188 (45%) 234 (55%)
2nd half 18th 260 (27%) 710 (73%)
1st half 19th 117 (18%) 538 (82%)
2nd half 19th 204 (16%) 1106 (84%)
1st half 20th 56 (3%) 2009 (97%)
(adapted from Kowalska 1976: 63)
Analogous estimates concerning the possible auxiliary positions are provided by Rittel
(1975: 91), who also indicates the ratio of the second (Wackernagel) position
violations.21 I recalculate Rittel’s data into percentage terms in tableau (73).
21 A similar chart is to be found in Andersen (1988: 29); however, I have recalculated Rittel’s data
in greater detail.
(73) The position of auxiliary verbs in Polish
Century Aux preceding V Aux (immediately) 2nd position violations

following V
14-15th 89% 10% 3%
14-15th Bible 34% 65% 12%
16th 77% 22% 2%
17th 50% 49% 5%
18th 31% 69% 4%
19th (prose) 27.5% 72% 13%
20th (prose) 15% 85% 15%
(cf. Rittel 1975: 91)
Two patterns of development can be observed here. First, Rittel’s calculations indicate
an increase in the ratio of the Wackernagel position deviations. The deviations were
considerably more prominent in the 14-15th century Bible translations than in the non-
religious (mostly legal) texts from the period. Since legal texts represent spoken
language more faithfully than religious writings, Kowalska (1976: 37) concludes that the
difference reflects the fact that in spoken language the Wackernagel law was observed
more diligently. Still, the rise in the second position violations had been rather
insignificant.
What is more telling is an increasing tendency for the auxiliary to appear adjacent
to the participle. This tendency most likely indicates a morphological change from a
clitic into a verbal affix, which continues in Modern Polish. It must be noted, however,
that the development of synthetic forms in Polish varies within the verbal paradigm.
The 1st person singular form of the auxiliary was the first one to fuse with the participle.
According to Kowalska’s (1976: 48) estimates, 86 % of the 1st person singular forms
attested in the corpus from the period between the 2nd half of the 18th and the 20th
century are synthetic. Bajerowa (1964: 17) claims that this is due to the fact that the 1st
person auxiliary -(e)m is homophonous with the instrumental case morpheme of a
nominal paradigms (cf. 74a), and also resembles a plural nominal ending -e (cf. 74c).
The examples in (74a and b) show that this may lead to ambiguity. However, when the
auxiliary is placed immediately after the l-participle rzucił (cf. 74d), the sentence is
disambiguated, as only the meaning in (74b) is available.
(74) a. Kamieni-em rzucił

stoneINSTR throwPART.M.SG
“It was a stone that he threw”
b. Kamieni-em rzucił
stone+AUX.1SG throwPART.M.SG
“It was a stone that I threw”
c. Kamienie-m rzucił
stones+AUX.1SG throwPART.M.SG
“It was the stones that I threw”
d. Rzucił-em kamieniem
throwPART.M.SG+AUX.1SG stoneINSTR
“I threw a stone” (Pl, Kowalska 1976: 55)
The spread and the reanalysis of the present perfect as the default past tense were not
the only reasons for the auxiliary impoverishment. It seems that a change in the word
stress played a major role as well. The rhythm patterns of medieval relics reveal that
there was no regular lexical stress in Old Polish. In the 14th-15th century the main stress
was established on the first syllable of a word, which according to Długosz-
Kurczabowa & Dubisz (2001: 307-308), fostered the encliticization of the auxiliary ‘to
be’. Along with the initial stress, some words received a secondary stress on the
penultimate syllable, which eventually prevailed as the main stress at the beginning of
the 18th century. It may be hypothesized that the penultimate stress pattern led to
further morphological impoverishment of post-verbal auxiliaries. The hypothesis is
supported by Czech, which has retained the initial word stress. As will be shown in the
subsequent section, the auxiliary clitics in Czech enjoy a far greater independence from
the l-participle than in Polish.
The reanalysis of the auxiliary clitic as a verbal affix on the l-participle continues in
Modern Polish, but the process has not yet been completed. A detailed analysis of the
compound tense forms in Modern Polish is presented in chapter 5.
1.3.4.2.2.3 The auxiliary versus copula distinction in Czech

I conclude my survey of the present perfect in the West Slavic languages with a
discussion of the impoverishment of the auxiliary ‘to be’ in Czech. I already mentioned
that in the South Slavic languages the copula and the auxiliary have the same forms and
distribution. In comparison with the South Slavic languages and Polish, Czech
represents an intermediate stage. As in South Slavic, the forms of the auxiliary ‘to be’
and the copula are morphologically almost the same. However, their distribution
differs. I investigate the distinctions using data from Toman (1980). See also Veselovská
(2004) for a more recent overview.
The chart in (75) presents the paradigm of the copula and the auxiliary forms in
Czech. It shows that the 3rd person forms must be omitted in compound tenses but
preserved in copula constructions.
(75) Forms of the copula and the auxiliary in Czech
SG PL
auxiliary copula auxiliary copula
1 jsem jsem jsme jsme
2 jsi jsi jste jste
3 - je - jsou
(cf. Toman 1980)
The contrast is exemplified in (76a) for the copula structure and in (76b) for the present
perfect.
(76) a. On *(je) učitel

he be3SG teacher
“He is a teacher”
b. On přisĕl
He comePART.M.SG
“He came” (Czech, Toman 1980)
Another distinction between the two forms concerns ellipsis. As shown in (77a), the 1st
person singular auxiliary may be deleted. The copula may not (cf. 77b).
(77) a. Já (jsem) už spal

I beAUX.1SG already sleepPART.M.SG
“I was already asleep”
b. Já *(jsem) už pĕt let učitel
I be1SG already five years teacher
“I have been a teacher for five years” (Czech, Toman 1980)
Furthermore, the auxiliary ‘to be’ may be reduced in the 2nd person singular (cf. 78),
while the copula may not (cf. 79).
(78) a. Ty jsi přišel

you beAUX.2SG comePART.M.SG
“You came”
b. Ty-s přišel
you +beAUX.2SG comePART.M.SG (Czech, Toman 1980)
(79) a. Ty jsi učitel

you be2SG teacher
“You are a teacher”
b. *Ty-s učitel
you +be2SG teacher (Czech, Toman 1980)
These distributional differences suggest that the reduction of the auxiliary has gone
further than the impoverishment of the copula. The conclusion is supported by the fact
that in spite of the morphological identity of the two elements, only the copula may
appear in the sentence initial position.
(80) a. *Jste vidĕli?

beAUX.2PL seePART.PL
b. Vidĕli jste?
seePART.PL be2PL
“Did you see?”
c. Jste dnes na řadě?
be2PL today on row
“Is it your turn today?” (Czech, Toman 1980)
Moreover, the two forms of the verb ‘to be’ behave differently with respect to negation.
As indicated in (81) and (82), the negative prefix ne attaches either to the l-participle or
to the copula. It may not be attached to the auxiliary.
(81) a. Přišel jsi

comePART.M.SG beAUX2.SG
“You have come”
b. Nepřišel jsi
NEG+comePART.M.SG beAUX2.SG
“You haven’t come”
c. *Nejsi přišel
NEG+beAUX2.SG comePART.M.SG (Czech, Toman 1980)
(82) a. Jsi hlupák / zdráv / na řadĕ

be2SG idiot / healthy / on row
“You are an idiot/healthy/… it’s your turn”
b. Nejsi hlupák / zdráv / na řadĕ
NEG+be2SG idiot / healthy / on row
“You’re not an idiot/healthy/… it’s not your turn”
c. *Jsi nehlupák/nezdráv/nena řadĕ (Czech, Toman 1980)
The gradual differentiation between the paradigms of the copula and the auxiliary is a
common pattern in the history of the Slavic languages. Chapter 4 will show that a
similar phenomenon occurs in Macedonian. I would like to relate the observed
distinction to the decline of the aspectual tenses in most of the West Slavic languages.
After the disappearance of the simple past tenses, the original present tense forms of
the verb ‘to be’ did not have any simplex counterparts. In other words, Polish and
Czech lost the discrimination of the is versus was-type in English. Once the tense
contrast is lost, the auxiliaries represent only φ-feature distinctions, and eventually are
reanalysed as person and number markers. The semantic impoverishment corresponds
to their morphological reduction, when they are finally reinterpreted as affixes.
1.3.4.2.3 A reanalysis of the present perfect in Serbo-Croatian

As was noted in section 1.3.4.2, the present perfect in Bulgarian and Macedonian may
express the renarrated mood. This is a type of modal meaning which was not available
in Old Church Slavonic. For this reason, the analysis of the present perfect in these
languages is given in section 1.3.4.5.1, and the present subsection discusses Serbo-
Croatian as the only representative of the South Slavic group.
The morphological form of the present perfect in Serbo-Croatian has remained the
same, since it is still constructed with the l-participle and the auxiliary ‘to be’. However,
its meaning has changed, and the construction is now used as the general narrative past
tense, which does not have to render resultativity or retrospective aspect. This is
demonstrated in the translation of the sentence in (83).
(83) Ana je napisala pismo

Ana bePRES.3SG writePART.F.SG letter
“Ana wrote/has written a letter” (S-C, Tomić 1989: 364)
According to Lindstedt (2000: 372), the change was triggered by the fact that the
opposition between perfectivity versus imperfectivity can be suitably rendered via
verbal prefixes, which are independent of tense specifications. Consequently, the
present perfect became available for expressing past tense.
To summarize, it has been shown that as a result of the decline of the simple past
tenses in the languages described in section 1.3.4.2, the present perfect began to be used
more often and took over the temporal meanings expressed by the aorist and the
imperfectum. Moreover, in East and West Slavic the forms of the auxiliary verb ‘to be’
were simplified and started to differ from the copula. The process of the auxiliary
impoverishment has gone further in Polish than in Czech. In Polish the auxiliary has
been largely reanalyzed as a suffix on the l-participle, whereas in Czech the auxiliary ‘to
be’ is still a free form. In Serbo-Croatian the morphological make-up of the present
perfect has remained the same, but its meaning has been shifted, on a par with the East
and West Slavic languages.
1.3.4.3 The pluperfect

The following section will characterize the form of the pluperfect in contemporary
Slavic. It will be shown that this tense has undergone different morphological
modifications in each of the Slavic groups.

Section 1.3.3.4.2 reported that in Old Church Slavonic the pluperfect was formed with
the l-participle accompanied by the auxiliary ‘to be’ in the imperfectum or in the
imperfective form of the aorist. Currently, the tense is commonly used only in
Bulgarian, where it is composed of the past tense (imperfectum) form of the auxiliary
beše and the l-participle (cf. chapter 2 for details).
(84) Beše pročel knigata

bePAST.3SG readPART.M.SG book-the
“He had read the book” (Bg)
In Macedonian, there exists a related construction that makes use of a form of the
auxiliary beše. However, it is rather uncommon and expresses a meaning that is not
directly related to temporal distinctions. It emphasizes that an event has taken place (cf.
Stieber 1973: 56).
(85) Toj beše izlegol koga dojdov jas kaj nego

he bePAST.3SG leavePART.M.SG when arriveIMP.1SG I to him
“He had gone out when I came to his place” (Mac, Friedman 1977: 100)
The pluperfect meanings are more commonly rendered through a type of have-perfects,
with the auxiliary ‘have’ in a past tense form (cf. section 1.3.4.5.2). This variant is given
in (86a), and is contrasted with the beše-construction in (86b).
(86) a. Žiži mi ja pokaža Keti, no jas

Žiži meDAT herACC showPAST.3SG Keti but I
veќe ja imav videno
already herACC haveAOR.1SG seePASS.N
b. Žiži mi ja pokaža Keti, no jas
Žiži meDAT herACC showPAST.3SG Keti but I
veќe ja bev videl
already herACC beAOR.1SG seePART.M.SG
“Žiži pointed Keti out to me, but I had already seen her”
In Serbo-Croatian, the pluperfect is usually formed with the present tense form of the
auxiliary ‘to be’ and the l-participle (cf. 87a). The auxiliary may also appear in the
imperfectum (cf. 87b), but this is rather uncommon.
(87) a. Mi smo bili vidjeli Mariju

we bePRES.1PL bePART.M.PL seePART.PL MaryACC
“We had seen Marija”
b. Mi bijasmo vidjeli Mariju
we beIMP.1PL seePART.PL MaryACC
“We had seen Marija” (S-C, cf. Browne 2002: 331-332)
1.3.4.3.2 West and East Slavic

In the contemporary East and West Slavic languages, the pluperfect has largely
disappeared. However, when it still existed, it was constructed in a different way than in
South Slavic. Namely, since the imperfectum had been lost in the East and the South
Slavic languages, it was impossible to follow the South Slavic pattern and form the
pluperfect with the auxiliary ‘to be’ in the imperfectum. Therefore, a new way of
constructing the pluperfect was adopted, with the l-participle of the main verb, the
auxiliary ‘to be’ in the present tense, and the l-participle of the verb ‘to be’. This is
illustrated for Old Polish in (88).
(88) Pisał jeśm był

writePART.M.SG bePRES.1SG bePART.M.SG
“I had written” (Old Polish, Kowalska 1976: 64)
The pluperfect is largely gone from contemporary Polish, although it may occasionally
be used for stylistic reasons.
In Old Russian manuscripts the pluperfect occurs very rarely, and certainly had
been lost by the 17th century. However, in some Russian dialects, especially in the
North, there are quasi-pluperfect structures. They are constructed with the l-participle
form of the auxiliary and the l-participle form of the main verb. They are not real
pluperfects, because they do not always relate a more remote past event to some other
occurrence that took place in the past. They signify very distant past events.22
(89) Jagody rosly byly

berries growPART.PL bePART.PL
“Berries used to grow” (dialectal Russian, Stieber 1973: 58)
Summarizing, we have seen that the decline of the compound tenses in East and West
Slavic affects both the present perfect and the pluperfect. Conversely, the pluperfect is
still used in South Slavic.
1.3.4.4 Ways of expressing the future

As was noted in section 1.3.3.4.1, there were two future tenses in Old Church Slavonic:
Future I and Future II. The Future I usually consisted of an infinitive preceded by the
auxiliary bõdõ ‘bePRF’, a modal verb like xotętъ ‘to want’ or imamъ ‘have, be destined to’,
or a phase verbs like načьnõtъ and vъčьnõtъ ‘to begin’ accompanied by the infinitive.
Future II was formed with bõdõ followed by the l-participle. It was used very rarely
and disappeared rather early in most of the Slavic languages, although it is still found in
the literary variants of Serbo-Croatian (cf. 90).
22 Maaike Schoorlemmer (p.c.) remarks that this structure is also used in set phrases and story-
telling.
(90) Kad budemo govorili s Marijom,

when bePRF.1PL speakPART.PL with Marija
sve će biti jasno
everything want3SG beINF clear
When/if we speak with Marija, everything will be clear”
(S-C, Browne 2002: 331)
The subsequent subsections will discuss the Future I. In most of the Slavic languages it
is constructed with a contemporary variants of the auxiliary bõdõ, which is followed by
the l-participle (Slovene), the l-participle or the infinitive (Polish), or exclusively the
infinitive (Czech, Slovak, Kashubian, Belarusian, Ukrainian, and Russian). Bulgarian,
Macedonian and some variants of Serbo-Croatian do not have the bõdõ-type of future at
all. Instead, they selected the modal verb ‘to want’ as the future auxiliary, which is
followed by a lexical verb in the form of the l-participle or the subjunctive.

A detailed syntactic analysis of the Slavic Future Tenses in the Balkan area is offered in
Tomić (2004), from which all examples in this section are taken, unless indicated
otherwise. She convincingly shows that the South Slavic languages represent a
continuum in the grammaticalization of the future tense formed with the modal verb
‘to want’.
In Old Macedonian and Old Bulgarian the Future tense was constructed with a
finite form of the verb xotěti ‘want/will’ followed by the infinitive.
(91) Xoščet počiti moj brat

will3SG dieINF my brother
“My brother will die” (14th c. Mac, Trojanska prica, Tomić 2004: 534)
Gradually, the forms of the modal verbs were reduced, as can be observed in the Old
Macedonian example in (92a) and the Old Bulgarian example in (92b). Both sentences
come from the 18th century.
(92) a. Koi ќet mislit, koi ќet iskat

who will3SG thinkINF who will3SG wantINF
“Who will think, who will want” (18th Mac)
b. Štem ostavi
will1PL leaveINF
“You shall leave” (18th Bg, Tomić 2004: 535)
In the 15th century the Balkan languages started to lose the infinitive, which was
replaced by subjunctive formations (cf. 93a). Around the 17th-18th century the future
auxiliaries in Bulgarian and Macedonian turned into clitics, and eventually lost their
person-number distinctions. In Bulgarian the person-number marking was lost later
than in Macedonian (late 19th-early 20th century) and nowadays the auxiliary šte is the
only uninflected auxiliary in this language.
(93) a. Jas ќe go storam toa utre

I willCL itCL.ACC doPRES.1SG that tomorrow
“I will do that tomorrow (18/19th c. Mac, Tomić 2004: 524)
b. Az šte napravja tova utre
I willCL do1SG that tomorrow
“I will do that tomorrow” (Bg, Tomić 2004: 523)
The future auxiliary can be complemented by a subjunctive also in South-Eastern

Serbian dialects (cf. 94a). However, in the northern parts of Serbia and in Croatia it is
more common to find future tense auxiliary clitic će followed by the infinitive (cf. 94b).
(94) a. Tvoj prijatelj će (da) stigne jutre

your friend will1SG SUBJ.COMP arriveSUBJ tomorrow
“Your friend will arrive tomorrow”
(South-Eastern Serbian, Tomić 2004: 521)
b. Petar će doći sutra
Peter will1SG comeINF tomorrow
“Peter will come tomorrow” (S-C, Tomić 2004: 520)
Apart from the clitic forms, Serbo-Croatian has lexical, non-clitic counterparts of the
future auxiliary. Both of them are presented in the tableau in (95).
(95) Paradigm of the future auxiliary in Serbo-Croatian
clitic non-clitic
SG PL SG PL
1 ću ćemo hoću hoćemo
2 ćeš ćete hoćeś hoćete
3 će će hoće hoće
(S-C, Tomić 2004: 520)
Analogously with the clitic forms, the lexical forms can be complemented by either an
infinitive or a subjunctive form. Only the lexical forms, however, distinguish between
aorist and perfect variants. The construction characterizes future events that are relative
to a past moment.
(96) a. Petar ht(j)ede doći / da dođe sutradan

Petar wantAOR.3SG comeINF / SUBJ.COMP comeSUBJ tomorrow
“Peter wanted to come the next day”
b. Petar je ht(j)eo doći / da dođe sutradan
Petar bePRES.3SG wantPART.M.SG comeINF/ SUBJ.COMP comeSUBJ tomorrow
“Peter wanted to come the next day” (S-C, Tomić 2004: 521)
Outside the Balkan Slavic group, the future tenses are formed with a perfective form of
the verb ‘to be’ as the auxiliary. In Slovene and Kajkavian Croatian the auxiliary tends
to stay in the second position and is complemented by the l-participle, either in the
imperfective or perfective form. Most likely, the construction is a direct descendant of
the Future II, whose meaning has been generalized and now covers all future events.
(97) Vsi bodo dosegli svoj cilj

everyone bePRF.1SG reachPART.PL self’s goal
“Everyone will reach his/her goal” (Slovene, Franks & King 2000: 33)
1.3.4.4.2 West and East Slavic

There are two ways of constructing future tenses in West and East Slavic. If the main
verb is imperfective, it is preceded by the perfective form of the auxiliary ‘to be’ (cf.
section 1.3.4.4.2.1). Perfective verbs, however, may not be preceded by this auxiliary,
and are interpreted as expressing the future meaning when they appear in the present
tense without any accompanying auxiliary (cf. section 1.3.4.4.2.2).
1.3.4.4.2.1 Compound future tense forms

In most of the West and East Slavic languages the future tense is constructed with the
perfective form of the verb ‘to be’ as the auxiliary, which is followed by the infinitive.
In Polish the auxiliary may also be complemented by the l-participle (cf. 98a). Only the
imperfective forms of the main verb are permitted (cf. 98b).
(98) a. Jan będzie pisać/pisał list

Jan bePRF.1SG writeINF.IMPF/writePART.IMPF.M.SG letterACC
“Jan will be writing a letter”
b. *Jan będzie napisać/napisał list
Jan bePRF.1SG writeINF.PRF/writePART.PRF.M.SG letterACC (Pl)
There have been some attempts in the literature to attribute the restriction concerning
the aspectual marking of the main verb to the fact that in Proto-Slavic/Old Church
Slavonic the future could be expressed with phase verbs (e.g. načьnõtъ and vъčьnõtъ ‘to
begin’), which imposed aspectual constraints on the main verb in this construction (cf.
Křižková 1960: 82-108). However, this does not explain the lack of this restriction in
the South Slavic languages, including Slovene. Whaley (2000a: 137ff.) proposes that in
the Northern Slavic languages, the descendant of the Old Church Slavonic verb bõdõ
shifted its meaning from a ‘change of state’ verb, which expressed the meaning
‘become’, into an inceptive verb, which denotes the meaning ‘begin’. She points out
that in East Slavic a similar shift has occurred to the change-of-state verb stat’ (Russian).
The verb has now two meanings, ‘become’ and ‘begin’, but only the form with the
meaning ‘become’ has the imperfective variant, stanovit’sja.
(99) Ja xoču stat’ vračom

I wantPRES.1SG becomeINF doctor
“I want to become a doctor” (Rus, Whaley 2000a: 141)
Whaley observes that the Russian verb stat’ developed in the same way as the Old
Church Slavonic bõdõ. It became polysemous over time, with the two distinct meanings
mentioned above. The form which took on the inceptive meaning of ‘begin’ could be
complemented by infinitives, but only in the imperfective forms. Currently, stati can be
used as an alternative form of a future tense auxiliary in all East Slavic languages, as in
the Russian example from a poem by Lermontov.
(100) Stanu skazyvat’ ja skazku

beginPRES.1SG tellINF I story
“I will (begin to) tell a story” (Rus, Whaley 2000a: 62)
I mentioned above that in Slovene the future auxiliary ‘to be’ is complemented by the l-
participle, whereas in most of the North Slavic languages the auxiliary occurs with an
infinitive. A striking fact about Polish, as well as some eastern dialects of Slovak
(Stieber 1973: 62-64), is that the future auxiliary can occur with both l-participles and
infinitives (cf. 98a). The selection of either variant does not change the meaning of the
clause. In the other North Slavic languages only the infinitive may complement the
auxiliary, as shown for Russian in (101).
(101) On budet čitat’/*čital

he bePRF.1SG writeINF/writePART.M.SG
“He will be reading” (Rus)
The origin of the l-participial form in this construction is subject to much controversy.
According to some researchers (e.g. Stieber 1973: 64), the participle is a descendant of
the Future II form. However, in Old Church Slavonic the Future II was compatible
with both perfective and imperfective forms of the l-participle (cf. section 1.3.3.4.1),
whereas in Polish only the imperfective variants are found. If Future II were the source
of the l-participle future constructions in Polish, this aspectual restriction would not be
observed.
Górecka and Śmiech (1972: 13) show that the combinations of the future auxiliary
with the l-participle are very rare in the oldest Polish texts; therefore this construction
seems to be a more recent innovation, rather than a continuation of the Future II.
Andersen (1988: 26-27) arrives at the same conclusion. Hence, it seems that the
construction is a recent development, whose source remains unclear.
1.3.4.4.2.2 Expressing the future with perfective verbs

In addition to the compound constructions mentioned above, Northern Slavic
languages (including Slovene) may also express future meanings with perfective forms
of present tense verbs. The use of the perfective form implies that the activity will be
completed.
(102) Napiszę list

writePERF.1SG letter
“I will write the letter” (Pl)
The strategy is not available in South Slavic languages, although it is certain that it was
present in Old Church Slavonic and in older variants of South Slavic languages (cf.
Whaley 2000a: 95). Due to the influence of neighboring languages of the Balkan
Sprachbund, which characterize future events with the verb ‘want’, the old way of
expressing the future was lost.
To summarize, it has been shown that both the Balkan Slavic and the Northern
Slavic languages use compound tenses to describe future events. The major distinction
between the two language groups concerns the selection of the auxiliary. Balkan Slavic
uses a descendant of the verb ‘want’, while the North Slavic languages opt for the
perfective form of the verb ‘to be’.
1.3.4.5 New types of compound tenses in Slavic

The following subsections will discuss compound structures which did not exist in Old
Church Slavonic and are recent innovations that have arisen under the influence of
non-Slavic languages. The newly developed forms include the renarrated mood (cf.
section 1.3.4.5.1) and the ‘have’-perfect (cf. section 1.3.4.5.2). The renarrated mood is
constructed in the same way as the present perfect in Old Church Slavonic, but it
expresses the meaning of “non-evidentiality”, which is not grammaticalized in any other
Slavic language apart from Bulgarian and Macedonian. The ‘have’-perfect renders the
meaning of resultativity, which in Old Church Slavonic was characterized by the
present perfect. In some Slavic languages it is now expressed by a compound tense
formed with the auxiliary ‘have’ and an invariant form of the passive participle.
1.3.4.5.1 The renarrated mood

As has been noted above, the renarrated mood has been grammaticalized in Bulgarian
and Macedonian, which are described in section 1.3.4.5.1.1. Constructions that express
a similar meaning are found in a few other languages mentioned in subsection
1.3.4.5.1.2.
1.3.4.5.1.1 Bulgarian and Macedonian

In Bulgarian and Macedonian the compound tense formed with the l-participle may still
function as a resultative perfect, but is increasingly being used to render the renarrated
mood.23 The construction developed on the basis of a Turkish model in the 15th
century. Its grammaticalization is more advanced in Macedonian than in Bulgarian.
The renarrated mood implies that the situation described in the clause has not
been personally witnessed by the speaker, but is only inferred or has been reported by
someone else; therefore the speaker is not responsible for the veracity of the statement.
The renarrated mood contrasts with the “narrated” (“non-reported”) mood, which is
characterized by a verb in the past tense (the aorist or the imperfectum) and indicates
that the event described is known to the speaker from his/her own experience.
The contrast between the two types of mood is illustrated in (103) for Bulgarian.
The sentence in (103a) contains a verb in the past tense, because the speaker’s
information about the event described in the clause comes from his/her own
experience. In sentence (103b) a past tense form would be ungrammatical, because the
speaker reports the information s/he has heard about from other people or inferred
through some other kind of indirect evidence.
23 The renarrated mood is also commonly termed ‘evidential’ or ‘reported’. The modal meaning it
expresses is referred to as ‘evidentiality’.

(103) a. Kogato bjax mlad horata ne pušexa po ulicite

when bePAST.1SG youngM people NEG smokePAST.3PL on street
“When I was young, people did not smoke on the streets”
b. Kogato bašta mi bil mlad horata
when father meDAT bePART.M.SG youngM people
ne pušeli po ulicite
NEG smokePART.PL on street
“When my father was young, people did not smoke on the streets”
(Bg, P. Vitkova, p.c.)
In renarrated constructions in Bulgarian the 3rd person singular and plural auxiliaries
tend to be deleted (cf. 103b). Some researchers (e.g. Andrejczin 1938 and Friedman
1978) have tried to establish a semantic contrast on the basis of the presence or absence
of the auxiliary, and argued that the paradigm with all forms of the auxiliaries present
has the “past indefinite” meaning, whereas the absence of the 3rd person auxiliaries
triggers the “past reporting” meaning. This claim has been challenged by Tomić (1983,
1989) and Lindstedt (1994: 44ff), who show that the auxiliary drop cannot be
consistently correlated with the semantics of non-witnessed events. In fact, the auxiliary
may be frequently present in renarrated contexts in Bulgarian, as indicated in (104).
(104) Čux če ošte nesa došli

hearPAST.1SG that still not bePRES.3PL comePART.PL
“I have heard that they haven’t come yet” (Bg, Tomić 1989: 369)
Moreover, the loss of the auxiliary in the 3rd person is a general feature of many Slavic
languages, not necessarily related to the renarrated mood. For example, in Czech and
Macedonian the 3rd person auxiliary is always missing (cf. section 1.3.4.2.2.3). It is also
possible to drop the 3rd person auxiliary in Serbo-Croatian, even though evidentiality is
not grammaticalized in this language.
(105) a. Bio je u žurbi

bePART.M.SG bePRES.3SG in haste
“He was in a hurry”
b. Bio, veli, u žurbi
bePART.M.SG says in haste
“He was in a hurry, he says” (S-C, Tomić 1989: 369)
The example in (105a) contains only one compound verbal form, and the auxiliary is
present. The auxiliary is dropped in the sentence in (105b), where the reference is made
to two different points in time. One of them is the speech time, the other one occurs
before it.
1.3.4.5.1.2 Beyond Bulgarian and Macedonian

Stieber (1973: 66-67) shows that some forms of the renarrated mood can also be found
in Czech, Polish, and Upper Sorbian, although they have a completely different origin
and are clearly not grammaticalized. For instance, in order to express the idea of non-
evidentiality, Czech uses the morpheme prý, which derives from the verb pravi ‘s/he
says’. The morpheme does not decline and cannot be used without a main verb, as
illustrated in (107).
(106) On prý je v Praze

he prý be3SG in Prague
“Reportedly, he is in Prague now”
(107) A. Byl Jan v Praze?

bePART.M.SG Jan in Prague
“Has Jan been to Prague?”
B. Prý *(byl)
Prý bePART.M.SG
“He probably was” (Czech, cf. Stieber 1973: 66-67)
In Polish, a related meaning can be expressed with the verb ‘to have’ or ‘have to’,
followed by an infinitive (cf. Świderska-Koneczna 1930 for more data).
(108) To miał/musiał być wielki pisarz

it have/mustPART.M.SG beINF great writer
“Reportedly, he was supposed to be a great writer”
(Pl, Stieber 1973: 66-67)
Likewise, in Upper Sorbian evidentiality can be rendered with a perfective form of the
verb ‘have’ in the present tense.24
(109) Nĕ wón drje z-mĕ-je je hižo hotowe

no he sure PFV-have3SG themACC already readyPL
“No, he’ll have finished it [them] already”
(Upper Sorbian, Tommola 2000: 453)
However, these are only indirect means of characterizing the meaning of evidentiality.
The renarrated mood cannot be expressed in a regular manner in other Slavic languages
than Bulgarian and Macedonian.
1.3.4.5.2 ‘Have’-perfects
The preceding section analyzed the construction formed with the auxiliary ‘to be’ and
the l-participle, which in Bulgarian and Macedonian expresses the renarrated mood.
The same form rendered resultativity in Old Church Slavonic. Macedonian and
Kashubian have developed a new structure in order to characterize the result of a past
action, which is composed of the auxiliary ‘have’ and an invariant form the passive
participle as the main verb. I will refer to it as the ‘have’-perfect, and I will describe it in
detail in chapter 3.
Apart from Macedonian, some other Slavic languages use a construction that
resembles the ‘have’-perfect. It will be discussed in section 1.3.4.5.2.2.
1.3.4.5.2.1 Macedonian
‘Have’ perfects will be analyzed in detail in chapter 3. Here I will present only a few
properties of the construction. It is formed with the auxiliary ‘have’, which is
complemented by the passive participle in the singular neuter form.
24 Recall from section 1.3.4.4.2 that in West and East Slavic perfective forms of the present tense
verbs have a future meaning.

(110) a. Imame kupeno knigi

have1PL buyPASS.N books
“We have bought books” (Mac, cf. Tomić 1996a)
The morphological form of the passive participle is invariant, and does not depend on
the feature specification of the subject of the clause. In this way the ‘have’-perfect
differs from the ‘be’-perfect, in which the l-participle obligatory agrees with the subject
in φ-features.
Similar forms involving the auxiliary ‘have’ are found in the neighbouring
languages in the area close to Macedonia. However, they differ in the gender of the
invariant passive participle. For example, it is feminine in Arumanian (cf. 111) and
masculine in Megleno-Romanian (cf. 112).25
(111) Am vidzută/vinită
have1SG seenPASS.F/comePASS.F
“I have seen/come” (Arumanian, Gallis 1960: 180)
(112) Nu l- am vizut di lunj

NEG himCL have1SG seePASS.M.SG from Monday
“I haven’t seen him since last Monday” (Tomić, 2006: 378)
A striking property of the construction in Macedonian is the fact that the auxiliary have
can be complemented by both transitive and unaccusative passive participles. This is
remarkable, because unaccusative verbs cannot be passivized. Thus, the form dojdeno in
(113) may appear exclusively as complement of the auxiliary have. It is never found as an
independent passive participle.
(113) Imam dojdeno

have1SG comePASS.N
“I have come” (Mac, Gołąb 1959)
The construction was acquired from non-Slavic languages of the area, most probably
from Arumanian, and was influenced by similar forms in Albanian and Greek (Gołąb
1959). It was registered for the first time in a manuscript from the monastery of Krnino
in 1706. The form is rarely used in literature, possibly because it is perceived as
colloquial. The spread of its usage varies across the country. The form is the least
common in the eastern areas, and it is the most widely used in south-western dialects,
particularly in the South-Western Macedonian dialects of Ohrid and Struga, where have
perfects can be formed even with the verbs ‘to be’ or to have‘ as participles.
25 Arumanian (or Macedoromanian) is a language spoken by the Vlach minority mainly in
Macedonia, but also Albania, Serbia, Bulgaria, and Greece. It is closely related to Romanian.
Megleno-Romanian is a Romanian dialect spoken in the Greek province of Meglen. It is more
closely related to Romanian than Arumanian, and is on the verge of extinction.
(114) a. Imam bideno vo ovoj grad

have1SG bePASS.N in thisM.SG town
“I have been to this town”
b. Imam imano vakov fustan
have1SG havePASS.N suchM.SG dress
“I have had a dress like this one”
(Ohrid and Struga dialects of Mac; Tomić 2006: 342)
As far as their meaning is concerned, ‘have’-perfects emphasize the result of an event

for the current state of affairs (Graves 2000: 483, Tomić 1989: 366). Thus, the example
in (113) carries the implicature of “I have arrived so I am present here”.
Correspondingly, the example in (115) carries the meaning of the perfect of experience;
that is, it indicates an event that occurred at least once until the moment of speaking.
(115) a. Dosega nemam jadeno vakva riba

until-now not+have1SG eatPASS.N such fish
“So far, I haven’t eaten such a fish” (Tomić 1989: 366)
The grammatical properties of the construction will be examined in chapter 3, where I

will also provide and analyze new data exemplifying ‘have’-perfects in Kashubian. In
the next section I will discuss properties of the ‘have’-perfects that have not been
grammaticalized yet.
1.3.4.5.2.2 Beyond Macedonian

The forms that resemble the ‘have’-perfect are common in a number of Slavic
languages, such as Polish (cf. 116a) and Czech (cf. 116b). However, these languages
never use ‘have’ as a true auxiliary, and the passive participle always agrees with the
object.
(116) a. Mam już upieczone ciasto

havePRES.1SG already bakePASS.N.SG cakeN.SG
“My cake is already baked” (Pl)
b. Mám úlohu napsanou
havePRES.1SG taskACC.F writePASS.F.SG
“I have my task written” (Czech, Maslov 1988: 80)
However, some researchers argue that these structures are slowly becoming reanalyzed
as compound tense forms. For instance, Pisarkowa (1984: 58) notices that initial traces
of the grammaticalization of have-perfects in Polish can be observed in the innovation
of non-obligatory case agreement between the passive participle and the complement of
the verb. Consider the dialogue in (117), with both variants of the answers equally
acceptable.
Conclusions 59
(117) A Słodził-eś herbatę?

SweetenPART.M.SG+AUX.2SG teaACC
“Have you put sugar in your tea?”
B Mam już posłodzoną (herbatę)
have1SG already sweetenPASS.ACC.F.SG teaACC.F.SG
B’ Mam już posłodzone (*herbatę)
have1SG already sweetenPASS.NOM/ACC.F.PL teaACC.F.SG (Pl)
In the answer in (117B), the passive participle agrees with an elided object in case and
φ-features. In (117B’) there is an agreement mismatch, and an overt realization of the
object results in ungrammaticality.
1.4 Conclusions
This chapter has overviewed the diachronic evolution of aspectual and temporal
distinctions in the Slavic languages. It has been shown that Proto Slavic inherited from
Proto-Indo-European a rather conservative way of rendering these distinctions
morphologically, further developed aspectual tenses, and in addition created a uniform
system of aspectual pairs of verbs. As a result, aspect was often “doubly marked”: via
aspectual past tenses and aspectual morphology on all verbal forms. The overlap in
aspect marking was the impetus for syntactic and semantic simplification of the system
of tenses.
All compound tenses in Old Church Slavonic except for Future I were constructed
with the l-participle as the main verb and the auxiliary ‘to be’. Both of them occurred in
aspectual pairs. Since all forms of the l-participle could appear in virtually all tenses, the
temporal interpretation of an event described by a compound tense depended entirely
on the aspect or tense of the auxiliary. For instance, when the auxiliary ‘to be’ was
specified for imperfective aspect, the tense was analyzed as pluperfect. When the
auxiliary ‘to be’ occurred in the perfective variant, it gave rise to a future perfect
interpretation.
The languages that evolved from Proto Slavic and Old Church Slavonic resorted to
different grammatical solutions in order to eliminate the excess of aspectual forms. The
South Slavic languages have largely retained the aspectual tenses, and kept the structure
of the compound tense formed with the auxiliary ‘to be’ and the l-participle intact. The
North Slavic languages have lost the aspectual tenses, and have reinterpreted the
present perfect as the default past tense. This process has been accompanied by a
morphological reduction of the auxiliary ‘to be’, which ranges from a diversification of
the copula and auxiliary paradigms in Czech, through a renanalysis of the auxiliary clitic
as an affix on the l-participle in Polish, to its complete disappearance in East Slavic.
Moreover, new structures have arisen: the “renarrated mood” in Bulgarian and
Macedonian, and the ‘have’-perfect in Macedonian and Kashubian. The subsequent
chapters will demonstrate that the development has syntactic consequences, which are
related to X0 versus XP-movement of the l-participle, and the emergence of a new type
of VP headed by the auxiliary ‘have’.
Chapter 2 The syntax of 'be'-perfects
and the l-participle
2.1 Introduction
This chapter develops an account of the syntax of the compound tense formed with the
l-participle. It has been mentioned that this is a structure peculiar to Slavic, which
exhibits a number of properties that are not often found among other Indo-European
languages. It is formed with the verb ‘to be’ used as the auxiliary in all contexts,
regardless of the transitivity of the l-participle. The l-participle is used exclusively in this
construction, so in contrast to the past participle in Germanic and Romance languages,
it never occurs as the passive participle. Moreover, it may appear in perfective and
imperfective variants, but it does not express any temporal specification on its own, as
it can be used in both past and future tenses. Morphologically, it always shows
agreement in number and gender with the subject of the clause.
The research questions that this chapter will deal with will include a comparison of
properties of the l-participle with the past participle in Germanic and Romance
languages, especially in relation to case and theta role assignment. The analysis will
determine the functions performed by the auxiliary ‘to be’ and the l-participle, and
show how they influence the structure of the VP in Slavic. Special attention will be paid
to the syntactic configuration that makes agreement between the l-participle and the
subject possible. The assumptions concerning the VP architecture will be used in an
alternative account of a widely-discussed l-participle movement across the auxiliary,
which is the main topic of this chapter.
The outline of the chapter is as follows. Section 2.2 examines some earlier
approaches to l-participle fronting, according to which the operation occurs via head
movement. Section 2.3 argues that the fronting should be treated as an instance of
locative inversion, and that the l-participle XP-moves into Spec, TP in order to check
the φ-features of T. The subsequent sections provide more support for this claim,
which comes from the behaviour of the l-participle in double participle constructions,
and the way it patterns with negation and the future auxiliary šte in Bulgarian.
The analysis focuses on Bulgarian and Serbo-Croatian. It was noted in chapter 1
that Macedonian also has a compound tense constructed with the auxiliary ‘be’ and the
l-participle. However, due to different patterns of cliticization, the l-participle undergoes
head movement in this language. I will address the issue in chapter 4, where I will also
show that although Macedonian lacks fronting of the l-participle, it exhibits a similar
process of inversion with passive participles, predicative nouns and adjectives.
Likewise, an analysis of compound tenses in Polish is postponed to chapter 5 for
the reasons mentioned in chapter 1. Namely, Polish has largely reanalyzed the l-
participle as a verbal root, whereas the forms of the auxiliary verb ‘be’ have been
morphologically reduced and are on the way to become affixes. This precludes the
possibility of l-participle fronting via locative inversion.
62 The syntax of 'be'-perfects and the l-participle
2.2 Head-movement accounts of participle

fronting in Bulgarian and Serbo-
Croatian
Chapter 1 showed that Slavic languages have developed a compound tense that does
not occur in any other Indo-European languages apart from the Tocharian and
Armenian groups. The tense is constructed with the auxiliary ‘be’ and the so-called
(resultative) l-participle. The l-participle can be formed from all types of perfective and
imperfective verbs, irrespectively of the fact whether they are unergative or
unaccusative. It always agrees in gender and number with the subject of a clause. For
instance, the l-participle of the verb stati ‘to stand’ in Serbo-Croatian has six possible
forms (cf. 1).
(1) Forms of the l-participle in Serbo-Croatian
SG PL
MASC stao stali
FEM stala stale
NEUT stalo stala
(S-C, Spencer 1991: 352)
The l-participle in Bulgarian has more variants than in Serbo-Croatian, because it

additionally distinguishes between the aorist and the imperfectum forms, which are
exemplified for the verb četa ‘read’ in (2).
(2) Forms of the l-participle in Bulgarian
aorist imperfectum
SG PL SG PL
MASC čel čeli četjal četjali
FEM čela čeli četjala četjali
NEUT čelo čeli četjalo četjali
(Bg, Tomić 2006: 351)
Moreover, both in Bulgarian and Serbo-Croatian the l-participle forms aspectual pairs.
For example, the imperfective variant of the aorist form of the verb četa ‘read’ in
Bulgarian is čel, while the perfective aorist is pročel.
The l-participle can be fronted across the auxiliary to the clause initial position.
This is shown for Bulgarian in (3) and for Serbo-Croatian in (4).
(3) a. Az sŭm čel knigata

I beAUX.PRES.1SG readPART.F.SG book-the
b. Čel sŭm knigata
readPART.F.SG beAUX.PRES.1SG book-the
“I have read the book” (Bg)
Head-movement accounts of participle fronting in Bulgarian and Serbo-Croatian 63
(4) a. Ja sam čitao knjigu

I beAUX.PRES.1SG readPART.M.SG book
b. Čitao sam knjigu
readPART.M.SG beAUX.PRES.1SG book
“I have read the book” (S-C)
The movement has been extensively discussed in the literature, but so far it has always
been analyzed as head raising, either as Long Head Movement from V to C (Lema &
Rivero 1989) or as head adjunction of the participle to C (Wilder & Ćavar 1994), to
Aux (Bošković 1997), or to a discourse-related focus projection Delta (Lambova 2003).
I have argued in previous work (Broekhuis & Migdalski 2003, Migdalski 2005) that the
head movement accounts face a number of empirical and theoretical problems and that
the fronting is in fact a case of remnant XP-movement. Before I elaborate on the XP-
movement proposal, I will first provide a brief evaluation of the previous accounts in
sections 2.2.1 and 2.2.2, which will also give me the opportunity to present the relevant
data.
2.2.1 Participle fronting as long head

movement
Lema and Rivero (1989) offer the first generative analysis of the participle-auxiliary
reorderings in Slavic. They argue that the operation consists in long head movement
(LHM) of the participle located in V across the auxiliary in I to the complementizer
projection C, as schematized in (5).26
(5) [CP [CParti] [IP Aux [VP [V ti] DP]]]
Lema and Rivero claim that the Part-Aux order is a result of head movement rather
than of XP-movement because the participle may be fronted only entirely on its own.
Thus, movement of a VP together with its complement is unacceptable.
(6) a. *Čel knigata e (Ivan)

readPART.F.SG book-the beAUX.3SG Ivan (Bg)
b. *Čitao knjigu je (Jovan)
readPART.M.SG book beAUX.3SG Jovan (S-C, cf. Rivero 1991: 322-323)
However, the analysis implies that the head movement crosses the head position
occupied by the intervening auxiliary head. Consequently, it violates the Head
Movement Constraint (Travis 1984, Baker 1988, Chomsky 1986), so it should be illicit.
Lema and Rivero try to solve this problem by deriving the HMC-restrictions from a
slightly complicated version of the ECP.
26 In her subsequent work, Rivero (1991, 1994) has extended the LHM analysis of verb raising to
other Southern and Western Slavic languages, as well as to Old Romance and Celtic languages
(see e.g. Borsley, Rivero, and Stephens 1996 for a discussion of Breton, as well as Roberts 1992,
1994). A detailed overview of head movement accounts of the participle-auxiliary structures in
Slavic can be found in Wilder & Ćavar (1994), Phillips (1996), Embick & Izvorski (1995, 1997),
and Lambova (2003).
Apart from the theoretical shortcoming, the LHM proposal also faces empirical
problems. First of all, the analysis wrongly predicts that the construction is restricted to
main clauses. This is what Lema and Rivero claim to be the case; however, Embick and
Izvorski (1995, 1997) report that the restriction is not valid, and mark the Bulgarian
example in (7a) as acceptable. The judgments are confirmed by Lambova (2003), who
provides the output in (7b).
(7) a. Rasbrax če pročel e knigata

understandPAST.1SG that readPART.M.SG beAUX.3SG book-the
“I understood you have read the book”
(Bg, Embick and Izvorski 1997: 216)
b. Decata kazvat če gledali sa filma
kids-the say3PL that watchPART.PL be3PL movie-the
“The kids say that they have seen the movie” (Bg, Lambova 2003)
Likewise, Embick and Izvorski (1997) notice that the restriction does not hold for
Serbo-Croatian either,27 because the l-participle may be fronted across the past tense
auxiliaries in this language.
(8) On tvrdi da istukao bejaše Jovan Petrovog prijatelja

he claims that beatPART.M.SG bePAST.3SG Jovan Peter’s friend
“He claims that Jovan had beaten Peter’s friend”
(S-C, Embick and Izvorski 1997: 216)
Secondly, the LHM approach wrongly predicts that the subject may be inserted
between the participle in C and the auxiliary in I. However, this is never the case, as has
been observed by Wilder & Ćavar (1994: 19-20) for Serbo-Croatian and Embick &
Izvorski (1995: 111) for Bulgarian.
(9) a. *[CP Čel [IP Ivan [I′ e [ ... knigata]]]

readPART.M.S Ivan be3SG book-the (Bg, cf. Embick & Izvorski 1995)
b. *[CP Čitao [IP Jovan [I′ je [ ... knjigu]]]
readPART.M.SG Jovan be3SG bookACC (S-C, cf. Wilder & Ćavar 1994)
To sum up, this section has shown that there are both theoretical and empirical
problems with the Long Head Movement analysis of participle fronting. They have led
27 Participle movement is impossible across the present tense auxiliaries in embedded clauses in
Serbo-Croatian, because they are clitics that must occur in the second position (see chapter 4 for
details). This means that the auxiliary je in (i) must be right-adjacent to the complementizer da.
The presence of the subject or the participle after the complementizer violates the second
position requirement, so such constructions are ruled out for independent reasons.
(i) a. Ivan kaže da je (Marija) čitala Krležu

Ivan says that bePRES.3SG Marija readPART.F.SG KrležaACC
“Ivan says that Mary has read Krleža”
b. *Ivan kaže da čitala je (Marija) Krležu
c. *Ivan kaže da Marija je čitala Krležu (S-C, Wilder & Ćavar 1994: 8)
Bulgarian does not exhibit second position requirements on the clitic auxiliaries; therefore the
examples comparable to (ia, b) are permitted.
Head-movement accounts of participle fronting in Bulgarian and Serbo-Croatian 65
to alternative head movement accounts involving adjunction, which will be addressed in

the next section.
2.2.2 Participle fronting as head adjunction

According to Wilder & Ćavar (1994), the main problem with the LHM account of
participle fronting is the fact that it is not able to predict that the preposed participle
must always be adjacent to the auxiliary that follows it. The adjacency requirement
holds for both Bulgarian and Serbo-Croatian, as shown in (10a and b), respectively.
(10) a. Pročel (*toj/pravilno ) e knigata

readPART.M.SG he/correctly bePRES.3SG book-the (Bg, Caink 1999)
b. Čitao (*Ivan/ rado) je knjigu
readPART.M.SG Ivan/ gladly bePRES.3SG book
(S-C, Wilder & Ćavar 1994)
Moreover, Wilder & Ćavar observe that in embedded clauses in Serbo-Croatian the
auxiliary clitic must be adjacent to the complementizer (cf. footnote 27). To account for
this they suggest that it is right-adjoined to C. Given that the fronted participle occurs
immediately to the left of the auxiliary, they conclude that it is also adjoined to C. The
derivation they propose is presented in (11b).
(11) a. Pio je Jovan pivo

drinkPART.M.SG bePRES.3SG Jovan beer
“Jovan drank beer”
b. [CP [C pioi je [AGRP Jovan ti pivo]]] (S-C, Wilder & Ćavar 1994)
Bošković (1995, 1997)28 adopts the adjunction analysis, but claims that the participle
cannot move as high as to C in Serbo-Croatian. For example, he observes that it is
impossible to raise the participle in front of the interrogative complementizer li, which
is standardly assumed to be in C across Slavic (cf. section 2.3.6.3.3 and chapter 4 for
details). By contrast, finite verbs may precede li, which means that this site is accessible
for finite verbs.
(12) a. [C Ljubi li] nju?

kissPRES.3SG Q her
“Does he kiss her?”
b. *[C Poljubio li] je nju?
kissPART.M.SG Q bePRES.3SG her
“Did he kiss her?” (S-C, Bošković 1995: 251)
Furthermore, Bošković (1995, 1997) investigates the positions taken by the l-participle
in the presence of different types of adverbs. He notices that although the participle
may precede VP-adverbs, such as juče ‘yesterday’ in (13a and b), it may not move across
sentential adverbs, such as nesumnjivo ‘undoubtedly’ in (13c).
28 Lambova (2003) extends the adjunction account to Bulgarian. I discuss her analysis in section
2.3.4.1.
(13) a. Vi ste mu je nesumnjivo

you beAUX.3PL himCL.DAT herCL.ACC undoubtedly
predstavili juče
introducePART.M.PL yesterday
“You undoubtedly introduced her to him yesterday”
b. Predstavilii ste mu je ti juče
introducePART.M.PL beAUX.3PL himCL.DAT herCL.ACC yesterday
c. *Predstavilii ste mu je nesumnjivo ti juče
introducePART.M.PL beAUX.3PL himCL.DAT herCL.ACC undoubtedly yesterday
(S-C, Bošković 1997: 148-149)
Following Watanabe (1993), Bošković assumes that sentential adverbs are universally
TP-adjoined. Since the l-participle may not cross sentential adverbs, he concludes that
the participle is not adjoined to C, but rather head-adjoined to the auxiliary, which
resides in Aux0 (cf. 14).
(14) [AuxP Poljubioi je] [VP ti Mariju]]

kissPART.M.SG beAUX.3SG MariaACC
“He kissed Maria” (S-C, Bošković 1997: 156)
In contrast to the Long Head Movement hypothesis, the adjunction analyses avoid the
theoretical problems related to the HMC violation. They also correctly predict that the
reordering is possible in embedded clauses (cf. the examples in 7 and 8), and that the
subject may not intervene between the fronted participle and the auxiliary (cf. the
sentences in 9). However, the adjunction accounts are unable to preclude the option of
locating the subject in Spec, IP, that is in front of the preposed participle. As (15)
indicates, this type of ordering is ungrammatical.
(15) *[IP Jovan [AuxP poljubioi je] [VP ti Mariju]]]

Jovan kissPART.M.SG beAUX.3SG Maria (S-C)
In fact, it seems that the fronted participle must normally be the left-most constituent
in its clauses and may not be preceded by other elements. The only elements that may
exceptionally appear in front of it are topics, which reside in the left periphery of the
clause above TP (cf. section 2.3.4.3).
(16) Filma gledali bjaxa decata

movie-theTOP watchFOC.PART.M.SG bePAST.3PL kids-the
“As for the movie, the kids had seen it” (Bg, cf. Lambova 2003)
The restriction in (15) in Serbo-Croatian could be explained via the requirements of

clitics, which must appear in the second position in this language. However, the
explanation does not hold for Bulgarian, where clitics are not subject to the second
position constraint.
In sum, I have established that neither the LHM nor the adjunction account offer a
satisfactory explanation of the properties of participle preposing in Bulgarian and
Serbo-Croatian. In the next section I will propose an alternative analysis, which will
show that participle fronting occurs as XP-movement.
Towards an alternative analysis 67
2.3 Towards an alternative analysis

This section will develop an XP-movement account of participle fronting in Bulgarian
and Serbo-Croatian. It will begin with a description of some core properties of the
compound tenses in South Slavic of a general nature that the analysis will rest on in
section 2.3.1. Section 2.3.2 will provide an analysis of participle movement in terms of
locative inversion. Section 2.3.3 will explain why the l-participle must move entirely on
its own and may never pied-pipe any lexical material. Finally, section 2.3.4 will
investigate discourse effects created by movement of the l-participle and other elements
to the left periphery.
2.3.1 Properties of the l-participle

A remarkable property of the l-participle is the fact that it always agrees in φ-features
with the subject of the clause. This happens irrespectively of whether the participle is
unaccusative or unergative. The examples in (17) illustrate this for the unergative
participle čel, whose gender and number specification depends on the subject of the
clause.
(17) a. Ivan e čel knigata

b. Polja e čela knigata
Polja bePRES.3SG readPART.F.SG book-the (Bg)
I take the obligatory agreement between the subject and the participle to be a crucial
property of the construction. It makes the Slavic languages significantly different from
the Romance languages, where the agreement obtains only in unaccusative and passive
structures; that is, when the subject is an internal argument of the verb. I would like to
propose that the contrast reflects a difference in the syntactic composition of
compound tenses that contain unergative participles. In the case of unergative
constructions in Slavic, which are formed with the auxiliary ‘be’, the subject is the
external argument of the l-participle. In the case of unergative constructions in
Germanic and Romance, which are formed with the auxiliary ‘have’, the subject is
generated as the external argument of the auxiliary verb. The templates representing the
compound tenses formed with the auxiliaries ‘be’ and ‘have’ are given in (18a) and
(18b), respectively.
(18) a. ...[ be [vp DPagent v [VP/PartP VPART DPtheme]]] Slavic

b. ...[vp DPagent v [VP have [VP VPART DPtheme]]] Romance and Germanic
The assumption I make here is related to the proposals by Hoekstra (1984, 1986),
Roberts (1987), and Broekhuis and Van Dijk (1995), who postulate that only the
auxiliary ‘have’ is able to introduce an agent and assign accusative case to the object.
The verb ‘be’ is an unaccusative auxiliary, so it may not perform these functions.
The structure in (18b) implies that the past and passive participles in Germanic and
Romance languages must be conflated, that is, analyzed as categories of the same type29
(cf. Hoekstra 1994; Emonds 2000). Hence, it is argued that the idea that passive
participle morphology suppresses the external theta role and absorbs structural case (cf.
Den Besten 1981; Jaegli 1986, and Roberts 1987) should be extended to past participles
as well. In this scenario, the function of the auxiliary ‘have’ is the reintroduction of the
agent and assignment of accusative case.
The l-participle is always accompanied by the verb ‘to be’. Since this is an
unaccusative auxiliary, it is unable to assign accusative case or project an external theta
role. This means that these functions are performed by the l-participle. The assumption
is reflected in the structure in (18a). It implies that in constructions with the auxiliary
‘be’ the subject is the external argument of the l-participle, rather than of the auxiliary.
The fact that the l-participle is a case assigner indicates that it has rather different
properties from past participles in Germanic and Romance languages. It also means
that the l-participle in Slavic cannot be conflated with the passive participle, and that
the two categories may not receive a uniform analysis. Recall from chapter 1, section
1.3.3.5 that this is a priori confirmed by the fact that the two types of participles are
morphologically different in the Slavic languages. Thus, the passive participle form of
the verb ‘read’ in Bulgarian is četan+AGR, as exemplified in (19).
(19) Knigata e četana ot Ivan

book-the bePRES.3SG readPASS.F.SG by Ivan
“The book is being read by Ivan” (Bg)
Summarizing, this section has shown that the l-participle has different properties from
the past participles in the Germanic and the Romance languages. Even though it always
occurs with the auxiliary ‘be’, it is a case assigner, and is able to project an external theta
role. In the next section I will demonstrate that these properties have direct
repercussions for participle fronting.
2.3.2 Participle fronting as locative inversion

In chapter 1, section 1.3.3.5.1.2 I claimed that the l-participle in Slavic derives from a
class of Proto-Indo-European *lo-adjectives, which denoted propensity for performing
certain actions. Even though in the contemporary Slavic languages l-participles are
never used as adjectives, they still exhibit some adjectival properties. Most notably, they
show agreement in gender and number with the subject, and in South Slavic they
appear with the auxiliary that has the same form as the copula ‘be’. This suggests that
the subject and the participle phrase (VP) in (18a) are in a canonical Small Clause
configuration (cf. Hale & Keyser 1993). The configuration is overtly manifested
through agreement on the l-participle. Since there is no Small Clause relation in (18b),
the agreement between the subject and the past participle is impossible.
I would like to argue that l-participle fronting is contingent on subject-participle
agreement. Since in the minimalist framework syntactic movement is conceived of as a
“last resort” procedure, the operation may only apply if it results in feature checking.
Given that both the subject and the l-participle are marked for φ-features, they are both
29This view will be discussed in more detail in chapter 3, where I develop an analysis of ‘have’-
perfects in Slavic.
eligible candidates for checking the φ-features of T by movement to Spec, TP. They
raise to a specifier position, which means that both of them undergo phrasal
movement.
(20) a. [TP Ivan[+φ] [T[+φ] e [ ... pročel[+φ] knigata]]

b. [TP Pročel[+φ] [T′[+φ] e [ ... knigata ...]]
The proposal adopted here is closely related to the analysis of locative inversion in
English pursued by Hoekstra and Mulder (1990). The construction is exemplified in
(21). Following the standard assumptions of the 1980’s, Hoekstra and Mulder assume
that movement of the subject to Spec, IP in (21a) is motivated by the Case Filter,
because I assigns nominative case to the subject hosted in its specifier. It is, however,
less clear what triggers the fronting of the PP in (21b). In fact, the output should be
ungrammatical, given that the subject follows the verb, so it cannot be directly assigned
nominative case by I.
(21) a. The baby carriagei rolled [SC ti down the hill]

b. Down the hilli rolled [SC the baby carriage ti]
Hoekstra and Mulder solve this problem by proposing that locative inversion makes it
possible for I to assign case to the subject of the clause in its base position. Their
solution relies on the assumption that the fronted PP and the subject are in a
predicative Small Clause relationship, and that this enables transmission of nominative
case. Case assignment to the subject in (21b) occurs in the following way: I assigns
nominative case in the normal way to Spec, IP so that the fronted PP is assigned
nominative case by I. This case is then transmitted via the movement chain to the trace
of the preposed PP. Finally, nominative case is transmitted to the subject of the clause,
which is the external argument of the Small Clause.
Hoekstra and Mulder claim that the transmission of case from the locative PP in
Spec, TP to the subject in postverbal position is possible, because all relevant relations
(case assignment, movement and predication) involve co-indexing. Given that each
element can have a single index at most, it follows that I is also co-indexed with the
postverbal subject, and thus can assign case to it: I is co-indexed with the fronted PP
under case assignment, the PP is co-indexed with its trace, and the PP trace is co-
indexed with the DP under predication.
(22) a. Down the hill rolled the baby carriage

b. [IP PPi [I I [VP V [SC DPi ti ]]]]
Even though co-indexing does not play a role in the current syntactic theory, I will
follow some of Hoekstra and Mulder’s assumptions and propose that l-participle
movement is an instance of locative inversion.30 However, in the case at hand, the
movement is not conditioned by case assignment, but rather by agreement.31 Since both
the subject and the l-participle carry the appropriate φ-features, either of them may
check the φ-features of T by XP-raising to Spec, TP.
30 There are some other accounts of verb movement in terms of locative inversion available in
the literature. See, for example, Massam (2001a and b) for a study of Niuean.
31 See Broekhuis (2005) for a reanalysis of Hoekstra and Mulder’s proposal in terms of agreement
feature sharing.
One of the properties of locative inversion is the fact that it is possible only across
the verb ‘be’ and a few other unaccusative verbs (cf. 23a and b). It is incompatible with
verbs assigning external theta-roles (cf. 23c).
(23) a. On that table was put a valuable book

b. Crashing through the woods came a wild boar
c. *Down the street walked the old nanny her dog (cf. Bresnan 1994: 77-79)
I suggest that in Bulgarian and Serbo-Croatian l-participle movement may occur via
locative inversion because the auxiliary is always the verb ‘to be’. More evidence for this
claim will be given in chapter 5, where I will show, taking the structure of compound
tenses in Polish as an example, that once the auxiliary is impoverished and starts to
morphologically differentiate from the verb/copula ‘to be’, the l-participle may not raise
via locative inversion any more.
There are two empirical arguments that directly support the XP-movement analysis
of participle fronting (more will follow later in this chapter). First, this proposal predicts
that the fronted participle and the subject are in complementary distribution, as they
compete for the same position. The examples in (24) show that the prediction is borne
out. This fact makes the new analysis superior to the head movement proposals,
because they are unable to explain the exclusive distribution of these elements.
(24) a. *Čel Ivan e knigata

readPART.M.SG Ivan bePRES.3SG book-the
b. *Ivan čel e knigata
Ivan readPART.M.SG bePRES.3SG book-the (Bg)
Secondly, the analysis accounts for the fact that the fronted participle must be rigidly
left-adjacent to the auxiliary, as shown in (25).
(25) a. Pročel (*ne /pravilno /kŭde) sŭm/ bjax knigata

readPART.M.SG NEG/correctly/where bePRES.1SG bePAST.1SG book-the
(Bg, cf. Caink 1999)
b. Čitao (*ne/ rado) sam knjigu
readPART.M.SG NEG/gladly bePRES.1SG book
(S-C, cf. Wilder & Ćavar 1994)
The adjacency requirement follows straightforwardly on the assumption that the

auxiliary verb must raise to T in order to check Tense. Since the l-participle lands in
Spec, TP, it must be left-adjacent to the auxiliary.
To summarize, it has been argued that the l-participle undergoes XP movement to
Spec, TP in order to check the φ-features of T0. The operation was claimed to be an
instance of locative inversion. The analysis is supported by the complementary
distribution between the subject and the fronted l-participle, and the adjacency between
the fronted participle and the auxiliary. More empirical arguments for the idea of l-
participle raising via XP-movement will be provided later in this chapter. In the
meantime I will explain how the operation proceeds in more detail.
2.3.3 Details of the present analysis

It has been established that the l-participle is able to assign accusative case and project
the external theta role. This means that the subject is underlyingly the external
argument of the l-participle and resides in Spec, vP, whereas the auxiliary ‘to be’ is
located above the subject. For convenience, I assume that the auxiliary is generated as
the Aux head (cf. 26). More specific assumptions concerning the position of the
auxiliary will be presented in chapter 4.
(26) [TP ... T[+φ] ... [Aux BE [vP subject[+φ] v [PartP Part[+φ] object]]]]
The participle moves into Spec, TP, which means that the movement is phrasal. This
raises the important question of why the participle is not able to pied-pipe the object,
prepositional phrases, or any other material to Spec, TP. This is certainly not a trivial
issue, because it was the main reason why participle fronting was analyzed as head
movement in all the previous accounts. The next subsections will address this problem
in detail. The motivation for the lack of the movement of adjuncts will be provided in
section 2.3.3.1, whereas the impossibility of the pied-piping of the internal arguments
will be investigated in section 2.3.3.2.
2.3.3.1 Movement of adjuncts

The examples in (27) show that the participle may not pied-pipe adjuncts when it
moves to the clause-initial position. Placement of the locative PP v Plovdiv ‘in Plovdiv’ in
front of the l-participle or to its right results in ungrammaticality.
(27) a. Učil e v Plovdiv

studyPART.M.SG beAUX.3SG in Plovdiv
“He studied in Plovdiv”
b. *Učil v Plovdiv e
c. *V Plovdiv učil e (Bg, S. Marinov, p.c.)
I would like to suggest that the restriction is related to the size of the moved
constituent. The template in (26) indicates that the PartP is generated very low in the
structure, so that the only elements that it contains apart from the participle are its
internal arguments. Consequently, the participle is not able to pied-pipe any adjuncts,
because all of them are generated above PartP. Thus, the adverbial PP in (27) is located
too high in the clause structure to be affected by the movement of the PartP. As an
illustration, the derivation of the sentence in (27a) is given in (28). The sentence
contains a pro subject, which is marked in the template accordingly.
(28) a. [TP [T[+φ] ... [Aux e [AdvP v Plovdiv [vP pro[+φ] v [PartP učil[+φ]]]]]]
b. [TP [PartP učil] [T ej [Aux tj [AdvP v Plovdiv [vP pro[+φ] v [tPartP ]]]]]
As the scheme in (28a) shows, the auxiliary e raises as a head from Aux to T, where it
checks Tense.32 Subsequently, the l-participle undergoes phrasal movement from PartP
to Spec, TP, where it checks the φ-features of T. As expected, the adverbial v Plovdiv is
left stranded at the end of the clause.
2.3.3.2 Movement of internal arguments

The situation is more complicated when the PartP contains an object, which can never
be pied-piped by the l-participle, either. I propose that the restriction is related to the
accessibility of the φ-features, which are carried by the l-participle, for checking the φ-
features of T. Thus, if we assume the clause structure in (26), it is evident that
movement of a constituent that is larger than PartP is precluded, because otherwise the
features on the l-participle would be too deeply embedded to check the φ-features of T.
Percolation of the φ-features from Part would not solve the problem, either, because
they are blocked by the lexical head v.
Furthermore, the participle may undergo movement to Spec, TP only if the object
has been moved out of PartP. I suggest that movement of the whole PartP is barred,
because this would raise the direct object across its case checking position and thus
leave the uninterpretable case feature unchecked. In order to avoid this, the object must
first be evacuated out of PartP and check its case, which results in the requirement of
the remnant movement of PartP to Spec, TP. Hence, the sentence in (29a) must be
derived in the way depicted in (29b-c).
(29) a. Gledali sa filma decata

watchPART.M.SG bePRES.3PL movie-the children-the
“The kids have watched the movie”
b. [TP [T[+φ] ... [Aux sa [AgrO [vP decata [+φ] v [PartP gledali[+φ] filma]]]]]
c. [TP [PartP gledali tk] [T saj …[Aux tj [AgrO filmak [vP decata [+φ] v [tPartP tk]]]]]
The head T in (29b) contains uninterpretable φ-features, which can be checked by the
elements that have them if they raise to Spec, TP. The only candidates available are the
subject decata and the l-participle gledali. They agree with each other in φ-features, which
as was argued earlier signifies that they form a Small Clause. Suppose that the l-
participle is selected as the element to be moved. Observe that we cannot move the
whole PartP to Spec, TP because that would raise the direct object filma across the
position where its case is checked, that is, Spec, AgrOP. Hence, it is necessary to raise
the object out of PartP into Spec, AgrOP first. Subsequently, the auxiliary clitic sa
moves to T and checks Tense. Finally, the remnant PartP raises to Spec, TP and checks
the φ-features of T.
Undoubtedly, there are more issues that need to be explained in some detail. First
of all, since I claim that the object shift occurs for case checking, it is necessary to
prove that the object undergoes A-movement, rather than A’-movement. Moreover, it
is desirable to find independent evidence of object shift in Bulgarian and Serbo-
Croatian also in other syntactic contexts.
Secondly, it has already been explained that PP adverbials cannot be fronted
together with the l-participle, because they are generated above PartP, so they are
32 Alternatively, the auxiliary may be generated directly in T. Neither of the options has any
bearing on the analysis developed here. The placement of the auxiliary and pronominal clitics in
the clause structure will be discussed in chapter 4.
outside the domain of the movement. However, the situation is more complicated with
indirect objects that are preceded by prepositions. It might be difficult to maintain that
they are evacuated out of the PartP for case checking on a par with direct objects,
because their case can be checked by the preposition. Note though that this is an issue
only for Bulgarian, which does not have case distinctions on non-pronominal forms,
therefore it has to introduce the indirect object with a preposition. Serbo-Croatian
overtly marks indirect objects with dative case, so it does not use prepositions in this
context.
These issues are complex enough to deserve an independent treatment; therefore I
will address them separately. Sections 2.3.3.2.1.2 and 2.3.3.2.1.3 will analyze direct and
indirect object shift, respectively. Section 2.3.3.2.1.4 will be devoted to movement of
the indirect object that is preceded by a preposition in Bulgarian. Section 2.3.3.2.1.5 will
discuss the evacuation of PP and small clause complements out of the PartP.
2.3.3.2.1.1 Object shift requirement

There is a lot of independent evidence for object shift in Bulgarian and Serbo-Croatian.
The examples in (30) and (31) indicate that both languages permit scrambling, and that
the objects may be moved in order to reflect different information structure relations.
The Bulgarian data in (30) show that both the direct and the indirect objects are able to
raise out of the VP, because they may move across the temporal adverb včera. Likewise,
the direct object knjigu in the Serbo-Croatian examples in (31) may be placed in front of
the participle, or even preposed into the clause initial position.
(30) a. Ivan e podaril včera kartina(ta) na Maria

Ivan bePRES.3SG givePART.M.SG yesterday painting(-the) to Maria
“Ivan has given a/the painting to Maria yesterday”
b. Ivan e podaril kartina(ta) včera na Maria
c. Ivan e podaril na Maria včera kartina(ta) (Bg, Lambova 2003)
(31) a. Petar je kupio knjigu

Petar bePRES.3SG buyPART.M.SG bookACC
“Petar bought a book”
b. Petar je knjigu kupio
c. Knjigu je kupio Petar (S-C, Stjepanović 1999)
However, it is necessary to establish whether the object shifts occurs as A or A’-

movement. Since I assume that the object is evacuated from the Participle Phrase for
case checking, it is expected that this be done by A-movement.33
2.3.3.2.1.2 Direct object shift

Bošković (1997: 121ff) provides evidence for object shift in Serbo-Croatian. He claims
that the object must raise out of VP, and that the movement is followed by raising of
the participle into AgrO0.
33 A-movement can normally be followed by A’-movement, so this does not imply that the word
orders in (30) and (31) are all derived by A-movement. It is, however, important to determine
that the object moves out of the PartP/VP via A-movement.
(32) Jovan je poljubioi [AgrOP Marijuj ti [VP ti tj ]]

Jovan beAUX.3SG kissPART.M.SG Marija
“Jovan kissed Maria” (S-C, Bošković 1997: 121)
Following Pesetsky (1989), Bošković draws evidence in favour of the object shift from
the interpretation of adverb scope. He argues that the scope of adverbs is sensitive to
the direction of their adjunction to the verb. Given the standard assumptions, scope is
calculated hierarchically, so a higher adverb should c-command a lower adverb. In (33a)
twice takes scope over intentionally, which means that there were two instances of
intentional knocking, and that the adverbs are right adjoined to the VP. In (33b)
intentionally takes scope over twice, and the sentence describes a single intention of
knocking twice. This suggests that the adverbs are left-adjoined to the VP.
(33) a. John [[[ knocked on the door] intentionally ] twice]

twice > intentionally
b. (?)John [intentionally [twice [knocked on the door ]]]
intentionally > twice
Bošković points out that the judgments concerning the relative scope interpretation of
adverbs give some clues for determining whether the elements that are base-generated
within VP have been evacuated out of the VP. He provides the following examples.
(34) a. Jovan je namerno dva put oborio Petra

Jovan beAUX.3SG deliberately twice failPART.M.SG Peter
namerno>dva put
b. Jovan je oborio Petra namerno dva put
Jovan beAUX.3SG failPART.M.SG Peter deliberately twice
namerno><dva put (S-C, Bošković 1997: 122)
In sentence (34a) the first adverb takes scope over the second. By contrast, (34b) is
ambiguous in terms of the adverb scope, because both the reading on which the first
adverb dva put ‘twice’ takes scope over the second adverb namerno ‘deliberately’, as well
as the reading on which namerno takes scope over dva put are available. Bošković submits
that on the latter interpretation, both the participle and the object must have raised
across the adverbs.
As far as the landing site of the moved object is concerned, Bošković (1997: 123)
claims that it must be an A-position. He concludes this on the basis of quantifier float
data. If Sportiche (1988) and Déprez (1989) are correct when suggesting that only A-
movement can float quantifiers, then the object must target an A-position. Bošković
proposes that it is Spec, AgrOP.
(35) a. Jovan je oborio studente sve

Jovan beAUX.3SG failPART.M.SG students all
“Jovan failed all the students”
b. Jovan je oborio sve studente (S-C, cf. Bošković 1997: 123)
Correspondingly, Stjepanović (1999: 81) remarks that binding facts also indicate that
the object may move out of the VP. This is shown in (36), where the direct object
Gorana i Petra ‘GoranACC and PetarACC’ is able to bind the anaphor jednog drugom ‘each
other’ inside the VP temporal adverbial.
(36) Slikao je Gorana i Petra za vrijeme

photographPART.M.SG beAUX.3SG Goran and Petar during
sudjenja jednog drugom
trials oneGEN anotherDAT
“He photographed Goran and Petar during each other’s trials”
(S-C, Stjepanović (1999: 81)
Assuming the standard idea that adverbs are base-generated higher than objects, and
following Lasnik’s (1995) proposal that feature movement does not feed binding,
Stjepanović argues that this indicates that the object has A-moved overtly out of the
VP.
Summarizing, this section has shown that direct objects must raise out of the VP
via A-movement. The next section will discuss indirect object shift.
2.3.3.2.1.3 Indirect object shift

We have established that the direct object undergoes movement out of VP in Serbo-
Croatian. The same seems to be true of indirect objects as well. Stjepanović (1999) uses
the familiar tests related to adverb scope and shows that indirect objects may raise
above VP adjuncts together with direct objects in ditransitive constructions.
(37) a. Marija je namjerno dva put pokazala Vesnu Igoru

Marija beAUX.3SG intentionally twice showPART.F.SG VesnaACC IgorDAT
“Marija intentionally twice showed Vesna to Igor”
namjerno > dva put
b. Marija je pokazala Vesnu Igoru namjerno dva put
Marija beAUX.3SG showPART.F.SG VesnaACC IgorDAT intentionally twice
“Marija showed Vesna to Igor intentionally twice”
namjerno > < dva put (S-C, Stjepanović 1999: 84)
As the examples illustrate, the first adverb namjerno in (37a) has scope over the second
adverbial dva put. However, example (37b) is ambiguous with respect to adverb scope,
because both the reading with dva put having scope over namjerno, as well as the reading
with namjerno having scope over dva put are available. This indicates that both the l-
participle and the two objects have moved out of the VP. Thus, the results of scope
interactions for constructions with two objects are the same as for the structures with
just a direct object in (34).
Likewise, the distribution of quantifier float with double objects matches the
behaviour of quantifier float in the single object constructions. As Stjepanović shows,
indirect objects can float quantifiers as well (cf. 38). This indicates that they also target
an A-position.
(38) a. Marija je podijelila kolace prijateljima svim

Marija beAUX.3SG givePART.F.SG cookiesACC friendsDAT allDAT
“Marija gave cookies to all friends”
b. Marija je podijelila kolace sve prijateljima
Marija beAUX.3SG givePART.F.SG cookiesACC allACC friendsDAT
“Marija gave all cookies to her friends” (S-C, Stjepanović 1999: 85)
Furthermore, Stjepanović demonstrates that the indirect object can also bind an
anaphor in an adverbial phrase, matching the behaviour of the direct object in (36).
This suggests that the indirect object may raise to an A-position above the PartP as
well.
(39) Marija je predstavila Vesni Petra i Markai

Marija beAUX.3SG introducePART.F.SG VesnaDAT PetarACC and MarkoACC
za vrijeme sudjenja jednog drugomi
during trial oneGEN anotherDAT
“Marija introduced Vesna to Petar and Marko during each other’s trials”
(S-C, Stjepanović 1999: 85)
To sum up, it has been demonstrated that both the direct and the indirect objects raise
out of their VPs in Serbo-Croatian via A-movement. However, it is still necessary to
find evidence for the movement in Bulgarian, which is different from Serbo-Croatian,
because it does not exhibit case distinctions on non-pronominal objects and requires
that indirect objects be introduced by a preposition. Since prepositions can be case
assigners, the movement of indirect objects might be excluded, because potentially
there is no need for it.
2.3.3.2.1.4 Movement of the indirect object PPs in Bulgarian

This section will provide arguments for the idea that the indirect object in Bulgarian
moves out of the PartP for case checking. At first sight this is unexpected, because the
indirect object is introduced by the preposition na from which potentially it should
receive case. However, on the basis of data from other Slavic languages I will show that
na is not a preposition, but a dative case realization.
According to Arnaudova (2003), both objects must move from their base positions
in Bulgarian. She analyzes clauses with subjects in the final position, such as the one in
(40).
(40) Včera pročete knigata Marija

yesterday readPAST.3SG book-the Marija
“MARIJA read the book yesterday” (Bg, Arnaudova 2003: 115-116)
Arnaudova assumes that the low subject is in Spec,vP, and that the object must move
across it. Following Ordóñez’s (1998) observations for Spanish, she argues that the
requirement is confirmed by quantifier binding facts. Recall that a pronoun may be
interpreted as a variable bound by a quantifier only if it is in the c-command domain of
that quantifier (cf. Reinhart 1983: 122). Accordingly, the sentence in (41a) is ill-formed,
because the indirect object quantified phrase na vsjako dete ‘every child’ does not bind
the possessive pronoun negovoto of the subject. However, once the quantified PP moves
into a position preceding the subject, the sentence becomes grammatical, which argues
in favour of A-movement of the indirect object.
(41) a. ??Kakvo zanese negovotoi drugarče [na vsjako dete]i?

what bringPAST.3SG POSS.PRON friendDEF to every child
“What did his friend bring for every child?”
b. Kakvo zanese [na vsjako dete]i negovotoi drugarče?
what bringPAST.3SG to every child POSS.PRON friendDEF
(Bg, Arnaudova 2003: 115-116)
The binding facts suggest that the object quantifier in FP c-commands the subject in
Spec,vP, as indicated in the phrase structure in (42); cf. Ordóñez (1998: 320). However,
the binding requirements may be satisfied only if the object moves across the subject in
Spec, vP (cf. 41b).
(42)
FP
Spec F’
F vP
[na vsjako dete]i

Spec v’
V ti
negovotoi drugarče
Embedded clauses exhibit the same type of asymmetry. The sentence in (43a) is
ungrammatical, because the indirect object quantifier vseki does not c-command the
subject bašta mu ‘his father’. The sentence in (43b), by contrast, is well-formed, because
vseki binds the subject.
(43) a. *Mislja če [bašta mui] na vsekii e dal

think1SG that father himDAT to everybody bePRES.3SG givePART.M.SG
po nešto
PARTIT something
“I think that his father has given something to everybody”
b. Mislja če na vsekii [bašta mui] e dal
think1SG that to everybody father himDAT bePRES.3SG givePART.M.SG
po nešto
PARTIT something (Bg, Arnaudova 2003: 115-116)
Note that when the relationship between the binder and the bindee is reversed, so that
the object contains the possessive pronoun and the subject is a quantifier binder, there
is no asymmetry of the type depicted in (41) and (43).
(44) a. Kakvo podari [vsjaka majka]i [na nejnoto dete]i ?

what givePAST.3SG every mother to her child
“What did every mother give to her child?”
b. Kakvo podari [na nejnoto dete]i [vsjaka majka]i ?
what givePAST.3SG to her child every mother
(Bg, Arnaudova 2003: 115-116)
Thus, the subject may bind the possessive pronoun of the object whether it follows or
precedes the subject binder. Ordóñez (1998) accommodates all the cases that do not
show the asymmetry by reconstruction. In (41a) and (43a) the binding conditions
concerning the occurrence of possessive pronouns are not met, and they cannot be
repaired by reconstruction. Conversely, when the subject c-commands the object and
contains the quantifier in the initial stage of the derivation (cf. 44b), reconstruction
takes place, because even when the indirect object raises, the tail of the chain is c-
commanded by the subject.
Recall that reconstruction is associated with A’-movement. Since there is no
reconstruction taking place in the examples in (41) and (43), the data indicate that the
indirect object raises via A-movement in Bulgarian.34
Summarizing, I have shown that both indirect and indirect objects are evacuated
out of the PartP in Bulgarian and Serbo-Croatian via A-movement. I have suggested
that the evacuation takes place for case checking in the Agr projections above PartP.
Surprisingly enough, the generalization holds even for indirect objects in Bulgarian,
which are introduced by the preposition na. They are not expected to raise out of PartP,
because their case can be checked by the preposition. However, they show the same
type of movement as the preposition-less indirect objects in Serbo-Croatian. In view of
this, I propose that na is not a true preposition, but rather it is an alternative realization
of dative case.35 The idea receives more support from clitic doubling in the presence of
na in Macedonian, which is a language very closely related to Bulgarian.36
In Macedonian definite direct objects and specific indirect objects are doubled by
clitics. This is exemplified in (45): the direct object pismoto ‘the letter’ is doubled by the
accusative clitic go, whereas the indirect object PP is doubled by the dative clitic mu.
(45) Jana mu go dade pismoto na deteto

Jana himCL.DAT itCL.ACC givePAST.3SG letter+the to child+the
“Jana gave the letter to the child” (Mac, Tomić forthcoming)
Na is a preposition that always introduces dative objects. It is also the only preposition
in Macedonian that is compatible with clitic doubling (cf. 46a). The elements that are
introduced by other prepositions, such as za in (46b) may not be doubled by clitics.
(46) a Im zboruvav na decata

themCL.DAT talkPAST.1SG to children-the
“I was talking to the children”
b. (*Im) zboruvav za decata
themCL.DAT talkPAST.1SG about children-the
“I was talking about the children” (Mac, Berent 1980: 152; 174)
It is standardly assumed in the literature that clitic doubling is a means of case checking
for the objects that are associated with the clitics (cf. chapter 4 for details). In this
scenario, the fact that the dative PPs are clitic-doubled in Macedonian implies that they
require case checking by doubling, and that they may not receive case from the
preposition na. This indicates that na is not a preposition, but the realization of dative
case.
34 Note that it has also been observed that A-movement need not destroy binding possibilities, as
there are instances of reconstruction associated with A-movement (cf. Hoekstra 1991, Fox 1999).
However, it is necessary to assume that movement of arguments in Bulgarian is subject to
reconstruction, as otherwise the contrast between (41) and (44) cannot be maintained.
35 See Asbury (2005) for a similar proposal for Hungarian.
36 Clitic doubling will be analyzed in detail in chapter 4. Here I only make a brief reference to the
phenomenon.
2.3.3.2.1.5 Movement of PP- and small clause complements

The preceding sections discussed raising of direct and indirect objects out of the PartP.
It has been demonstrated that these elements always undergo A-movement, even when
the indirect object is introduced by the preposition na. However, it is still necessary to
account for the evacuation of PP complements, such as na vlaka ‘for a train’ and na riba
‘for a fish’ (cf. 47 and 48), as they may not be pied-piped by the l-participle when it
moves to Spec, TP.
(47) a. Ivan e čakal na vlaka

Ivan bePRES.3SG waitPART.M.SG on train
“Ivan has waited for the train”
b. *Čakal na vlaka e Ivan
waitPART.M.SG on train bePRES.3SG Ivan (Bg, P. Vitkova, p.c.)
(48) a. Otišŭl beše na riba

goPART.M.SG bePAST.3SG on fishing
“He had gone fishing”
b. *Otišŭl na riba beše
goPART.M.SG on fishing bePAST.3SG (Bg, cf. Lambova 2004: 239)
In principle, since the PP complements are introduced by the same preposition as the
indirect object, it might be possible to claim that here na is not a preposition, but a case
realization. The problem with this assumption is that PP complements do not have to
be doubled by clitics in standard Macedonian (cf. 49a), or they are doubled when they
are not introduced by a preposition (cf. 49b). Therefore, it is difficult to determine
whether na in (47) and (48) represents a case realization, or whether it is a true
preposition.
(49) a. Otidov na odmor/odmorot

goPAST.1SG on holiday/holiday-the
“I went on holiday”
b. Go čekav vozot/devojčeto
himCL.ACC waitPAST.1SG train-the/girl-the
“I waited for the train / the girl” (Mac, L. Grujoska; O. Tomić p.c.)
Olga Tomić (p.c.) informs me that the distribution of the Macedonian equivalents of
the PP complements found in other languages is further complicated by the fact that in
general they are DPs used without any preposition. In some dialects, though, PP
complements are available. They are compatible with clitic doubling, which suggests
that in these contexts na is a case realization as well.37
(50) Go čekav na devojčeto

himCL.ACC waitPAST.1SG on girl-the
“I waited for the girl” (South-Eastern Mac, O. Tomić p.c.)
37Devojčeto ‘the girl’ is a diminutive, and as such it is doubled by the masculine singular clitic go,
even though its natural gender is feminine. See chapter 4, section 4.4.2.4.2.1 for a detailed
explanation of this phenomenon.
More research is required in order to conclusively establish the nature of PP-

complements in South Slavic. However, irrespectively of the future findings, it is worth
pointing out that according to some recent proposals PP complements are always
generated outside the VP (PartP), so there is no need for movement of these elements
out of the PartP, because they are never inside it. For instance, Barbiers (1995) argues
that PP complements are base-generated external to the VP in Dutch, and that the
thematic relation between the verb (phrase) and the PP is established via overt or
covert movement of the VP into the specifier of the PP. One of the arguments for this
proposal is that it provides a natural explanation of PP extraposition.
(51) a. Jan heeft [PP op vader] [VP gewacht]

John has for father waited
b. Jan heeft [PP [VP gewacht] [PP op vader]] (Dutch, Barbiers 1995)
Another option would be to generalize Kayne’s (2004) suggestion that at least some
prepositions are merged as probes external to VP to all prepositions of PP-
complements. These prepositions would then be generated external to PP and attract
the DP-complement of the verb into their specifier. The movement would be
obligatorily followed by raising of P to the head of a WP.
(52) [WP ... W+P [PP DP tP [VP V tDP]]]
The discussion of these two alternative proposals is clearly beyond the scope of this
thesis. The remarks above are made to show that the assumption that PPs are generated
VP-internally and do not undergo A-movement is not as uncontroverial as it is often
believed. For the time being, I assume that one of the alternative proposals is on the
right track, and that as far (parts of) PP complements are generated PartP (VP)-
internally, they must be moved to a PartP-external position.
Apart from the PP-complements, small clause predicates such as žuto ‘yellow’ in
(53) must raise out of the PartP as well, as they may not be pied-piped by the l-
participle when it moves to Spec, TP.
(53) a. Ofarbao je zid žuto

paintPART.M.SG bePRES.3SG wallM.SG yellowN.SG
“He painted the wall yellow”
b. *Ofarbao žuto je zid (S-C, N. Milićević, p.c.)
I am not in the position to provide a motivation for this movement. A potential

solution might be to follow Broekhuis (2005), who suggests that in similar cases in
Dutch the Small Clause complement raises in order to check the φ-features of V. A
serious problem with this idea is the fact that in Slavic these complements do not agree
with any other elements in the clause, and are often introduced by prepositions. For
instance, the adjective žuto in (53) always occurs in the singular neuter form. It is also
possible to have adverbial Small Clause complements, without any φ-feature
specification whatsoever. I leave the explanation of this issue for future research.
Summarizing, on the basis of crosslinguistic evidence from the South Slavic
languages I have shown that both direct and indirect objects must raise out of the PartP
for case checking via A-movement. The movement occurs in all contexts, even when
indirect objects are introduced by prepositions. Moreover, it has been demonstrated
that PP complements and small clause complements are evacuated out of the PartP as
well. However, the trigger for this operation is unknown.
2.3.4 Word order in the Slavic left periphery

The preceding parts of this chapter have outlined the analysis of participle fronting
across the auxiliary ‘to be’ in terms XP-movement. However, this is not the only type
of movement that may take place across the auxiliary. As is well-known, the Slavic
languages have a very lax word order, and different categories may be preposed in order
to reflect the ways information is structured in discourse. As a rule, elements that carry
old information appear clause-initially, while the ones that express new information
occur towards the right periphery of the sentence. This relative freedom of word
placement may potentially bear on the analysis developed in this chapter. It has been
argued that participle fronting is an instance of locative inversion, which is contingent
on the subject agreement marking on the l-participle. Since the auxiliary may be
preceded by different categories, it is necessary to demonstrate that only the l-participle
may raise via locative inversion, while the displacement of other elements involves
different operations. This will be done in sections 2.3.4.1 through 2.3.4.4, whereas the
remaining parts of the chapter will give more support for the XP-movement approach
to participle fronting by providing more relevant data.
Section 2.3.4 is organized as follows. Section 2.3.4.1 discusses discourse effects
associated with participle movement across clitic and non-clitic auxiliaries. Section
2.3.4.2 examines syntactic and semantic effects triggered by the placement of elements
other that the l-participle in the clause initial position. Section 2.3.4.3 addresses the
question of whether all constituents that precede the auxiliary target the same position
as the fronted l-participle. Section 2.3.4.4 demonstrates that this is not the case and
discusses different types of focus movement.
2.3.4.1 Two types of l-participle fronting

The examples in (54) and (55) present minimal pairs exemplifying two types of
participle fronting in Bulgarian and Serbo-Croatian, respectively. The sentences in (54a)
and (55a) contain the present tense auxiliaries, while the ones in (54b) and (55b) are
formed with an auxiliary in the past tense.
(54) a. Gledali sa filma

watchPART.PL bePRES.3PL movie-the
“They have watched the movie”
a’. *Sa gledali filma
b. Gledali bjaxa filma
watchPART.PL bePAST.3PL movie-the
“They had WATCHED the movie”
b’. Bjaxa gledali filma
“They had watched the movie” (Bg, Lambova 2003)
(55) a. Sreo je Petra

meetPART.M.SG bePRES.1SG Peter
“He (has) met Peter”
a’. *Je sreo Petra
b. Sreo bejaše Petra
meetPART.M.SG bePAST.1SG Peter
“He had MET Peter”
b’. Bejaše sreo Petra
“He had met Peter” (S-C, Embick & Izvorski 1997)
As can be inferred from the data in (54a’) and (55a’), the present tense auxiliaries are
enclitics, so they need a phonological support to the left. The movement of the l-
participle fulfils this requirement. By contrast, the past tense auxiliaries in (54b) and
(55b) are not enclitic, and they need not be preceded by any overt material.
Some previous analyses have argued for a relationship between the clitic status of
the auxiliary and participle fronting. For instance, Lema and Rivero (1989), Rivero
(1991), and Wilder and Ćavar (1994) have claimed that the movement is motivated by
the need to provide a phonological host for the auxiliary enclitic. However, this view
has been proved to be inadequate on both theoretical and empirical grounds (cf.
Bošković 1995).
From a theoretical perspective this claim is problematic, because it presupposes a
“look-ahead” in the derivation. It suggests that the displacement does not occur in the
interest of the moved element, but rather for altruistic reasons, to circumvent the
phonological deficiency of another constituent. Moreover, the assumption cannot be
on the right track for empirical reasons. It will be shown in section 2.3.4.4 (cf. also
footnote 42), that a number of different categories, such as adverbs and DP objects can
be preposed to the position in front of the auxiliary clitic for semantic reasons, such as
focus or topicalization. The same type of semantic interpretation is observed when the
raising occurs across a clitic and a non-clitic auxiliary. The fact that they may provide
phonological support for the auxiliary is thus only a side-effect of their movement.
Crucially, Bošković (1995: 250ff) explicitly shows that providing a host for an
enclitic is not sufficient to trigger participle movement, even if it means that the clitic
remains otherwise stranded in the clause initial position. The case in question is the
interrogative particle li, which is an enclitic, and is commonly argued to be in C (cf.
section 2.3.6.3.3 and chapter 4). As demonstrated in (56a) for Serbo-Croatian, li may
not appear sentence initially, and must be supported by another element, such as the
finite verb ljubi, which undergoes head movement in (56b). However, the verb needs to
be finite to be able to raise to this position. As shown in (56c), the movement of the l-
participle poljubio is barred, which is unexpected if it occurs in order to provide support
for the enclitic. If this were the case, example (56c) should be as grammatical as (56b).38
38 Observe that the Bulgarian variant of (56c) is grammatical.
(i) Celunal li ja e?
kissPART.M.SG Q herCL.ACC be3SG
“Did he kiss her? (Bg, S. Marinov, p.c.)
This indicates that from a crosslinguistic point of view li does not have any inherent property that
blocks participle fronting. It is rather a feature of Serbo-Croatian, which disallows movement of
the l-participle to a position higher than Spec, TP.
(56) a. *Li ljubi nju?

Q kissPRES.3SG her
b. [C Ljubi li] nju?
kissPRES.3SG Q her
“Does he kiss her?”
c. *[CP Poljubio li] je nju?
kissPART.M.SG Q beAUX.3SG her
“Did he kiss her?” (S-C, Bošković 1995: 251)
This conclusively shows the clitic status of the auxiliary is not a sufficient condition to
drive participle movement. Bošković (1995: 251) argues that it is also not a necessary
condition. This is demonstrated by the examples containing the non-clitic, past tense
auxiliary bjaxa/bejaše in (54) and (55), which are partly repeated as (57).
(57) a. Gledali bjaxa filma

watchPART.PL bePAST.3PL movie-the
“They had WATCHED the movie”
a’. Bjaxa gledali filma
“They had watched the movie” (Bg, Lambova 2003)
b. Sreo bejaše Petra
meetPART.M.SG bePAST.1SG Peter
“He had MET Peter”
b’. Bejaše sreo Petra
“He had met Peter” (S-C, Embick & Izvorski 1997)
The l-participle in (57a) and (57b) is preposed across the past tense auxiliaries, even
though they are not enclitic and do not need to be prosodically supported. Still,
although the movement is not required in (57), it does not mean that it is “optional”.
Embick and Izvorski (1995), as well as Lambova (2003), observe that the reordering
across bjaxa/bejaše gives rise to a contrastively focused, or ‘non-neutral’ interpretation of
the predicate. This is also marked in the translations of examples (57a and b).
It is evident that the participle movement across the non-clitic auxiliary cannot be
driven only by the need to check the φ-features of T. It always results in a “non-
neutral” interpretation, which in general is not associated with the TP layer. Following
Lambova (2003), I will assume that the l-participle lands higher when it is preposed
across the non-clitic auxiliary, and the movement is triggered by a focus feature.
However, since the l-participle shows subject agreement, it must move via Spec, TP,
the way it does in the case fronting across the clitic auxiliary.
The derivation of (57a) is given in (58b-c). As suggested by Lambova (2003), I will
term the focus projection that is the target of l-participle movement ∆P (“Delta
Phrase”).
(58) a. Gledali bjaxa filma decata

watchPART.M.SG bePAST.3PL movie-the children-the
“The kids have watched the movie”
b. [∆P [TP [T[+φ] ... [Aux bjaxa … [AgrO [PartP gledali[+φ] filma]]]]]
c. [∆P [PartP gledali tk] [TP ti’ [T bjaxaj …[Aux tj … [AgrO filmak [tPartP tk]]]]]]
As in the case of participle fronting across the enclitic auxiliary, it is necessary to raise
the object filma ‘movie’ out of PartP into Spec, AgrO for case checking. Subsequently,
the auxiliary bjaxa moves to T and checks Tense. Next, the remnant PartP raises to
Spec, TP and checks the φ-features of T. It still needs to check the focus feature, and it
does so by landing in Spec, ∆P.
Summarizing, it has been demonstrated that Bulgarian and Serbo-Croatian have
two types of participle fronting: across the clitic and non-clitic forms of the auxiliary.
The latter type gives rise to a focus interpretation of the l-participle, and hence is
triggered by a focus feature. In spite of the different semantic effects associated with
the two types of participle movement, it was shown that the clitic form of the auxiliary
is neither a sufficient nor a necessary condition for the operation.
2.3.4.2 Constituents in the left periphery

It has been remarked that the word order in Slavic languages is often determined by the
information structure of the clause or a desire to focus or topicalize a certain
constituent. As a result, the auxiliary may be preceded not only by the subject or by the
l-participle, but also by a number of other constituents. Yet, since the basic word order
is SVO, the sentence in (59) represents the most neutral pattern, with the subject
preceding the auxiliary.
(59) Mislja če Ivan/toj e kupil knigata

think1SG that Ivan/he beAUX.3SG buyPART.M.SG book-the
“I think that Ivan/he bought the book” (Bg, Rudin 1986: 24-25)
Section 2.2.1 showed that the l-participle may be moved to the position in front of the
auxiliary as well.
(60) Mislja če kupil e knigata

think1SG that buyPART.M.SG beAUX.3SG book-the
“I think that he has bought the book” (Bg)
However, the auxiliary, as well as other types of clitics that cluster together with it, may
also be preceded by a direct object (cf. 61a), an adverb (cf. 61b and c), as well as the
subject accompanied by the object (cf. 61d) or a wh-word (cf. 61e).
(61) a. Mislja če knigata e kupil Ivan object

think1SG that book-the beAUX.3SG buyPART.M.SG Ivan
“I think it’s the book that Ivan has bought” (Bg, Rudin 1986: 24)
b. Mislja če pravilno e otgovoril
think1SG that correctly beAUX.3SG answerPART.M.SG
na vŭprosa im adverb
to question their
“I think that he has answered their question correctly” (Bg, Caink 1999)
c Mislja če mnogo sŭm dovolen adverbial modifier
think1SG that very be1SG gladM
“I think that I’m very glad” (Bg, M. Lambova, p.c.)
d. Razbrax če knigata (Ivan) ja object+subject
find-outPAST.1SG that book Ivan CL.ACC.F
beše pročel (Ivan)
bePAST.3.SG readPART.M.SG Ivan
“I had found out that Ivan had read the book indeed”
(Bg, Dimitrova-Vulchanova 1998: 17)
e. Čudja se (Ivan) na kogo e subject+wh-word

wonderPRES.1SG CL.REFL Ivan to whom bePRES.3SG
kupil podarŭk (Ivan)
buyPART.M.SG present Ivan
“I wonder for whom Ivan has bought a present”
(Bg, Dimitrova-Vulchanova 1998: 17)
The same observations hold for main clauses, where the clitic auxiliary may be preceded
by many different types of categories as well. In most cases these constituents can be
premodified by another element, which clearly shows that they are XPs.
(62) a. Mnogo bŭrzo e pročel knigata AdvP

very quickly bePRES.3SG readPART.M.SG book-the
“He has read the book very quickly” (Bg, S. Marinov p.c.)
b. Počti vsičko sme vzjeli ot xladilnika AdvP
almost everything bePRES.1PL takePART.PL from fridge
“We took almost everything from the fridge” (Bg, V. Tchonova, p.c.)
c. Dovolen sŭm Adjective
gladM bePRES.1SG
“I am glad”
d. Mnogo sŭm dovolen39 modifier of a predicative adjective
very bePRES.1SG gladM
“I am very glad” (Bg, Caink: 1999; P. Vitkova p.c.)
In embedded clauses in Serbo-Croatian the auxiliary clitic must be in the second

position, so it always follows the complementizer (cf. the example in footnote 27,
repeated as 63).
(63) a. *Ivan kaže da čitala je (Marija) Krležu

Ivan says that readPART.F.SG bePRES.3SG Marija KrležaACC
b. Ivan kaže da je (Marija) čitala Krležu
Ivan says that bePRES.3SG Marija readPART.F.SG KrležaACC
“Ivan says that Mary/she has read Krleža” (S-C, Wilder & Ćavar 1994: 8)
However, just as in Bulgarian, in main clauses in Serbo-Croatian the auxiliary clitic may
also be preceded by constituents of many different types. This is shown in (64) for the
equivalents of the Bulgarian sentences in (62).
(64) a. Veoma brzo je pročitao knjigu

very quickly bePRES.3SG readPART.M.SG bookACC
“He has read the book very quickly”
39 It is impossible to front the adjective together with its adverbial modifier, as *Mnogo dovolen sŭm
is ungrammatical. I suggest that this is due to a requirement that movement carries as little
material as possible (cf. Chomsky 1995: 264-265) A comparable restriction seems to hold for
preposition stranding in English, in which it is preferred to strand a preposition and raise just a
wh-word (H. Broekhuis, p.c.).
(i) a. Who did you talk to?

b. To whom did you talk?
b. (Skoro) sve smo uzeli iz frižidera

almost all bePRES.1PL takePART.PL from fridge
“We took almost everything from the fridge”
c. (*Mnogo) zadovoljan sam
very gladM bePRES.1SG
“I am glad”
d. Mnogo sam zadovoljan
very bePRES.1SG gladM
“I am very glad” (S-C, N. Milićević, p.c.)
Summarizing, this section has enumerated a number of categorially unrelated elements

that may be placed in the clause initial position in front of the auxiliary. In the next
section I will investigate whether this means that they all land in the same position as
the l-participle.
2.3.4.3 Is it a case of stylistic fronting?

In principle, it might be possible to argue that all the elements preceding the auxiliaries
in (61) and (62) target Spec, TP, just as the subject or the l-participle. In fact, this is an
assumption made by Holmberg (2000) in his analysis of stylistic fronting in Icelandic
and Faeroese.
The relevant data are given in (65) through (68). Holmberg argues that each of the
italicized elements in the examples below, such as negation in (65), the sentence adverb
in (66), the PP in (67) and the DP in (68) targets Spec, TP, and that their displacement
does not change the meaning of the clauses in any way.
(65) a. Þetta er tilboð [sem er ekki hægt að hafna]

this is offer that is not possible to reject
“This is an offer that cannot be rejected”
b. Þetta er tilboð [sem ekkii er ti hægt að hafna]
(66) a. Hver sagðir þu’ [að hefði sennilega skrifað þessa bók]?
who said you that has probably written this book
“Who did you say has probably written this book?”
b. Hver sagðir þu’ að sennilegai hefði ti skrifað þessa bók?
(67) a. Þeir sem hafa verið í Ósló segja að…

those that have been in Oslo say that
“Those who have been to Oslo say that…”
b. Þeir sem í Óslói hafa verið ti segja að…
(68) a. Þeir sem verða að taka þessa erfiðu ákvörðun

those that have to take this difficult decision
“Those who have to take this difficult decision…”
b. Þeir sem þessa erfiðu ákvörðuni verða ti að taka
(Icelandic, Holmberg 2000: 448-449)
The movement is assumed by Holmberg to be triggered by a variant of the EPP

feature, which requires that Spec, TP be filled by some phonological material. The exact
grammatical category of the element that undergoes movement is irrelevant, because it
functions as a pure expletive, and only the phonological feature matrix lands in this
position, whereas the semantic and formal features can be left in situ. In most cases it is
the subject that raises to Spec, TP. If for some reason the subject is not available for the
movement, the closest phonologically overt category must raise. In this way any
element may function as an expletive, whose sole role is to fill in Spec, TP.
Holmberg (2006) suggests that this analysis might potentially be extended to
participle fronting in the Slavic languages, given that the constructions have several
properties in common, such as clause boundedness and lack of semantic effects on the
sentence interpretation. In spite of these similarities, I reject this option, and I argue
that only those elements that have the appropriate φ-features, that is the subject or the
l-participle, can target Spec, TP. Even though all the preposed elements in examples
(61) through (64) uniformly precede the auxiliary, they will be argued to be located
higher than Spec, TP. Let me consider some motivations for this claim.
In the Bulgarian examples in (61) and (62) the auxiliary verb is immediately
preceded by adverbials, prepositional phrases, and other categories. Their placement
with respect to the auxiliary does not necessarily imply that they are hosted in Spec, TP,
though. First of all, they do not agree with the subject, so they are not eligible
candidates for checking the φ-features of T. Secondly, in certain marked contexts they
can precede the subject (cf. 69a) or the fronted participle (cf. 69b and c), which
indicates that they are located higher than Spec, TP.40
(69) a. (Bŭrzo) Ivan e pročel knigata

quickly Ivan bePRES.3SG readPART.M.SG book-the
“Ivan has read the book (really) quickly” (Bg, S. Marinov p.c.)
b. Tŭkmo polučil beše izvestieto
just-then receivePART.M.SG bePAST.AUX.3PL letter-the
“Just then he had received the letter” (Bg, Lambova 2004: 254)
c. Za izpita čel e bil Ivan
for exam readPART.M.SG beAUX3.SG bePART.M.SG Ivan
“Concerning the exam, Ivan has supposedly studied for it”
(Bg, M. Lambova p.c.)
In Serbo-Croatian, the evidence is harder to find, because the present tense auxiliary is a
Wackernagel clitic, which can be preceded by at most one constituent (cf. 70a).
However, the past tense auxiliary is not a clitic, so it does not have to appear in the
second position. In this context either the subject (cf. 70b) or the l-participle (cf. 70c)
may be preposed over the auxiliary.
(70) a. *Brzo Jovan je čitao knjigu

quickly Jovan beAUX.3SG readPART.M.SG bookACC
“Jovan read the book quickly”
b. Brzo Jovan bješe čitao knjigu
quickly Jovan bePAST.AUX.3SG readPART.M.SG bookACC
“Jovan had read the book (really) quickly”
c. Brzo čitao bješe Jovan knjigu (S-C, N. Milićević p.c.)
The items that occur in front of the l-participle or the subject always receive a focused
or topicalized interpretation. This type of interpretation is not associated with the TP
layer, but is rather typical of the left periphery. Therefore, it is likely that these elements
40 The sentence in (69c) exemplifies a double participle construction. See section 2.3.5.2 for an
analysis.
are in a Topic or Focus projection above TP. The issue will be addressed more
thoroughly in the subsequent section, where I describe types of foci associated with
different word orders in Slavic.
2.3.4.4 Focus and word order

Stjepanović (1999 ch. 3) provides a detailed account of focus effects triggered by
placement of different constituents in front of the auxiliary in Serbo-Croatian.41 She
distinguishes among three types of foci: new information, contrastive and emphatic
foci.
New information focus obtains in out-of-the-blue contexts. For instance, in an
answer to the question What happened? – John saw Mary the whole clause is new
information. Example (71) shows that new information focus (given in capitals) may
also constitute the answer to a wh-question.42
(71) A: Ko je kupio knjigu?

who beAUX.3SG buyPART.M.SG bookACC
“Who bought a book?”
B: Knjigu je kupio PETAR
bookACC beAUX.3SG buyPART.M.SG Petar
“The book was bought by Petar” (S-C, N. Milićević p.c.)
New information focus is always in the clause final position. Stjepanović argues that it
is assigned prosodically and that it is not the result of syntactic movement.
Contrastive focus is related to the negation of a presupposition (as in PETER
bought a pen, not Mary), whereas emphatic focus is associated with an assertion of a
41 I will discuss only Serbo-Croatian here, because Bulgarian seems to pattern in a very similar
way (cf. Arnaudova 2003 ch. 7). I extend Stjepanović’s analysis by studying discourse effects
triggered by placement of manner, sentential, and temporal adverbs in front of the auxiliaries. For
a detailed investigation of topic and focus in South Slavic see Arnaudova (2003) and Lambova
(2001, 2003) on Bulgarian; Čamdžić (1999) on Serbo-Croatian; and Tomić (1996b) on Bulgarian,
Macedonian, and Serbo-Croatian.
42 All the examples in this section contain clitic auxiliaries. However, with the exception of the l-
participle, the interpretation of the elements preceding the auxiliary does not depend on its clitic
or non-clitic status. This is illustrated for (i), which is the equivalent of (71) in the pluperfect, and
contains the non-clitic form bješe.
(i) A: Ko bješe kupio knjigu?

who bePAST.AUX.3SG buyPART.M.SG bookACC
“Who had bought a book?”
B: Knjigu bješe kupio PETAR
bookACC bePAST.AUX.3SG buyPART.M.SG Petar
“The book had been bought by Petar” (S-C, N. Milićević p.c.)
The ordering of all the constituents in (i) is the same as in (71), and the subject Petar has the new
information focus reading. The only semantic difference between (i) and (71) is temporal: the
former is in the pluperfect, whereas the latter is in the past tense. This is to be expected, and
shows once again that the movement of constituents in the structure is related only to the
semantic interpretation of the clause, and never occurs in order to provide phonological support
for the auxiliary clitic. Since the choice of the clitic versus non-clitic auxiliary does not have any
bearing on the clause information structure, all the examples discusses in this section will contain
the clitic forms.
presupposition (as in Yes, JOHN did it). According to Stjepanović, in Serbo-Croatian

contrastive focus is licensed in a preverbal position. For instance, the direct object
Mariju may be interpreted as contrastively focused in (72a), where it precedes the
auxiliary; as well as in (72b), where it precedes the l-participle. However, this reading is
unavailable when the object is located at the end of the clause (cf. 72c), because this
position is associated with new information focus. Stjepanović claims that the elements
that are contrastively focused move to one of the Focus projections (cf. Stjepanović
1999: 188ff), but since they also carry heavy stress, they are licensed prosodically as
well. In the examples below new information focus is capitalized; contrastive/emphatic
focus is bold-faced.
(72) a. Mariju je Petar zagrlio

MarijaACC beAUX.3SG Petar hugPART.M.SG
“It was Marija that Petar hugged”
b. Petar je Mariju zagrlio
Petar beAUX.3SG Marija hugPART.M.SG
c. ??Petar je zagrlio Mariju
Petar beAUX.3SG hugPART.M.SG Marija (S-C, Stjepanović 1999:73)
With these generalizations in mind, let me consider sentences that contain categories
other than the l-participle or the subject in front of the auxiliary. The canonical pattern
for transitive sentences in Serbo-Croatian, as well as all other Slavic languages is SVO.
Hence, the sentence in (73b) is the most natural answer to the question in (73a).
(73) a. Šta se desilo?

what REFL happenPART.N.SG
“What happened?”
b. Mačka je uhvatila miša
cat beAUX.3SG catchPART.F.SG mouse
“A cat caught a mouse”
c. #Miša je uhvatila mačka (S-C, N. Milićević p.c.)
The SVO order is felicitous when the whole sentence is a new information focus.
Recall, though, that since Bulgarian and Serbo-Croatian are pro-drop languages, the
subject is preferably omitted. However, Stjepanović (1999: 94) claims that the subject
must appear clause-initially when it is presupposed. Therefore, the sentence in (73b) is
also the most felicitous reply to the question What has the cat done?
The OVS order is found less often and always occurs in semantically marked
contexts. According to Stjepanović (1999: 92, 97), it may arise when both the verb and
the object are presupposed, and when the subject receives the main sentence stress.
This is exemplified in (74b), where the subject Marko appears at the end of the clause,
because it constitutes new information focus.
(74) a. Ko je udario Petra?

who beAUX.3SG hitPART.M.SG PeterACC
“Who hit Peter?”
b. Petra je udario MARKO
PeterACC beAUX.3SG hitPART.M.SG Marko
“Marko hit Peter” (S-C, Stjepanović 1999: 97)
Furthermore, the OVS order may also correlate with the contexts in which only the
object, such as Peter in (75b), is presupposed, and the remaining part of the sentence
represents new information focus.
(75) a. Šta se desilo Petru?

what REFL happenPART.N.SG PeterDAT
“What happened to Peter?
b. Petra JE UDARIO AUTO
PeterACC beAUX.3SG hitPART.M.SG car
“A car hit Peter” (S-C, Stjepanović 1999: 98)
Let us turn to examples with clause-initial adverbs. Just as the other elements that are
placed at the beginning of a sentence, they represent old information. Thus, the
sentence in (76b) is a felicitous reply to the question What happened yesterday?
(76) a. Šta se desilo juče?

what REFL happenPART.N.SG yesterday
“What happened yesterday?”
b. Juče JE PETAR KUPIO KNJIGU
yesterday beAUX.3SG Peter buyPART.M.SG book
“Yesterday Peter bought a book” (S-C, N. Milićević p.c.)
The event time of the predicate in (76b) is presupposed, so the temporal adverb juče
‘yesterday’ appears at the beginning of the clause. However, the string that follows it
constitutes “new information” and correspondingly receives new information focus.
As far as manner (cf. 77b, c) and sentential adverbs (cf. 77d) are concerned, native
speakers report that they are usually positioned at the beginning of a clause. The
remaining parts of the sentences following the adverbials in (77) are new information
foci.
(77) a. Šta se desilo?

what REFL happenPART.N.SG
“What happened?”
b. Brzo je Petar bacio knjigu
quickly beAUX.3SG Peter packedPART.M.SG book
“Peter packed the book quickly”
c. Potpuno smo ispraznili frižider
completely beAUX.1PL emptyPART.M.PL refrigerator
“We emptied the refrigerator completely”
d. Neočekivano smo dobili pismo
unexpectedly beAUX.1PL receivePART.M.PL letter
“We received a letter unexpectedly” (S-C, N. Milićević p.c.)
Hence, a clause containing a manner or a sentential adverb in which a subject or an l-

participle occurs in the initial position will not be the most felicitous answer to the
question What happened?, because it will put the adverb in the new information focus
position.
(78) Petar je bacio knjigu BRZO

Peter beAUX.3SG packedPART.M.SG book quickly
“Peter packed the book really quickly” (S-C, N. Milićević p.c.)
To summarize, we have seen that in Serbo-Croatian the constituents whose referents

are presupposed are placed at the beginning of a clause, while the new information foci
are located in the right periphery. This pattern is not surprising, because it has been
known since Mathesius’ work in the early 20th century (see Mathesius 1975 for an
English version of his publications) that the ordering of sentence constituents in Slavic
generally reflects the theme-rheme distinction: the elements representing old
information are followed by those carrying new information.
Finally, consider the contexts in which the l-participle or the subject occurs at the
beginning of the clause. They involve the most neutral word orders, but it is the
discourse information structure that decides whether the subject or the l-participle is
preposed. For example, (79a) is the most felicitous reply to the question What happened?,
that is, in the context when the whole sentence constitutes new information focus. If
the subject Petar has been previously mentioned and hence its referent is presupposed,
there is no need to repeat it, which may result in fronting of the participle to the initial
position, as in (79b). Placement of the subject towards the right periphery of the clause
gives rise to special discourse effects. For instance, in the VSO pattern in (79c) the
fronted participle is understood as emphatically focused: the event of buying the book
has been presupposed and is reasserted.
(79) a. Petar je kupio knjigu

Peter beAUX.3SG buyPART.M.SG bookACC
“Peter has bought a book”
b. Kupio je knjigu (juče)
buyPART.M.SG beAUX.3SG book (yesterday)
“He bought the book (yesterday)”
c. Kupio je Petar knjiguACC
buyPART.M.SG beAUX.3SG Peter bookACC
“Oh yes, Peter did buy the book” (S-C, N. Milićević, p.c.)
Stjepanović (1999) does not discuss transitive examples like the ones in (79). However,
she mentions sentences with unaccusative participles, such as umro ‘die’, and argues that
the sentence in (80a), with the subject Truman in the initial position would be uttered in
a situation where people had been aware of Truman and his illness before his death. In
this way his death is interpreted as new information. Conversely, if Johnson’s death
came unexpectedly, the only way to express the information under neutral focus is to
put the subject in the final position, as in (80b).
However, if the subject Johnson is preposed to the initial position and receives
stress, as in (80c), it is interpreted as contrastively focused. That is, the meaning of the
sentence is that it is Johnson and not anybody else who died.
(80) a. Truman je UMRO

Truman beAUX.3SG diePART.M.SG
“Truman died”
b. Umro je JOHNSON
diePART.M.SG beAUX.3SG Johnson
“Johnson died”
c. Johnson je umro
Johnson beAUX.3SG diePART.M.SG
“Johnson died” (S-C, Stjepanović 1999: 89)
Summarizing, the preceding sections have shown that although the Slavic languages
have very free word order, the most natural one is SVO. As a rule, placement of
temporal adverbs or objects at the beginning of a sentence requires special information
structure contexts, which indicates that these elements are located in the Topic/Focus
domain, above the TP layer. By contrast, movement of the subject or the l-participle to
the initial position usually does not result in a focused interpretation of a clause.43 This
fact suggests that only these two elements may raise to Spec, TP.
2.3.5 L-participle fronting as XP movement -

elaboration
The preceding parts of this chapter have demonstrated that only the subject or the l-
participle may target Spec, TP. The following sections will provide more arguments for
the analysis of participle fronting in terms of XP-movement. In particular, more
evidence will drawn from the properties of short participle movement discussed in
section 2.3.5.1. The operation will be shown to be obligatory in Bulgarian (cf. section
2.3.5.1.2), but not in Serbo-Croatian (cf. section 2.3.5.1.1). The conclusions reached
there will be confirmed by patterns of double participle constructions investigated in
section 2.3.5.2. Finally, participle fronting will be juxtaposed with finite verb movement
in the contexts of future tense constructions in section 2.3.5.3 and negation in section
2.3.6.
2.3.5.1 Auxiliary-participle adjacency

The following section will point out some differences between participle movement in
Bulgarian and Serbo-Croatian concerning the adjacency between the auxiliary and the
participle. The contrast between the two languages will be argued to be related to the
richness of aspectual marking in Bulgarian.
The Bulgarian and Serbo-Croatian examples in (81a) and (81b) indicate that the
fronted participle must always be left-adjacent to the auxiliary.
(81) a. Čel (*bŭrzo) e/ beše knigata (bŭrzo)

readPART.M.SG quickly bePRES/PAST.3SG book-the quickly
“He has/had read the book (quickly)” (Bg, cf. Lambova 2003)
b. Zaboravio (*potpuno) je/bješe Petra potpuno
forgetPART.M.SG completely bePRES/PAST.3SG Peter completely
“Jovan has/had completely forgot Peter” (S-C, cf. Bošković 1995)
43 Note, though, that the exact discourse function of the clause-initial l-participle also depends on
the position of the constituents in the right periphery, as in (79c).

This is expected on the assumption that the fronted participle is in Spec, TP, whereas
the auxiliary lands in T0, therefore no phrasal material may intervene between the two
constituents.44 Thus, the only elements that may occur between the fronted participle
and the present tense auxiliary are clitics forming a clitic cluster.
(82) Dal mu go e
givePART.M.SG himCL.DAT himCL.ACC bePRES.3SG
“He has given it to him” (Bg)
However, Bulgarian and Serbo-Croatian exhibit divergent adjacency patterns when the
l-participle follows the auxiliary. Namely, Bulgarian requires the auxiliary to be adjacent
to the participle in such contexts, and neither an adverb (cf. 83a) nor an object (cf. 83b)
may split their sequence.45
(83) a. Ivan e/beše (*bŭrzo) čel bŭrzo knigata

Ivan bePRES/PAST.3SG quickly readPART.M.SG quickly book
“Ivan has/had the book quickly”
b. Ivan e/beše (*knigata) čel knigata
Ivan bePRES/PAST.3SG book-the readPART.M.SG book
(Bg, cf. Lambova 2003)
The fact that the intervening elements bŭrzo and knigata are of different categories
implies that the constraint is related to movement of the main verb, rather than object
shift or a restriction on adverb placement in this position. Therefore, I will refer to the
operation as ‘short verb/participle movement’.
In contrast to Bulgarian, Serbo-Croatian does not impose any adjacency
restrictions in the auxiliary-participle contexts and allows both adverbs (cf. 84a) and
objects (cf. 84b) to intervene between the two constituents.
(84) a. Jovan je (potpuno) zaboravio (potpuno) Petra

Jovan beAUX.3SG completely forgetPART.M.SG completely Peter
“Jovan completely forgot Peter”
b. Petar je (Mariju) zagrlio (Mariju)
Petar beAUX.3SG MarijaACC hugPART.M.SG MarijaACC
“Petar hugged Marija” (S-C, Stjepanović 1999:73)
Importantly, the adjacency patterns do not hold exclusively for the l-participle. The
same variation is observed with infinitives (cf. 85a) and finite verbs (cf. 85b), which may
be followed or preceded by an adverb in Serbo-Croatian, although the latter option is
strongly preferred by native speakers with all types of verbs.
44 The option with the clitic auxiliary in (81b) is also ruled out, because clitics must follow the
first constituent in Serbo-Croatian (cf. chapter 4).
45 Some native speakers require only the present tense auxiliary to be left-adjacent to the l-
participle and allow the past tense variant to be separated from the participle by some material.
See Krapova (1999a) for details.
(85) a. Jovan će (potpuno) zaboraviti (potpuno) Petra46

Jovan willCL.3SG completely forgetINF completely Petar
“Jovan will forget Petar completely”
b. Petar (ludo) voli (ludo) Mariju
Petar madly love3SG madly Marija
“Petar loves Marija madly” (S-C, Stjepanović 1999: 39)
Correspondingly, it has been observed in the literature that finite verbs must move
across VP adverbs in Bulgarian (cf. Lambova 2003: 7-8). This indicates that the
requirement of short verb movement concerns not only l-participles (cf. 86), but finite
verbs in the past and the present tense (cf. 87) as well. Moreover, this suggests that the
movement is motivated by the same feature irrespectively of whether it is phrasal
movement in the case of l-participles, or head movement in the case of finite verbs.47
(86) Ivan e/beše (*često) navestjaval (često) roditelite si

Ivan bePRES/PAST.3SG often visitPART.M.SG often parents selfCL
“Ivan has/had often visited his parents” (Bg, Lambova 2003: 7-8)
(87) a. Ivan (*često) navestjava (često) roditelite si

Ivan often visitsPRES.3SG often parents selfCL
“Ivan often visits his parents”
b. Ivan (*obiknoveno) e (obiknoveno) v kabineta si
Ivan usually bePRES.3SG usually in office selfCL
po tova vreme
at this time
“Ivan is usually in his office at this time” (Bg, Lambova 2003: 7-8)
It is plausible that the short participle movement is an intermediate step in “long”

participle fronting in Bulgarian discussed in the previous sections. Since “long”
participle fronting targets an argument position, it represents A-movement. This
implies that short participle movement must be of the A-type, too. If the landing site of
the short participle movement were an A’-position, the “long” participle fronting would
be illicit as a case of improper movement, because it would then involve raising from an
A’ to an A-position.
To summarize, I have presented evidence for an intermediate step in the
movement of the l-participle and verbal heads, which is obligatory in Bulgarian. In the
next section I will try to determine the trigger of the operation in both languages.
2.3.5.1.1 Short verb/participle movement in Serbo-Croatian

This section will argue that in spite of some superficial similarities between the short
participle fronting in Bulgarian and Serbo-Croatian, these movements are of different
types and occur for a different reason. The difference is related not only to the apparent
“optionality” of the movement in Serbo-Croatian, but also to divergent semantic effects
that the reordering brings about.
46 The future auxiliary će nay not be preceded by the adverb potpuno, because it is a second
position clitic.
47 This feature will be identified in section 2.3.5.1.2. Movement of finite verbs will be contrasted
with l-participle fronting in sections 2.3.5.3 and 2.3.6.

Native speakers report that in neutral contexts the participle follows the adverb in
Serbo-Croatian, as in (88a) and (89a). They also state that movement of the participle in
front of the adverb gives rise to a somewhat focused interpretation of the adverb (N.
Milićević, p.c.).
(88) a. Jovan je potpuno zaboravio Petra

Jovan beAUX.3SG completely forgetPART.M.SG Peter
“Jovan completely forgot Peter”
b. #Jovan je zaboravio potpuno Petra (S-C)
Progovac (2005a: 31) states that the string in (89b) nearly requires a comma intonation
before and after the adverb. This suggests that even though the verb movement is
possible, it is dispreferred, and occurs only for special discourse effects.
(89) a. Petar ludo voli Mariju

Petar madly love3SG Marija
“Petar loves Marija madly”
b. #Petar voli ludo Mariju (S-C, Stjepanović 1999: 39)
Correspondingly, given that Serbo-Croatian has SVO as the basic word order, the
object follows the participle in neutral contexts. Preposing of the object in front of the
participle leads to a contrastive focus reading, which was claimed to be licensed in
preverbal positions (cf. section 2.3.4.4).
(90) a. Petar je zagrlio Mariju

Petar beAUX.3SG hugPART.M.SG MarijaACC
“Petar hugged Marija”
b. Petar je Mariju zagrlio
Petar beAUX.3SG Marija hugPART.M.SG
“It was Marija that Peter hugged” (S-C, Stjepanović 1999:73)
Thus, it seems that short participle movement in Serbo-Croatian is related to the

information structure of the clause, and requires a special context to be (marginally)
acceptable. This suggests that the operation is not necessarily an intermediate step in
the “long” participle fronting to Spec, TP. By contrast, as shown by (87), in Bulgarian
the movement is not related to topic or focus considerations, because the lack of it
always leads to ungrammaticality. This means that it is triggered by the necessity to
establish some required feature checking configuration, as I will argue in the next
section.
2.3.5.1.2 Short verb/participle movement in Bulgarian

Example (86) shows that in Bulgarian participles must be adjacent to the auxiliary. This
fact is regarded here as an indication of short verb movement. However, there is an
exception to this requirement. Krapova (1999a and b), Billings (2002), and Franks
(forthcoming) observe that when the l-participle follows the auxiliary (or the cluster
formed by the auxiliary together with pronominal clitics), the two elements may be
separated from each other by aspectual adverbs, such as (vse) ošte ‘still’ and veče ‘already’,
as in (91).
(91) a. Ivanne mu go e [vse ošte] vŭrnal

IvanNEG himCL.DAT itCL.ACC beAUX.3SG still returnPART.M.SG
“Ivan still has not returned it to him” (Bg, Franks, forthcoming)
b. Ivana ne e [ošte] napisala domašnoto si
Ivana NEG beAUX.3SG still writePART.F.SG homework-the her
“Ivana has not finished her homework yet”
c. Da utre šte sŭm gi [veče] pratila
by tomorrow FUT beAUX.1SG themCL.ACC already sendPART.F.SG
“By tomorrow I will have already sent them” (Bg, Krapova 1999b)
By contrast, when the participle is fronted, the aspectual adverb veče may not occur
between the preposed participle and the auxiliary. Instead, it follows the clitic cluster
that contains the auxiliary.
(92) Dala (*veče) sŭm ti ja veče

givePART.F.SG already beAUX.1SG youCL.DAT itCL.ACC already
“I have already given it to you!” (Bg, Franks, forthcoming)
Bošković (2001: 181) claims that these adverbs might be incorporated into the verb, on
a par with adverbial clitics in some other languages (cf. Dobrovie-Sorin 1994 for
Romanian).48 However, Billings (2002) remarks that the explanation is incorrect,
because adverbs are always stressed in this position. Moreover, in (91a) the adverb
consists of two separate words, so it cannot be a clitic.
The fact that the only elements that may intervene between the auxiliary and the
participle are aspectual adverbs suggests to me that the short verb movement takes
place in order to check an aspect feature in a projection I will dub AspP. A similar
proposal has already been made for Bulgarian by Krapova (1999a), who follows Giorgi
& Pianesi’s (1997) hypothesis of splitting TP into separate Tense- and Aspect-related
projections.49 Krapova claims that Bulgarian has two Tense projections: T1P, which
dominates T2/AspP. T1 relates the reference time to the moment of speech, whereas
T2/AspP relates the reference time to the event time. Following these insights, I
suggest that aspect can be checked in two ways in Bulgarian: either by verb movement
into Asp, or by merging an aspectual adverb in this position, as in (91).
The proposal gains additional support if we consider adjacency conditions outside
the compound tenses formed with the l-participle. Lambova (2003: 179-180) observes
that in copular constructions the verb ‘to be’ does not have to be adjacent to the
subject or the nominal or adjectival predicate. Both clitic (cf. 93a) and non-clitic (cf.
93b) forms of the copula may be separated from the subject or the predicate by any
type of adverb, not necessarily an aspectual one. This indicates that neither of these
constituents must raise to the position immediately to the right of the copula.
48 Alexandra Cornilescu (p.c.) points out to me that all adverbial clitics in Romanian express
aspectual meanings, which makes a correlation between the Bulgarian and the Romanian patterns
even stronger.
49 The idea that verbs may move to check an aspect feature is not new. For example, Bok-
Bennema (2001) suggests that verbs in French and Spanish may undergo short movement, which
is triggered by an Aspect feature. She argues that the projection termed Agr by Pollock (1989),
which is targeted by finite verbs in French, encodes an aspectual feature. Furthermore, see
Svenonius (2004) for a claim that the locus of the aspectual tenses in Bulgarian is immediately
below T.
(93) a. Ivan (veče) e (veče) doktor na fizičeskite nauki

Ivan already bePRES.3SG already doctor of physics
“Ivan is already a doctor of physics”
a’. Doktor na fizičeskite nauki (veče) e Ivan
b. Ivan (opredelno) beše (opredelno) dovolen
Ivan certainly bePAST.3SG certainly satisfiedM
“Ivan was certainly satisfied”
b’. Dovolen (opredelno) beše Ivan (Bg, Lambova 2003: 179-180; p.c.)
I would like to argue that the lack of adjacency between the copula and the predicative
elements in (93) is due to the fact that the AspP projection is accessible only to
categories that are specified for aspect, such as finite verbs and participles. Since
adjectives and nouns do not mark aspectual distinctions, they do not pass through
AspP. Consequently, they do not have to be right-adjacent to the copulas e and beše,
which reside in T, immediately above AspP. They may move directly to Spec, TP,
where they check the φ-features of T.
The postulation of AspP below TP is also relevant for the position of the subject
with respect to the auxiliary. In (93) the subject may be split from the copula with an
adverb. This is possible not only in copula constructions, but also in compound tenses
formed with the l-participle, but only when the subject precedes the auxiliary (cf. 94a).
If the l-participle is fronted, it must be adjacent to the auxiliary (cf. 94b).
(94) a. (Nesŭmneno) Ivan (nesŭmneno) e pročel knigata

undoubtedly Ivan undoubtedly beAUX.3SG readPART.M.SG book-the
“Ivan has undoubtedly read the book”
b. (*Nesŭmneno) pročel (*nesŭmneno) e knigata (Bg, S. Marinov, p.c.)
The data in (94) indicate that unlike the l-participle, the subject can be topicalized and
raise across sentential adverbs, such as nesŭmneno ‘undoubtedly’. Alexandra Cornilescu
(p.c.) informs me that the impossibility of l-participle movement in this context might
be related to the fact that verbs are the only grammatical category that may never be
topicalized, but only focused (cf. also Cinque 1993).50
Furthermore, the examples in (94) also indicate that the l-participle in Bulgarian
must always first move to Spec, AspP, and only then may it raise further to Spec, TP.
By contrast, the subject does not need to target this intermediate landing site, and it
may raise directly to Spec, TP. In Migdalski (2005) I suggested that the difference is
related to φ-feature specification: the subject is marked for a full set of φ-features,
whereas the l-participle carries only the gender and number features.
(95) [TP T[+Person/Number/Gender] ... [AspP Part[+Number/Gender] ... [vP Subject[+Person/Number/Gender] ]]]
50 Bošković (1997: 144-146) shows that the subject is able to move higher than the l-participle in
Serbo-Croatian as well. For instance, in contrast to the l-participle, the subject may cross
sentential adverbs.
(i) a. Jovan je nesumnjivo istukao Petra

Jovan beAUX.3SG undoubtedly beatPART.M.SG PeterACC
“Jovan undoubtedly beat Peter”
b. *Istukao je nesumnjivo Petra (S-C, cf. Bošković 1997: 144ff)
Assuming with Chomsky (2001) that T is specified for a full set of φ-features, I argued
that the subject is raised when the feature [Person] is selected as the attractor. When
[Gender] or [Number] are the attractors on T, the l-participle moves.
However, the claim makes use of a rather mechanical solution, which is ad hoc
related to the richness of φ-features. It seems more reasonable to argue that the subject
may not raise via AspP, because it does not have the aspectual morphology that is
eligible for checking the feature of Asp.
The proposal developed here, which relates short participle movement to the
presence of AspP receives typological support. Recall from chapter 1 that Bulgarian has
retained the aspectual tenses inherited from Old Church Slavonic, imperfectum and
aorist. Thus, it has two options of marking aspectual distinctions: via aspectual
morphemes (usually prefixes) and via tense morphology. By contrast, Serbo-Croatian
has lost the aspectual tenses, which in syntactic terms may mean that it does not project
AspP or that this projection is weak. Consequently, there is no requirement of short
verb movement in Serbo-Croatian. If a comparable movement does occur in this
language, it is triggered by information structure considerations, rather than the
necessity to check the aspect feature in Spec, Asp.
2.3.5.2 Double participle constructions

This section will analyze complex structures formed with a present tense form of the
verb ‘be’, the l-participle of the verb ‘be’, and the l-participle of the main (thematic)
verb. These structures will provide additional arguments for the existence of the AspP
projection in the phrase structure, which is the target of the short participle movement.
The sentences in (96a and b) exemplify the double participle construction for
Bulgarian and Serbo-Croatian, respectively. It is formed in the same way in both
languages, but has a different meaning. In Bulgarian, it expresses the renarrated mood
(cf. chapter 1, section 1.3.4.5.1), whereas in Serbo-Croatian it functions as the
pluperfect tense (cf. chapter 1, section 1.3.4.3.1).
(96) a. Az sŭm bil četjal knigata

I bePRES.1SG bePART.M.SG readPART.M.SG book-the
“I am said to have been reading the book” (Bg)
b. Ja sam bio pročitao knjigu
I bePRES.1SG bePART.M.SG readPART.M.SG bookACC
“I had finished reading the book” (S-C, cf. Tomić 1996a: 853-854)
Both participles agree in φ-features with the subject and either of them may be moved
to the clause-initial position in Bulgarian (cf. 97) and Serbo-Croatian (cf. 98).51
51 There is some discrepancy among native speakers concerning acceptability of fronting the
thematic participle. According to Lema & Rivero (1989), the movement is excluded, but the
judgments are disconfirmed by Rivero herself in her later work (1991), as well as by Embick and
Izvorski (1995) and Lambova (2003). Tomić (1996a: 853) marks (97b) as ‘?’. Correspondingly,
Embick and Izvorski (1995) reject a Serbo-Croatian example that is similar to the one in (98b),
but their data are contested by Bošković (1995). What this suggests to me is that fronting of the
thematic participle requires extra focusing, which is not accepted by all native speakers. This is
confirmed by Lambova’s (2003: 174) observation that in the absence of the subject the most
neutral pattern is BePART-BeAUX-VPART, while the ordering VPART-BeAUX-BePART requires “non-
neutral” intonational contours. The generalization is expected from a syntactic point of view:
since the l-participle form of the verb ‘be’ is generated higher in the structure, it should be easier
(97) a. Bili sŭm četjal ti knigata

bePART.M.SG bePRES.1SG readPART.M.SG book-the
b. Četjalj sŭm bil tj knigata (Bg, cf. Tomić 1996a: 853-854)
(98) a. Bioi sam ti pročitao knjigu

bePART.M.SG bePRES.1SG readPART.M.SG bookACC
b. Pročitaoj sam bio tj knjigu (S-C, cf. Tomić 1996a: 853-854)
However, it is impossible to move the two participles at the same time, as indicated in
(99) for Bulgarian and in (100) for Serbo-Croatian.
(99) a. *Četjal bil sŭm knigata

readPART.M.SG bePART.M.SG bePRES.1SG book-the
b. **Bil četjal sŭm knigata (Bg)
(100) a. *Pročitao bio sam knjigu

readPART.M.SG bePART.M.SG bePRES.1SG bookACC
b. **Bio pročitao sam knjigu (S-C, cf. , Bošković 1997: 159-160)
As shown in (101), the direct object may follow two participles in Serbo-Croatian.
Neither the object (cf. 101b) nor the subject may intervene between the two participles
when they are preceded by the auxiliary.
(101) a. Vas dvoje ste bili čekali Marijinu prijateljicu

you two beAUX.2PL bePART.M.PL waitPART.M.PL Maria’s friend
b. *Vas dvoje ste bili Marijinu prijateljicui čekali ti
c *Marijinu prijateljicui ste bili vas dvoje čekali ti (S-C, cf. Bošković 1997: 157)
However, the two participles may be separated by an auxiliary (cf. 98). This leads
Bošković (1997: 157) to conclude that they are head-adjoined to the auxiliary verb. One
of them is adjoined to the left of it, whereas the other one is adjoined to the right.
Bošković suggests that fronting of the thematic participle čekali across the auxiliary in a
double participle construction proceeds as in (102).
(102) Čekali ste bili Marijinu prijateljicu

waitPART.M.PL beAUX.2PL bePART.M.PL Maria’s friend
a. [AuxP ste [VP bili [VP čekali Marijinu prijateljicu]]]
beAUX.2PL bePART.M.PL waitPART.M.PL Maria’s friend
b. [AuxP [Aux [Aux ste] bilii ] [VP ti [VP čekali Marijinu prijateljicu]]]
c. [AuxP [Aux čekalij [Aux [Aux ste] bilii ]] [VP ti [VP tj Marijinu prijateljicu]]]
First, bili right-adjoins to the auxiliary and checks the [+aux] feature (cf. 102b). Next,
čekali, crosses over the trace of bili and adjoin to the left of the auxiliary ste (cf. 102c).52
In principle, the direct adjunction of čekali to ste violates the Minimize Chain Link
to raise it. See also Čamdžić (2004 ch. 1) for an overview of double participle constructions
across Slavic.
52 Bošković is aware that his account is against Kayne’s (1994) claim that rightward adjunction is
disallowed. However, since he analyses participle fronting as head adjunction, his proposal is in
line with Chomsky’s (1995) suggestion, which excludes rightward adjunction only in the case of
XP-movement.
Principle, but Bošković argues that this is a licit operation, because the trace ti in (102c)
and the landing site of čekali, which is a position adjoined to ste, belong to the same
minimal domain of the chain (bilii, ti). Hence, they are equidistant from the base
position of čekali, and the direct adjunction of čekali to ste does not lead to a violation of
the locality conditions.
Bašić (2003) points out that Bošković’s adjunction account is empirically
inadequate, because an adverb, such as potpuno in (103a), may be inserted between the
auxiliary verb and a thematic participle, such as zaboravio in (103). This demonstrates
that the participle does not need to immediately follow the auxiliary, so it may not be
right-adjoined to it.
(103) a. Bio je potpuno zaboravio na sastanak

bePART.M.SG beAUX.3SG completely forgetPART.M.SG about meeting
“He had completely forgotten about the meeting” (S-C, Bašić 2003)
Furthermore, Bašić remarks that only a thematic participle may be preceded by an

adverb in this position. The l-participle of the verb ‘be’ does not allow adverb
premodification (cf. 104). I suggest that the restriction is syntactic in nature, and I will
account for it later in this section.
(104) ?*Zaboravio je potpuno bio na sastanak

forgetPART.M.SG beAUX.3SG completely bePART.M.SG about meeting
(S-C, Bašić 2003)
In Bulgarian the adjacency requirements related to the position of the auxiliary and the
participles are stricter that in Serbo-Croatian. As the data in (105) illustrates, both the
thematic and non-thematic participle must always be adjacent to the auxiliary, whether
they follow or precede it.
(105) a. Čel (*Ivan) e (*Ivan) bil (Ivan) za izpita

readPART.M.SG Ivan be3SG Ivan bePART.M.SG Ivan for exam
“Supposedly, Ivan must have STUDIED for the exam”
a’. Bil (*Ivan) e (*Ivan) čel (Ivan) za izpita
b. Čel (*intenzivno) e (*intenzivno)
readPART.M.SG hard beAUX.3SG hard
bil (intenzivno) za izpita
bePART.M.SG hard for exam
“He must have studied hard for the exam”
b’. Bil (*intenzivno) e (*intenzivno) čel (intenzivno) za izpita
(Bg, cf. Lambova 2003; p.c.)
To summarize the observations that have been made so far, let me enumerate the
possible patterns of double participle constructions. The auxiliary clitic cannot be
clause-initial (cf. 106a), but must be preceded by one of the l-participles (cf. 106b/c).
However, fronting of both participles at the same time is excluded (cf. 106d/e).
Furthermore, Bulgarian and Serbo-Croatian differ in that in the former the non-fronted
participle must be right-adjacent to the auxiliary, whereas in the latter the thematic
participle may be separated from it by an adverb.
(106) a. *BeAUX-BePART-VPART
b. BePART-BeAUX-VPART
c. VPART-BeAUX-BePART
d. *VPART-BePART-BeAUX
e. **BePART-VPART-BeAUX
The analysis of participle fronting developed in this chapter presupposes that the
double participle constructions are also formed by means of locative inversion. For this
reason I will assume that the l-participle form of the verb ‘be’ is a copula that makes
locative inversion possible. I propose it occupies the specifier of the phrase I label BioP
for convenience. Bearing this in mind, the clause in (107a) will have the base structure
given in (107b).
(107) a. Az sŭm bil četjal knigata

I bePRES.1SG bePART.M.SG readPART.M.SG book-the (Bg)
b. [TP [T[+φ] [AspP [AgrOP [Aux sŭm [BioP bil[+φ] [vp az[+φ] [v [PartP četjal[+φ] knigata]]]]]]]]]
The subject and both of the participles carry φ-features, so each of them can be
attracted by T and move into Spec, TP. However, since the copula participle bil is
generated in the closest proximity to T, it is the most suitable candidate for the
operation. Still, bil does not carry any aspectual morphology, so it may not check the
aspect feature of Asp. This can be only done by the thematic participle, which always
specifies aspectual distinctions. For instance, it may appear in aspectual pairs (cf. čel
‘readIMPF’ and pročel ‘readPRF’ in Bulgarian).
Assuming these generalizations, let us consider two cases of participle fronting.
The default ordering is “BePART-BeAUX-VPART”, which I suggest is derived in the
following way. As in the constructions with a single participle, the object must be
evacuated from the PartP prior to the movement of the thematic participle. I posit that
it raises to Spec, AgrOP, where it checks case. The thematic participle raises via XP-
movement across the copula participle and lands in Spec, AspP, where it checks
Aspect, while the finite auxiliary sŭm targets T and checks Tense. As was noted above,
Spec, AspP is inaccessible for bil, which may only check the φ-features of T by raising
into Spec, TP. The derivation is schematized in (108).
(108) Bil sŭm četjal knigata

bePART.M.SG bePRES.1SG readPART.M.SG book-the (Bg)
a. [TP [T[+φ] [AspP [Aux sŭm [BioP bil[+φ] [AgrOP [vp [PartP četjal[+φ] knigata]]]]]]]]
b. [TP [BioP bil] [T sŭmk [AspP [PartP četjal ti] [AgrOP knigatai [Aux tk [tBioP [vp [tPartP
ti]]]]]]]]
The other instance of participle fronting involves movement of the thematic participle.
It always gives rise to a focused interpretation of this element. I suggest that this means
that the operation is triggered by a Focus feature and occurs successive-cyclically via
Spec, AspP, where the participle checks the Aspect feature; Spec, TP, where the φ-
features of T are checked, and ends up in Spec, FocP.
(109) Četjal sŭm bil knigata

readPART.M.SG bePRES.1SG bePART.M.SG book-the (Bg)
a. [TP [T[+φ] [AspP [Aux sŭm [BioP bil[+φ] [AgrOP [vp [PartP četjal[+φ] knigata]]]]]]]]
b. [FocP [PartP četjal ti] [TP [tPartP ti] [T sŭmk [AspP [tPartP ti] [AgrOP knigatai [Aux tk [BioP bil
[vp [tPartP ti]]]]]]]]]
The templates in (108) and (109) provide derivations of double participle constructions
in Bulgarian. In Serbo-Croatian they proceed in a similar way, and the main difference
concerns the intermediate movement of the thematic participle to AspP, which is not
obligatory in this language, the way it is also not required in the case of short participle
movement (cf. section 2.3.5.1.1). Hence, the thematic participle does not have to be
right-adjacent to the auxiliary, and may be preceded by some lexical material, such as
the adverb potpuno ‘completely’ in (103).
Summarizing, this section has overviewed the double participle construction in
Bulgarian and Serbo-Croatian. It has been demonstrated that even though it is formed
in a similar way in both languages, Bulgarian requires the thematic participle to move
via Spec, AspP, where the aspect feature is checked. This is related to the richness of
aspectual specifications in this language, which in this way has been shown to be
relevant for both the short participle movement and the double participle formations.
2.3.5.3 The future auxiliary šte in Bulgarian

This section will briefly examine the behaviour of the future auxiliary šte in Bulgarian. It
occurs both with finite verbs and the l-participle, but with divergent patterns. It will be
shown that its distribution may be straightforwardly explained on the assumption that
whereas l-participles move as phrases, finite verbs undergo head movement.
Šte is the only auxiliary in Bulgarian that has an invariant form throughout the
whole paradigm. I take this to mean that it is generated above T. Following Rivero’s
(1994a) suggestion, I propose that it heads a Modal Phrase.
(110) Az/Ivan šte dojda

I/Ivan FUT comeSUBJ.3SG
“I/Ivan will come” (Bg, Hauge 1999: 105-106)
Šte is a proclitic, so it may appear clause-initially. It may be followed by finite verbs,

including the present tense auxiliaries, and must be adjacent to them.53
(111) Ivan šte (*bŭrzo) napiše pismoto

Ivan FUT quickly writeSUBJ.3SG letter-the
“Ivan will write the letter (quickly)” (Bg, Krapova 1999a: 76-77)
Šte may also be followed by a sequence consisting of an auxiliary clitic and the l-
participle. Since it is a clitic, it must cluster with the other clitics in the clause. The
sequence of clitics may not be interrupted by any other elements, such as the l-participle
izpil in (112b).
53 The only exceptions are other clitics, which may intervene between šte and the finite verb. See
section 4.4.3 in chapter 4 for details.

(112) a. Šte e izpil konjaka

FUT bePRES.3SG drinkPART.M.SG brandy-the
“He will have drunk the brandy”
b. *Šte izpil e konjaka (Bg, Embick & Izvorski 1995: 106)
A focused l-participle may be fronted across šte (cf. 113a). However, finite verbs, such
as izpie in (113b) may never be preposed across the future auxiliary, even when they are
focused.
(113) a. Izpil šte e konjaka

drinkPART.M.SG FUT bePRES.3SG brandy-the
“He will have DRUNK the brandy”
b. Šte izpie konjaka
FUT drinkSUBJ.3SG brandy-the
“S/he will drink up the brandy”
b’. *Izpie šte konjaka
drinkSUBJ.3SG FUT brandy-the
“S/he will DRINK up the brandy” (Bg, Labova 2003: 124)
On the assumption that finite verbs and l-participles have the same categorial status and
are both heads, the contrast between (113a and b’) is quite mysterious. Locality
principles cannot play a role here, because the participle in (113a) crosses more
elements on its way than the finite verb in (113b’), and yet only the former output is
well-formed. However, when the XP-movement approach to l-participle fronting is
adopted, the contrast receives a straightforward account. The finite verb izpie in (113b’)
may not raise across the future auxiliary, because this would lead to violation of the
Head Movement Constraint. Conversely, the l-participle may move over šte, because it
undergoes XP movement. First it raises to Spec, TP to check φ-features, and
subsequently to a focus projection in the left periphery of the clause to check a focus
feature. This results in a non-neutral interpretation of the preposed participle.
(114) a. Izpil šte e konjaka

b. [FocP [PartP izpilk ti] [Mod šte [TP [tPartP ti] [T e ... [AgrOP konjakai [vp Sub v [tPartP ti
]]]]]]]
Thus, it has been shown that fronting of the l-participle across the future auxiliary šte
provides more evidence for the proposal that the movement is phrasal. More
conclusive support for this claim will come from the way negation patterns with
participle and finite verb movement. This issue will be addressed in the next section.
2.3.6 Interaction between participle movement

and negation
The present section will examine interactions between participle movement and
negation. It will be demonstrated that in South Slavic negation attracts certain
grammatical categories, which incorporate into it. This property will be used as a
criterion to determine the position of other constituents in the clause structure, as well
as to decide whether the attracted element, such as the finite verb or the l-participle,
undergoes X0 or XP movement.
Negation has received a lot of attention in the literature, but a detailed overview of
the theories relevant to this phenomenon is far beyond the scope of this dissertation. In
the analysis pursued here I will assume that the negative marker ne is a head which
projects NegP, and that a constituent must be under the scope of NegP to be negated.
The position of the NegP in the clause structure is a matter of debate, but I will follow
Zanuttini (1997), who argues that there can be more than one NegP projections located
at different heights in the clause and that each of them may host the negative marker.
However, irrespectively of the number of NegPs, the polarity status of the clause (i.e.
its negation or assertion) is always interpreted at LF in a Polarity Phrase, which
dominates TP.
The subsequent sections will make a distinction between sentential negation and
constituent negation, therefore it is important to define these two terms. Sentential
negation takes scope over the whole predicate (cf. Acquaviva 1995: 84; Błaszczak 2001:
117), whereas constituent negation takes scope only over the elements it negates, so it is
generated on the constituent that is negated (cf. Iatridou 1990: 574). For instance,
(115a) involves sentential negation, because it scopes over the whole proposition, and
the sentence carries the meaning “it is not the case that I read the book”. By contrast,
(115b) exemplifies constituent negation, because only the internal argument is negated.
(115) a. Ja nisam čitao knjigu

I NEG+beAUX.1SG readPART.M.SG bookACC
“I didn’t read the book”
b. Ja sam čitao ne knjigu nego pismo
I beAUX.1SG readPART.M.SG NEG bookACC but letterACC
“I didn’t read the book but a letter” (S-C, N. Milićević, p.c.)
The distinction between sentential and constituent negation can sometimes be very
subtle, but a few tests have been devised to distinguish between these two types. One
of them is the not even test due to Klima (1964), who observes that the not even tag is
possible only in the case of sentential negation.54
(116) a. John doesn’t drive a car, not even a Fiat 500 sentential negation
b. *Not long ago Bill drove a car, not even a Fiat 500 constituent negation
The not even test can be used in South Slavic as well. As shown in (117), the not even tag is
compatible only with sentential negation, but not with constituent negation.
(117) a. Ja nisam čitao knjige, čak ni detektivske priče

I NEG+beAUX.1SG readPART.M.SG booksACC even not detective stories
“I didn’t read books, not even detective stories”
54 See Zeijlstra (2004: 47-51) for other tests that differentiate between sentential and constituent
negation. It will be shown in chapter 5 that in Polish the two types of negation trigger different
prosodic effects, and that only sentential negation licenses genitive of negation. These tests
cannot be applied in the South Slavic languages, though.
b. Ja sam čitao ne knjigu nego pismo

I beAUX.1SG readPART.M.SG NEG bookACC but letterACC
“I didn’t read the book but a letter”
b’. *Ja sam čitao ne knjigu nego pismo , čak ni
I beAUX.1SG readPART.M.SG NEG bookACC but letterACC even not
detektivske priče
detective stories
“*I didn’t read the book but a letter, not even detective stories”
(S-C, N. Milićević, p.c.)
The analysis presented in the next sections is organized as follows. Section 2.3.6.1
overviews the relation between participle movement and negation. Section 2.3.6.2
discusses properties of negation in Serbo-Croatian, and proposes that there are two
NegP sites available in this language: one above TP, referred to as “high negation”, and
another above VP, which will be dubbed “low negation”. Section 2.3.6.2.2 is concerned
with an interaction between participle fronting and negation in Serbo-Croatian. Section
2.3.6.3 describes negation in Bulgarian, arguing that it has only one NegP available.
Section 2.3.6.3.2 discusses participle movement in the presence of negation in
Bulgarian. On the basis of these findings, section 2.3.6.3.3 develops an alternative
account of negated questions in Bulgarian.
2.3.6.1 Introduction
The examples in (118) for Bulgarian and in (119) for Serbo-Croatian demonstrate that
the l-participle may not raise when it is preceded by the negative particle ne,
irrespectively of whether it undergoes short (cf. 118b/119b) or long movement (cf.
118/119c and d). Moreover, negation must always precede the auxiliary, rather than the
participle. However, placement of the subject in front of negation, as in (118a/119a), is
possible.
(118) a. (Az) ne sŭm pročel knigata

I NEG bePRES.1SG readPART.M.SG book
“I have not read the book”
b. *Az sŭm ne pročel knigata
I bePRES.1SG NEG readPART.M.SG book
c. *Pročel ne sŭm knigata
readPART.M.SG NEG bePRES.1SG book
d. *Ne pročel sŭm knigata
NEG readPART.M.SG bePRES.1SG book (Bg, cf. Rivero 1991: 331)
(119) a. (Ja) nisam čitao knjigu

I NEG+ bePRES.1SG readPART.M.SG book
”I haven’t read the book”
b. *Ja sam ne čitao knjigu
I bePRES.1SG NEG readPART.M.SG book
c. *Čitao nisam knjigu
readPART.M.SG NEG+bePRES.1SG book
d. *Ne čitao sam knjigu
NEG readPART.M.SG bePRES.1SG book (S-C, cf. Rivero 1991: 334)
According to Rivero (1991 & 1994: 90ff), this means that l-participle fronting is
blocked by negation. She suggests that negation heads NegP in Bulgarian and Serbo-
Croatian and takes TP as complement. On Rivero’s approach the participle raises as a
head via Long Head Movement from V to C, skipping the auxiliary located in I (cf.
section 2.2.1 for an evaluation). The presence of Neg0 is argued to block the operation.
The data in (118) and (119) require a more careful analysis, though, because
negation does not block participle movement in all contexts. For instance, what has not
been observed so far is that the fronting is blocked in Serbo-Croatian only by sentential
negation. The sentence in (119c) improves to perfection when a “correction phrase”,
such as nego pismo ‘but a letter’ is added (cf. 120a). It exemplifies constituent negation, as
it is incompatible with the not even tag.
(120) a. Čitao nisam knjigu nego pismo

readPART.M.SG NEG+bePRES.1SG book but letter
“I didn’t read the book but a letter” (S-C, A. Čamdžić, p.c.)
b. *Čitao nisam knjigu nego pismo,
čak ni detektivske priče
even not detective stories
“*I didn’t read the book but a letter, not even detective stories”
The situation with constituent negation in Bulgarian is more complicated, because

participle fronting seems to be sensitive to the type of the element that is negated. If
the negated element is an argument, participle fronting is barred; if it is an adjunct,
participle movement is possible. This is shown in (121) and (122), respectively.55
(121) a. *Pročel ne sŭm knigata, a pismoto

readPART.M.SG NEG beAUX.1SG book-the but letter-the
b. ?Pročel sŭm ne knigata, a pismoto (Bg, S. Marinov, p.c.)
(122) a. Čel sŭm ne za izpita a za udovolstvie

readPART.M.SG beAUX.1SG NEG for exam but for pleasure
“I didn’t read for the exam but for pleasure”
b. Učil e ne v Plovdiv, a v Sofia
studyPART.M.SG beAUX.3SG NEG in Plovdiv but in Sofia
“He didn’t study in Plovdiv, but in Sofia!”
c. Umrjal e ne na 24 maj a na 5 april
diePART.M.SG beAUX.3SG NEG on 24 may but on 5 april
“He didn’t die on 24 May, but on 4 April” (Bg, S. Marinov, p.c.)
It seems that Rivero’s account does not explain the patterns described above precisely
enough. For instance, her proposal predicts the example in (120a) to be ungrammatical,
contrary to fact. Correspondingly, it is not able to address the argument-adjunct
asymmetry with respect to negation in Bulgarian in any way.
In the subsequent sections I will analyze interactions of verb movement in
Bulgarian and Serbo-Croatian with negation in detail. I will demonstrate that sentential
negation does not block participle fronting. Rather, the fronting is made invisible by the
55Note that placement of negation in front of the auxiliary or the l-participle results in
ungrammaticality.
obligatory movement of verbal or pronominal heads to the Neg projection that

immediately dominates TP.
(123) [NegP [Neg NEG+V0/D0i [TP PartPk [T ti [PartP tk ]]]]]
Furthermore, I will show that the differences between Bulgarian and Serbo-Croatian
with respect to constituent negation are due to a lack of the low NegP in Bulgarian. The
analysis will also provide more evidence for the idea that the l-participle moves as a
phrase and lands in Spec, TP, whereas finite verbs undergo head movement.
2.3.6.2 Serbo-Croatian
2.3.6.2.1 Properties of negation

The negative marker ne is commonly assumed to be the head of NegP in Serbo-
Croatian. The evidence for its X0 nature comes from the behaviour of n-words such as
nikoga in (124), which may land in Spec, NegP to check the negative feature against
Neg. The example in (124b) indicates that the movement need not be overt, but if an n-
word is left in the post-verbal position, it receives a focus interpretation. The analysis of
the negative feature checking has been first proposed by Brown (1999) for Russian, and
extended to Serbo-Croatian by Progovac (2005b).
(124) a. On [NegP nikogai [Neg ne [voli ti]]]

he nobody NEG love3SG
“He loves no one”
b. On ne voli nikoga (S-C, cf. Progovac 2005b)
Progovac (2005b) posits that positive polarity items, such as the existential quantifier
neko ‘someone’56 are attracted to this position as well. Therefore, she claims that NegP
is not only related to licensing negation, but polarity items in general. For this reason,
she decides to term the projection PolP.
(125) Petar nekoga nije video

Peter someone NEG+beAUX.3SG seePART.M.SG
“Peter did not see someone” (S-C, Progovac 2005b)
N-words may participate in negative inversion. For example, the n-word might be
followed by an adverb, such as ranije ‘before’ in (126), and the whole complex may raise
to the clause initial position. I suggest the constituent raises via Spec, NegP, and that
the movement is triggered by a polarity feature in the Neg head.
56 According to Ladusaw (1980) and Linebarger (1981), existential quantifiers such as someone
should be classified as positive polarity items, because they can only be interpreted outside the
scope of clausal negation.
(126) Nikada ranije nisam pušio

never before NEG+beAUX.1SG smokePART.M.SG
“Never before have I smoked” (S-C, N. Milićević p.c.)
I will define the exact position of NegP in the clause structure in section 2.3.6.2.2. For
the time being, observe that negation obligatorily attracts verbs in Serbo-Croatian (cf.
Rivero 1991, Leko 1996,57 Progovac 2005b). As a result, no overt material may
intervene between the negative marker ne and the verb in (127).
(127) a1. Ne čitam knjigu

NEG readPRES.1SG bookACC
“I don’t read the book”
a2. *Ne knjigu čitam
b1. Ne čitam (nikad) knjigu (nikad)
NEG readPRES.1SG never bookACC never
“I never read a book”
b2. *Ne nikad čitam knjigu (Bosnian, Leko 1996: 15)
Likewise, pronominal or reflexive clitics are also prohibited in this position, and they
must follow the verb instead.
(128) a1. Ne diraj me

NEG touch2SG meCL.ACC
“Don’t touch me”
a2. *Ne me diraj
b1. Ne vracam se kuci
NEG return1SG REFL homeDAT
“I don’t return home”
b2. *Ne se vracam kuci (Bosnian, cf. Leko 1996: 15)
Rivero (1991: 338) argues that Neg attracts finite verbs in Serbo-Croatian, which
incorporate into it. Her argument is motivated by the position occupied by pronominal
clitics, which must always occur in the second position in this language. The example in
(129) conclusively shows that the sequence neg+verb forms a single word, because it
may be followed by the pronominal clitic ga. If this sequence were analyzed as two
words, the clitic in (129) would occur in a non-second position, which is excluded in all
other contexts in Serbo-Croatian (cf. chapter 4 for details).
(129) Ne (*ga) vidim ga

NEG himCL.ACC seePRES.1SG himCL.ACC
“I don’t see him” (S-C, Rivero 1991: 338)
The sentences in (130) indicate that negation forms a single word only with verbs. It
may not be placed clause-initially in front of a DP, such as Jovan in (130a) to render
constituent negation on the subject, because then the clitic ga will appear in the third
position. This proves that negation may only incorporate into verbs, and that the
incorporation is possible only in the case of sentential negation.
57 Leko’s analysis is concerned with Bosnian, but native speakers inform me that negation in
Serbo-Croatian patterns in exactly the same way.

(130) a. *Ne Jovan ga vidi

NEG Jovan himCL.ACC seePRES.3SG
b. Ne vidi ga Jovan
NEG seePRES.3SG himCL.ACC Jovan
“It is not Jovan who sees him” (S-C, N. Milićević p.c.)
However, it is important to observe that it is not only finite verbs that are attracted by
Neg in Serbo-Croatian. Infinitives, such as jesti ‘to eat’ in (131) are attracted to Neg too,
and they cannot be separated from it by any overt material, either.
(131) a. Ne jesti tesku hranu je preporučljivo

NEG eatINF heavy food be3SG recommendableN
“Not eating heavy food is recommendable”
b. *Ne tesku hranu jesti je preporučljivo (Bosnian, Leko 1996: 17)
By contrast, negation is unable to attract the l-participle in Serbo-Croatian. It must

attract the auxiliary verb instead.
(132) a. *Ne čitao sam knjigu

NEG readPART.M.SG beAUX.1SG book
b. (Ja) nisam čitao knjigu
I NEG+beAUX.1SG readPART.M.SG book
“I haven’t read the book” (S-C, N. Milićević p.c.)
The auxiliary in (132) is a clitic, so it might be possible to argue that the incorporation
of the auxiliary into negation is a way of compensating for its structural deficiency.
However, the example in (133), which contains a non-clitic, past tense auxiliary bješe
shows that this reasoning is not on the right track. Negation always attracts the auxiliary
verb, and may never adjoin to the l-participle. This indicates that the clitic status of the
attracted element is irrelevant, and that the attraction is not motivated prosodically. It is
rather the highest verbal head that is incorporated.
(133) a. *Ne čitao bješe knjigu

NEG readPART.M.SG beAUX.PAST.3SG book
b. (On) ne bješe čitao knjigu
he NEG bePAST.AUX.3SG readPART.M.SG book
“He hadn’t read the book” (S-C, N. Milićević p.c.)
Hence, I would like to claim that the relevant condition for the Neg attraction is neither
finiteness nor the clitic status of the verb. Rather, the verb must be a head in order to
incorporate into negation. This is expected, given that only heads may incorporate into
other heads (cf. Kayne 1994). Consequently, the l-participle is not eligible for
incorporation into negation, because it always moves as an XP in Serbo-Croatian. If it
were to undergo head movement, the contrast between (132) and the previous
examples could not be explained.
2.3.6.2.2 Negation and participle movement

I submit that in Serbo-Croatian NegP (or a projection that encodes polarity) can be
located either above TP (in “high” negation) or above VP (in “low” negation).
Furthermore, I also suggest that the low negation attracts n-words (e.g. nikoga ‘nobody’)
to its Specifier, as demonstrated for the sentence in (124), repeated below as (134).
(134) On [NegP nikogai [Neg ne [voli ti]]]

he nobody NEG love3SG
“He loves no one” (S-C, Progovac 2001: 239)
I have argued that verbal heads must obligatorily incorporate into negation. I will
assume that this holds for both the high and the low negation.
As is well-known, negation takes scope over the constituents it c-commands. Since
the low negation dominates VP, it is expected that it will trigger constituent negation.
This is exactly what happens in (125), repeated as (135), where the auxiliary je
incorporates into the low Neg. The sentence means that there is a certain person that
Peter did not see. It does not mean that Peter did not see any person.

Accordingly, a similar meaning is expressed by example (120), repeated below in (136a).

The sentence is acceptable only on the constituent negation reading, which means that
the auxiliary je incorporates into the lower Neg. The derivation is schematized in (136b).
(136) a. Pročitao nije knjigu nego pismo

“He didn’t read the book but a letter” (S-C, A. Čamdžić, p.c.)
b. [TP pročitao ... [NegP [Neg ne + jei [AuxP [Aux ti ... [AgrOP knjigu ]]]]]
By contrast, the examples in (119), repeated below in (137), instantiate sentential

negation.
(137) a. Ja nisam čitao knjigu

I NEG+beAUX.1SG readPART.M.SG book
”I haven’t read the book”
b. Nisam čitao knjigu
NEG+beAUX.1SG readPART.M.SG book
c. *Čitao nisam knjigu
readPART.M.SG NEG+beAUX.1SG book (S-C, Rivero 1991: 334)
Since negation takes scope over the entire clause, the negation marker ni must c-
command TP, and thus, dominate it. The derivation of the example in (137b) is given
in (138). Ni attracts the auxiliary clitic, which must move from T and adjoin to it in
Neg. This gives the impression of the blocking effect of negation on participle
movement. The impression is only apparent, because the participle raises to Spec, TP as
usual, but the movement is made invisible by the subsequent incorporation of the
auxiliary verb into negation.
(138) [NegP [Neg ni + sami [TP čitao [T ti ... [AgrOP knjigu]]]]]

Example (139) shows that the l-participle may not raise across the auxiliary that has
incorporated into negation. The movement is barred on the sentential negation reading,
because then negation must scope over the whole VP, rather than just the auxiliary
verb.
(139) *Čitao nisam knjigu

readPART.M.SG NEG+beAUX.1SG book (S-C, Rivero 1991: 334)
However, it is possible to move the complete TP dominated by negation for focus

reason, which results in an emphatic interpretation of the fronted constituent, as in
(140).
(140) a. Pio vina nisam

drinkPART.M.SG wine NEG+beAUX.1SG
“Drink wine, I did not!”
b. Čitao knjigu nisam
readPART.M.SG book NEG+beAUX.1SG
“Read the book, I did not!” (S-C, Tomić 1996a: 857)
The derivation proceeds as follows. The object vina moves out of the PartP for case
checking. The auxiliary sam raises as a head from Aux to T and checks Tense in this
position. Subsequently, the l-participle pio raises as a remnant XP to Spec, TP in order
to check φ-features. This stage of the derivation resembles the familiar case of l-
participle fronting argued for earlier in this chapter and is illustrated in (141a). Next, the
auxiliary is attracted by the head of the higher NegP and incorporates into it (cf. 141b).
Finally, the whole TP, which is dominated the high NegP, raises via A’-movement to
Spec, FocP, as schematized in (141c).58
(141) a. [TP [PartP pio ti] [T samj... [AgrOP vinai [Aux tj [vp [tPartP ti]]]]]]
b. [NegP [Neg ni + samj [TP [PartP pio ti] [T tj ... [AgrOP vinai ]]]]]
c. [FocP [TP [PartP pio ti] [T tj ... [AgrOP vinai ]]] [tTP [NegP [Neg ni + samj]]]]
The analysis proposes that the whole TP moves to Spec, FocP. Consequently, it
predicts that it should be possible to pied-pipe all the elements contained by TP across
negation. This is indeed the case. For example, adverbs, such as nikad ‘never’ or juče
‘yesterday’, must raise together with TP (cf. 142a), and the result is ungrammatical if the
adverb is not pied-piped by TP (cf. 142b).
(142) a. Pio vina nikad/juče nisam

drinkPART.M.SG wine never/yesterday NEG+beAUX.1SG
“Drink wine, I never did!”/”Drink the wine yesterday, I didn’t”
b. *Pio vina nisam nikad/juče (S-C, N. Milićević, p.c.)
Since the TP is raised across the higher NegP, it is expected that only the sentential
negation interpretation is available. The prediction is borne out, as is evidenced by the
ill-formedness of (143), which necessarily implies constituent negation.
58 The derivation partly resembles Kayne & Pollock’s (2001) analysis of stylistic inversion in
French, which also presupposes movement of the whole TP.

(143) *Pio vina nisam nego rakiju

drinkPART.M.SG wine NEG+beAUX.1SG but rakija
The intended meaning: “Drink wine, I did not, but rakija, I did
The proposed analysis can be straightforwardly extended to other polarity items if

Progovac (2005b) is correct when claiming that NegP is not only related to licensing
negation, but polarity in general. For example, she shows that positive polarity items,
such as the existential quantifier neko ‘someone’ in (125), repeated below as (144), move
to the specifier of the lower NegP as well.

In a similar vein, Wilder & Ćavar (1994: 22ff) observe that the assertive morpheme je
shows the same distribution as the negation particle ne. The morpheme is
homophonous with the 3rd person singular auxiliary form je59 and the merger of je and
the clitic auxiliary results in an emphatic assertive form such as jesam/be1SG, ‘I AM’.
Since the emphatic auxiliaries induce an affirmative interpretation of a clause (cf. 145a),
I would like to suggest that semantically they correspond to positive polarity items.
As expected, the permutations in (145) demonstrate that it is possible to front TP
across the affirmative auxiliary for focus reasons, on a par with the movement over the
negative auxiliaries. Movement of a bare participle in (145c) is excluded, because this
would only include the auxiliary under the scope of positive polarity, and not the l-
participle (cf. 139 for a related case of movement across nisam).
(145) a. Jesam čitao knjigu

je+beAUX.1SG readPART.M.SG bookACC
“I have read the book”
b. Čitao knjigu jesam
readPART.M.SG bookACC je+beAUX.1SG
“Read the book, I did
c. *Čitao jesam knjigu (S-C, Wilder & Ćavar 1994)
Apart from the auxiliaries related to assertion and negation, Serbo-Croatian has the
future/modal auxiliary ću (cf. chapter 1, section 1.3.4.4.1), which is an enclitic, so it may
not occur clause initially (cf. 146b), but must always appear in the second position (cf.
146c).
(146) a. Čitati ću knjigu

readINF wantCL.1SG bookACC
“I want to read the book”
b. *Ću čitati knjigu
c. *Čitati knjigu ću (S-C, Wilder & Ćavar 1994)
Ću may incorporate into the morpheme ho, which produces the focused variant hoću ‘I
WILL/WANT’. The modal auxiliary is followed by the infinitival form of a main verb
59 In chapter 4 I will show that the assertive je is not the same element as the 3rd person auxiliary,
but is rather a spell-out of a focus-related feature.

rather than the l-participle, but the structures with the modal verb exhibit the same
fronting patterns as the constructions with nisam and jesam. Thus, hoću may be preceded
by a focused TP, as shown in (147b), but the movement of a bare infinitive across it is
prohibited (cf. 147c).
(147) a. Hoću čitati knjigu

want1SG readINF bookACC
“I want to read the book”
b. Čitati knjigu hoću
“Read the book, I want!”
c. *Čitati hoću knjigu (S-C, N. Milićević, p.c.)
The data presented so far demonstrate that the auxiliary clitic may be attracted not only
by polarity items, such as the negative particle ne or the affirmative particle je, but also
by elements that express a modal meaning, such as ho in (147). In chapter 4 I will argue
that the attraction is an instance of Force licensing by the head Σ (cf. Laka 1994), which
hosts elements that are operators over propositions (TPs). Moreover, I will discuss
similar types of movement which involve imperatives in Macedonian (cf. chapter 4,
section 4.4.2.4.2.3) as well as subjunctive mood and topicalization in Polish (cf. chapter
5, sections 5.3.4.1.1 and 5.3.4.1.2).
The constructions with non-clitic affirmative and negative auxiliaries have been
analysed in the literature by Wilder & Ćavar (1994), among others. However, their
account of these constructions is different from mine. They argue that the morphemes
ni/ne, je, and ho occupy the AST (“assertion”) head,60 which is located between CP and
TP and which may adjoin to verbs, such as the auxiliary sam. This results in the creation
of “strong” (that is, non-clitic) auxiliaries. Recall from section 2.2.2 that Wilder & Ćavar
analyze l-participle fronting as head movement. Hence, on their approach the head
movement of the participle is blocked by the presence of the AST head.
(148) a. *[C VPart] … [ASTP [AST ni+sam] [TP t [VP tV ]]]

b. [C VPart] … [TP [T sam] [VP tV ]] (S-C, Wilder & Ćavar 1994: 23)
However, in this way their analysis faces the same problem as the one by Rivero (1991),
as it does not explain why the presence of the AST-head does not block participle
movement in the context of constituent negation.
(149) Pročitao nije knjigu nego pismo

“He didn’t read the book but a letter” (S-C, A. Čamdžić, p.c.)
Moreover, on the assumption that the strong auxiliaries are derived via adjunction of an
auxiliary clitic to the AST head above TP, it is predicted that the strong forms should
always appear in the second position or in the left periphery of the clause. However,
this is not the case, because the strong forms may occur in the same positions as other
finite verbs, that is following other clitics (cf. 150a) or clause-initially (cf. 150b).
60 In their later work (Ćavar & Wilder 1997) they dub the projection ΣP, following Laka (1994),
but place it below TP.

(150) a. Ja [mu se] nisam predstavio

I himCL.DAT REFL NEG+beAUX.1SG introducePART.M.SG
“I have not introduced myself to him” (S-C, Rivero 1991:336)
b. Nije mi ga dao
NEG+bePRES.3SG meCL.DAT itCL.ACC givePART.M.SG
“He didn't give it to me” (S-C, Tomić 1996a: 844)
This suggests that movement of the auxiliary clitics to NegP above TP is only one of
the options to pick up a polarity item. The auxiliary clitics may also do this by
incorporating into a polarity morpheme located above the VP, into the low Neg head.
This is what happens in the examples in (150). The fact that the strong auxiliary forms
may appear outside the left periphery supports the idea of the existence of a lower and
a higher Negation/Polarity Phrase in Serbo-Croatian.
Summarizing, it has been shown that Serbo-Croatian has two NegP projections in
the clause structure: above VP and TP. The head of NegP obligatorily attracts verbal
heads, which become incorporated and form a single word with it. Neg may never
attract the l-participle, which is to be expected on the assumption that the l-participle
always undergoes XP movement.
The subsequent section will analyze the relationship between negation and verb
movement in Bulgarian.
2.3.6.3 Bulgarian
2.3.6.3.1 Properties of negation

Just as in Serbo-Croatian, ne is claimed to head NegP in Bulgarian (cf. Dimitrova-
Vulchanova 1995; Dimitrova-Vulchanova & Hellan 1999). It attracts n-words, such as
ništo ‘nothing’ in (151), which may raise into Spec, NegP, as shown in (151b). The
movement is not obligatory, but if the n-word is left in the post-verbal position (cf.
151a), it receives a focus interpretation.
(151) a. Petŭr ne čete ništo

Peter NEG readPRES.3SG nothing
“Peter doesn’t read anything”
b. Ništo ne čete Petŭr (Bg, Pavlov 2000: 45)
The assumption that n-words target a specifier position in Bulgarian is confirmed by the
fact that they may pied-pipe other words, such as the adverb predi ‘before’. I will assume
that in (152) predi forms a constituent with the n-word ništo ‘never’, and that they move
together to Spec, NegP.
(152) Ništo predi ne sŭm go viždal

never before NEG beAUX.1SG himCL.ACC seePART.M.SG
“Never before have I seen him” (Bg, S. Marinov p.c.)
Negation attracts finite verbs, the way it also does in Serbo-Croatian. Therefore, no
overt material may intervene between ne and the auxiliary clitic in (153a) or between ne
and the non-clitic auxiliary beše in (153b).
(153) a. Vlakŭt ne (*togava) sme go čuli da pristiga

train-the NEG then bePRES.3PL itCL.ACC heardPART.PL that arriveSUBJ.3SG
“As for the train, we did not hear it coming”
b. Vlakŭt ne (*togava) beše pristignal
train-the NEG then bePAST.3SG arrivePART.M.SG
“The train had not arrived (then)” (Bg, Pavlov 2000: 42)
However, in contrast to Serbo-Croatian, negation in Bulgarian must attract not only

verbal heads, but pronominal and reflexive clitics as well. In fact, in the presence of
both clitics and the finite verb, the clitics must be attracted, rather than the verb (cf.
154b3).
(154) a1. Ne me pipaj

NEG meCL.ACC touch2SG
“Don’t touch me”
a2 *Ne pipaj me
b1. Nikŭde ne se sreštat takiva primeri
nowhere NEG REFL encounter3PL such examples
“Such examples cannot be encountered anywhere”
b2. *Nikŭde se ne sreštat takiva primeri
b3. *Nikŭde ne sreštat se takiva primeri (Bg, Pavlov 2000: 80)
Rivero (1994b: 113) argues that this means that negation in Bulgarian imposes no
categorial restrictions on the clitics it attracts. It is always the highest clitic available, in
line with the template given in (155; cf. chapter 4 for details concerning cliticization),
and irrespectively of the fact whether it is a pronominal, reflexive or an auxiliary clitic.
(155) li> šte > AUX (except 3rd SG e) > DAT > ACC > REFL > e
(Bg, Tomić 1996a; Franks & King 2000: 61)
However, negation attracts not only clitics, but non-clitic elements such as the past
tense auxiliary beše (cf. 153b) as well. Therefore, I would like make a generalization that
negation in Bulgarian attracts the first available head, regardless of its clitic or non-clitic
status.
A well-known property of the negation particle in Bulgarian is that it never bears
stress by itself, but it is always followed by a stressed constituent, such as the verb vali
in (156).
(156) a. Ne vaLI
NEG rain3SG
“It doesn’t rain” (Bg, Rudin et al 1999: 562)
Obviously, clitics do not carry any lexical stress of their own. However, they do receive
stress in the presence of negation. This is indicated (157) by capitalization. In (157a) the
stress falls on the accusative clitic me, which precedes the verb boli. In (157b and c) the
verbs are preceded by a whole clitic clusters, but it is only the first clitic that receives
stress from negation.
(157) a. Ne ME boLI
NEG meCL.ACC hurt3SG
“It doesn’t hurt me”
b. Ne MI se STRUva, če ...
NEG meCL.DAT REFL seem3SG that
“It doesn’t seem to me that…”
c. Ne SŬM ti go DAla
NEG bePRES.1SG youCL.DAT itCL.ACC givePART.F.SG
“I haven’t given it to you” (Bg, Rudin et al 1999: 562)
In view of this property, the consensus is that in Bulgarian negation forms a prosodic
word with the constituent that follows it (cf. Dimitrova-Vulchanova 1995; Tomić
1996a; Rudin et al 1999).
(158) a. [Ne ME] [boLI]

b. [Ne MI] se [STRUva]
c. [Ne SŬM] ti go [DAla] (Bg, Rudin et al 1999: 563)
Since prosodic constituency is often assumed to reflect syntactic constituency (cf.

Progovac 2000), I take this to mean that negation attracts only the first clitic in the
cluster (cf. section 2.3.6.3.3 for more evidence).
Still, even though negation in Bulgarian attracts finite verbs, verbal clitics and
pronominal clitics alike, it is not able to attract the l-participle. Again, this is to be
expected, because the l-participle moves as an XP. Since negation is a head, it may
adjoin to verbal heads, such as the past tense auxiliary beše or the clitic e in (159a), but
not to the l-participle pristignal in (159b).
(159) a. Vlakŭt ne e/ beše pristignal na vreme

train-the NEG bePRES/PAST.3SG arrivePART.M.SG on time
“The train had not arrived on time”
a. *Vlakŭt e/ beše ne pristignal na vreme
train-the bePRES/PAST.3SG NEG arrivePART.M.SG on time (Bg, Pavlov 2000: 41)
Hence, the main descriptive difference between negation in Serbo-Croatian and

Bulgarian is that in the former language only verbal heads are attracted by Neg. In the
latter language the closest head available is attracted by Neg. In the subsequent section I
will try to account for this contrast by suggesting that this is due to the presence of only
one NegP layer in Bulgarian.
2.3.6.3.2 Negation and participle movement

Pavlov (2000: 73) points out that the differences in the distribution of negative
elements in Bulgarian and Serbo-Croatian indicate that the former lacks the low NegP
above VP. I take his observation to be correct. Moreover, Pavlov shows that negation
in Old Church Slavonic patterned in the same way as in Serbo-Croatian, so verbs had
to obligatorily incorporate into the negative particle ne. This is exemplified in (160),
where negation immediately precedes the finite verb in the past tense, poklonijõ ‘bowed’.
The orderings corresponding to the ones in (157a and b), with negation preceding
pronominal clitics, are unattested in Old Church Slavonic.
(160) I ne poklonijõ sę imъ bъxma

and NEG bowed3PL REFL themCL.DAT at all
“And they did not bow to them at all”
(OCS, Suprasliensis 261.2; Pavlov 2000: 76)
Once the lower NegP is lost, verbs are not able to adjoin to it any more. Presumably,
they are not able to adjoin to the higher Negation dominating TP, either, because it is
too distant. In chapter 4 I will show that pronominal clitics in Bulgarian must leave the
VP in order to check number and person features. Because of this obligatory
movement, they may precede the verb, and thus are eligible for attraction by negation.
The loss of the lower NegP in contemporary Bulgarian has consequences for the
interaction between negation and participle movement. Recall from section 2.3.6.1 that
participle fronting in Bulgarian is blocked by both sentential and constituent negation,
which is exemplified in (161).
(161) a. ?Pročel sŭm ne knigata, a pismoto

readPART.M.SG beAUX.1SG NEG book-the but letter-the
b. *Pročel ne sŭm knigata, a pismoto
readPART.M.SG NEG beAUX.1SG book-the but letter-the
(Bg, S. Marinov, p.c.)
Conversely, participle movement is possible in Serbo-Croatian when constituent

negation is involved, so the Serbo-Croatian equivalent of the sentence in (161b) is
grammatical (cf. 136). The contrast is due to the presence of the lower NegP in Serbo-
Croatian, which c-commands the VP and is thus able to negate it without negating the
whole clause. This option is unavailable in Bulgarian, because of the lack of NegP
above VP. However, the crucial question is how Bulgarian is able to express constituent
negation once it has lost the lower NegP. It seems that the only option is to apply
contrastive negation by placing negation in front of any element that is supposed to be
negated. In this context the negative particle is not the head of NegP, but rather
functions like a negative adverb. Hence, this strategy does not involve any restriction on
the position of negation in the phrase structure. For example, the subject in (162) can
be negated whether it is clause-initial, or clause-final.
(162) a. Ne az sŭm izmislil tova

NEG I beAUX.1SG conceivePART.M.SG this
“I haven’t come up with this (thing)”
b. Tova sŭm go izmislil ne az (a Ivan)
this beAUX.1SG itCL.ACC conceivePART.M.SG NEG I but Ivan
“This hasn’t been conceived by me (but by Ivan)” (Bg, Pavlov 2000: 74-75)
As far as sentential negation is concerned, Bulgarian patterns in the same way as Serbo-
Croatian. The movement of the l-participle is made invisible by the obligatory
incorporation of the auxiliary into the negative marker above TP. The derivation of the
sentence in (163a) is presented in (163b-c). First, the l-participle pročel raises to Spec,
TP, in the same manner as described earlier in this chapter. Next, the auxiliary verb sŭm
is attracted by negation and incorporates into the Neg head located immediately above
TP.
(163) a. Ne sŭm pročel knigata

NEG bePRES.1SG readPART.M.SG book
“I haven’t read the book”
b. [TP [PartP pročel ti] [T sŭmj... [AgrOP knigatai [Aux tj [vp Sub v [tPartP ti ]]]]]]
c. [NegP [Neg Ne + sŭmj [TP [PartP pročel ti] [T tj ... [AgrOP knigatai ]]]]] (Bg)
The assumption that NegP is located above TP in Bulgarian is additionally supported

by the position of subjects in negative clauses. Arnaudova (2003: 93) shows that the
subject may then occur stranded at the right edge of the clause.
(164) Ništo ne e napravil Ivan

nothing NEG beAUX.3SG doPART.M.SG Ivan
“Ivan didn’t do anything” (Bg, Arnaudova 2003: 93)
I suggest that the post-verbal subject in (164) is in Spec,vP, as is also proposed by

Arnaudova (2003) on independent grounds. The l-participle napravil is located in Spec,
TP, where it has moved to check the φ-features of T. The auxiliary e incorporates into
the negation in Neg. The negative adverb ništo opens the clause in Spec, NegP (cf. 165).
(165) [NegP ništo [Neg ne + ei [TP napravil [T ti ... [vP Ivan ]]]]]
The fact that the subject often appears towards the end of the clause in negative clauses
has led some researchers (e.g. Dimitrova-Vulchanova & Hellan 1999; Pavlov 2000) to
assume that the subject-final order is the basic one in Bulgarian.61 However, native
speakers inform me that the object ništo is somewhat focused in (164). At any rate, this
suggests that negation must reside very high in the Bulgarian clause structure.
Summarizing, this section has shown that negation in Bulgarian may only attract
pronominal and verbal heads. Just as in Serbo-Croatian, it may not attract elements that
undergo XP-movement, such as the l-participle. More evidence for this distinction will
come from the analysis of negated focus questions in Bulgarian, which is presented in
the next section.
2.3.6.3.3 Interaction between the interrogative complementizer li and

negation in Bulgarian
This section will discuss a problematic issue of negated questions in Bulgarian, which
has given rise to a lot of controversy in the literature. It will be demonstrated that the
phenomenon receives a principled explanation once it is assumed that the l-participle
moves as an XP element.
Focus questions in Bulgarian are formed with the particle li. Li is an interrogative
complementizer that is found in most of the Slavic languages. It is commonly assumed
to be hosted in C, because it is in complementary distribution with other
complementizers (cf. chapter 4, section 4.4.3.2 for details).62
61 See Rudin (1986: 14ff) for compelling evidence showing that this is not true.
62 Moreover, li is right-adjacent to the wh-word in wh-questions, which also suggests that it is in C.
(i) a. Koj li e vzel knigata?

who Q beAUX.3SG takePART.M.SG book-the
“Who(ever) could have taken the book?”
b. *Koj e vzel li knigata? (Bg, Rudin 1986: 67)
Li is an enclitic and may be preceded by many different constituents, both heads

and phrases alike. Rudin (1986: 64) claims that when the preceding word is a verb, the
whole sentence is interpreted as being questioned.
(166) a. Kupil li e vestnika?

buyPART.M.SG Q beAUX.3SG newspaper-the
“Did he buy the newspaper?”
b. Pitaxa me kupil li e vestnika
asked3PL meACC buyPART.M.SG Q beAUX.3SG newspaper-the
“They asked me whether he (had) bought the newspaper”
(Bg, Rudin 1986: 63)
However, when li is attached to a non-verbal constituent, the scope of interrogation

falls on this constituent. In such cases li licenses a focus feature on the fronted element
that is located in Spec, CP (cf. Rudin et al. 1999).
(167) a. Kŭštata li namerixte (vie)?

house-the Q found2PL you
“Was it the house that you found?” (Rudin 1986: 64)
b. Prez gardinata li xodeše?
through garden-the Q walked2SG
“Were you walking THROUGH THE GARDEN?” (Rudin et al. 1999:546)
Furthermore, li can be used for questioning single words in isolation, which might be
heads and XPs alike (cf. 168a and c).
(168) a. Az li?
I Q
“Me?”
b. Kŭštata li?
house-the Q
“The house?”
c. Na masata li?
on table-the Q
“On the table?” (Bg, Rudin 1986: 65)
Li is an enclitic, so it needs a phonological host to its left, but it must also precede all
the other clitics when it clusters together with them. The ordering in which it occurs is
indicated in the template in (169) and the examples in (170).
(169) li> Mod > Neg > šte > AUX (except 3rd SG) > DAT > ACC > REFL > e
(3rd SG AUX)
(cf. Tomić 1996a, Franks & King 2000; cf. chapter 4 for details)
(170) a. (Vie) davate li mu go?

youPL give2PL Q himCL.DAT itCL.ACC
“Do you give it to him?”
b. (Vie) dali li ste mu go?
youPL givePART.PL Q beAUX.2PL himCL.DAT itCL.ACC
“Have you given it to him?” (Bg, Hauge 1999: 117)
As shown in (171), clitics may be attracted by negation. However, they are arranged
hierarchically with respect to each other, and only the highest clitic is accessible for the
attraction. Hence, negation may attract the accusative clitic go (cf. 171a) only in the
absence of the dative clitic mu (cf. 171c). If both pronominal clitics are present, only the
dative one may be preposed (cf. 171b). Moreover, it is impossible to move more than
one clitic at the same time (cf. 171d).63
(171) a. Ne go li iskate
NEG itCL.ACC Q want2PL
“Don’t you want it?”
b. Ne mu li go dadoxte?
NEG himCL.DAT Q itCL.ACC gave2PL
“Didn’t you give it to him?”
c. *Ne go li mu dadoxte?
d. *Ne mu go li dadoxte? (Bg, Hauge 1999: 108; S. Marinov, p.c.)
Apart from pronominal clitics, auxiliary clitics and non-clitic finite verbs can be moved
as well (cf. 172a and b). However, since the 3rd person singular auxiliary clitic follows all
the others in the cluster (cf. the template in 169), it may not be preposed across the
pronominal clitics (cf. 172c).
(172) a. Ne iskate li?

NEG want2PL Q
“Don’t you want to?”
b. Ne ste li mu go dali?
NEG beAUX.2PL Q himCL.DAT itCL.ACC givePART.PL
“Haven’t you given it to him?”
c. *Ne e li mu go dal?
NEG beAUX.3SG Q himCL.DAT itCL.ACC givePART.M.SG
(Bg, Hauge 1999: 108; S. Marinov, p.c.)
The pattern has been a long-standing problem of Bulgarian syntax and has received a
number of different analyses. For example, Rivero (1993) claims that in the context of
neutral yes-no questions, such as (166a) and (170), the verb moves as a head to C to host
the clitic li, which may not appear in the clause-initial position. In negated clauses,
however, the verbal head may not move to C, because negation is a barrier for the
movement. Consequently, li must lower (“hop”) in the structure. It adjoins to I, and as
a result it precedes the finite verb and the participle, but follows the clitics.
Izvorski et al (1997) notice empirical problems with Rivero’s account. First of all,
negation cannot be a barrier for all verbal heads, because non-clitic finite verbs may
move in front of li (cf. 172a). Second, Rivero’s analysis predicts that if there are several
clitics in a negated question, all of them will precede li, given that li must lower and
adjoin to I. In reality this never happens, and li always appears after the first clitic,
irrespectively of the number of clitics that are present in the cluster.
According to Izvorski et al, the crucial property of li is that it requires a
phonologically overt host capable of bearing stress to its left. Therefore, they suggest
that if a host of this type is missing, li will undergo prosodic inversion and will cliticize
63 All the sentences in (171) have a sentential negation reading. This is to be expected, as the
“high” negation is involved, so ne scopes over the entire clause.

onto the right edge of the first word that bears stress. The process is claimed to take
place at PF.
(173) Li izpratix mu kniga
Q sendPAST.1SG himCL.DAT book

“Did I send him the book?” (Bg, Izvorski et al. 1997: 193)
Negation does not carry stress on its own in Bulgarian, so it is not a suitable candidate
for giving support to li. However, it has the property of shifting stress over to the word
that follows it, even if it is a clitic. Hence, in negated clauses li will move across the first
clitic to its right (capitalized in 174), which will carry stress in this context.64
(174) Ne MU li go dadoxte?
NEG himCL.DAT Q itCL.ACC gave2PL
“Didn’t you give it to him?” (Bg, Izvorski et al. 1997: 193)
Without appealing to the mechanism of prosodic inversion, Bošković (2001 ch. 4.3.1.1)
analyzes the data under a “scattered deletion” approach. What is crucial for his account
is the fact that li must occur after the first stressed constituent in the clause. He argues
that the string ne go vidja (cf. 175a) forms a complex head, which arises through
adjunction of the pronominal clitics and negation to the verb. Next, the complex head
left-adjoins to li (cf. 175b).
(175) a. [Ne go vidja]

NEG himCL.ACC saw2/3SG
“He/she/you didn’t see him”
b. [ne go vidja+li] ne go vidja (Bg, Bošković 2001: 208)
This results in a phonologically infelicitous structure, because ne assigns stress to the

word that follows it. In the case at hand, the clitic go is stressed. The problem is that
vidja is stressed, too, because it is a lexical verb with its own lexical stress. However, li
must occur after the first stressed word in the clause, while in (175b) there are two
stressed elements preceding it. Bošković argues that the derivation is rescued by a
phonological filter,65 which enforces spell-out of the lower copy of the verb vidja, rather
than the head of the chain.
(176) [ne go vidja+li] ne go vidja
In this way the phonological requirements of both the negative marker ne and the
interrogative complementizer li are met.
64 A similar account of the facts, which also relies on prosodic inversion, is presented in Franks
(1998) and Rudin et al. (1999).

65 According to Bošković, syntactic movement is evaluated by a phonological filter. It may
enforce pronunciation of a lower of copy of the movement chain if the pronunciation of the
head of a chain results in a violation of phonological constraints. See Bošković (2001) for an in-
depth description of his pronounce-a-copy account of cliticization in South Slavic.
To sum up what has been established so far, all the previous approaches to the
interaction between li and negation state that li may be preceded by clitics (cf. 177a) and
finite verbs (cf. 177b) in the presence of ne.
(177) a. Ne go li iskate
NEG itCL.ACC Q want2PL
“Don’t you want it?”
b. Ne iskate li?
NEG want2PL Q
“Don’t you want to?” (Bg, Hauge 1999)
However, a major flaw of these prosodic analyses is that they do not take into account
the fact that neither the l-participle nor the subject may precede li in this context.66
(178) a. *Ne kupil li e knigata?

not buyPART.M.SG Q bePRES.3SG book-the
b. Ne e li kupil knigata?
“Hasn’t he bought the book?” (Bg, S. Marinov, p.c.)
(179) a. *Ne az li (sŭm)?

NEG I Q bePRES.1SG
b. Ne sŭm li az?
“Not me?” (Bg, S. Marinov, p.c.)
This is quite striking, given that the non-negated variants of (178a) and (179a) are
completely grammatical.
(180) a. Kupil li e knigata?

buyPART.M.SG Q beAUX.3SG book-the
“Has he bought the book?”
b. Az li?
I Q
“Me?” (Bg)
Likewise, the sentence corresponding to the one in (178a) is completely acceptable if a

past tense form of the verb is used instead of the l-participle, as shown in (181).
(181) a. Ne kupi li knigata?

not bought3SG Q book-the
“Didn’t he buy the book?” (Bg, S. Marinov, p.c.)
Moreover, there is no phonological reason why (178a) and (179a) should be excluded.
The stress patterns that these examples would potentially show should be the same as
in the constructions with past tense verbs or clitics. In principle, negation may place
66 The only syntactic account of li placement in the presence of negation that I am aware of is due
to Tomić (1996a), who argues that the ‘ne + first clitic’ sequence excorporates from the clitic
cluster and raises as unit in order to lend support to li. However, she does not mention the way
this sequence patterns in the presence of the subject and the l-participle (cf. 178a and 179a). She
also does not provide a syntactic motivation for this movement.
Conclusions 123
stress on the l-participle or the subject, so there is no reason why this option should be
ruled out.
I conclude that the contrasts indicate that the process is constrained syntactically,
rather than prosodically. The impossibility of the subject or the l-participle insertion
between negation and li is due to the phrasal status of these elements.
I propose that the focus questions are constructed in the following way: the
highest head available in the structure is attracted by negation and subsequently, the
‘negation+head’ complex left-adjoins to li. Each of the attracted elements has a focus
interpretation, whereas li is the spell-out of the focus feature. I suggest that these
elements check focus via adjunction into li. Since negation is a head, it may only attract
X0 elements, such as finite verbs and clitics. It does not attract l-participles or subjects,
because they undergo XP movement. As an example, I provide the derivation of (171b)
in (182).
(182) a. Ne mu li go dadoxte?
NEG himCL.DAT Q itCL.ACC givePAST.2PL
“Didn’t you give it to him?”
b. [TP [T dadoxtei] [VP [V ti] mu go]]
c. [TP [T muj + gok + dadoxtei][VP [V ti ] tj t k]]
d. [NegP [Neg ne+muj] [TP [T t’j + gok + dadoxtei][VP [V ti ] tj t k]]]
e. [CP [C <ne+muj>m + li] [NegP [Neg tm] [TP [T t’j + gok + dadoxtei][VP [V ti ] tj t k]]]]
The derivation proceeds as follows. Since the verb dadoxte is tensed, it must raise to T in
order to check Tense (cf. 182b). Pronominal clitics must be adjacent to the verb in
Bulgarian, which I assume implies that they raise from their argument positions within
the VP and adjoin to the verb in T (cf. 182c and section 4.4.3.4.3 in chapter 4 for
details). NegP is merged immediately above TP. Negation in Neg attracts the highest
head available below it in the structure, which is mu in (182), which incorporates into it.
Finally, li is merged as the C head. Li is a spell-out of the Focus feature; besides it is
also an enclitic, so it needs a phonologically overt element to its left. The complex head
ne+mu raises out of Neg and left-adjoins to li in C (cf. 182e). This results in the focus
question Ne mu li go dadoxte?.
To summarize, the analysis of the negated questions developed in this section has
shown that the operation is fully syntactic. Moreover, it has demonstrated that the
impossibility of the subject or the l-participle insertion between negation and li is due to
their phrasal status. This provides more evidence for the claim that the l-participle
undergoes XP movement.
Furthermore, this section has made two hypotheses concerning the nature of
negation in Slavic. First, sentential negation always involves incorporation of a verbal or
pronominal head into the negative particle. Hence, it is a useful criterion for deciding
about the X0/XP status of other elements in the clause. Second, it has been suggested
that NegP may be located in two positions in Slavic: above TP, as in Bulgarian, or
above both TP and VP, as in Serbo-Croatian. The logical extension of this idea is that
there should be a language with just one NegP above VP. It will be shown in chapter 5
that this option is represented by Polish.
2.4 Conclusions
This chapter has analyzed the structure of the compound tenses formed with the l-
participle and the auxiliary ‘to be’. The l-participle was argued to be able to assign
accusative case and project an external theta role. Moreover, it was claimed that the l-
participle and the subject are in a Small Clause configuration, which is overtly
manifested through morphological agreement between the two constituents.
The main topic of the chapter was the widely-discussed l-participle fronting across
the auxiliary to the clause-initial position. It was pointed out that the previous accounts
of the operation, which argue that it proceeds via head movement, face theoretical and
empirical problems. For example, they are unable to explain why the subject and the
preposed l-participle are in complementary distribution, or why the l-participle must be
the left-most constituent in the clause.
The present proposal suggests that the l-participle raises into Spec, TP, and that
the movement is phrasal. It is contingent on agreement, as the l-participle checks the φ-
features of T in this position. The phrasal status of the movement was confirmed on
the basis of the properties of double participle constructions, short participle
movement, as well as the interactions between the l-participle and the future auxiliary šte
in Bulgarian and negation.
Chapter 3 The syntax of ‘have’-perfects
and passive participles
3.1 Introduction
So far the thesis has been concerned with the syntax of compound tenses formed with
the auxiliary ‘be’ and the l-participle. The present chapter will discuss properties of the
tenses constructed with the auxiliary ‘have’ and the past and passive participles. These
are the default compound tense constructions in Germanic and Romance, but as was
noted in chapter 1, section 1.3.4.5.2, they are diachronically very recent in Slavic, and
have been completely grammaticalized only in Kashubian and Macedonian. They have
received very scant attention in the generative literature so far. This chapter aims at
analyzing them in detail, and is organized as follows. Sections 3.2 and 3.3 will describe
properties of passive participles. It will be shown that they vary in the degree of their
verbal or adjectival properties. In spite of the variation, following the ideas developed in
chapter 2 I will claim that they are categorially the same as the past participle in the
‘have’-perfect construction. I will also argue that a high degree of verbiness of the
participle is a hallmark of the grammaticalization of the ‘have’-perfect. Section 3.4 will
discuss ‘have’-perfects in Kashubian and Macedonian. To my knowledge, the syntax of
Kashubian has not been studied in the generative literature so far, thus section 3.4.1
introduces the language into this framework. Special attention will be given to past
participle fronting across the auxiliary ‘have’ in Macedonian (cf. section 3.4.2.2), which
will be contrasted with l-participle movement across the auxiliary ‘be’ in Bulgarian and
Serbo-Croatian investigated in chapter 2. Subsequently, section 3.5 will analyze
impersonal participles in Polish. They resemble past participles in Kashubian and
Macedonian in their morphological structure. However, it will be shown that they
exhibit a higher degree of verbiness, because they are able to assign accusative case and
project an external theta role. Section 3.6 will describe the process of
grammaticalization of ‘have’-perfects in some older variants of Germanic and Romance
languages. Section 3.6.3 will demonstrate that the ‘have’-perfect is currently undergoing
the same process of grammaticalization in some Slavic languages.
3.2 A typology of passive participles

It has often been observed in the literature that passive participles do not form a
uniform grammatical category. As a result, a number of taxonomic distinctions have
been proposed to capture their variation. For instance, Wasow (1977) suggests a
division into verbal and adjectival participles. In English they are morphologically
identical in most contexts, but there are a few criteria that are used to distinguish
between the two types. For example, Siegel (1973) observes that the negative prefix un-
attaches to adjectives, such as unfriendly, unhappy, and unspectacular, as wells as to some
passive participles, such as unopened, unshaven, and untouched. However, un- may not be
126 The syntax of ‘have’-perfects and passive participles
prefixed to verbs (cf. 1).67 Therefore, those participles that permit prefixation with un-
are regarded as adjectival (cf. 2).
(1) a. *Human hands untouch our products

b. *Humans uninhabited the island
c. *Data have unsupported all his claims (Anagnostopoulou 2003)
(2) a. Our products are untouched by human hands

b. The island was uninhabited by humans
c. All his claims have been unsupported by the data (Anagnostopoulou 2003)
Furthermore, Wasow (1977) notices that only adjectives and adjectival participles may
be selected as complements by linking verbs such as act, become, remain, seem, and sound
(cf. 3a). The contrast is due to the fact that linking verbs carry the meaning of a
completed activity or “having a property”, which is compatible with adjectives, but not
with verbs.
(3) a. The door remained closed during the noon hour

a.’ John seems very (un)satisfied
b. *Many polluted cities remain (un)avoided/escaped during the summer
b.’ *New York seems (very much) approached/left in the tourist season
(Emonds, 2000: 174-175)
Finally, it has been pointed out that only adjectival participles may act as DP-modifiers.
Since verbs do not modify nouns, verbal passive participles may not appear in a
prenominal adjectival position.
(4) The broken/filled/painted/cherished box sat on the table

(Anagnostopoulou 2003)
Levin and Rappaport (1986: 625) argue that the two types of passive participles differ
also in their semantics. Thus, adjectival participles describe a state which results from a
previous event and do not introduce an implicit agent. Conversely, verbal participles
characterize an event that has taken place and imply the existence of an agent.
The variation in the distribution of adjectival and verbal participles is often
assumed to reflect contrasts in their syntactic structures. For instance, Wasow (1977)
claims that while verbal participles are derived in syntax, adjectival passive participles
must be formed in the lexicon, because they involve a category change when a verb is
reinterpreted as an adjective. Jackendoff (1977) and Abney (1987) challenge this view,
arguing that both types of participles are built in syntax. However, in the case of
adjectival passives the participial morphology is a sister to V (cf. 5). In verbal passives
participial morphology is adjoined to VP (cf. 6).
67Un- can be attached to some verbs, such as zip/unzip, load/unload, button/unbutton. However, in
these cases the prefix carries the meaning of a reversal of an action, rather than negation. For
instance, John unloaded the truck does not mean that John did not load the truck. Rather, it means
that somebody loaded the truck first and then John reversed the action (cf. Allen 1978).
A typology of passive participles 127
(5) lexical affixes (adjectival passive participle)
V Z
participle affix
(6) phrasal affixes (verbal passive participle)
VP Z
participle affix
(Abney 1987, as illustrated in Anagnostopoulou 2003)
The contrast between XP versus X0 adjunction of the participial morphology captures

the fact that only verbal passives may contain DP complements, which are italicized in
(7).
(7) a. Those workers were allowed a lot of vacations

a.’ Peter was (being) forgiven his sins
b. *Those workers seemed allowed a lot of vacations
b.’ *Peter felt forgiven his sins (Emonds 2000: 177)
Summarizing, I have enumerated a number of differences between adjectival and verbal

passive participles. The adjectival vs. verbal dichotomy has been modified a number of
times in the literature. Some researchers (e.g. Kratzer 1994, 2000; Embick 2004) have
argued for more fine-grained distinctions, but the exact details of these approaches are
not relevant for the present purposes. What is crucial, though, is that passive participles
may vary in the degree of their verbal or adjectival properties. The verbal properties
that I will be concerned with are mainly syntactic and include the ability to assign
structural case and project an external theta role. Adjectives and adjectival passive
participles are not able to perform these functions. However, through a process of a
diachronic reanalysis they may increase their verbiness, become reanalyzed as verbal
forms and in some cases even reinterpreted as finite verbs. The present chapter will
demonstrate that the Slavic languages display an array of passive participles which are
grammaticalized as ‘verbal’ to various degrees. Before I turn to Slavic, I will discuss
properties of ‘have’-perfects, which are formed with a verbalized variant of the passive
participle.
3.3 Generalizations concerning ‘have’-

perfects
The Germanic and the Romance languages form compound tenses with the auxiliary
‘have’ or ‘be’, which select the past participle as a main verb. The past participle is
morphologically very similar or identical to the corresponding passive participle. As an
example, consider the Dutch sentences in (8).
(8) a. Een geschreven manuscript

a written manuscript
b. Het manuscript werd geschreven
the manuscript was written
c. Hij heeft het manuscript geschreven
he has the manuscript written
“He has written the manuscript” (Dutch, Hoekstra 1986: 97)
The participle geschreven can be used as an attributive adjective (cf. 8a), a passive
participle (cf. 8b), or as a past participle (cf. 8c). All the three variants are identical in
form. The participle derives from the verb schrijven ‘to write’, which assigns two theta
roles: Agent and Theme. However, it has been assumed since Chomsky (1981: 54-55
and 117-127) that the thematic role Agent is absorbed by the passive morphology.
Moreover, the passive morphology disables the verb from assigning structural case to
its complement. Consequently, in order to comply with the Case Filter requirement the
object het manuscript must move to the subject position in order to receive nominative
case from INFL. This results in the placement of the object het manuscript in the subject
position in (8b).
The situation is more complex in the compound tense construction in (8c). Even
though the participle geschreven is morphologically identical to the passive participle, the
subject hij is present in the sentence, which means that its thematic role has not been
suppressed. Likewise, the object het manuscript has not moved to the subject position,
which indicates that it was assigned accusative case. In chapter 2 this irregularity was
ascribed to the presence of the auxiliary ‘have’, whose role is to reintroduce the external
theta role and assign structural case to the internal argument of the participle. On this
approach there is nothing exceptional about the behaviour of the past participle in (8c).
Just as the passive participle in (8b) it suppresses the agent and is not able to assign
accusative case to the object.
Consequently, I would like to pursue the idea that the past and the passive
participle are the same elements and should receive a uniform analysis. This proposal
receives additional support from diachronic development of the compound tense
formed with the auxiliary ‘have’ and the past participle. In many languages the tense
derives from a possessive construction, which combined the verb ‘have’ with a passive
participle. It is exemplified for Latin in (9), and from now onwards it will be referred to
as the ‘stative perfect’.
(9) a. Habeo scriptum librum

havePRES.1SG writtenACC bookACC
“I possess the book which is written” (Latin, Smith 1995: 271)
Properties of ‘have’-perfects in Slavic 129
The sentence in (9) does not render an eventive meaning, because Latin did not form a
perfect tense with the auxiliary ‘have’. However, the construction was reinterpreted as a
compound tense in the Romance languages, which are descendants of Latin.
(10) a. J’ai écrit le livre (French)

b. Ho scritto il libro (Italian)
c. He escrito el libro (Spanish, Smith 1995: 271)
“I have written the book”
The reinterpretation involved a number of grammatical changes, such as the loss of

agreement between the participle and the object (cf. section 3.6). A similar reanalysis is
taking place in the Slavic languages, but so far the process has been completed only in
Kashubian and Macedonian, which will be discussed in the subsequent section.
3.4 Properties of ‘have’-perfects in Slavic

All the South Slavic and most West Slavic languages use a compound tense which
consists of the auxiliary ‘be’ and the l-participle. This construction was extensively
investigated in chapter 2. In chapter 1, section 1.3.4.5.2 I pointed out that two Slavic
languages, Kashubian and Macedonian have an additional option of forming a
periphrastic tense with the auxiliary ‘have’. The main difference between ‘be’-perfects
and ‘have’-perfects is that the auxiliary ‘have’ is accompanied by an invariant form of
the passive participle, which always occurs in the same morphological form
irrespectively of the gender and number specification of the subject or the object. Thus,
even though the subject is masculine singular and the object is feminine singular in the
Macedonian example in (11), the participle završeno is neuter singular.
(11) Petar ja ima završeno taa rabota

PetarM itCL.ACC have1PL finishPTP.N thatF workF
“Petar has finished that work” (Mac)
Section 3.4.2.1 will show that ‘have’-perfects are possible with unergative and
unaccusative participles, that is the types of verbs that never undergo passivization. For
this reason, I will refer to the main verb in (11) as the past participle.
Constructions similar to the one in (11) are found in many other Slavic languages,
including Bulgarian, Czech, Polish, and Serbian. However, these languages have not
grammaticalized ‘have’-perfects yet. The differences between grammaticalized and non-
grammaticalized ‘have-perfects’ will be investigated in section 3.6.3.
3.4.1 Kashubian
Kashubian (or Cassubian) is a language spoken in Northern Poland near the city of
Gdańsk. It has approximately 150 thousand speakers (Stone 2002) and has some literary
tradition, but only 200-300 people are able to write it (www.naszekaszuby.pl). Little
attention has been paid to the syntax of compound tenses in this language so far, but
the issue certainly deserves more detailed research, because together with Macedonian
Kashubian is the only Slavic language that exhibits fully grammaticalized ‘have’-
perfects. This section overviews the construction and presents some new linguistic data
from this language.
Unless indicated otherwise, the data come from a survey I conducted among native
speakers of Kashubian on the website www.naszekaszuby.pl. The orthography of
Kashubian has not been completely standardized. The examples presented in this
section are taken from different sources but for consistency the spelling of some of
them has been modified in line with the convention adopted in Stone (2002).
The compound tenses in Kashubian show auxiliary alternation that depends on the
type of participle that the auxiliary appears with. This alternation is also common in
many Germanic and Romance languages. In these languages the auxiliary ‘be’ may only
select unaccusative past participles, while the auxiliary ‘have’ is accompanied by
transitive and unergative participles. In the languages that have morphological means of
showing agreement, the unaccusative past participles agree with the subject in gender
and number. This is the way it also happens in the Kashubian example in (12), where
the unaccusative past participle jidzenô is in the feminine singular form and occurs with
the auxiliary ‘be’.
(12) Ta białka je precz jidzenô

this womanF.SG beAUX.3.SG away goPTP.F.SG
“This woman has gone away” (Csb, Stone 2002: 777)
The auxiliary ‘be’ may also be used in pluperfect constructions. They are formed with
an l-participle of the verb ‘be’ as the auxiliary, which is followed by the main verb. The
main verb can be either an l-participle (cf. 13a) or a past/passive participle (cf. 13b).
Both of the participles agree with the subject.
(13) a. Jô jem béł pisôł

I bePRES.3SG bePART.M.SG writePART.M.SG
“I had written” (Csb, Stone 2002: 777)
b. Jô jem béł jachony
I bePRES.3SG bePART.M.SG goPASS.M.SG
“I have left” (Csb, www.naszekaszuby.pl)
The auxiliary ‘have’ selects transitive (cf. 14) and unergative (cf. 15) past participles.
They occur in the singular neuter form and never agree with the subject or the object in
φ-features.
(14) a. Të măš to wszétko zrob’iõné/zrob’iõny 68

you havePRES.2SG this all doPTP.N.SG/doPTP.M.SG
“You have made all of this”
b. Jô móm tą białkã bité
I havePRES.1SG thisF womanACC.F.SG beatenPTP.SG.N
“I have beaten this woman” (Csb, www.naszekaszuby.pl)
68 According to some sources (Breza & Treder 1981: 133; Gogolewski 1963; Lorentz 1919: 45;
74), the past participle may also appear in the masculine variant, irrespectively of the subject or
object feature content (cf. 14a). However, the native speakers I have consulted claim that only the
neuter form is possible, and that the confusion may have resulted from divergent orthographic
conventions.
c. Măš të vizõné?
havePRES.2SG you seePTP.N.SG
“Have you seen [it]?” (Csb, Gogolewski 1963: 71)
d. Mój woejc mô ten čôłn zbudovõné
my father havePRES.3SG this shipACC.M.SG buildPTP.N.SG
“My father has built that ship” (Csb, Elliott 2001: 137)
(15) a. On mô sponé
he havePRES.3SG sleepPTP.N.SG
“He has slept”
b. Ona mô leżoné
she havePRES.3SG liePTP.N.SG
“She has lain” (Csb, www.naszekaszuby.pl)
‘Have’-perfects in Kashubian are possible with non-human and inanimate subjects.

Section 3.6.3 will show that this semantic property is a strong indication that the
construction is fully grammaticalized.
(16) a. To aùto mô rozjachoné kùrã

thisN carN havePRES.3SG run-overPTP.N.SG henF.SG
“This car has run over the hen”
b. Nen pòjk mô wëpité mlékò
thisM catM havePRES.3SG drinkPTP.N.SG milkN
“This cat has drunk milk” (Csb, www.naszekaszuby.pl)
Morphologically, the past participles are the same as passive participles. This can be
observed in the form of the passive participle bité in (17), which is identical to the past
participle in (14b).
(17) To dziecko je bité

thisN childN be3SG beatPASS.N.SG
“This child is beaten” (Csb, Breza & Treder 1981: 134)
The past participle must always appear together with the auxiliary ‘have’. Omission of
the auxiliary results in ungrammaticality.
(18) a. Móm to wszétko zrob’iõné

havePRES.1SG it all donePTP.N.SG
“I have done all of this”
b. *Zrob’iõné to wszétko (Csb)
Bearing in mind that the past participle is morphologically the same as the passive
participle, which is assumed to be unable to assign structural case and project the
external theta role, the restriction on the auxiliary omission suggests that ‘have’
performs these two functions. Hence, the example in (18b) is ill-formed because the
object to wszétko lacks a case assigner. I will provide more support for this claim in
section 3.5 while discussing impersonal participles in Polish.
Just as the verb ‘be’, the auxiliary ‘have’ renders the pluperfect meaning when it is
used in the form of the l-participle.
(19) Jô miôł to wszëtko zrob’iõné69

you havePART.M.SG this all doPTP.N.SG
“You have made all of this” (Csb, www.naszekaszuby.pl)
The auxiliary ‘have’ also occurs with reflexive verbs. As expected, they appear in the
singular neuter form irrespectively of the φ-feature specification of the subject.
(20) a. On mô sã dowiedzóné
he have3SG REFL learnPTP.N
“He has found out”
b. Ona mô sã pitóné
she have3SG REFL askPTP.N
“She has asked” (Csb, www.naszekaszuby.pl)
The past participle in (21a) is in the perfective aspect. However, it may appear in the
imperfective aspect as well (cf. 21b and c), but then it expresses the meaning of the
“perfect of experience”.
(21) a. Jô móm zjadłé połnie

I have1SG eatPRF.PART.N.SG dinner
“I have (just) had dinner”
b. Jô móm jadłe pomuchla
I have1SG eatIMPF.PART.N.SG cod
“I have (already) eaten cod”
c. Mój woejc mô budovõné čôłna
My father have1SG buildIMPF.PTP.N.SG ships
“My father has built ships” (Csb, www.naszekaszuby.pl)
In (21a and b) the auxiliary ‘have’ is accompanied by the l-participle. This is because
Kashubian tends to conflate the l-participle with the past participle. Likewise, in
contrast to other Slavic languages, Kashubian may use the l-participle as a passive
participle. For instance, (22) exemplifies two sentences in the passive voice. The one in
(22a) is formed with the passive participle, but the one in (22b) contains the l-participle
(cf. also 13 above).
(22) a. To aùto je pòstawioné przed chëczą

this car be3SG placePASS.N.SG before house
“This car is parked in front of the house”
b. Mój czôłn je òsôdły na mielëznie
My ship be3SG come-downPART.M.SG on shallows
“My ship is on shallows” (Csb, www.naszekaszuby.pl)
This strategy is not available in any other Slavic language. According to Piotrowski
(1981: 13), this shows that Kashubian has lost a categorial distinction between l-
participles and passive participles. Possibly, this has happened under the influence of
German, which has the same type of participle in passive and compound tense
constructions. As expected, when the l-participle is used in ‘have’-perfects, it appears in
its invariant neuter form.
69 The finite auxiliary is missing in this example, because in contemporary Kashubian the l-
participle may occur without the present tense auxiliary ‘to be’ (cf. Breza & Treder 1981: 133).
(23) a. Jô môm to widziałé

I have1SG it seePART.N.SG
“I have seen it” (Csb, Stone 2002: 777)
It has been demonstrated that most verbs, transitive and unergative alike, are
compatible with the auxiliary ‘have’. However, modal verbs are excluded in this
construction (cf. 24). The l-participle is used to render modal meaning in the past
instead (cf. 25).70
(24) a. *Jô miôł muszoné

I havePART.M.SG mustPTP.N.SG
b. *Jô miôł rozmioné
I havePART.M.SG canPTP.N.SG (Csb, www.naszekaszuby.pl)
(25) a. Muszelé to zrobiec, bò przébliżiwôł sã termin

mustPART.N.SG it doINF because approachPART.M.SG REFL deadlineM.SG
“They had to do this, because the deadline was approaching”
b. Nie rozmielé wétłomaczéc sã z nygò
NEG canPART.N.SG explainINF REFL from this
“They couldn’t explain themselves” (Csb, www.naszekaszuby.pl)
The restriction on the usage of past participles as modal verbs will be explained in
section 3.7, where I will argue that it is related to the degree of verbalization of passive
participles.
Summarizing, the section has outlined properties of ‘have’-perfects in Kashubian.
It has been shown that the past participle selected by the auxiliary ‘have’ is
morphologically the same as the passive participle. However, the construction permits
unergative verbs such as ‘sleep’, which are never passivized. This suggests that the past
participle is more verbal than the passive participle. Nevertheless, in contrast to finite
verbs, the past participle is unable to project an external theta role or assign structural
case, and it needs to be accompanied by the auxiliary ‘have’, which performs these
functions.
3.4.2 Macedonian
‘Have’-perfects in Macedonian have been described in the literature more extensively
than the related construction in Kashubian, which permits a more detailed analysis.
Section 3.4.2.1 will describe properties of this construction. Section 3.4.2.2 will compare
past participle fronting across the auxiliary ‘have’ in Macedonian with l-participle
fronting across the auxiliary ‘be’ in Bulgarian and Serbo-Croatian, which was
investigated in chapter 2.
70 This restriction may be unexpected, because modal verbs do not have a defective paradigm in
Slavic.
3.4.2.1 Properties of the construction

Unlike Kashubian, Macedonian does not make a distinction in the auxiliary selection
related to transitivity of a verb, so both unaccusative (cf. 26a), unergative (cf. 26b) and
transitive participles (cf. 26c) take the auxiliary ‘have’.
(26) a. Gostite imaat dojdeno

guests-the have3PL arrivePTP.N
“The guests have arrived” (Mac, Elliott 2001: 39)
b. Goce Delčev ima spieno tuka
Goce Delčev have3SG sleepPTP.N here
“Goce Delčev has slept here” (Mac, Friedman 1977: 91)
c. Imam dobieno edno pismo do sega od Violeta
have1SG receivePTP.N one letter up till now from Violeta
“Up till now I have received one letter from Violeta”
Likewise, reflexive verbs may be selected by the auxiliary ‘have’, too.
(27) a. Veќe se imam izmieno

just REFL have1SG washPTP.N
“I have just washed (myself)”
b. Veќe se imam izbriceno
just REFL have1SG shavePTP.N
“I have just shaved (myself)” (Mac, L. Grujoska, p.c.)
Just as in Kashubian, human, non-human, and inanimate subjects are permitted in this
construction.
(28) a. Mačkata go ima ispieno mlekoto

cat itCL.ACC have3SG drinkPTP.N milk-the
“The cat has drunk milk”
b. Brodot se ima udreno vo karpite
ship-the REFL have3SG hitPTP.N in rocks
“The ship hit rocks” (Mac, L. Grujoska, p.c.)
However, not all verbs can be used as participles in ‘have’-perfects. For instance, on a
par with Kashubian, modal verbs such as ‘must’ (cf. 29a) and ‘can’ (cf. 30a) are
excluded. They may only appear in a simple past tense form (cf. 29b and 30b).
(29) a. *Imam morano/trebano da gi napravam ovie raboti

have1SG mustPTP.N/mustPTP.N da themCL.ACC doSUBJ.1SG these workPL
b. Morav da ja napravam ovaa rabota
mustPAST.1SG that itCL.F doSUBJ.1SG thisF work
“I have had to do this work” (Mac, L. Grujoska, p.c.)
(30) a. *Nemam moženo da se objasnam

NEG+have1SG canPTP.N that REFL explainSUBJ.1SG
b. Ne možev da se objasnam
NEG canPAST.1SG that REFL explainSUBJ.1SG
“I couldn't explain myself” (Mac, L. Grujoska, p.c.)
Correspondingly, the usage of the verbs ‘have’ and ‘be’ as past participles in ‘have’-
perfects is quite limited, as well. It is possible only in the Western dialects of the
language, that is in the area where the construction is the most widespread and where it
was grammaticalized the earliest (cf. chapter 1, section 1.3.4.5.2.1).
(31) a. Imam bideno tamu

havePRES.1SG bePTP.N there
“I have been there” (Radožda-Vevčani dialect, Hendriks 1976: 226)
b. Imam imano
havePRES.1SG havePTP.N
“I have had” (Ohrid dialect, Graves 2000: 489)
However, this option is considered ill-formed in the standard dialect, and a past tense
form or a compound tense with the l-participle must be used instead (cf. 32).71
(32) a. Toj beše vo Skopje

he bePAST.3SG in Skopje
“He was in Skopje (I vouch for it)”
b. Toj bil vo Skopje
he bePART.M.SG in Skopje
“Supposedly, he was in Skopje” (Mac, Friedman 2002: 272)
It has been shown that the past participle always occurs in the invariant form, which is
morphologically the same as the singular neuter passive participle. For instance, skinato
in (33) is a neuter variant, which can be both a passive participle and a past participle
that complements the auxiliary ‘have’. The passive and ‘have’-perfect constructions are
compared in (33) through (35). These examples demonstrate that even though the
passive form always agrees with the object, the past participle remains the same.
(33) a. Novoto palto mu e skinato

new-theN coatN himCL.DAT be3SG tearPASS.N
“His/her new coat is torn”
b. Go imam skinato moeto novo palto
himCL.ACC have1SG tearPTP.N my-the new coat
“I have torn my new coat” (Mac, L. Grujoska, p.c.)
71 The two sentences differ in evidentiality. As noted in chapter 1, section 1.3.4.5.1.1, the
constructions formed with the l-participle express the renarrated mood in Macedonian.
(34) a. Novata košula mu e skinata

newF shirtF himCL.DAT be3SG tearPASS.F
“His/her new shirt is torn”
b. Ja imam skinato mojata nova košula
herCL.ACC have1SG tearPTP.N my-the new shirtF
“I have torn my new shirt” (Mac, L. Grujoska, p.c.)
(35) a. Noviot kaput mu e skinat

new-theM jacketM himCL.DAT be3SG tearPASS.M
“His/her new jacket is torn”
b. Go imam skinato mojot nov kaput
himCL.ACC have1SG tearPTP.N my-the new jacketM
“I have torn my new jacket” (Mac, L. Grujoska, p.c.)
Just as in Kashubian, the construction is ungrammatical when the auxiliary ‘have’ is

absent (cf. 36). Following the hypothesis concerning auxiliary selection assumed in
chapter 2, this suggests that the past participle is not a case assigner by itself, and that
the auxiliary ‘have’ is needed to assign case to the object and project an external theta
role.
(36) Jas gi *(imam) kupeno knigite

I themCL.ACC have1SG buyPTP.N.SG books-the
“I have bought the books” (Mac)
Summarizing, it has been shown that ‘have’-perfects are formed in a similar way in
Kashubian and Macedonian. The main difference between the two languages concerns
the requirement of the selection of the auxiliary ‘be’ with unaccusative participles,
which is absent in Macedonian. Furthermore, it seems that the participle selected by the
auxiliary ‘have’ has been verbalized to a greater degree in Macedonian than in
Kashubian. This is evidenced by the fact that the verbs ‘have’ and ‘be’ are possible as
past participles in some dialects of Macedonian.
3.4.2.2 Participle fronting across the auxiliary ‘have’

This section analyzes fronting of the past participle across the auxiliary ‘have’ in
Macedonian. The movement is contrasted with l-participle movement in Bulgarian and
Serbo-Croatian, which was discussed in chapter 2.72
The example in (37b) illustrates preposing of the past participle kupeno across the
auxiliary ‘have’. The operation always gives rise to a focused or topicalized reading of
the fronted constituent; therefore I will refer to it as ‘VP topicalization’.
(37) a Nie gi imame kupeno knigite

we themCL.ACC have1PL buyPTP.N books-the
“We have bought the books”
b. Kupeno knigite (nie) gi imame
buyPTP.N books-the we themCL.ACC have1PL
“Buy the books, we did!” (Mac, S. Spasovska, p.c.; cf. Tomić 1996a)
72 Recall from chapter 2 that l-participle movement is impossible in Macedonian. This restriction
will be explained in chapter 4, section 4.4.2.4.2.2.

It is also possible to front the past participle when the auxiliary ‘have’ is negated, as in
(38).
(38) Kupeno knigite (nie) gi nemame

buyPTP.N books-the we themCL.ACC NEG+have1PL
“Buy the books, we didn’t!” (Mac, L. Grujoska, p.c.)
VP topicalization differs from l-participle fronting across the auxiliary ‘be’ in several
ways. First of all, the past participle does not agree with the subject in φ-features. In
(37), the subject pronoun nie is marked for plural, whereas the past participle kupeno is
singular neuter. In chapter 2 I claimed that the agreement on the l-participle bears
evidence for its movement to Spec, TP, where the φ-features of T are checked. The lack
of agreement on the past participle indicates that the movement to Spec,TP does not
take place.
The past participle is a potential candidate for raising to Spec, TP, so it is necessary
to explain why the movement is excluded. The answer is relatively simple on the
assumptions that l-participle movement across the auxiliary ‘be’ is an instance of
locative inversion and that the auxiliary ‘have’ is an external theta-role assigner. As is
well-known, locative inversion is incompatible with verbs assigning external theta-roles
(cf. 39). It is possible only with the verb ‘be’ and a small selection of other unaccusative
verbs (cf. 40).
(39) a. *Into the room rolled John the ball

b. *Down the street walked the old nanny her dog (Rochemont 1978)
(40) a. Crashing through the woods came a wild boar

b. On that table was put a valuable book
(cf. Bresnan 1994: 77-79; Coopmans 1989)
In chapter 2 I argued that the verb ‘have’ is a transitive auxiliary, which assigns
structural case and projects an external argument. In this scenario, the past participle
may not participate in locative inversion, because it raises across the transitive verb
‘have’.
Another difference between l-participle fronting and VP-topicalization concerns
the landing site of the moved element. The example in (37b) shows that the past
participle raises higher in the structure than the l-participle, because it may cross the
subject nie. As the translation indicates, the movement results in a focused
interpretation of the preposed element. I take this to mean that the operation is
triggered by a Focus feature. The participle heads PartP and is complemented by the
object knigite. The whole PartP lands in Spec, FocP above Spec, TP.
(41) [FocP kupeno knigitei [TP nie [vp [v gi+imame [PartP ti ]]]]]
This means that in contrast to l-participle fronting, which takes place in order to check
the φ-features of T and thus involves A-movement, the preposing of the past participle
exemplifies A’-movement.
VP-topicalization also differs from l-participle movement with respect to the size
of the constituent that undergoes movement. As demonstrated in (37b), the past
participle may pied-pipe the object when it raises to the clause initial position. However,
it may also leave the object stranded behind, as in (42a), or it may move together with
manner adverbs such as brzo in (42b).
(42) a. Kupeno gi imame knigite

buyPTP.N.SG themCL.ACC have1PL books-the
“We did buy the books!” (Mac, S. Spasovska, p.c.)
b. Brzo pročitano gi imame knigite
quickly readPTP.N.SG themCL.ACC have1PL books-the
“We have read the books really quickly” (Mac, L. Grujoska, p.c.)
By contrast, recall from chapter 2 that the l-participle may raise across the auxiliary ‘be’
only entirely on its own. Pied-piping of an object (cf. 43a) or an adverb (cf. 43b) results
in ungrammaticality.
(43) a. Pročel e knigata

readPART.M.SG beAUX.3SG book-the
“He has read the book”
b. *Pročel knigata e
c. *Bŭrzo pročel e knigata
quickly readPART.M.SG beAUX.3SG book-the (Bg)
In chapter 2 I suggested that the restriction is due to the way the compound tense
formed with the auxiliary ‘to be’ and the l-participle is structured (cf. 44). Since ‘to be’ is
an unaccusative verb, it is not able to assign accusative case to the object and project an
external argument, so the l-participle must perform these functions. This implies that
the subject must be the external argument of the l-participle, rather than of the auxiliary
‘be’. Therefore, in the template in (44) the subject is located in Spec, vP, while the
auxiliary ‘be’ heads an independent projection AuxP above the subject.
(44) [TP ... T[+φ] ... [Aux BE [vP subject[+φ] v [PartP Part[+φ] object]]]]
The ban on object pied-piping by the l-participle (cf. 43b) was attributed to the
assumption that the movement of the whole PartP raises the object across its case
checking position. Therefore, it was argued that the object must first raise to Spec,
AgrOP, and only then may PartP raise to Spec, TP via remnant movement.
The restriction on fronting of adjuncts or the subject together with the l-participle
was claimed to be due to the size of the constituent that is preposed. It is impossible to
move vP together with these elements, because this would leave the l-participle too
deeply embedded to check the φ-features of T. Therefore, the largest constituent that
can be fronted is PartP. This precludes raising of adjuncts or the subject, which are
generated above PartP.
Correspondingly, the following structure was proposed in chapter 2, section 2.3.1,
for the compound tense formed with the auxiliary ‘have’.
(45) [TP ... T[+φ] ... [vp subject[+φ] v [VP have [PartP Part [-φ] object]]]]
The past participle selected by the verb ‘have’ in Kashubian and Macedonian is
morphologically the same as the singular neuter form of the passive participle. This
suggests that it is analyzed by speakers as a verb which is unable to project an external
theta role or to assign accusative case to its complement. In line with the argumentation
developed in chapter 2, these functions must be performed by the auxiliary ‘have’ The
assumption is confirmed by the data in (18) and (36), repeated below as (46) and (47),
respectively, which indicate that ‘have’-perfects in Kashubian and Macedonian are
ungrammatical if the auxiliary is dropped.
(46) a. Móm to wszétko zrob’iõné

havePRES.1SG it all donePTP.N.SG
“I have done all of this”
b. *Zrob’iõné to wszétko (Csb)
(47) Jas gi *(imam) kupeno knigite

I themCL.ACC have1SG buyPTP.N.SG books-the
“I have bought the books” (Mac)
Furthermore, the past participle does not share φ-features with T0. In both Kashubian
and Macedonian it always occurs in the invariant form, regardless of the φ-features
carried by the subject. Moreover, since it appears with the transitive auxiliary ‘have’, it is
not eligible for movement via locative inversion. These two properties preclude raising
of the past participle to Spec, TP and indicate that the φ-features of T may be checked
only by movement of the subject. As a result, unlike the l-participle in Bulgarian and
Serbo-Croatian, the past participle in Macedonian does not compete with the subject
for the same position. Therefore, the fronted past participle kupeno may co-occur with
the subject nie (cf. 37b; repeated below as 48).
(48) Kupeno knigite nie gi imame

buyPTP.N books-the we them havePRES.1PL
“Buy the books, we did!” (Mac)
It has been established that preposing of the past participle in Macedonian (cf. 37 and
42) is triggered by a Focus feature that resides in the Focus head located above TP.
Obviously, the size of the element that needs to be focused in a clause depends on
discourse structure requirements, and as the data show, it may comprise a bare past
participle (cf. 42a), a past participle accompanied by an object (cf. 37b) or by an adverb
(cf. 42b). This suggests that the constituent that undergoes VP-topicalization in
Macedonian is much larger than the PartP that is moved to Spec,TP via locative
inversion in Bulgarian and Serbo-Croatian.
The fact that the topicalized PartP may be accompanied by some other elements
gives some insights about the structure of the VP that contains ‘have’-perfects (cf. 45).
For instance, it implies that PartP is not immediately dominated by the VP headed by
the auxiliary ‘have’. There might be a number of functional elements, marked as XP and
ZP in (49), which intervene between the two projections.
(49) [TP ... T[+φ] ... [vp subject[+φ] v [VP have [XP … [ZP [PartP Part [-φ] object]]]
Presumably, these projections host different types of adverbs that can be fronted
together with PartP. However, they may also contain an Agreement projection in which
the object, such as knigite in (48), can check its case. Hence there is no need for
evacuation of an object out of the fronted phrases prior to the movement of PartP to
Spec, FocP, and this is why it is possible for the past participle to carry the object on
the way to Spec, FocP.73
The assumption that there is a range of extended projections of the past participle
below the auxiliary ‘have’ is independently supported by the base positions of adverbs
in ‘have’-perfects. As indicated in (50), in Macedonian the auxiliary ‘have’ may be
separated from the past participle by a number of different adverbs and other
constituents.
(50) a. (Sama) gi imam (sama) napraveno ovie raboty

oneself itCL.ACC have1SG oneselfF doPTP.N.SG these workPL
“I have done this work by myself”
b. Gi imam denes veќe napraveno ovie raboty
itCL.ACC have1SG today already doPTP.N.SG these workPL
“I have already done this work today”
c. (Nie) go imame potpolno/kompletno isprazneto frižiderot
we itCL.ACC have1PL completely emptyPTP.N fridge-the
“We have completely emptied the fridge”
d. Petar (brzo) gi ima (brzo) spakuvano knigite (brzo)
Petar quickly themCL.ACC have3SG quickly packPTP.N books quickly
“Petar has packed the books quickly” (Mac, L. Grujoska, p.c.)
By comparison, consider the minimal pairs for (50) in (51), which consist of ‘be-
perfects’. Here the auxiliary clitic ‘be’, which clusters together with pronominal clitics
such as gi or go, must immediately dominate the l-participle.
(51) a. (Jas) sama sam gi (*sama)

I oneselfF beAUX.1SG themCL.ACC.PL oneselfF
napravila ovie raboti
doPART.F.SG these workPL
“I have done this work by myself, they say”
b. (Nie) potpolno/kompletno sme go
we completely be1PL itCL.ACC
(*potpolno/kompletno) ispraznile frižiderot
completely emptyPART.PL fridge
“Supposedly, we have emptied the fridge completely”
c. Petar (brzo) gi (*brzo) spakuval knigite (brzo)
Petar quickly themCL.ACC quickly packPART.M.SG books-the quickly
“Petar has packed his books quickly, they say” (Mac, L. Grujoska, p.c.)
Fronting of the l-participle across the auxiliary ‘be’ is excluded in Macedonian (see
chapter 4, section 4.4.2.4.2.2 for an explanation), so it is impossible to juxtapose the
two types of movement operations within one language. However, the contrast
between the data in (50) and (51) suggests that unlike the past participle, the l-participle
must raise to a position immediately below the auxiliary ‘be’, as it may not be separated
from the auxiliary with any non-clitic material. Since the past participle does not need to
undergo this movement and be right-adjacent to the auxiliary ‘have’, it may be fronted
together a number of functional projections that dominate it.
73 An alternative might be to assume that case on the object is checked by reconstruction, which
is possible with A’-movement and thus also with VP topicalization, but is not available with l-
participle fronting, which exemplifies A-movement.
Past participle fronting in Macedonian exhibits the same properties as raising of

the past participle across the auxiliary ‘have’ in some Germanic languages, such as
Dutch and German. The construction has received a lot of attention in the literature
(cf. Thiersch 1985, Den Besten and Webelhuth 1987, Koster 1987, Den Besten and
Webelhuth 1990, and Müller 1998), and is exemplified in (52) for Dutch.
(52) a. Jan heeft het boeki niet [ VP ti gelezen ]

Jan has the book not read
b. [VP gelezen ] heeft Jan het boek niet tVP
c. [VP het boek gelezen ] heeft Jan niet tVP
d. [dat boek te snel gelezen]i [C′ heeft hij niet ti]]
that book too quickly read has he not
(Dutch, H. Broekhuis, p.c.)
Just as in Macedonian the past participle in Dutch may move entirely by itself (cf. 52b),
it may raise together with an object (cf. 52c) and it may even pied-pipe a VP-external
element, such as a VP-adverb te snel ‘too quickly’ in (52d). The fact that (52d) is possible
indicates that virtually an infinitely large phrase may be preposed.
The landing site of the fronted element in (52) is usually assumed to be Spec, CP
or Spec, FocP. Given that finite verbs are standardly claimed to land in C in V2
contexts in Dutch and German, the auxiliary heeft in (52) must be in C. Moreover, the
fronted phrase dat boek te snel gelezen found to the left of the auxiliary has crossed the
subject pronoun hij. This indicates that exactly as in the case of VP topicalization in
Macedonian, the fronted constituent in Dutch raises higher than the l-participle in
Bulgarian and Serbo-Croatian.
Summarizing, this section has shown that fronting of the l-participle in Bulgarian
and Serbo-Croatian and fronting of the past participle in Macedonian instantiate two
different types of raising. The former is A-movement, the latter is A’-movement. Both
of them involve XP-displacement, but the former is triggered by the φ-features of T,
whereas the latter is driven by a Focus feature in Foc. They also vary in the size of the
preposed constituent.
Both operations are found outside Slavic. The l-participle fronting exemplifies
locative inversion, while the past participle fronting in Macedonian is a Slavic
counterpart of the Germanic VP topicalization.
L-participle fronting in Slavic has been frequently contrasted with VP
topicalization in Germanic (cf. especially Borsley, Rivero & Stephens 1996) in order to
show that it does not involve XP-movement. This section has demonstrated that
indeed l-participle fronting is different from VP-topicalization, but this does not
preclude an XP-movement account. These movements instantiate two different types
of syntactic operations, which differ in the size of the moved element, as well as the
A/A’ properties, therefore it is misguided to postulate any similarities between them.
This concludes the analysis of grammaticalized ‘have’-perfects in Kashubian and
Macedonian. The next section will discuss impersonal participles in Polish. It will be
shown that they exhibit the highest degree of verbalization of passive participles in
Slavic.
3.5 Impersonal participles in Polish

The preceding sections investigated past participles that are used in compound tense
constructions formed with the auxiliary ‘have’. The past participles historically originate
from passive participles. Even though they may be derived from verbs that normally
reject passivization, such as unergatives or unaccusatives, they are morphologically the
same as their passive variants. Moreover, on a par with passive participles they are
unable to project an external theta role or assign accusative case, therefore they must
always appear with the auxiliary ‘have’. In terms of the distinction between ‘adjectival’
and ‘verbal’ types of passive participles made in section 3.2, they are still largely on the
adjectival side.
The present section will analyze impersonal participles in Polish. They never occur
with the verb ‘have’ or any other overt auxiliary. Morphologically they are very similar
to the past participles in Macedonian, because they also end in the invariant -no/-to
morpheme. Moreover, just as the past participles they are the product of a diachronic
reanalysis of passive participles. However, in contrast to the past participles, they
demonstrate a greater degree of verbiness. It will be shown that they are able to assign
structural case and project a (covert) external theta role.
Even though impersonal participles have been subject to numerous linguistic
analyses (cf. the references mentioned in the following section, as well as Baker 1988,
Baker, Johnson & Roberts 1989, Goodall 1993, and Boeckx 1998), they have never
been contrasted with past participles in Kashubian and Macedonian, in spite of the
obvious morphological and structural similarities. The subsequent section is an attempt
in this direction.
3.5.1 Properties of the impersonal participles

Polish, Ukrainian, Byelorussian, and some dialects of Northern Russian have developed
a construction that is often referred to as the ‘impersonal passive’74 or a ‘no/to
construction’. For the reasons that will become immediately clear, I will avoid the
notion ‘passive’, and following Siewierska (1988), I will adopts the term ‘impersonal
participle’. The investigation in this section will focus on Polish, but see Lavine (2000)
for a comparison of Polish with Ukrainian, Byelorussian and Northern Russian.
The construction is exemplified in (53). It consists of a participle followed by an
object marked for accusative case.75 The participle does not show agreement with the
object, and it always remains in the same morphological form. Its morphological
structure is similar to the passive participle, because it ends in the -n or -t suffix (cf.
chapter 1, section 1.3.3.5). Semantically, it always describes a past event, which is
understood as finished if the verb appears in the perfective variant (cf. 53a), or as
unfinished if the verb is marked for imperfective aspect (cf. 53b).
74The term was first used by Comrie (1977).

75 Obviously, the case of the object is not always accusative. Some verbs, such as kierować ‘to
drive’ assign inherent case, which is also preserved in impersonal participle constructions.
(i) Kierowano samochodem

driveIPT carINSTR
“The car was driven”
Impersonal participles in Polish 143
(53) a. Napisano listy

writeIPT.PRF lettersACC
“The letters were written down”
b. Pisano listy
writeIPT.IMPF lettersACC
“The letters were being written” (Pl)
In spite of the similarity, the impersonal participle is not part of the inflectional
paradigm of passive participles in modern Polish.76 Thus, the examples in (54) indicate
that the 3rd person neuter morphology used in adjectival passives is -ne or -te, whereas in
impersonal participles it is always -no or -to.
(54) a. Kochano dziecko

loveIPT childACC.N
“The child was loved”
a’. Dziecko było kochane przez matkę
childNOM.N bePART.N.SG lovePASS.N by mother
“The child was loved by the mother”
b. Bito dziecko
beatIPT childACC.N
“The child was beaten”
b’. Dziecko było bite przez matkę
childNOM.N bePART.N.SG beatPASS.N by mother
“The child was beaten by the mother” (Pl, cf. Dziwirek 1994: 185)
In this way impersonal participles stay in a stark contrast with past participles in
Kashubian and Macedonian, which are morphologically the same as passive participles
in the singular neuter form.
Impersonal participles are never used with any overt auxiliary, neither ‘be’ (cf. 55a)
nor ‘have’ (cf. 55b).
(55) a. Pisano (*są/*były/*było) listy

writeIPT bePRES.3PL/PART.NV.PL/PART.N.SG letters
b. Pisano (*mają/*miałe/*miało) listy
writeIPT havePRES.3PL/PART.NV.PL/PART.N.SG letters (Pl)
The construction is dubbed “impersonal”, because it never allows the subject to be

expressed overtly, not even via a “by-phrase” (cf. 56a), the way it is possible in passive
structures (cf. 56b).
(56) a. Pisano listy (*przez dzieci)

writeIPT letters by children
“The letters were written”
b. Listy są pisane przez dzieci
letters bePRES.3PL writePASS.NV.PL by children
“The letters are written by children” (Pl)
76 However, in the older stages of Polish it had the same form as the neuter singular passive
participle (cf. section 3.5.2).

However, even though the subject may never be expressed, there is some evidence that
it is present covertly. For example, (57) illustrates that the construction permits subject-
oriented anaphors such as swój ‘one’s own’ and the reflexive się, which must be bound
by a subject.
(57) a Sprzedano swoje samochody

sellIPT one’s cars
“People sold their (own) cars” (Pl, Kupść & Marciniak 1997: 2003)
b. Oglądano się w lustrze
lookIPT REFL in mirror
“One was looking at oneself in the mirror”
(Pl, Bondaruk & Charzyńska-Wójcik 2002: 87)
Moreover, Bondaruk & Charzyńska-Wójcik (2002) point out that the presence of a
covert subject in sentences with impersonal participles (cf. 58a) is confirmed by control
patterns, which are analogous to those found in the sentences with referential subjects
(cf. 58b).
(58) a. Ewa chce wyjechać za granicę

Ewa wantPRES.3SG goINF abroad
“Ewa wants to go abroad”
b. Chciano wyjechać za granicę
wantIPT goINF abroad
“One wanted to go abroad” (Pl, Bondaruk & Charzyńska-Wójcik 2002: 88)
The sentences in (58) involve the same predicate, but the one in (58b) does not contain
an overt subject. However, PRO in (58b) is interpreted as identical with the implied
subject of the main impersonal clause, on a par with the example in (58a), in which the
subject Ewa controls PRO.
Furthermore, Bondaruk & Charzyńska-Wójcik (2002) observe that the covert
subject must bear a θ-role, because impersonal constructions are incompatible with
‘weather’ verbs, which require an expletive subject (cf. 59a). It is necessary to use the l-
participle in the singular neuter form instead in order to render the intended meaning
(cf. 59b; see chapter 5 for a discussion of the syntax of the l-participle in Polish).
(59) a. *Wiano śniegiem

blowIPT snowINSTR
b. Wiało śniegiem
blowPART.N.SG snowINSTR
“It was blowing with snow”
(Pl, cf. Bondaruk & Charzyńska-Wójcik 2002: 88)
Rozwadowska (1992) and Kibort (2004: 257ff) show that impersonal participles can be
formed from unaccusative verbs. As indicated in the primed examples in (60),
unaccusative verbs never passivize.
Impersonal participles in Polish 145
(60) a. Umierano z wycięczenia

dieIPT from exhaustion
“(They) died/used to die of exhaustion”
a’. *Było umierane
bePART.N.SG diePASS.NV.PL
b. Często bywano w Warszawie
often beIPT.IMPF in WarsawLOC
“(They) used to be often in Warsaw”
b’. *Było bywane
bePART.N.SG bePASS.IMPF.NV.PL (Pl, Kibort 2004: 258)
c. Tonięto w morzu, a nie w wannie
drownIPT in sea and NEG in bathtub
“People drowned in the sea, not in a bathtub” (Pl, Rozwadowska 1992: 62)
Finally, Jabłońska (2006) observes that modal verbs are also eligible candidates for
impersonal participles. This contrasts them with passive participles (cf. 61a’ and b’),77 as
well as the past participles in Kashubian and Macedonian.
(61) a. Musiano to wykonać, bo zbliżał się termin

mustIPT this doINF because approachPART.M.SG REFL deadline
“(They) had to do this, because the deadline was approaching”
a’. *Ktoś był musiany to zrobić
someone bePART.M.SG mustPASS.M.SG this done
b. Nie umiano się z tego wytłumaczyć
NEG canIPT REFL from this explainINF
“(They) didn’t know how to explain themselves”
b’. *Ktoś był nie umiany
someone bePART.M.SG NEG canPASS.M.SG (Pl, cf. Jabłońska 2006)
To sum up, it has been shown that constructions with the impersonal participle contain
a covert subject, which bears the external θ-role, and an object, which carries accusative
or inherent case. Since there is no other verb available in the structure, the external θ-
role and accusative case are evidently assigned by the impersonal participle. This
suggests that it cannot be classified as passive. In this way the impersonal participle is
different from the past participle in Kashubian and Macedonian. Moreover, in contrast
to the past participles, the impersonal participles can be formed from modal verbs.78
Thus, they are less restrictive in their lexical selection. These two facts indicate that
impersonal participles are more verbal than past participles.
In spite of the synchronic differences between past participles and impersonal
participles in Slavic, both forms derive from passive participles. The diachrony of
impersonal passives will be presented in the next section.
77 The passive forms of ‘must’ and ‘can’ are hypothesized morphological variants, because they
do not exist in the language.

78 In syntactic terms this may mean that the impersonal participle raises higher in the structure
than the past participle. For example, it may target a Mood projection, which is inaccessible for
the past participle. See section 3.7 as well as Jabłońska (2006), who claims that the degree of
verbalization of a category corresponds to the number of functional heads spelt out by the
participial morpheme: the higher head it lexicalizes, the more verbal it is.
3.5.2 Historical development of the impersonal

participles
Impersonal participles in Polish originate from passive participles. Recall from chapter
1 (section 1.3.3.5) that passive participles in Old Church Slavonic were frequently used
as adjectives, and that adjectives and passive participles occurred in two declensions: a
nominal declension and a pronominal declension (cf. Klemensiewicz et al 1964: 323-
326). The passive participles from the nominal declension paradigm ended in -no/to in
their neuter forms. Gradually, the nominal paradigm was eliminated in Polish, with the
exception of the neuter -no/to variants, which were the only ones that remained. Since
no/to participles were passive, they were incompatible with nominative subjects, which
appeared in instrumental case or were introduced by the preposition od, such as od
nyeprzyaczol and od poganov in (62). The no/to participle were accompanied by the copula
‘be’ in the present tense or in the form of the l-participle, such as było in (62), to render
the past tense (cf. Oesterreicher 1926: 57; Siewierska 1988: 270).
(62) By thesz krolesthwo tho nakonyecz abo skazano

so also kingdom this end or sentencePASS.N
od nyeprzyaczol abo posyandzyono od poganov nye było
by enemies or possessPASS.N by pagans not bePART.N.SG
“And also so that the kingdom will never be overrun by (its) enemies or
possessed by pagans” (Old Polish, Posłanie króla Aleksandra; Doros 1975: 92)
The no/to passive participle was marked for the singular neuter agreement, so it
appeared only with singular neuter nouns. However, Lavine (2000) observes that some
of these nouns, such as rucho in (63) were ambiguous between nominative and
accusative case forms. They could be interpreted as nominative subjects, as in “the
garment is woven” or accusative objects, as in “someone has woven the garment”.
(63) Nye obleczesz syø w rucho, jesto z velni

NEG dress2SG REFL in garmentN.SG whichNOM/ACC.N.SG from wool
a z lnu tkano jest
and from linen weavePASS.N.SG be3SG
“You shall not wear a garment that is woven of wool or of linen”
(Deuteronomy, 21,11; Biblia Królowej Zofii 15th c; Oesterreicher 1926: 55)
Lavine argues that the interpretation was disambiguated when the 3rd person auxiliary
was lost in Polish between the 15th and the 17th century (cf. chapter 1, section
1.3.4.2.2.1). Since the no/to forms were the only remaining representatives of passive
participles from the nominal paradigm, once the 3rd person auxiliary disappeared, it was
no longer possible to interpret them as passive participles or adjectives. Rather, they
were reanalyzed as a new type of active predicate that selects the object in accusative
case. As a result, once the new form emerges, the noun rucho in (63) is unambiguously
interpreted as an accusative object.
Lavine’s hypothesis is supported by diachronic considerations. Oesterreicher
(1926) notices that the development of impersonal participles in Polish coincides with
the decline of the 3rd person auxiliary. Even though neuter singular nouns, such as syerce
in (64b) remain morphologically ambiguous between nominative and accusative, the
missing auxiliary indicates that they are direct objects carrying accusative case.
Grammaticalization of ‘have’-perfects 147
(64) a. Y smyotano nan’ wyeliką gromadø kamyenya

and raisePTP on-him great bunchACC.F stoneGEN
“And they raised over him a great heap of stones”
(Old Polish, Jos, 7, 26; Biblia Królowej Zofii 15th c; Oesterreicher 1926: 55)
b. A popødzono syerce lyuczske ku dzalanyu
and drivePTP heartACC.N people’s towards work
“For the people’s heart were driven to work”
(Old Polish, Nehem 4,6; Biblia Królowej Zofii 15th c; Oesterreicher 1926: 55)
To conclude, even though the impersonal participles appear without the auxiliary
‘have’, they developed in the same way as past participles in Kashubian and
Macedonian, as they are also a result of a reanalysis of passive participles. However, the
impersonal participles became ‘verbalized’ to a greater degree, as they are able to assign
accusative case, project an external theta role, and thus, they pattern like finite verbs.
Moreover, they cover a wider spectrum of verbal forms, because they can be also
constructed from modal verbs.
The chart in (65) summarizes properties of the past participles and the impersonal
participles.
(65)
past participles impersonal participles

invariant morphological form yes yes
morphologically the same as yes no
passive participles
the auxiliary ‘have’ present yes no
able to project an external no yes
argument and assign
structural case
can be formed from the no in Kashubian; yes in some yes
verbs ‘be’ and ‘have’ dialects of Macedonian
can be formed from modal no yes
verbs
can be constructed from both yes yes
perfective and imperfective
forms
Impersonal participles are not compound formations; so a syntactic analysis of their

structure is beyond the scope of this thesis. However, they have received a lot of
attention in the literature; see Blevins (2003), Bondaruk & Charzyńska-Wójcik (2002),
Jabłońska (2006), Kibort (2004), Lavine (2000), and Rozwadowska (1992) for an
extensive discussion.
3.6 Grammaticalization of ‘have’-perfects

Sections 3.4.1 and 3.4.2 described the ‘have’-perfect in Kashubian and Macedonian.
The construction developed a full paradigm and accepts almost all types of verbs as
past participles. These facts indicate that it has been fully grammaticalized in these
languages.79
The present section will analyse constructions formed with the verb ‘have’ and a
passive participle in some other Slavic languages. They will be termed ‘stative-perfects’.
Semantically, these structures resemble the ‘have’-prefects, but they differ in the way
they are formed. For example, the passive participle always agrees with its complement
in φ-features, and may only appear in the perfective variant. It will be shown that they
represent an early stage of grammaticalization of the compound tense formed with the
auxiliary ‘have’.
In Germanic and Romance ‘have’-perfects’ developed in the same way as in Slavic.
They originated from a possessive construction formed with ‘have’ as the main verb,
which was followed by a DP complement and a passive participle. Given the similarity,
I will first outline grammaticalization of the structure in these languages.
3.6.1 Romance
As was noted in section 3.3, ‘have’ perfects did not exist in Latin. This language
expressed perfectivity synthetically, as indicated in (66).
(66) Obsedimus oppidum

besiegePF.1PL town
“We have besieged the town” (Latin, Hoekstra 1986: 98)
Nevertheless, there was a related construction, which I will dub ‘the stative perfect’. It
combined the verb habere ‘have’ with a passive participle, as in (67).
(67) Habemus oppidum obsessum

have1PL townACC.M.SG besiegedACC.M.SG
“We have a town besieged” (Latin, cf. Hoekstra 1986: 98)
The stative perfect paved the way to the analytic ‘have’-perfect in the contemporary
Romance languages (cf. 68 for Italian). At first blush, the main difference between the
compound tense formed with the auxiliary ‘have’ (cf. 68a) and the stative perfect
construction (cf. 68b) is the lack of agreement between the object and the participle.
Thus, the object and the participle are both feminine in the stative perfect in (68b),
whereas in the ‘have’-perfect in (68a) the participle appears in an invariant form.
(68) a. Abbiamo assediato una cittá

have1PL besieged aF townF.SG
“We have besieged the town”
b. Abbiamo una cittá assediata
have1PL aF townF.SG besiegedF.SG
“One of the towns we have is besieged” (Italian, N. Grillo p.c.)
Before I discuss grammaticalization of ‘have’-perfects, let me analyze the Latin sentence

in (67) in more detail. The verb habemus takes two complements: the direct object
79 See Oubouzar (1974) and Breitbarth (2005), who argue that a construction is completely
grammaticalized once it has developed a full paradigm.

oppidum and the object complement obsessum. Obsessum is not a modifier of the direct
object oppidum. They are two separate constituents, which according to Salvi (1987) is
evidenced by the fact that the direct object can be pronominalized independently of the
object complement (that is, the passive participle).
(69) (Equitatum) quem ex omni provincia… coactum habebat

cavalry whichACC from each province gatheredACC had3SG
“(The cavalry) that he had (having) gathered (it) from all provinces”
(Latin, Caesar, De bello gallico I.15.1; Salvi 1987: 226)
If the participle were a modifier of the object, it should not be possible to express it
overtly after pronominalization of the object, as in the contemporary Italian example in
(70).
(70) a. Leggo il libro svedese

read1SG the book Swedish
“I read the Swedish book”
b. Lo leggo (*svedese)
him read1SG Swedish
“I read it” (Italian, Salvi 1987: 326)
The passive participle agrees in φ-features and case with the object (cf. 67). Moreover, it
may appear in the comparative form. Since verbs do not show comparison inflection,
this suggests that the passive participle is adjectival in nature.
(71) Comitiorum dilationes occupatiorem me habebant

meetingsGEN delaysNOM occupyPASS.ACC.COM me had3PL
“The delays of the meetings kept me rather occupied”
(Latin, Celius in Cicero, Epistulae ad familiars VIII.4.3; Salvi 1987: 227)
Salvi proposes that the verb ‘have’ selects a Small Clause, which is headed by an
adjective. The direct object is located in Spec, AP. Thus, the sentence in (72a) is
assigned the syntactic representation in (72b).
(72) a. Habeo epistulam scriptam

have1SG letterACC writtenACC
“I have the result of the fact that a letter has been written”
“I have a letter that has been written”
b. [VP habeo [SC(AP) [DP epistulam] [A scriptam]]
(Latin, adapted from Salvi 1987: 228)
Moreover, according to Salvi the DP-subject of the Small Clause epistulam is

underlyingly the direct object of the (adjectival) passive participle scriptam. In this way
the construction is very similar to a passive sentence, which consists of the same
participle and the copula ‘be’ instead of the verb ‘have’ (cf. 73a). The object and the
passive participle form a Small Clause. The subject position of the sentence is empty,
but it can be filled by the direct object, which in Latin moves together with the passive
participle and becomes the subject of the clause, as in (73).
(73) a. Epistula scripta (est)

letterNOM writtenNOM is
“A letter has been written”
b. [TP [SC [DP epistula] [scriptaPASS]]i [T est [SC ti]]]80
(Latin, adapted from Salvi 1987: 231)
As is standardly assumed in the analyses of passive constructions, the movement occurs

for case checking. The direct object cannot receive accusative case in its base position,
because the passive participle is not a case assigner. It may not receive case from the
copula, either, because it is an unaccusative verb. Therefore, it must move to the
subject position, where it checks nominative case of T in Spec, TP. However, the direct
object does not need to move to Spec, TP in the presence of the verb ‘have’, which is
transitive and thus assigns accusative case to its object. Hence, the object follows ‘have’
in (72).
I argued in chapter 2 that since the passive participle is unable to assign a thematic
role to the subject, this role is performed in ‘have’-perfects by the auxiliary. A corollary
of this claim is the assumption that ‘have’ is always the same verbal category,
irrespectively of its function as an auxiliary or a lexical verb meaning ‘possess’ (see also
Kayne 1993). Salvi (1987: 232ff) argues that grammaticalization of the lexical verb
‘have’ into an auxiliary involves semantic bleaching, because ‘have’ does not express the
meaning of possession. In his view, this semantic process has structural consequences,
because ‘have’ is no longer capable of projecting an external theta role. However, it may
“help” the participle to do so.
Furthermore, Salvi claims that since the auxiliary is not able to assign accusative
case to the object by itself, the object must stay in situ. In effect, the participle-direct
object word order prevails (cf. 74a). This also results in the impossibility of raising the
direct object to the subject position of the Small Clause (cf. 72b). Since this position
cannot be filled any more, it loses its function and the Small Clause is eliminated.
Consequently, the compound tense formed with the auxiliary ‘have’ will have a
structure as in (74b).
(74) a. Piero ha scritto la lettera

Piero has written the letter
b. [TP Piero [T ha [VP [V scritto [DP la lettera]]]]]
(Italian, adapted from Salvi 1987: 232-233)
I agree with Salvi that grammaticalization of ‘have’-perfects is accompanied by the

elimination of the Small Clause. I also follow his assumption that the verb ‘have’ is
semantically bleached when it becomes an auxiliary. However, I do not subscribe to the
idea that the auxiliary loses its ability to assign an external theta role. It is not clear to
me how the auxiliary ‘have’ should be able to “help” the passive participle to do that,
either. Thus, I suggest that ‘have’ has the same properties related to case and theta role
assignment whether it is a lexical verb or an auxiliary. Likewise, following the ideas
developed in chapter 2, I will continue to assume that the passive participle is the same
80 The derivation is slightly simplified in order to render the Latin word order correctly. In the
modern Romance languages only the object moves to the subject position, and the passive
participle is left in situ. See Salvi (1987: 235 fn 9)
(i) [TP [DP Subjecti] [T BE [VP Participle [DP ti]]]

element, whether it occurs in a passive construction or as a past participle in a

compound tense formed with the auxiliary ‘have’. However, due to the process of
verbalization, the paradigm of the past participle is extended to all verbs, and it also
covers verbs which disallow passivization, such as the one-place predicate ‘laugh’.
(75) Ho riso
have1SG laughPTP
“I have laughed” (Italian)
This concludes the presentation of the grammaticalization of the ‘have’-perfect in

Romance. Before I turn to Slavic, I will discuss the way the process proceeded in the
Germanic languages in the next section.
3.6.2 Germanic
The compound tense formed with the auxiliary ‘have’ evolved in the Germanic
languages in largely the same way as in the Romance languages. Thus, it is argued in
Behaghel (1928) that the passive structure in (76a) was the source of the modern ‘have’-
perfect in German (cf. 76b).81 The passive participle agrees with the direct object in
(76a). However, once the verb ‘have’ is reinterpreted as an auxiliary, the agreement on
the participle is lost (cf. 76b).
(76) a. Ich habe ihn geschlagenen

I have1SG him beatenPASS.M.N
“I have him as a beaten one”
b. Ich habe ihn geschlagen
I have1SG him beatenPTP
“I have beaten him”
Turning to English, it is a matter of debate whether ‘have’-perfects reached

grammaticalization already in Old English, or only in Middle English (cf. Mitchell 1985:
292-295; Wischer 2004). At any rate, it is clear that in some Old English relics the
passive participle, such as getynedne in (77) agrees with the object it modifies. This tends
to be an exception, rather than a regular pattern, though.
(77) Oðrum þe hiora dæl getynedne hæbben…

others who their part enclosedPASS.ACC have
“Others who have their part enclosed” (Old English, Wischer 2004)
In (77) the participle precedes the verb ‘have’. However, Mitchell (1985: 287) points out
that the most common ordering was “have-participle-direct object”, as in Modern
English.
The verb ‘have’ did not have to carry the meaning of possession in Old English,
which suggests that it had undergone semantic bleaching. For instance, the subject of
the clause in (78) has drunk poison, so s/he cannot possess it any more.
81 I thank Anne Breitbarth for pointing out this example to me.

(78) Gif he ær hæfþ attor gedruncen...

if he before has poison drunk…
“If he has drunk poison before…” (Old English, Wischer 2004)
As far as the aspectual content of the participle is concerned, it has often been claimed
that in Old English ‘have’ may only combine with participles of telic verbs (cf.
Hoffmann 1934). For instance, the ‘have’-structure in (79a) marks completion of an
action in a series of events and characterizes the event as perfective. However, Wischer
(2004) provides a few examples with atelic participles, as in (79b).
(79) a. Hafa þe ær geworht clam of beor … & of …

haveSG youSG.DAT previously made paste of beer … & of …
“Previously, have yourself made a paste of beer … & of …”
b. Æfterðæmðe hie gesyngod habbað
after-that they sinned have
“…after they have sinned” (Old English, Wischer 2004)
Section 3.6.3 will show that the use of atelic/imperfective participles is one of the
criteria for grammaticalization of ‘have’-perfects.
Finally, I will conclude with some remarks on the grammaticalization of ‘have’-
perfects in Dutch. Following Kern (1912), Hoekstra (1984: 268) points out that older
stages of Dutch had adjectival participles, which could be formed from both
intransitive verbs and two-place predicates. They were combined with verbs worden
‘become’ and zijn ‘be’. Constructions with transitive verbs expressed the meaning of a
perfect or a pluperfect, whereas structures with two-place predicates were interpreted as
passive. These formations paved the way to the stative perfect, which was formed with
the verb ‘have’ and a passive participle that agreed with the direct object. Thus, its
structure was similar to the Latin construction in (67), repeated below as (80).
(80) Habemus oppidum obsessum

have1PL townACC.M.SG besiegedACC.M.SG
“We possess a town that is besieged” (Latin, cf. Hoekstra 1986: 98)
At first the new perfect was possible only with transitive verbs, as in the Modern Dutch
example in (81a), which contains the two-place predicate ‘read’. Subsequently, the
construction was extended to one-place predicates, such as lachen ‘laugh’ in (81b).
(81) a. dat Jan het boek gelezen heeft

that John the book read has
“…that John has read the book”
b. dat Jan gelachen heeft
that John laughed has
“…that John has laughed” (Dutch, Hoekstra 1984: 268)
Hoekstra (1984, 1986) argues that the reinterpretation of the ‘have’-perfect involved a
syntactic reanalysis. He claims that in the stative perfects (cf. 80) ‘have’ is the main verb,
which takes a Small Clause complement. However, a Small Clause may not consist
solely of a predicate. This is why the English examples in (82) are ungrammatical.
(82) a. *I want laughed

b. *I want off my ship
This suggests that the verb ‘have’ in (81b) cannot be complemented by a Small Clause
any more, because ‘laugh’ is a one-place predicate. In section 3.6.1 I have claimed that
the elimination of the Small Clause structure marks the emergence of the ‘have’-perfect.
In view of this, the acceptability of the structures such as (81b) serves as a reliable
criterion for the presence of the ‘have’-perfect in a language. The construction becomes
available only when the verb ‘have’ is not complemented by a Small Clause. This is a
condition for the verb ‘have’ to function as an auxiliary.
Summarizing, the preceding two sections analyzed the development of the stative
perfect into the ‘have’ perfect in Romance in Germanic. The chart in (83) compares
some of the properties of the two constructions.
(83)
the stative perfect the ‘have’ perfect

participle-direct object yes no
agreement (cf. 67, 68, 76)
the telic/perfective variant of yes no
the participle predominates
(cf. 79)
grading of the participle is yes no
possible (cf. 71)
the verb ‘have’ may carry the yes no
meaning of possession (cf.
78)
the participle may be a one- no yes
place predicate
In the next section I will investigate grammaticalization of ‘have’-perfects in the Slavic

languages. I will also provide a syntactic account of this process.
3.6.3 Slavic
As was noted in chapter 1, section 1.3.4.5.2 and in section 3.4 of the present chapter,
‘have’-perfects have been grammaticalized only in two Slavic languages, Kashubian and
Macedonian. It is difficult to trace the origin of the construction in Kashubian, because
the first description of the grammatical system of the language comes from 1879
(Ceynova 1879), and there are very few literary works available. However, the evolution
of ‘have’-perfects in Macedonian is quite well documented. The earliest example that is
reminiscent of the contemporary ‘have’ perfect was found in a manuscript from the
monastery of Krnino in 1706. The sentence contains a passive participle that agrees in
number and gender with the object clitic, so it represents the stative perfect.
(84) ...imamъ go aforesanъ…

havePRES.1SG himCL.ACC excommunicatePASS.M.SG
“I [will] have him excommunicated…” (18th c. Mac, cf. Koneski 1987)
In the contemporary version of the sentence the participle does not agree with the
object, but it occurs in the singular neuter form. Thus, the structure in (85) exemplifies
a grammaticalized ‘have’-perfect.
(85) Go imam aforesano

himCL.ACC havePRES.1SG excommunicatePTP.N
“I [will] have excommunicated him…” (Mac, Elliott 2001: 39)
In Macedonian the stative perfect has been completely replaced by the ‘have’-perfect.
However, it is still available in many other Slavic languages, as will be shown in section
3.6.3.1. Section 3.6.3.2 will contrast the ‘have’-perfect with the stative perfect and will
provide a syntactic account of its grammaticalization.
3.6.3.1 Properties of the stative perfect

A few other Slavic languages that have not grammaticalized the compound tenses
formed with the auxiliary ‘have’ display a related construction that resembles the stative
perfect construction attested in older variants of Germanic and Romance. The Slavic
languages exemplified here include Polish (cf. 86), Czech (cf. 87), Serbian (cf. 88), and
Bulgarian (cf. 89). As shown in (86) and (88), both affirmative and negated forms are
possible. The most noticeable difference between stative perfect structures and ‘have’-
perfects is agreement between the object and the participle, which obtains only in the
former type of constructions.
(86) a. Mam już zapięte pasy

have1SG already fastenPASS.ACC.NV.PL seatbeltsACC.NV.PL
“I have already fastened the seatbelts”
b. Nie mam jeszcze spakowanych książek
NEG have1SG still packPASS.GEN.NV.PL booksGEN.NV.PL
“I haven’t packed my books yet” (Pl)
(87) Mám úlohu napsanou

have1SG taskACC.F.SG writePASS.ACC.F.SG
“I have my task written” (Czech, Maslov 1988: 80)
(88) On nema položen nijedan ispit

he NEG+have1SG passPASS.M.SG NEG+single examM.SG
“He has not passed a single examination/He does not have a single exam
passed” (Serbian, Dimitrovski 1957: 246 quoted in Friedman 1976: 97)
(89) Toj ima dve nivi izoreni

he have1SG two fieldPL plowPASS.PL
“He has two fields ploughed/He has two ploughed fields/He has ploughed
two fields” (Bg, Dimitrovski 1957: 246 quoted in Friedman 1976: 97)
However, the two constructions differ in more respects, which will be illustrated by
contrasting stative perfect structures in Polish with ‘have’-perfects in Macedonian.
Moreover, since Polish also has impersonal participles, which are morphologically very
similar to ‘have’-perfects (cf. section 3.5), they will be exemplified as well.
I will begin the description of the stative perfect by establishing a syntactic relation
between the direct object and the passive participle. The two elements agree in φ-
features. However, this does not mean that the participle is an adjectival modifier of the
direct object. This is can be evidenced through pronominalization of the direct object,
which leaves the passive participle intact82.
(90) a. Mam już wszystkie ciasta upieczone

have1SG already all cakesNV.PL bakePASS.NV.PL
“I have already baked all the cakes”
b. Mam już je wszystkie upieczone
have1SG already themNV all bakePASS.NV.PL
“I have baked all of them already” (Pl)
Conversely, when a noun is (pre-)modified by an adjective, pronominalization affects

both the noun and the adjective.
(91) a. Mam wiele ciekawych książek

have1SG manyGEN.NV.PL interestingGEN.NV.PL booksGEN.NV.PL
“I have a lot of interesting books”
b. Mam ich wiele (*ciekawych)
have1SG themGEN.NV.PL manyGEN.NV.PL interestingGEN.NV.PL
“I have many of them” (Pl)
The contrast shows that the passive participle is not an adjectival modifier of the direct
object in (90). I will assume that the two constituents form a Small Clause.
The sentence in (92a) demonstrates that in the case of stative perfects, the agent of
the action described by the participle need not be the same as the subject of the entire
clause. In Macedonian the subject of the ‘have’-perfect clause must be the same as the
agent of the event characterized by the past participle (cf. 92b). In impersonal
participles the agent is always undetermined (cf. 92c).
(92) a. Mamy już zarezerwowane miejsca

have1.SG already bookPASS.ACC.NV.PL seatsACC.NV.PL
“We have already booked our seats”
“We have already had our seats booked” (Pl)
b. Gi imame veќe rezervirano sedistata
them have1PL already reservePRF.PTP.N.SG seats-the
“We have already booked our seats”
“*Someone has already booked the seats for us” (Mac, O. Tomić, p.c.)
c. Zarezerwowano już miejsca
bookIPT already seatsACC.NV.PL
“The seats have already been booked” (Pl)
The examples in (93b and c) illustrate that impersonal participles as well as compound
structures formed with the auxiliary ‘have’ permit both perfective and imperfective
variants of the main verb. By contrast, stative-perfects are possible only with perfective
forms (cf. 93a).
82 The same test was applied by Salvi (1987) in his analysis of the ‘have’-structures in Latin (cf.
example 69 above).
(93) a. Mam już przeczytane dwie książki

have1SG already readPRF.PASS.NV.ACC.PL two booksNV.ACC.PL
“We have already finished reading two books”
a’. *Mam już czytane dwie książki
have1SG already readIMPF.PASS.NV.ACC.PL two books (Pl)
b. Gi imame pročitano knigite
themCL.ACC have1PL readPRF.PTP.N.SG books-the
“We have (finished) read(ing) the books” (Mac, L. Grujoska, p.c.)
b’. Gi imame čitano knigite
themCL.ACC have1PL readIMPF.PTP.N.SG books-the
“We been reading the books” (Mac, L. Grujoska, p.c.)
c. Przeczytano dwie książki
readPRF.IPT two books
“Someone has finished reading two books” (Pl)
c’. Czytano dwie książki
readIMPF.IPT two books
“Someone has been/was reading two books” (Pl)
It was noted in section 3.6.1 that the participle-direct object order predominates in
English once the ‘have’-perfect is grammaticalized. Given the free word order in the
Slavic languages, the pattern is not so clear-cut in the stative-perfect constructions, but
structures with the object preceding the participle are usually emphatic. Likewise,
placement of the direct object in front of the past participle requires additional focus on
the object. In fact, Bubenik (2001: 81-82) even reports the example in (94b’) to be
ungrammatical in standard Macedonian, but the native speakers I have consulted
disconfirm his judgments and claim that the sentence is acceptable when the object taa
rabota is focused.
(94) a. Mam już [te wszystkie sprawy] załatwione

have1SG already these all thingsNV.PL arrange-forPASS.NV.PL
“All these things have been taken care of”
a’. Mam już załatwione [te wszystkie sprawy] (Pl)
b. Ja imam završeno taa rabota
sheCL.ACC have1SG finishPTP.N thisF workF
“I have finished this work”
b.’ %Ja imam taa rabota završeno (Mac, cf. Bubenik 2001: 81-82)
I have argued that one of the criteria of grammaticalization of ‘have’-perfects is

availability of one-place predicates as past participles (cf. 95a). One-place predicates
may also appear as impersonal participles (cf. 95c), but they are excluded in stative
perfects (cf. 95b).
(95) a. Goce Delčev ima spieno tuka

Goce Delčev havePRES.3SG sleepPTP.N here
“Goce Delčev has slept here” (Mac, Friedman 1977: 91)
b. *Jan ma już tutaj spane
Jan have1SG already here sleepPASS.N.SG (Pl)
c. W tym łóżku już spano
in this bed already sleepPTP
“Someone has already slept in this bed” (Pl)
Correspondingly, adverbs are incompatible with adjectival passives in stative-perfect

clauses, because they may modify only verbal elements, such as past participles (cf. 96b)
or impersonal participles (96c).
(96) a. Imam često pieno mleko

have1SG often drinkPTP.N milk
“I have often drunk milk” (Mac, L. Grujoska, p.c.)
b. *Mam często pite mleko
have1SG often drinkPASS.N.SG milkN.SG (Pl)
c. Często pito mleko
often drinkPTP milk (Pl)
It has been shown that the reanalysis of the verb ‘have’ as an auxiliary in Germanic and
Romance leads to a loss of the meaning of possession expressed by the verb. The data
in (97) indicate that the verb ‘have’ in the stative-perfect constructions has been
semantically bleached as well. For instance, the events of selling apples and losing
umbrellas imply that the agent does not possess these objects any more. The fact that
the sentences in (97) are felicitous indicates that the verb ‘have’ has lost part of its
semantic content in the stative perfects in Slavic as well.
(97) a. Mam wszystkie jabłka sprzedane

have1SG all applesNV.PL sellPASS.NV.PL
“All my apples have been sold” (Pl)
b. Tazi godina imam zagubeni pet čadăra
this year have1SG losePASS.PL five umbrellas
“This year I have lost five umbrellas” (Bg, Lindstedt 2000: 41)
However, the stative perfect imposes semantic restrictions on the subject. The example
in (98a) shows that the subject may not be inanimate. By contrast, the ‘have’-perfect
permits inanimate subjects (cf. 98b). This confirms full grammaticalization of this
construction in Macedonian.
(98) a. *Statek ma uderzone w skały

ship have1SG hitNV.PL in rocks
“The ship hit rocks” (Pl)
b. Brodot se ima udreno vo karpite
ship-the REFL have1SG hitPTP.N in rocks
“The ship hit rocks” (Mac, L. Grujoska, p.c.)
3.6.3.2 Towards an analysis

Both Salvi (1987) and Hoekstra (1986) propose that the grammaticalization of ‘have’-
perfects consists in reduction of the Small Clause selected by the verb ‘have’ in the
stative perfect. Hoekstra is not clear about the structure of the Small Clause he
assumes, but Salvi suggests that the Small Clause is headed by an adjectival passive
participle, while the direct object is located in Spec, AP. In this way he seems to follow
Stowell’s (1981) concept of the Small Clause architecture, in which the Small Clause is a
projection of the category which heads the predicate. For example, in (99) the Small
Clause is AP, which is a complement of the verb have. It is headed by the adjectival
passive participle hidden, while the subject of the Small Clause fugitives is located in Spec,
AP.83
(99) a. We have fugitives hidden

b. [TP we [VP [V have [(SC)AP [NP fugitives [A hidden]]]]]]
If Slavi’s proposal is adopted for Slavic, the stative perfect in (86a) will have the
structure as in (100).
(100) a. Mam pasy zapięte

have1SG seatbeltsACC.NV.PL fastenPASS.ACC.NV.PL
b. [VP [V mam [(SC)AP [NP pasy [A zapięte]]]]] (Pl)
The subject of the Small Clause pasy ‘seatbelts’ is in the predicate relationship with the
adjectival passive participle zapięte ‘fastened’. The fact that the two elements form a
Small Clause is overtly manifested through agreement on the participle.
I suggest that the stative perfect structure is grammaticalized into a ‘have’-perfect
construction when the adjectival passive participle is reinterpreted as a verbal category.
In syntactic terms this means that the passive participle is no longer the head of the
Small Clause, but is reanalyzed as the head of the PartP, which takes the former subject
of the Small Clause as a complement. This eliminates the Small Clause configuration,
which results in the lack of agreement between the participle and the object. As an
illustration, a template representing the ‘have’-perfect is given in (101).
(101) a. Jas imam kupeno knigi

I have1SG buyPTP.N.SG booksF.PL
b. [vP jas [VP imam [PartP kupeno [DP knigite]]]] (Mac)
One of the intriguing properties of the stative perfect that still needs to be accounted
for is the prerequisite that all the participles must appear in the perfective form in this
construction (cf. 93a’ for Polish and 79a for Old English). I would like to explain this
requirement by referring to Embick’s (2004) analysis of passive participles.
Embick (2004: 361ff) suggests that adjectival passive participles differ from verbal
passive participles in a structural way.84 The root of the verbal passive participle is
dominated by v, a verbalizing head, which in turn is dominated by an Asp[ect]
projection. Adjectival passive participles lack the v projection above them, so they
attach directly to Asp in the course of derivation.
83See Hoekstra (2004) for an overview of various approaches to the structure of Small Clauses.
84 I am slightly simplifying Embick’s analysis here, because he proposes a ternary distinction of
participles. Namely, he examines the traditional division of passive participles in English into
‘verbal’ and ‘adjectival’ ones (cf. section 3.2), employs the term ‘eventive passive’ for the former
group and proposes a distinction between ‘stative’ and ‘resultative’ in the latter.
(102) a. Verbal passives
AspP
Asp vP
v √ROOT
b. Adjectival passives
AspP
Asp √ROOT
I assume that the Asp head hosts perfective prefixes.85 Since adjectival passives are not
dominated by the v-head, they must directly attach to Asp in the course of the
derivation. This is why only perfective forms of participles are possible in the stative
perfects in Slavic. Given that the (verbal) past participles in ‘have’-perfects are not
immediately dominated by the Asp head, they may appear in both perfective and
imperfective variants.
Verbal passives are dominated by v, which is a verbalizing head that encodes
eventivity and agentivity. One of the consequences of the presence of v is the
possibility of adverbial modification, which is compatible with eventive, but not with
stative readings.
(103) a. The package remained carefully opened

b. *The package remained carefully open
(104) a. The carefully opened package

b. *The carefully open package (Embick, 2004: 357; cf. also Kratzer 1994)
Correspondingly, since the adjectival passives in the stative perfect constructions lack
the verbalizing v head above their roots, they never allow any adverbial modification
(cf. 96, repeated below as 105a).
(105) a. *Mam często pite mleko

have1SG often drinkPASS.N.SG milkN.SG (Pl)
b. Imam često pieno mleko
have1SG often drinkPTP.N milk
“I have often drunk milk” (Mac, L. Grujoska, p.c.)
To summarize, this section has investigated the grammaticalization of ‘have’-perfects in

the Slavic languages. The analysis has been carried out by contrasting the ‘have’-perfect
85 The assumption follows from the commonly accepted idea that imperfective aspect is the
default (unmarked) form in Slavic. The perfective aspect requires an aspectual prefix, whereas the
imperfective aspect does not.
with the stative perfect. It has been shown that the process involves elimination of the
Small Clause selected by the verb ‘have’ in the stative perfect. The passive participle
becomes verbalized, which means that it is no longer the complement of the empty
head of the Small Clause, but instead it starts to occupy this position, which is
reinterpreted as V0.
To conclude, I contrast the properties of the stative perfect with the characteristics
of the ‘have’-perfect in the chart in (106).
(106)
the stative perfect the past perfect

participle-object agreement yes no
“subject” of the participle = not necessarily yes
subject of the clause
only perfective forms of the yes no
participle are permitted
the have-participle-object no yes
order is more common
the verb ‘have’ expresses often yes no
possession
the verb ‘have’ may be no yes
complemented by one-place
predicates
adverbials may modify the no yes
constructions
3.7 Conclusions
This chapter has investigated the evolution of the compound tense formed with the
auxiliary ‘have’ and the past participle in Slavic. Since the grammaticalization of this
construction has been completed only in Kashubian and Macedonian, whereas some
other Slavic languages use its rudimentary variant termed the “stative perfect”, it was
possible to study the diachronic development of this structure from a synchronic point
of view.
The chapter has also discussed movement of the past participle across the auxiliary
‘have’ in Macedonian. It was claimed that the operation corresponds to the well-known
case of “VP-topicalization” in Germanic. The movement was contrasted with l-
participle fronting across the auxiliary ‘be’ discussed in chapter 2. It was argued that
even though they both involve XP-movement, these are different types of syntactic
displacement. The l-participle raises to Spec, TP via A-movement in order to check the
φ-features of T0. The past participle raises much higher than the l-participle, and it may
pied-pipe a number of other constituents. It lands in Spec, FocP via A’-movement,
which is triggered by a Focus feature.
Both the l-participle in the ‘be’-perfect as well as the past participle in the ‘have’-
perfect derive from forms which were adjectival in nature. The l-participle has been
uniformly verbalized across Slavic, and was claimed in chapter 2 to be able to project an
external theta role and assign case to its arguments. The past participle, which originates
from the passive participle, has been verbalized to various degrees in different
languages. Slavic data discussed in this chapter indicate that if the past participle is
morphologically the same as the passive participle, it is unable to assign case or an
Conclusions 161
external theta role. However, once its morphological make-up is differentiated, it may
become further verbalized and eventually reanalyzed as a finite verb form, as in the case
of the impersonal participle in Polish. This observation provides more support for the
assumption made in chapter 2 that the past and passive participles represent the same
grammatical category and should be analyzed in a uniform way.
To conclude, let me show how the verbalization of the passive participle may be
represented syntactically. Following Jabłońska (2006), I would like to argue that the
degree of verbalization of a category corresponds to the height of movement of this
verbal form in the extended projection of the VP. The idea is schematized in (107) for
the participial forms analyzed in this chapter.
(107)
TP
finite verbs MoodP

and the l-participle
the impersonal participle vP
the past participle VP
the past participle AP
the passive participle

in the stative perfect
The passive participle is located the lowest in the tree, because it is the most adjectival
category among the ones marked in the phrase structure. The past participle, which is
the main verb in ‘have’-perfects, has been argued to be non-distinct from the passive
participle, because it is unable to project an external theta role and assign accusative
case. However, it accepts a wider spectrum of verbs than the passive participle, so it
targets a higher position. Moreover, recall from section 3.6.3.2 that the past participle is
generated higher (i.e. in V0) than the passive participle in the stative perfect. The
impersonal participle must be able to move above vP, because it assigns structural case
and the theta role to the covert subject. It can be derived from modal verbs, which
means that is able to reach the MoodP level. Nevertheless, it always appears in a
morphologically invariant form, which indicates it does not reach T0. Finally, l-
participles and finite verbs display morphological agreement with the subject, are able
to project an external theta role and assign structural case to the object. Hence, they
represent the highest degree of verbalization and are able to target the TP layer.
Chapter 4 Clitics in South Slavic
4.1 Introduction
It was mentioned in the introduction to this dissertation that auxiliaries in Germanic
and Romance languages often differ from main verbs in phonological and
morphological impoverishment. Correspondingly, in many Slavic languages auxiliaries
are clitics, whose distribution is subject to various phonological conditions. However,
since they always cluster with pronominal clitics, it is necessary to examine all types of
clitics together. The present chapter will discuss the properties of these elements in the
three contemporary South Slavic languages analysed in this thesis. Since Slavic clitics
have already received considerable attention in linguistics and the literature devoted to
the topic is extensive (cf. Bošković 2001, to appear; Franks 1998; Franks & King 2000
for a detailed overview), the discussion is relatively brief. However, the investigation of
the positions taken by clitics is relevant for the study of compound tenses not only
because some of the auxiliaries are clitics. The Slavic languages have very free word
order, but the distribution of clitics is exceptionally rigid, as they must always appear in
designated positions. Therefore, clitics can be used as a criterion to determine the
placement of other elements in the clause.
The chapter is organized as follows: after some expository remarks concerning the
nature of clitics in section 4.2, section 4.3 will discuss the evolution of clitic positions in
Old Church Slavonic. Section 4.4 will present the patterns of clitic placement in three
contemporary South Slavic languages. The major division will be drawn between Serbo-
Croatian (cf. section 4.4.1), in which clitics appear in the second position, and
Macedonian (cf. section 4.4.2) and Bulgarian (cf. section 4.4.3), which have verb-
adjacent clitics. The chapter will also show, taking recourse to Macedonian data, that
the distribution and the inventory of auxiliary clitics may influence the nature of l-
participle fronting investigated in chapter 2.
4.2 What are clitics?

Clitics are usually understood as small lexical elements that cannot form an independent
prosodic domain.86 They never bear stress lexically, and for reasons of prosody they
always form an accentual unit with the preceding or the following word. The word they
attach to is termed the “clitic host”. If a clitic is dependent on an element that follows
it, it is referred to as “proclitic”. Clitics that are attached to the material preceding them
are dubbed “enclitic”. Moreover, clitics may also adjoin to each other and form “clitic
clusters”.
In some languages clitics are not specified for the direction of attachment, and they
can be both enclitic and proclitic. For instance, in Macedonian the directionality of
cliticization depends on the categorial nature of the clitic host: the clitics procliticize on
tensed verbs (cf. 1), but encliticize on imperatives (cf. 2).87
86 For a detailed overview of properties of clitics based on non-Slavic data, see Van Riemsdijk
(1999).
87 Clitics are italicized in all examples included in this chapter.
164 Clitics in South Slavic
(1) a. Mi go dade Vera včera

meCL.DAT itCL.ACC givePAST.3SG Vera yesterday
“Vera gave it to me yesterday”
b. *Dade mi go Vera včera (Mac, Franks 1998)
(2) a. Penkaloto kupuvaj mi go!

Pen-the buyIMPV.2SG meCL.DAT itCL.ACC
“Buy me the pen!”
b. *Penkaloto mi go kupuvaj! (Mac, Franks 1998)
Given that clitic placement may be sensitive to the grammatical category of the host,
and sometimes even to the specification of some grammatical features such as
finiteness, the requirements of clitics are related not only to phonology, but also to
other levels of representation, such as syntax, morphology and semantics.
Chomsky (1995: 249) argues that clitics are ambiguous categories, which share XP
and X0 properties. This is indicated by the fact that they move from argument positions
within the VP, which are phrasal, and climb in order to attach to the inflectional head
T0. Since the movement violates the Head Movement Constraint, Chomsky claims that
the clitics raise as XPs and only the last step in the derivation involves head-adjunction.
Bošković (2002) states that the ambiguous XP/X0 status of clitics can be defined in
syntactic terms by arguing that they are non-branching elements.
Ever since Zwicky (1977), two types of clitics have been distinguished: “simple”
and “special” ones. Simple clitics have the same distribution as their non-clitic
counterparts. For example, simple pronominal clitics pattern in the same way as
pronouns and may exhibit no special phonological properties except for being
unstressed. By contrast, the syntactic and the phonological behaviour of special clitics
often differs from the behaviour of their non-clitic counterparts. For instance, special
pronominal clitics may be located in the positions that are inaccessible to non-clitic
arguments. Most of the clitics found in the South Slavic languages are of the “special”
type.
More recently, Cardinaletti & Starke (1999) have scrutinized the traditional division
of pronouns into clitic and strong pronouns and concluded that the binary distinction is
not fine-grained enough. Therefore, they suggested a three-way distinction of pronouns
into strong pronouns, weak pronouns, and clitics. Moreover, they proposed a
“Minimise Structure” generalization, which states that the realization of a clitic is the
most economical solution. The realization of a weak or a strong pronoun should be a
last resort procedure. Section 4.4.2.4.2 will provide an analysis of cliticization in
Macedonian, which is based on this proposal.
4.3 Diachrony of clitics in Slavic

This section will briefly outline the historical development of clitics in Slavic with
respect to their morphological forms and their positions in the clause structure. The
data presented here will be used later in the thesis to verify some claims concerning
cliticization in contemporary Slavic languages.
Diachrony of clitics in Slavic 165
4.3.1 Reduction of the copula/auxiliary ‘to be’

For convenience, the chart in (3) repeats the paradigm of the auxiliary ‘to be’ in Old
Church Slavonic from chapter 1 (section 1.3.3.5.2.1). These are full, orthotonic forms.
(3) Orthotonic forms the auxiliary ‘to be’ in Old Church Slavonic

1 jesmь jesvě jesmъ
2 jesi jesta jeste
3 jestъ jeste sõtъ
(OCS, cf. Lunt 1974)
Section 1.3.3.5.2 in chapter 1 showed that in Old Church Slavonic it was quite common
for the 3rd person singular and plural forms to appear as reduced enclitics je and sõ,
respectively. When the reduced forms emerged, the orthotonic forms were maintained
for an emphatic usage. The two forms differed in their distribution in the clause. The
reduced variants had to appear in the 2nd position. The orthotonic forms could be
found in other positions, also clause-initially. Their actual placement depended on the
information structure of the clause (cf. Andersen 1987: 24-25).
Further reduction of the auxiliary ‘to be’ occurred in different ways in particular
Slavic languages. According to Decaux (1955: 187-188), in Serbo-Croatian the change
took place in two steps. First, a reduced vowel (a jer) was inserted to split the -sm-
cluster in the Old Church Slavonic form jesmь. This gave the form jesъmь, which later
changed into the variant jesam. Afterwards, the 1st and the 2nd person forms in the
singular and the plural were reduced into clitics: jesam, jesi, jesmo, jeste > sam, si, smo, ste.
However, the old full forms were retained alongside, and consequently, there are two
paradigms of the auxiliary ‘to be’ in contemporary Serbo-Croatian (cf. section 4.4.1.1).
Most of the other South and West Slavic languages (including Bulgarian,
Macedonian, Czech, Slovene, and Sorbian) followed the same path of the change, but
the process went much further, because it additionally involved vowel reduction, as in
the contemporary Bulgarian forms of sŭm, si, sme, ste. The reduction of the auxiliary ‘to
be’ in Polish was the most complex in all the Slavic languages. It consisted in the
reanalysis of the reduced variants as affixes on the l-participle and a creation of a new
copula paradigm (cf. section 1.3.4.2.2.1 in chapter 1).
The reduction never took place in the East Slavic languages. The disyllabic forms
were continuously preserved until they suddenly completely disappeared (cf. section
1.3.4.2.1 in chapter 1).
4.3.2 Reduction of the pronominal clitics

Pronouns appeared in 6 morphological cases in Old Church Slavonic, but only the
dative and the accusative had clitic variants. The chart in (4) lists only the 1st and the 2nd
person forms. For the 3rd person, suppletive variants of the demonstrative j- and tъ were
used (cf. Lunt 1974: 65; Schmalstieg 1983: 62-65). Just as the auxiliary forms, the clitic
variants were derived via reduction of the full pronouns. However, Sławski (1946: 74)
points out that pronominal clitics were reduced earlier than the auxiliary clitics.
(4) Pronominal clitics in Old Church Slavonic
1SG 2SG 1DUAL 2DUAL 1PL 2PL REFL

ACC mene/mę tebe/tę na/ny va/vy nasъ/ny vasъ/vy sebe/sę
DAT mьně/mi tebě/ti nama/- vama/- namъ/ny vamъ/vy sebě/si
(OCS, Huntley 2002: 144)
4.3.3 Changes in the clitic positions

This section will analyse the positions occupied by clitics in the Old Church Slavonic
clause structure. It will be shown that even though they had to occur in the second
position, some violations of the requirement could be observed at later stages of the
language development. According to traditional linguistic accounts (cf. Sławski 1946),
these violations were due to diachronic changes in phonology. However, I will
demonstrate that phonology cannot be the only reason why they occurred, because it is
possible to define contexts in which only some types of clitics must move to the second
position. Therefore, even though the placement of clitics is influenced by their prosodic
deficiency, they still behave like syntactic units and undergo syntactically-constrained
movements.
4.3.3.1 Wackernagel’s law

A number of diachronic studies have shown that clitics had to appear in the clause
second position in the early Indo-European languages. Jakob Wackernagel was the first
linguist to describe this tendency in his 1892 paper. He investigated Ancient Greek
texts and observed that some accentless elements, such as personal pronouns, indefinite
adverbs, indicative forms of ‘be’ and ‘say’ in the present tense, as well as some particles,
had to appear immediately after the initial word in the sentence. He concluded that this
was a general principle of word order in Proto-Indo-European, which is now referred
to as “Wackernagel’s law” (cf. Anderson 1993 for an overview of Wackernagel’s
proposal).
There are very few comprehensive diachronic studies of clitic positions in Slavic
available. Sławski (1946) offers the most detailed investigation of the position of
enclitics ranging from Old Church Slavonic to Modern Bulgarian.88 Following
Wackernagel’s insights, he stresses the influence of prosody on word order and argues
that in Old Church Slavonic the clitics had to always follow the first word bearing
stress. Moreover, he investigates potential clitic hosts and concludes that even though
many elements of different categories could lend support to the clitics, they had to be
part of the same clause structure. For example, extra-sentential lexemes, such as
conjunctions could not be clitic hosts (Sławski 1946: 25). Although the rule was by no
means a Slavic invention, because exactly the same restrictions held in Sanskrit and
Ancient Greek, it contrasts with cliticization in contemporary Bulgarian and Polish, in
which coordinating conjunctions are eligible as phonological clitic hosts (cf. 4.4.3.2 and
chapter 5, section 5.2.2).
Verbs were the most likely candidates for clitic hosts in Old Church Slavonic.
Huntley (2002: 165) and Willis (2000) demonstrate that the clitics had to immediately
88 A more recent diachronic account of clitic placement in Bulgarian can be found in Pancheva
(2005).
follow the l-participle when it occurred at the beginning of a clause (cf. 5). Importantly,
this pattern is not necessarily observed in the Ancient Greek sources of the Old Church
Slavonic translations (Sławski 1946: 17, 29).
(5) I dalъ ti bi vodõ živõ

And givePART.M.SG youDAT would3SG water living
“And he would give you the water of life” (OCS, John 4: 10, Willis: 2000: 333)
However, there was no adjacency requirement for the participle and the auxiliary when
the participle followed an auxiliary clitic, as exemplified in (6), where the adverb sьde
intervenes between the conditional auxiliary clitic bi and the l-participle bylъ.
(6) Ašte bi sьde bylъ, ne bi bratrъ moi umrlъlъ

if would2SG here bePART.M.SG NEG would3SG brother my diePART.M.SG
“If you had been here, my brother would not have died”
(OCS, John 11: 21, Willis 2000: 328)
Consequently, the following combinations were possible.
(7) a. Part – Aux

b. XP/X – Aux – (YP/Y) – Part
Outside the left periphery of the clause, the word order was relatively free. As shown in
(8), the participle could be either followed or preceded by the auxiliary.
(8) a. Jako varilъ jestъ

as go-on-aheadPART.F.SG beAUX.3SG
“That he has gone ahead” (OCS, Suprasliensis 204.29, Huntley 2002: 165)
b. Jako-že jestъ obyklъ
as+FOC beAUX.3SG become-accustomedPART.M.SG
“As he has become accustomed”
(OCS, Suprasliensis 382.24, Huntley 2002: 165)
The data presented in this section lead to two conclusions. First, the word order in Old
Church Slavonic was rather free, but the clitics had to appear in designated positions. It
could be argued that these positions correspond to different functional heads, whose
features are spelt out by clitics. Second, the clitics often followed the l-participle, which
served as the verbal host. Following the analysis developed in chapter 2, it might be
postulated that the clause-initial l-participle targeted Spec, TP via locative inversion to
check the φ-features of T.
4.3.3.2 Violations of Wackernagel’s law in Old Church

Slavonic
Even though the clitics in Old Church Slavonic had to appear in the second position,
Sławski (1946: 29) observes that the clitics in some Old Church Slavonic texts showed a
very strong tendency to appear immediately preceding or following the verb, even if
this at times led to a violation of Wackernagel’s law. Sławski suggests that this tendency
was a Slavic innovation, which in particular could be observed in the case of the
reflexive clitic sę (cf. 9a) and the conditional auxiliary clitic (cf. 9b).
(9) a. Da byxъ pokajalъ sę kъ bogu

that wouldCL.1SG repentPART.M.SG CL.REFL to God
“In order that I might repent before God”
(OCS, Suprasliensis 167.2, Willis 2000: 333)
b. Da ne prĕdanъ bimь Ijudĕomъ
that NEG handed-overPASS.M wouldCL.1SG JewsDAT
“That I might not be handed over to the Jews”
(OCS, John 18: 36, Willis 2000: 332)
Sławski’s study illustrates two directions in the development of the clitic positions in
Old Church Slavonic. On the one hand, the clitics targeted the second position in the
clause. On the other hand, the clitics aimed to stay adjacent to the verb. Even though
the latter tendency was at odds with the strategies of clitic placement found in the
original Ancient Greek sources of the Old Church Slavonic translations, it was
widespread in many other Indo European languages, for instance in all Romance
languages, Albanian, and Greek. Sławski (1946: 79) attributes it to the increased
prominence of word stress as opposed to sentence stress.
However, the phonological change mentioned by Sławski was certainly not the
only reason for the violations of Wackernagel’s law. Willis (2000) points out that the
placement of clitics in Old Church Slavonic was often determined by their syntactic
category. For instance, when the l-participle or some other predicative element showing
subject agreement was clause-initial, all types of clitics would immediately follow it. This
is illustrated for the pronominal clitic mi ‘meDAT’ in (10a) and for the sentential focus
clitic že in (10b), which are right-adjacent to the passive participles dana ‘given’ and rečeno
‘said’, respectively.
(10) a. Dana mi estъ vьsĕka vlastь

givePASS.F.SG meCL.DAT bePRES.3SG every power
na neb[e]se i na zemi
on heaven and on earth
“Every power over heaven and earth is given to me”
(OCS, Matt. 28: 18, Willis 2000: 325)
b. Rečeno že bystъ
sayPASS.N.SG CL.FOC beAOR.3SG
“But it was said” (OCS, Matt. 5: 31, Willis 2000: 325)
By contrast, when some other constituent was placed in front of the predicative
element, the sentential clitics would move to the Wackernagel position, preceding the
participle in this way, but the pronominal clitics would stay in situ. This is exemplified
by the sentential clitic že, which occurs in the second position in the embedded clause
in (11), and by the pronominal clitic ě ‘them’ in (12), which must follow the l-participle
sъtvorilъ ‘create’.
(11) G[lago]ljõ vamъ ěko nikotory že pror[o]kъ

say1SG youPL.DAT that no CL.FOC prophet
prijętenъ estъ vъ otečъstvii svoemъ
acceptPASS.N.SG be3SG in homeland his
“I say to you that no prophet is accepted in his homeland”
(OCS, Luke 4: 24, Willis 2000: 326)
(12) A otъ načęla sъzъdaniju,...

and from beginning creationDAT
a. mõža i ženõ sъtvorilъ ě etsъ b[og]ъ
man and woman createPART.M.SG themCL be3SG God
“And from the beginning of creation, God created them man and woman”
b. *mõža i ženõ ě sъtvorilъ etsъ b[og]ъ
c. *mõža ě i ženõ sъtvorilъ etsъ b[og]ъ (OCS, Mark 10: 6, Willis 2000: 326)
Following the ideas concerning l-participle fronting developed in chapter 2, I will

assume that the l-participle sъtvorilъ in (12a) fills Spec, TP. This implies that the
pronominal clitic ě is internal to the TP. However, the focus clitic že in (11) must be
higher than the TP layer, because it precedes the subject. This indicates that sentential
and pronominal clitics could target different projections in Old Church Slavonic.
It is also possible to find examples in which the same clitic is hosted in different
positions in the clause structure. For instance, Willis observes that the conditional
auxiliary clitic had to be always right adjacent to the complementizer a (cf. 13),
irrespectively of the ordering of the elements following the clitic (cf. 13b).89 Conversely,
the complementizer da did not have to be adjacent to the conditional clitic. This is
shown in the two different translations of Luke 4: 42. In the first one in (14a) bi is
separated from da by negation. In the second one in (14b) the two constituents are
adjacent, and negation occurs to the right of the conditional auxiliary bi.
(13) a. A by bylъ sьde

if COND.3SG bePART.M.SG here
“If he had been here”
b. A by sьde bylъ
c. A by bylъ prorokъ
if COND.3SG bePART.M.SG prophet
“If he had been the prophet” (OCS, Vaillant 1977: 219)
(14) a. Drъžaaxõ i da ne bi otъšelъ

held3PL him that NEG COND.3SG leavePART.M.SG
otъ nixъ
from them (OCS, Codex Marianus, Willis 2000: 330)
b. Drъžaaxõ i da bi ne otъšlъ
held3PL him that COND.3SG NEG leavePART.M.SG
otъ nixъ
from them (OCS, Codex Zographensis, Willis 2000: 330)
“And they held him, so that he would not leave them”
Willis concludes that depending on their semantics, certain types of complementizers

could attract clitics. The attraction was obligatory in the case of a, which introduced
conditional sentences. Conversely, the complementizer da, which usually introduced
indicative (declarative) clauses (cf. Bräuer 1957), did not attract clitics, so they did not
have to be adjacent to it.
I would like to suggest that the variation in the position of the clitics indicates that
some of them may be attracted by a Force feature, which specifies the clause type. I
propose that this feature is in the functional head Σ (cf. Laka 1994). Clitics always have
89Willis (2000) does not provide any data with the complementizer a, so the examples in (13) are
taken from Vaillant (1977).
to move to Σ whenever the sentence deviates from declarative. Thus, in (13) the clitic
by has to raise to Σ because the sentence is marked as conditional. In (11) že raises to
the second position in the embedded clause in order to licence a focus feature on the
word nikotory ‘not a single one’, which is interpreted as emphatic. Correspondingly, the
conditional auxiliary byxъ in (9a) raises to Σ in order to licence the subjunctive
interpretation of the clause, leaving the reflexive clitic sę stranded behind, while (9a) is
an instance of attraction of the passive participle by negation (cf. section 2.3.6.3.2 in
chapter 2 for some discussion of negation in Old Church Slavonic).
I will return to the relation between the Wackernagel position of clitics and the
Force marking in section 4.4.4.2. I will also show that clitic movement to Σ for Force
licensing occurs in many contemporary Slavic languages as well (cf. 4.4.2.4.2.3 for a
discussion of imperatives in Macedonian, and section 5.3.4.1 in chapter 5 for an analysis
of enclitization in Polish).
Summarizing, this section has demonstrated that although clitics had to appear in
the second position in Old Church Slavonic, there were contexts in which the rule was
not obeyed. It is an empirical question whether the violations of Wackernagel’s law
were due to a phonological or a syntactic condition. Irrespectively of the motivation, it
has also been shown that although clitics are phonologically weak and need support
from other elements in the structure, they behave like syntactic units: they can be
attracted by formal features and undergo syntactic movements. This claim will be
further substantiated in the discussion of cliticization in Serbo-Croatian presented in
the next section.
4.4 Clitic positions in contemporary South

Slavic languages
The following sections will overview the properties and positions of clitics in the three
contemporary South Slavic languages discussed in this thesis. Even though these
languages are very closely related, their clitic systems are surprisingly distinct: Serbo-
Croatian has Wackernagel clitics; Bulgarian has enclitics that do not have to occur in
the second position, whereas clitics in Macedonian are proclitic or enclitic depending
on the type of host. In both Bulgarian and Macedonian clitics must be verb-adjacent,
but this requirement is absent in Serbo-Croatian. It will be shown that these contrasts
correspond to the ways clitics are arranged in the clause structure. The Wackernagel
clitics in Serbo-Croatian will be argued to occupy specifiers of relevant functional
projections, while the clitics in Bulgarian and Serbo-Croatian will be claimed to be
adjoined to a single head.
In spite of the differences mentioned above, the ordering of clitics with respect to
each other is the same in all South Slavic languages. The dative precedes the accusative,
while the auxiliary clitics show an intriguing split concerning the positions of the 3rd
person singular form, which must always appear as the last one in the clitic cluster.
(15) li90> Mod > AUX (except 3rd SG) > REFL > DAT > ACC > 3rd SG AUX
( Tomić 1996a, Rivero 2005, Franks&King 2000: 45)
90Recall from chapter 2 that li is an interrogative complementizer, and is commonly assumed to

be in C.
Clitic positions in contemporary South Slavic languages 171
4.4.1 Serbo-Croatian
I will begin the overview with an analysis of Serbo-Croatian, which is a prototypical
example of a language observing Wackernagel’s law. Along with all the other South
Slavic languages, it shows a distinction between clitic and non-clitic forms in both the
pronominal and the auxiliary systems.
4.4.1.1 The clitic paradigms

Serbo-Croatian distinguishes three cases in the pronominal paradigm: accusative,
genitive, and dative. All of them have full and clitic forms, exemplified in (16).
(16) Pronominal forms
Acc (full/clitic) Gen (full/clitic) Dat (full/clitic)

1SG mene/me mene/me meni/mi
2SG tebe/te tebe/te tebi/ti
3SG M/N njega/ga njega/ga njemu/mu
3SG F nju/je nje/je njoj/joj
1PL nâs/nas nâs/nas nama/nam
2PL vâs/vas vâs/vas vama/vam
3PL njih/ih njih/ih njima/im
REFL sebe/se sebe/- sebi/si
(S-C, Franks & King 2000: 19-24)
Likewise, the present tense auxiliaries may appear either in full or clitic forms, with a
further division into affirmative and negative variants, as shown in (17). Both variants
have been argued to be created in syntax by incorporation into a negative or a positive
polarity head (cf. chapter 2, section 2.3.6.2.2).
(17) Auxiliary forms
affirmative negative
SG (full/clitic) PL (full/clitic) SG PL
1 jesam/sam jesmo/smo nisam nismo
2 jesi/si jeste/ste nisi niste
3 jest(e)/je jesu/su nije nisu
(S-C, Franks & King 2000: 19-24)
Some other functional verbs, such as the conditional form of the verb biti ‘to be’ and
the future auxiliary ht(j)eti ‘want’, show a similar division into clitic and full forms. In
addition, ht(j)eti has negative variants, which are always strong.
(18) Forms of the verb ht(j)eti ‘want’/‘will’
positive negative
SG (full/clitic) PL (full/clitic) SG negative PL negative
1 hoću/ću hoćemo/ćemo neću nećemo
2 hoćeš/ćeš hoćete/ćete nećeš nećete
3 hoće/će hoće/će neće neće
(S-C, Franks & King 2000: 19-24)
(19) Forms of the conditional auxiliary
SG (full/clitic) PL (full/clitic)
1 bïh/bih bïsmo/bismo
2 bï/bi bïste/biste
3 bï/bi bï/bi
(S-C, Franks & King 2000: 19-24)
Unlike clitics, the non-clitic auxiliaries are insensitive to any phonological constraints on
their placement in the sentence; therefore they may appear clause-initially. Moreover,
they do not have to occur in the same sentence positions as their clitic counterparts.
For example, the non-clitic 2nd person plural variant jeste may precede the interrogative
complementizer clitic li (cf. 20a), while its corresponding clitic form ste has to appear to
the right of li (cf. 20b).
(20) a. Jeste li joj se predstavili u sali?

beAUX.2PL Q herCL.DAT REFL introducePART.M.PL in hall
“Did you introduce yourselves to her in the hall?” (S-C, Spencer 1991: 354)
b. Dali ste joj se predstavili u sali?
COMP+Q beAUX.2PL herCL.DAT REFL introducePART.M.PL in hall
Furthermore, strong auxiliary forms do not need to cluster with clitics. This is
exemplified for jeste and nije, which are split from the pronominal clitics by the subject
Petar in the embedded clauses in (21b).
(21) Stefan tvrdi...

Stefan claim3SG
a. da mu ga je Petar poklonio
that himCL.DAT itCL.ACC beCL.AUX.3SG Petar givePART.M.SG
“Stefan claims that Petar has given it to him as a present”
b. da mu ga Petar jeste/nije poklonio
that himCL.DAT itCL.ACC Petar beAUX.3SG/NEG+AUX.3SG givePART.M.SG
“He claims that Petar did/didn’t give it to him as a present”
(S-C, cf. Caink 2000: 63)
4.4.1.2 Properties and positions of the Serbo-Croatian

clitics
The clitics in Serbo-Croatian are not selective about the category of their hosts; the only
categories from which they may not receive phonological support are prepositions and
conjunctions.91 This is illustrated for the conjunction i in (22).
(22) a. Ivan je vidio auto i kupio je ga

Ivan beAUX.3SG seePART.M.SG car and buyPART.M.SG beAUX.3SG itCL.ACC
“Ivan saw the car and bought it”
b. *Ivan je vidio auto i je ga kupio (S-C, Ćavar & Wilder 1999: 454)
Moreover, they do not have to be adjacent to elements of any specific grammatical

category. For instance, they do not have to be adjacent to the l-participle (cf. 23) or the
finite verb (cf. 24). The only requirement that they must observe is that they appear in
the second position.
(23) Vi ste mu je nesumnjivo

you beAUX.2PL himCL.DAT herCL.ACC undoubtedly
predstavili juče
introducePART.M.PL yesterday
“You undoubtedly introduced her to him yesterday” (S-C, Bošković 1995)
(24) a. Sutra ga deca ne-će videti

tomorrow himCL.ACC children NEG+FUT.3PL seeINF
“The children will not see him tomorrow” (S-C, Progovac 1996: 425-426)
b. Koliko im ko daje?
How-much themCL.DAT who givePRES.3SG
“Who gives how much to them?” (S-C, Rivero 1991: 335)
In yes-no questions (cf. 25a), as well as in embedded clauses (cf. 25b-c), the clitics must
be right adjacent to the complementizer da.
(25) a. Da li mi ga daješ?
that Q meCL.DAT itCL.ACC givePRES.2SG
“Are you giving it to me?”
b. Kaže da mi ga je Petar dao
sayPRES.3SG that meCL.DAT itCL.ACC beAUX.3SG Petar givePART.M.SG
“He says that Petar has given it to me”
c. *Kaže da Petar mi ga je dao (S-C, Tomić 1996a: 818-819)
Placement of the clitics in any other position than the second, as well as splitting them
from each other results in ungrammaticality.
91 Conjunctions are possible clitic hosts in Bulgarian and Polish, as shown in sections 4.4.3 and in
chapter 5 (section 5.2.2), respectively.

(26) a. *Mi Marijinoj prijateljici smo ga dali

we Marija’s friend beAUX.1PL itCL.ACC givePART.M.SG
We gave it to Mary’s friend”
b. *Mi smo Marijinoj prijateljici ga dali (S-C, Stjepanović 1998: 528)
At first blush, the data might suggest that all the clitics target a designated clitic site
together as a unit. In fact, this is what was proposed in the earliest generative accounts
of cliticization in Serbo-Croatian, such as Wilder & Ćavar (1994), Progovac (1996), and
Tomić (1996a), who suggest that clitics are all right-adjoined to C0 in the case of
embedded clauses, or to the highest head in the structure available in the case of main
clauses (cf. Franks 1998). However, there are theoretical and empirical shortcomings
related to this proposal.
From a theoretical point of view, it is problematic to suggest that pronominal
clitics raise to C0 directly. If they were to move as heads, this would mean crossing
other heads on the way, such as the auxiliary nije in (27), and inducing the Head
Movement Constraint violation.
(27) …da gaj Ivan nije udario tj

that himCL.ACC Ivan NEG+bePRES.3SG hitPART.M.SG
“…that Ivan didn’t hit him” (S-C, Wilder & Ćavar 1994: 54)
In all fairness, it must be noted that each of the proponents of this analysis recognizes
this theoretical problem; see Franks (1998) and Progovac (1996, 1999) for a discussion
and potential solutions.
However, empirical facts also argue against the idea that both the auxiliary and
pronominal clitics are located in the same position. For instance, this is demonstrated
by the interpretation of certain adverbs in the presence of the two types of clitics.
(28) a. Oni su pravilno odgovorili Mileni

they beAUX.3PL correctly answerPART.M.PL MilenaDAT
“They did the right thing in answering Milena”
“They gave Milena a correct answer”
b. Oni su joj pravilno odgovorili
they beAUX.3PL herCL.DAT correctly answerPART.M.PL
“*They did the right thing in answering her”
“They gave her a correct answer” (S-C, Bošković to appear)
The adverb pravilno ‘correctly’ is ambiguous and may have a sentential or a manner
reading. The sentence in (28a), which contains only the auxiliary clitic su, is acceptable
under both interpretations of the adverb. However, the string in (28b), which contains
the auxiliary su followed by the dative clitic joj, permits only the manner-oriented
reading of the adverb. Given the standard assumption that sentential adverbs reside
higher in the structure than manner adverbs, this means that the auxiliary clitic su
moves higher when it occurs on its own, as in (28a), than when it is accompanied by a
pronominal clitic, as in (28b).
Stjepanović (1998, 1999) provides more evidence against the idea that the clitics
always cluster in the same position. She observes that a part of the clitic cluster may be
deleted under VP ellipsis, as illustrated in (29).
(29) a. Mi smo mu ga predstavili, a

we beAUX.1PL himCL.DAT himCL.ACC introducePART.M.PL and
i vi ste mu ga predstavili, (takodje)
also you beAUX.2PL himCL.DAT himCL.ACC introducePART.M.PL too
“We introduced him to him, and you did, too”
b. Mi smo mu ga predstavili, a i vi ste mu ga predstavili, (takodje)
c. *Mi smo mu ga predstavili, a i vi ste mu ga predstavili, (takodje)
d. *Mi smo mu ga predstavili, a i vi ste mu ga predstavili (takodje)
e. *Mi smo mu ga predstavili, a i vi ste mu ga predstavili (takodje)
(S-C, cf. Stjepanović 1998, 1999)
The examples in (29) contain two conjoined sentences. In both of them the subject is
immediately followed by the auxiliary clitic, which in turn is followed by two
pronominal clitics: the indirect object mu and the direct object ga. The second part of
the conjunct is affected by ellipsis. As indicated in (29a), both of the pronominal clitics
can be deleted with the auxiliary clitic remaining overt. The sentences in (29b, c)
demonstrate that it is also possible to elide one of the pronominal clitics, but it must be
the one that is higher in the structure.
Adopting the standard assumption that only constituents may be deleted (cf.
Lasnik 1995), the data in (29) show that the clitics in Serbo-Croatian do not cluster, but
are positioned hierarchically with respect to each other: the auxiliary clitic su is located
higher than the pronominal clitics, and the dative clitic dominates the accusative clitic.
If the accusative clitic were higher than the dative clitic, there should be a constituent
that contains the dative clitic (in addition to the l-participle), but not the accusative
clitic. This is not the case, as indicated by the ungrammaticality of (29c). These
examples also suggest that the clitics may not be adjoined to each other, but rather each
of them must be located in a separate maximal projection.
More support for the idea that clitics in Serbo-Croatian occupy maximal
projections comes from Progovac’s (1993) investigation of clitic movement across
different types of verbs. By applying a number of different syntactic criteria, Progovac
draws a distinction between verbs that take “subjunctive-like” complements, and verbs
that take “indicative-like” complements. For example, she shows that topics can raise
from “subjunctive-like” complements (cf. 30b), but not from “indicative-like”
complements (cf. 30a). Likewise, wh-movement is prohibited out of “indicative-like”
complements (cf. 31a), but it freely occurs out of “subjunctive-like” complements (cf.
31b). Similarly, negative polarity items may extend their domain only in “subjunctive-
like” complements (cf. 32b).
(30) a. *Pismo ne kažem da sam potpisao

letter NEG sayPRES.1SG that beAUX.1SG signPART.M.SG
“??The letter, I don’t say that I have signed”
b. Pismo ne želim da potpišem
letter NEG wantPRES.1SG that signPRES.1SG
“The letter, I don’t want to sign”
(31) a. ?*Koga ne kažeš da voliš?

who NEG sayPRES.2SG that lovePRES.2SG
“Whom don’t you say that you love?”
b. Koga ne želiš da voliš?

whom NEG wishPRES.2SG that lovePRES.2SG
“Whom don’t you want to love?”
(32) a. *Ne kažem da vidim nikoga

NEG sayPRES.1SG that seePRES.1SG nobody
“I do not say that I see anyone”
b. Ne želim da vidim nikoga
NEG wishPRES.1SG that seePRES.1SG nobody
“I do not wish to see anyone” (S-C, Progovac 1993)
Importantly, Progovac shows that clitic placement is sensitive to the same dichotomy
between the two types of verbs. The examples in (33b) and (34b) demonstrate that
clitics may climb out of “subjunctive-like” complements, but not out of “indicative-
like” complements. The movement occurs for focus reasons.
(33) a. Milan kaže da ga vidi

Milan says that himCL.ACC sees
“Milan says that he can see him”
b. *Milan ga kaže da vidi
(34) a. Milan želi da ga vidi

Milan wishes that himCL.ACC sees
“Milan wishes to see him”
b. ?Milan ga želi da vidi (S-C, Progovac 1993)
The outputs in (33) and (34) lead to two conclusions concerning the position of the
clitics in the clause structure. First, they confirm the assumption made in section 4.3.3.2
that although clitics are phonologically deficient, they behave like syntactic units. Their
placement adheres to syntactic locality conditions, which cannot be given a
phonological or morphological explanation, because it is subject to the same constraints
as the uncontroversially syntactic operations like wh-movement or topicalization.
Second, the sentence in (34b) exemplifies clitic climbing from an embedded clause
to the main clause. The fact that this is possible supports the idea that the pronominal
clitics in Serbo-Croatian occupy specifier positions and undergo XP movement. If they
were heads, the movement of the dative clitic across the complementizer da would
disobey the Head Movement Constraint.
Stjepanović (1999) evokes Progovac’s (1993) data and notices that if an embedded
subjunctive clause contains two pronominal clitics, only the higher one may climb to
the matrix clause. This is shown in (35), where only the dative clitic may raise.
(35) a. Marija želi da mu ga predstavi

Marija wishPRES.3SG that himCL.DAT himCL.ACC introducePRES.3SG
“Marija wants to introduce him to him”
b. ?Marija mu želi da ga predstavi
Marija himCL.DAT wishPRES.3SG that himCL.ACC introducePRES.3SG
c. *Marija ga želi da mu predstavi (S-C, Stjepanović 1999)
The contrast between (35b and c) proves that the dative clitic is located higher than the
accusative clitic. According to Stjepanović, the accusative clitic may not move across
the dative clitic, because this would lead to a violation of the Relativized Minimality
condition.
Summarizing, the Serbo-Croatian examples discussed in this section suggest that
the pronominal clitics target different XP-projections. There is no clitic doubling in this
language, which means that the pronominal clitics are pronouns, rather than spell-outs
of case or agreement features. They differ from full pronouns in their prosodic
requirements. However, just as pronouns and other object DPs they must be generated
in argument positions and receive Θ-roles. I propose that they raise from their
argument positions to specifiers of agreement projections in order to check case: the
dative clitic moves to Spec, AgrIOP to check dative case, and the accusative clitic
moves to Spec, AgrOP to check accusative case. Their arrangement is sketched in the
phrase structure in (36).92
(36)
TP
T’
T AgrIOP
Dat.Cl. AgrIO’
Aux
AgrOP
Acc.Cl. AgrO’
Aux
je PartP
The skeleton correctly predicts that each of the pronominal clitics may be affected by
deletion, as each of them is a constituent on its own (cf. the examples in 29). It also
implies that if the pronominal clitics are part of an embedded clause, only the higher
clitic may raise from Spec, AgrIO and move to the main clause (cf. the example in 35b).
Climbing of the lower clitic would lead to a violation of Relativized Minimality
condition. Since the dative clitic is an XP-element, the movement will observe the Head
Movement Constraint.
92The third person singular auxiliary je heads its own projection AuxP, and is located lower than
the remaining auxiliaries. I account for its exceptional distribution in the South Slavic languages
in section 4.4.4.1. The structure in (36) follows general ideas concerning cliticization in Serbo-
Croatian presented in Stjepanović (1999) and Bošković (2001).
4.4.1.3 The second position effect in Serbo-Croatian:

syntax or phonology?
The preceding sections have discussed the syntactic organization of the clitics in Serbo-
Croatian. It was claimed that each of the pronominal clitics undergoes XP-movement
and targets the specifier of the relevant Agreement Phrase. Moreover, it was shown that
the clitics are not selective about the category of their host. The only requirement
concerns their linear position in the structure, as they must immediately follow the first
element in the clause. This raises important questions concerning the properties of the
initial element: is it a syntactic constituent? If so, does it move to this position or is it
base-generated there? And more generally, what is the motivation for the second
position requirement – is it dictated by rules of syntax or prosody?
The next sections will provide tentative answers to these questions. Section
4.4.1.3.1 will investigate properties of the elements that lend support to the clitics. It
will be argued that although the clitics may seem to follow an incomplete constituent,
the placement of the initial element is always the result of a syntactic operation. Section
4.4.1.3.2 will discuss potential reasons for Wackernagel’s position requirement.
4.4.1.3.1 Apparent splitting of syntactic constituents by clitics in Serbo-

Croatian
Even though the clitics in Serbo-Croatian must appear in the second position in the
clause, it is not always entirely clear what counts as the “first position”. Tomić (1996a:
817) remarks that the first element that supports the clitics may be the first constituent
of the clause (cf. 37a), the first phrase of the first clausal constituent (cf. 37b), or the
first phonological word (cf. 37c).
(37) a. Veoma lepu haljinu si mi kupio

very beautifulACC dressACC beAUX.2SG meCL.DAT buyPART.M.SG
b. Veoma lepu si mi haljinu kupio
c. Veoma si mi lepu haljinu kupio
“You’ve bought me a very beautiful dress” (S-C, Tomić 1996a: 817)
The ability to split constituents is a problematic characteristic of the Serbo-Croatian

clitics, which has led some researchers to propose that the clitic placement is
determined by phonology. For example, Radanović-Kocić (1988) observes that
‘heaviness’ of a constituent may influence the position of a clitic cluster. She argues that
in the case of DP-V sentence initial orders, clitics can follow the verb, such as voleo in
(38) if the DP is “heavy” (cf. 38a and b), but not if it is “light” (cf. 38c).
(38) a. Taj čovek voleo je Mariju

that man lovePART.M.SG be3SG Marija
“That man loved Mary”
b. Petar Petrović voleo je Mariju
Petar Petrović lovePART.M.SG be3SG Marija
“Petar Petrović loved Mary”
c. *Petar voleo je Mariju
Petar lovePART.M.SG be3SG Marija (S-C, Zec & Inkelas 1990)
Radanović-Kocić claims that the notions of “heaviness or “lightness” cannot be

defined syntactically, so the position of the clitic clusters in examples such as the ones
in (38) may not be explained by syntactic means. However, Zec & Inkelas (1990) and
Schütze (1994) argue that the elements preceding the clitics in (38a and b) are in fact
topicalized or left-dislocated, and involve comma intonation. This implies that the
elements taj čovek ‘this man’ and Petar Petrović are external to the clause, and this is also
the reason why Wackernagel’s law is exceptionally not observed.
A more compelling argument for a phonological approach concerns the possibility
of clitic placement after the first prosodic word; that is, inside a syntactic constituent.
Some examples were already given in (37). Below I quote sentences in which the
auxiliary clitic is placed between a demonstrative and a noun.
(39) a. Taj čovjek je volio Milenu

this man beAUX.3SG lovePART.M.SG Milena
“That man loved Milena”
b. Taj je čovjek volio Milenu
this beAUX.3SG man lovePART.M.SG Milena (S-C, Bošković 2001: 12)
Halpern (1992) assumes that in (39b) the clitic appears at the beginning of the clause in
syntax. Since this position is prohibited by phonological constraints, it is moved after
the first stressed word through the operation of “prosodic inversion”, which preposes
taj in front of je at PF.
Wilder and Ćavar (1994) and Progovac (1996) challenge Halpern’s assumptions
and argue that the demonstrative is separated from the noun as the result of left-branch
extraction, which is widely available in the Serbo-Croatian syntax.93 In fact,
demonstratives in the Slavic languages are morphologically adjectival, so it is most likely
that the clitic in (39b) is positioned after a phrase that has been extracted from a larger
constituent. This proposal gains more support from the fact that the noun and its
modifier can be split by non-clitic material, such as the object Milena and the verb
voljela/voli in (40).
(40) a. Togi je Milena voljela ti čovjeka

that beAUX.3SG Milena lovePART.F.SG man
“Milena loved that man”
b. Tog Milena voli čovjeka
that Milena love3SG man
“Milena loves that man” (S-C, Bošković 2001: 12)
Moreover, Progovac (1996) argues that clitics may not follow the first word, unless this
word is a constituent that is independently able to undergo syntactic movement. This
implies that there are prosodic words which can never support clitics. This is indeed the
case, and is exemplified by prepositions, which may never be clitic hosts (cf. 41b), even
when they are contrastively stressed (cf. 41c).94
93See Bošković (2005) for an extensive discussion of the left-branch extraction contexts in Slavic.
94Progovac (2005a: 138-139) tentatively suggests that the reason why prepositions must always
immediately precede nouns is related to case assignment, which always takes place under
adjacency.
(41) a. Na sto ga ostavi

on table itCL.ACC leave
“Leave it on the table”
a’. *Na ga sto ostavi
b. Prema Miodragu ga je Marija bacila,
toward MiodragDAT itCL.ACC beAUX.3SG Marija throwPART.F.SG
a ne od njega
and not from heGEN
“Marija threw it toward Miodrag, and not away from him”
b’. *Prema ga je Miodragu Marija bacila, a ne od njega
(S-C, Progovac 2005a: 137)
The restriction on the position of clitics in (41) cannot be due to prosodic reasons, but
it rather follows from an independent syntactic principle which prohibits prepositions
from being displaced from the complement NPs in Serbo-Croatian.
Furthermore, there are also other contexts in which the placement of clitics can be
given only a syntactic, but not a prosodic account. The pair in (42) contains complex
NPs in clause initial position.
(42) a. [NP Roditelji uspešnih studenata] su se razišli

parents successfulGEN studentsGEN beAUX.3PL REFL dispersePART.M.PL
“The parents of the successful students dispersed”
b. *Roditelji su se uspešnih studenata razišli (S-C, Progovac 1996: 418)
The contrast between the two sentences in (42) indicates that clitics may only follow
the full NP, but not the first word. This is unexpected if the clitic placement is
motivated prosodically, because the head noun roditelji is a stress-bearer. However, from
a syntactic point of view there is nothing exceptional about the ill-formedness of (42b),
given that roditelji can neither raise independently, nor can it be questioned.
(43) a. *Roditelji su se razišli uspešnih studenata

parents beAUX.3PL REFL dispersePART.M.PL successfulGEN studentsGEN
b. *Ko su se uspešnih studenata razišli?
who beAUX.3PL REFL successfulGEN studentsGEN dispersePART.M.PL
(S-C, Progovac 1996: 418)
Summarizing, it seems that there are serious empirical problems with phonological
accounts of clitic placement (see also Bošković 2001 ch. 2; 2005, and Progovac 2005a
for a more extensive discussion), which do not arise if syntactic solutions are adopted.
It has been shown that all of the movements that have been proposed to take place in
phonology can be given a straightforward syntactic explanation. This is a welcome
result, because it is theoretically problematic to suggest that clitics raise from
syntactically defined positions to phonologically defined landing sites. It is also
unnecessary to equip the phonological component of grammar with movement
operations that have never been ascribed to PF.
4.4.1.3.2 Motivation for the second position requirement

The previous section has proved that phonology cannot govern the placement of the
elements that precede clitics in Serbo-Croatian. However, it is still necessary to show
why clitics must target Wackernagel’s position. Is the movement triggered by their
phonological deficiency, or does it take place for feature checking? This is a

controversial issue, and I am not in the position to provide a conclusive solution, but at
least I am able to give directions for future research. I will also come back to the
problem in section 4.4.4.2 at the end of this chapter.
According to some approaches, the requirement that clitics appear in the second
position is motivated syntactically. For instance, Franks (2000: 17-21) claims that
Wackernagel’s law follows from the hypothesis that all languages are verb second (V2)
at some level of representation. This is possibly a property of Universal Grammar. The
actual crosslinguistic differences boil down to the position where the finite verb is spelt
out. Therefore, any account of V2 should be valid for the second position clitic
placement. Since V2 is normally analysed as movement of the finite verb into the
highest functional head in the phrase structure (cf. Den Besten 1989, Holmberg and
Platzack 1995, Zwart 1997, and many others), the second position clitics should target
the same projection as V2. However, there are important differences between V2
languages and Wackernagel languages such as Serbo-Croatian. For example, clitics are
phonologically deficient, whereas finite verbs in V2 languages are not. Moreover, the
second position requirement in V2 languages concerns only finite verbs. In the Slavic
languages, the requirement holds not only for the auxiliary clitics, but also for the
pronominal clitics, which have to move to the Wackernagel position even when they
are not accompanied by an auxiliary or any other finite verb, as shown in (44).95
(44) a. Sada ga Nada gleda

now himCL.DAT Nada watchPRES.3SG
“Nada is watching him now”
b. Da ga sada kupi Nada
that himCL.DAT now buyPRES.3SG Nada
“That Nada is buying it now” (S-C, Fontana 1997: 208)
Therefore, it is difficult to pinpoint a syntactic feature that drives the movement,

because the pronominal, auxiliary, and interrogative clitics do not form a uniform
semantic class, and yet they all climb and cluster together.
It is certain that the phonological deficiency of the clitics cannot be the reason for
their movement, because they can be suitably supported in their base positions
(Progovac 1996: 425ff). However, Bošković (2000) argues that although the elements
that precede clitics reach this site via a syntactic operation, the second position
cliticization is phonological in nature. His main argument for this statement comes
from sensitivity of clitic placement to intonation boundaries. He shows that clitics may
not occur after a comma boundary (cf. 45a), so when it is present, they seem to appear
in the “third” position (cf. 45b).
95 Henk van Riemsdijk (p.c.) remarks that on a distributed spell-out account it might be possible
to argue that in the Wackernagel clitic languages finite verbs raise together with clitics to the
second position, but for reasons of prosody, only the clitics are pronounced there, whereas the
verb is spelt out lower. However, if this idea is adopted, it is still necessary to specify the prosodic
principles that prohibit the pronunciation of the finite verb in the Wackernagel position.
(45) a. *Svojim rodjacima po majci, će Rada prodati knjige

self’s relatives after mother willCL.3SG Rada sellINF books
b. Svojim rodjacima po majci, Rada će prodati knjige
“To his/her maternal relatives, Rada will sell the books”
(S-C, Progovac 2005a: 140)
Consequently, Bošković (2001: 65) argues that clitic placement in Serbo-Croatian

should be defined in prosodic terms, as the clitics must appear in the second position of
their intonational phrase. This statement, however, is far from uncontroversial. For
instance, Progovac (2000) claims that intonation patterns often reflect syntactic
structure. If this is the case, the constituent followed by the comma boundary in (45)
might be external to the rest of the clause, so that Rada in (45b) is actually the first
constituent, which lends support to the Wackernagel clitic će. This hypothesis is
confirmed by the fact that when comma intonation after the preposed phrase is absent,
the clitic can be inserted immediately following the first word in this phrase, as shown
in (46).
(46) Svojim će rodjacima po majci Rada prodati knjige

self’s willCL.3SG relatives after mother Rada sellINF books
“To his/her maternal relatives, Rada will sell the books” (no pause)
(S-C, Progovac 2000: 254)
Summarizing, this section has presented an analysis of cliticization in a Wackernagel

clitic language, exemplified by Serbo-Croatian. It has been shown that although the
clitics must always appear adjacent to each other, they land in separate projections: the
auxiliary clitics raises to T or Aux, whereas the pronominal clitics target specifiers of
agreement projections. They must always appear in the second position, but it is not
clear whether this requirement is dictated by syntax or phonology.
4.4.2 Macedonian
The following section will provide an analysis of cliticization in Macedonian. Unlike
Serbo-Croatian, Macedonian is not a Wackernagel position language, but it requires that
clitics be verb-adjacent. The patterns of cliticization in Macedonian are fairly complex,
but this section will offer some new insights into the issue. Section 4.4.2.1 will present
the clitic paradigm, which will be followed by a description of clitic doubling in section
4.4.2.2. Section 4.4.2.3 will show that the type of cliticization is related to the case-
assigning ability of the host. A new analysis of the cliticization patterns will be given in
section 4.4.2.4

Macedonian distinguishes only three cases on pronouns: nominative, dative and
accusative. Just as in the other South Slavic languages, they have both clitic and non-
clitic variants.
singular plural
Acc (full/clitic) Dat (full/clitic) Acc (full/clitic) Dat (full/clitic)
1 mene/me mene/mi nas/nè nam/ni
2 tebe/te tebe/ti vas/ve vam/vi
3M nego/go nemu/mu niv/gi nim/im
3F nea/ja nejze/ì
REFL sebe (si)/se sebe (si)/si
(Mac, Franks & King 2000: 71)
Even though Macedonian has lost most of the morphological cases on nouns, it still
distinguishes nominative, oblique, and vocative on non-clitic DPs. Admittedly, the non-
nominative variants are always optional; they are found only with masculine human
nouns, and there is a tendency to eliminate them altogether (cf. Friedman 2002: 263),
but the loss of nominal declension in Macedonian is not as complete as it is in
Bulgarian. This is also evident in the clitic paradigm: in Macedonian the dative full
forms are still part of colloquial speech, whereas they are considered archaic in
Bulgarian. The dative can be also replaced by the preposition na ‘to’ followed by the
accusative full forms.
The present tense copula/auxiliary forms of the verb ‘to be’ are clitics, whereas the
past tense forms are not.
(48) The copula/auxiliary forms of the verb ‘to be’
Present tense clitic forms Past tense forms

SG PL SG PL
1 sum sme bev bevme
2 si ste beše bevte
3 (e) (se) beše bea
(Mac, cf. Franks & King 2000: 71; Friedman 2002: 281)
The distribution of the 3rd person singular and plural clitic forms depends on whether
they function as an auxiliary or as a copula. As shown in (49), the copula is
morphologically expressed, whereas the auxiliary is not. For this reason, e and se are
given within parentheses in the paradigm in (48).
(49) a. Mu *(e) skinato paltoto

himCL.DAT be3SG tornPASS.N.SG coatN+the
“His coat is torn”
b. Mu se (*e) skinalo paltoto
himCL.DAT REFL beAUX.3SG tornPART.N.SG coatN+the
“Reportedly, his coat got torn” (Mac, Tomić, forthcoming)
Furthermore, the clitic template for Macedonian differs slightly from the one
representing the other South Slavic languages in (15), as both the singular and plural
clitic forms of the verb ‘to be’ occur at the end of the cluster, following all the other
clitics.
(50) li Mod > AUX (except 3rd SG&PL) > DAT > ACC > e/se
(Mac, Franks & King 2000: 81)
4.4.2.2 Clitic doubling

Macedonian requires objects to be doubled by clitics, but the environments in which
the doubling occurs are different for direct and indirect objects. Direct objects are
obligatorily doubled when they are definite, therefore the object šefot ‘the boss’ in (51)
must be doubled by the pronominal clitic go.
(51) *(Go) vidov šefot

himCL.ACC seePAST.3SG boss-the
“I saw the boss” (Mac, Tomić, forthcoming)
Indirect objects are doubled only if they are specific. Therefore, if the same definite
noun šefot appears as the indirect object, it is doubled only when it has a specific
interpretation, as in (52a).
(52) a. Mu dadov cveќa na šefot

himCL.DAT givePAST.1SG flowers to boss-the
“I gave flowers to (the specific person who is) the boss”
b. Dadov cveќa na šefot
givePAST.1SG flowers to boss-the
“I gave flowers to (whoever is) the boss” (Mac, Tomić, forthcoming)
Clitic doubling is required not only for full object DPs, but also for strong pronouns,
such as nego in (53). This is to be expected, given that pronouns are always definite.
(53) a. Ljupka go poznava nego

Ljupka himCL.ACC knowPRES.3SG himACC
“Ljupka knows him”/“It is him that Ljupka knows”
b. *Ljupka poznava nego
c. Ljupka go poznava
Ljupka himCL.ACC knowPRES.3SG
“Ljupka knows him” (Mac, cf. Berent 1980; L. Grujoska p.c.)
Nego is a strong pronoun, which can be used to indicate focus or emphasis (cf. 53a). It
must occur with the corresponding doubled clitic go (cf. 53b). However, the accusative
clitic go can also function as the object of the verb, but then its reference is interpreted
as any noun of the same gender and number in a given context.
DPs and strong pronouns can be used as objects of prepositions, and then they are
not doubled by clitics, even when they are definite (cf. 54a). The only exception is the
preposition na ‘to’, which always occurs with dative objects (cf. 54b).96
96 In section 4.4.2.4.2 I propose that clitic doubling is a means of case checking. In this scenario
clitic doubling is not required in (54a and b) because case is assigned to the object by the
preposition. The fact that it is obligatory in (54c) shows that na is not a preposition, but rather an
alternative case realization (cf. section 2.3.3.2.1.4 in chapter 2).
(54) a (*Im) zboruvav za decata

themCL.DAT talkPAST.1SG about children-the
“I was talking about the children”
b. Im zboruvav na decata
themCL.DAT talkPAST.1SG to children-the
“I was talking to the children” (Mac, Berent 1980: 174)
So far it has been standardly assumed that the doubled clitics are in AgrIO and AgrO
(cf. Rudin 1997, Tomić 2000, forthcoming). I will present an alternative account of
clitic doubling in Macedonian in section 4.4.2.4.2. Before I do that, I will examine the
way Macedonian clitics interact with different clitic hosts.
4.4.2.3 Direction of cliticization in Macedonian

The patterns of cliticization in Macedonian are fairly complicated. Depending on the
type of hosts, clitics are proclitics (cf. section 4.4.2.3.1) or enclitics (cf. section 4.4.2.3.2).
They are unable to cliticize on certain categories, and then they appear in postposition
(cf. section 4.4.2.3.3).
4.4.2.3.1 Proclisis
Proclisis is required if the clitic host is instantiated by a finite (tensed) verb, such as
raduvame in (55), or an l-participle, such as dale in (56). As indicated in (55/56a),
proclitics do not have to appear in the second position. Moreover, they do not need to
be preceded by any overt material and can freely occur clause initially (cf. 55/56b).
However, they must be immediately left-adjacent to their host. Placement of an adverb
or any other category between the clitics and the finite verb (cf. 55c) or the l-participle
(cf. 56c) results in ungrammaticality. Furthermore, since they are proclitic, they may
only precede the tensed verb and the l-participle (cf. 55d and 56d), and in contrast to
Bulgarian (recall the data concerning participle fronting in Bulgarian in chapter 2)
neither of the verbs may move in front of the clitics. Yet, the subject or any other non-
verbal form may precede the clitics, as in (55a and b) and (56a and b).
(55) a. (Nie) mnogu si se raduvame na vnučevo

we very much REFL.DAT REFL.ACC rejoicePRES.1PL to grandson-thePROX
“This grandson of ours is giving us a lot of pleasure”
b. (Nie) si se raduvame na vnučevo mnogu
c. *Nie si se mnogu raduvame na vnučevo
d. *Raduvame si se mnogu na vnučevo (Mac, Tomić 1999: 10)
(56) a. (Vie) včera ste im go dale proektot

youPL yesterday beAUX.2PL himCL.DAT itCL.ACC givePART.PL project-the
“As reported, you gave them the project only/already yesterday”
b. (Vie) ste im go dale proektot včera
c. *Vie ste im go včera dale proektot

d. *Dale ste im go včera proektot (Mac, Tomić 1999: 10)
4.4.2.3.2 Enclisis
Clitics must encliticize on imperatives (cf. 57) and gerunds (cf. 58). Proclisis on these
categories is prohibited, even when the clitics are supported by another word to their
left (cf. 57/58b).
(57) a. Penkaloto kupuvaj mi go!

Pen-the buyIMPV.2SG meCL.DAT itCL.ACC
“Buy me the pen!”
b. *Penkaloto mi go kupuvaj! (Mac, Franks 1998)
(58) a. Nemarno pišuvajќi go pismoto,...

carelessly writingGER itCL.ACC letter-the
“Carelessly writing the letter, …”
b. *Nemarno go pišuvajќi pismoto (Mac, Franks & King 2000: 84)
In Standard Macedonian the pattern does not change in the presence of negation, as
exemplified in (59a) for imperatives and in (59b) for gerunds. However, some speakers
of the Skopje dialect allow proclisis on imperatives in the presence of negation, but only
in colloquial speech.
(59) a. Ne (%mi go) nosi mi go!

NEG meCL.DAT itCL.ACC bringIMPV.2SG meCL.DAT itCL.ACC
“Don’t bring it to me!”
b. Ne (*mi go) donesuvajќi mi go toa na vreme,...
NEG meCL.DAT itCL.ACC bringingGER meCL.DAT itCL.ACC that on time
“Not bringing it to me on time…” (Mac, cf. Franks & King 2000: 83-84)
4.4.2.3.3 Postposition
The third strategy of clitic placement in Macedonian, which I will refer to as
“postposition” following Alexandra Cornilescu’s suggestion (p.c.), applies in copula
constructions with non-verbal predicative XPs, such as DPs (cf. 60 and 61), APs (cf.
62), and passive participles (cf. 63). In these environments clitics require a phonological
host to their left, which can be either a single word or a phrasal element (cf. 61b).97
(60) a. *Mu e tatko (na deteto)

himCL.DAT be3SG father to child-the
b. Tatko mu e na deteto
father himCL.DAT be3SG to child-the
“He is the father of this child (so he has to take care of him)!”
97 All native speakers reject clause-initial clitics in the presence of nouns (cf. 60). The distribution
of clitics with adjectives and passive participles is subject to speaker and dialectal variation. In the
Western dialects clitics may both precede and follow adjectives and passive participles. The latter
strategy prevails in the Eastern dialects (Olga Tomić, p.c.).
c. Toj mu e tatko
he himCL.DAT be3SG father
“He is his father (and not anyone else)!”
(Mac, cf. Tomić 2000: 295-296; L. Grujoska, p.c.)
(61) a. *Si ubava žena

be2SG pretty woman
b. Ubava žena si
c. Ubava si žena
d. Ti si ubava žena
You be2SG pretty woman
“You are a pretty woman” (Mac, Franks & King 2000: 86-87)
(62) a. ??Si mu mil

be2SG himCL.DAT dearM.SG
“He likes you”
b. Mil si mu
“You are dear to him”
c. Ti si mu na našion sin mnogu mil
you be2SG himCL.DAT to our-thePROX son very dearM.SG
“You are very dear to our son!” (Mac, Tomić 2001a: 664)
(63) a. ?Mu e rečeno da bide točen poveќe pati

himCL.DAT be3SG tellPASS.N to beSUBJ.3SG punctual more times
“He was told to be punctual more than once”
b. Rečeno mu e da bide točen poveќe pati
tellPASS.N himCL.DAT be3SG to beSUBJ.3SG punctual more times
c. Na Petreta mu e poveќe pati rečeno
to PeterDAT himCL.DAT be3SG more times tellPASS.N
da bide točen
to beSUBJ.3SG punctual
“Peter was told to be punctual more than once” (Mac, Tomić 2000: 296)
Unlike in the contexts with finite verbs and the l-participle, the clitics do not have to be
adjacent to the adjective (cf. 62d) or the passive participle (cf. 63c), as they can be
separated from these categories by some overt material. Furthermore, the examples in
(61) show that the clitics may be preceded by a full DP (cf. 61b), an adjectival head (cf.
61c), or the subject pronoun (cf. 61d). Baerman & Billings (1998: 20) state that these
facts indicate that they do not procliticize or encliticize on these categories, and that
they need them only for phonological support.
The position of the postpositive clitics with respect to their hosts needs to be
addressed in detail. In the presence of adjectives and nouns, they always appear in the
second position (cf. Tomić 2000: 300). Yet, their distribution is more complex with the
passive participle, such as rečeno in (63). Sentence (63c) demonstrates that the clitic do
not have to be left-adjacent to it. Moreover, when they are preceded by elements other
than the passive participle, they need not appear in the Wackernagel position, and in
fact may occur rather low in the structure, as long as they are to the left of the participle
(cf. 64). Native speakers inform me that the phrases preceding the cluster mu e are not
separated by pauses, which indicates that the clitics may move quite freely in the clause.
(64) a. Na Petreta (mu e) od strana na komisijata

to PeterDAT himCL.DAT be3SG from side of commission-the
(mu e) poveќe pati (mu e) rečeno
himCL.DAT be3SG more times himCL.DAT be3SG tellPASS.N
da bide točen
“Peter was more than once told by the commission to be punctual”
(Mac, cf. Tomić 2000: 299)
b. Na Petreta (mu e) poveќe pati (mu e) jasno
to PeterDAT himCL.DAT be3SG more times himCL.DAT be3SG clear
i glasno (mu e) rečeno da dojde
and loudly himCL.DAT be3SG tellPASS.N to comeSUBJ.3SG
“Peter was loudly and clearly told to come more than once”
(Mac, cf. Franks & King 2000: 86; L Grujoska, p.c.)
However, when the passive participle is the clause initial element, the clitics must
immediately follow it, and appear in the second position.
(65) a. Rečeno mu e poveќe pati (*mu e)

tellPASS.N himCL.DAT be3SG more times himCL.DAT be3SG
da bide točen
“He was told to be punctual more than once” (Mac, O. Tomić p.c.)
Observe that the passive participle always agrees with the subject of the clause. The
sentence in (65) has an impersonal meaning; therefore the participle is specified for
singular neuter. In view of the contrast between (64) and (65), and the arrangement of
the clitics with nouns and adjectives, I would like to make the generalization that the
distribution of the postpositive clitics is quite free. However, when the clause opens
with a non-verbal predicative element that shows subject agreement (a DP, an AP, or a
passive participle), the clitics must appear in the second position.
Summarizing, it has been established that there are three ways in which clitics are
positioned with respect to their host in Macedonian. They must procliticize on l-
participles and tensed verbs. They are encliticized on imperatives and gerunds. They do
not cliticize on non-verbal predicative XPs, such as adjectives, nouns, and passive
participles, but must then occur in postposition.
4.4.2.3.4 The relation between types of cliticization and stress assignment

The present section will demonstrate that the patterns of cliticization described above
have a phonological reflex, as they may affect the distribution of lexical stress.
In Standard Macedonian word stress falls on the antepenultimate syllable (cf. 66a
and b). If a word has fewer than three syllables, the first syllable is stressed (cf. 66c).
The stressed syllables are capitalized in the examples below.
(66) a. DOnesi!
bringIMPV.2SG
“Bring!”
a’. doNEsuvaš
bringPRES.2SG
“You are bringing” (Mac, Rudin et al 1999: 551-552)
b. (TOJ) PROdal MNOgu JAbolka
he sellPART.M.SG many apples
“He has reportedly sold a lot of apples” (Mac, Tomić 2001a: 648)
Placement of clitics at the end of a word may have an influence on stress assignment.
For instance, in (67) the addition of each monosyllabic clitic after the imperative shifts
the stress rightward by one syllable.
(67) a. DOnesi!
“Bring!”
b. doNEsi goCL.ACC!
“Bring it!”
c. doneSI miCL.DAT goCL.ACC!
“Bring it to me!” (Mac, Rudin et al 1999: 551; Baerman & Billings 1998: 20)
The situation is less clear with gerunds, as there are additional complications concerning
discrepancies between the prescribed literary norm and the actual practice, which
follow from diachronic phonological reinterpretations. Franks (1998) reports that
gerunds have fixed lexical stress, which remains the same even when clitics are added.
He contrasts cases of imperatives, in which clitics always enter the calculation of stress
(cf. 68a) with gerunds (cf. 68b), which supposedly resist the stress shift even when
enclitics are added.98
(68) a. KupuVAJ miCL.DAT goCL.ACC!

“Buy it for me!”
b. KupuVAJќi miCL.DAT goCL.ACC ...
“Buying it for me…” (Mac, cf. Franks 1998)
However, Baerman & Billings (1998: 20) claim that the stress recalculation may affect
gerunds as well. This pattern is beyond the normative standard, but has been
recognized by normative linguists.
Proclisis, which occurs with l-participles and tensed verbs, does not affect stress
placement. Hence, the clitic go in (69) is not stressed.
(69) (JAS) sum ti go KAžal

I beAUX.1SG youCL.DAT itCL.ACC sayPART.M.SG
“I have told it to you” (Mac, Rudin et al 1999: 553)
A similar distribution is observed when clitics occur with non-verbal predicative XPs,
such as adjectives, nouns (cf. 70a), and passive participles (cf. 70b). Clitics never form
prosodic units with these categories, so they do not alter stress placement. This is to be
expected, because clitics do not cliticize onto these categories.
98The gerund in (68) is stressed on the penultimate syllable. Baerman & Billings (1998: 21) claim
that this is the result of the collapse of two vowels into a diphthong in the history of Macedonian:
ku.pu.VA.e. ќi → ku.pu.VAJ. ќi.
(70) a. TAtko si mu
father be2SG himCL.DAT
“You are his father (so you have to take care of him)!”
a’. *TatKO si mu
b. REčeno mu e da DOjde
tellPASS.N himCL.DAT be3SG to comeSUBJ.3SG
“He was told to come”
b’. *RečeNO mu e da DOjde (Mac, cf. Tomić 2001a: 664-665)
The pattern becomes more complex in the presence of negation. In (71) the negation
operator ne forms a single antepenultimately stressed unit with the clitics and the verb
to its right. As a result, the clitic gi in (71b) carries stress, too. This means that negation
is able to shift stress in the same way as imperatives do.99
(71) a. (TOJ) ne gi PROdava jaBOLkata

he NEG themCL.ACC sellPRES.3SG apples-the
“He is not selling the apples”
b. (TI) ne si mu GI dala jaBOLkata
you NEG beAUX.2SG himCL.DAT themCL.ACC givePART.F.SG apples+the
“Reportedly, you haven’t given him the apples” (Mac, Tomić 2001a: 649)
As can be expected, stress is shifted with negated imperatives as well, as shown for the
singular form in (72a) and the plural one in (72b).
(72) a. Ne daVAJ mu go!

NEG giveIMPV.2SG himCL.DAT itCL.ACC
“Don’t give it to him!”
b. Ne davajTE mu go!
NEG giveIMPV.2PL himCL.DAT itCL.ACC (Mac, cf. Tomić 2001b: 165)
To sum up, it has been demonstrated that only imperatives, negation, and (in some
registers) gerunds trigger stress shift in Macedonian. The presence of l-participles or
tensed verbs as clitic hosts leaves the stress arrangement unaffected. Correspondingly,
non-verbal predicative XPs do not alter the stress assignment, either. I take these facts
to be syntactically significant, and I will return to the issue in section 4.4.2.4.2.3.
4.4.2.4 Towards an analysis of cliticization in

Macedonian
The subsequent sections will develop an alternative analysis of cliticization and its
directionality in Macedonian, and is organized as follows. Section 4.4.2.4.1 will briefly
overview some previous accounts. Section 4.4.2.4.2 will present a new proposal, which
argues that cliticization is possible only with case-assigning hosts. Section 4.4.2.4.2.4 will
postulate a division of pronominal clitics in Macedonian into clitics proper and weak
forms.
99 Recall from chapter 2, section 2.3.6.3.3 that a similar pattern is observed in Bulgarian, in which
negation assigns stress to the element that follows it, even when it is a clitic.
4.4.2.4.1 A note on previous scholarship

There have been a few attempts to capture the direction of cliticization in Macedonian
in the literature. For example, Joseph (1983) argues that proclisis occurs on finite verbs.
However, there are a few problems with this generalization. It suggests a distinction
between finite and non-finite forms of verbs, even though contemporary Macedonian
does not have infinitives. Moreover, it stipulates that l-participles are finite, which
cannot be correct. Another problem is the fact that Joseph does not address the
impossibility of cliticization with passive participles, adjectives, and nouns (see also
Tomić 1997 for an overview of Joseph’s account).
Undoubtedly, the behaviour of the Macedonian clitics ties in with the
morphological properties of the different types of hosts. Tomić (1997) handles the
variation by decomposing the clitic hosts into feature pairs [±V] and [±N], as in (73).
(73) V N
tensed verbs + –
l-participles + –
passive participles + +
adjectives + +
nouns – +
Tomić claims that the direction of cliticization depends on the saliency of verbal
properties of the host. Hence, nouns, which are described as [–V] categories, may never
serve as hosts for proclitics (cf. 60). L-participles and tensed verbs have positive values
for V, negative for N, so clitics may procliticize on them (cf. section 4.4.2.3.1).
Adjectives and passive participles, which are specified for [+V, +N] features, exhibit
indeterminate distribution (cf. 61, 62, and 63). Only some speakers accept them as
potential hosts for proclitics.
Tomić’s insights with respect to the relation between degree of verbal properties
and cliticization are certainly correct. However, the problem is that she does not
provide a feature decomposition of gerunds and imperatives, which always trigger
enclisis (cf. section 4.4.2.3.2). Consequently, it is not clear how they fit into her system.
Moreover, she assumes that nouns, adjectives, and passive participles are clitic hosts. It
has been shown that they are not, because clitics do not have to be adjacent to these
categories.
Furthermore, Tomić describes the cliticization patterns, but as far as I can see, she
does not explain why they hold. For instance, it seems crucial to find a reason why
proclisis is possible only with tensed verbs and l-participles, and enclisis only with
imperatives and gerunds. Correspondingly, it is necessary to explain why cliticization
does not occur with non-verbal predicative XPs.
As far as the syntactic arrangement of clitics in the clause is concerned, Tomić
(1997, 2000) argues that they head relevant functional projections. For instance,
sentence (74a), which contains the l-participle dadela as the host, has the structure given
in (74b): the negative particle heads NegP, the modal clitic is generated in Mod, and the
auxiliary clitic resides in Tense/AgrS. The dative and accusative clitics head AgrIOP
and AgrOP, respectively.
(74) a. Ne ќe sum mu go dadela proektot

not willCL.MOD beAUX.1SG himCL.DAT itCL.ACC givePART.F.SG project+the
“Rumour has it, I would be unwilling to give the project to him”
(Mac, Tomić 2000: 301)
b. [Neg Ne [Mod ќe [T/AgrS sum [AgrIO mu [AgrO go [VP dadela [DP proektot]]]]]]]
If a clitic cluster contains the third person auxiliary clitic in the singular or in the plural,
it is hosted below all the other clitics in the head of AuxP (cf. 75).
(75) a Ќe mu e izpraznet stanot

willCL.MOD himCL.DAT beAUX.3SG emptyPASS.M.SG apartment+the
“His apartment will be vacated” (Mac, Tomić 2000: 302)
b. [Mod ќe [T/AgrS [AgrIO mu [Aux e [VP izpraznet [DP stanot]]]]]]
On Tomić’s approach, the first person singular auxiliary clitic projects a different head
in (74b) than the third person auxiliary clitic in (75b). The proposal reflects the order of
the clitics in the cluster, but the reason why the auxiliaries pattern in this way remains
unclear.
Tomić follows Bošković (1997 ch. 5), who argues that crosslinguistically participles
must check the [Aux] feature. This is done by head adjunction of the participle to a
Mood (Mod) or Tense (T) head. In the spirit of this proposal, Tomić (1999: 17)
suggests that the clauses with passive and past participles are derived as in (76b): the
passive participle rečeno raises as a head from VP and lands in T/AgrS.
(76) a. Rečeno mu e da bide točen poveќe pati

tellPASS.N himCL.DAT beAUX.3SG to beSUBJ.3SG punctual more times
“He was told to be punctual more than once”
b. [T/AgrS Rečenoi [AgrIO mu [Aux e [VP ti [XP da bide točen poveќe pati]]]]]
Notice that this derivation induces multiple Head Movement Constraint violations,
because the passive participle crosses head positions occupied by the clitics on its way
to T/AgrS.
Tomić (1997) assumes that if a clause contains predicative adjectives or nominals,
they are left-adjoined to Mod or Tense/AgrS and check the [Aux] feature. Accordingly,
she proposes the derivation (77b) for the sentence with the adjective mil in (77a).
(77) a. Mil mi e
dearM.SG meCL.DAT be3SG
“He is dear to me” (Mac, Tomić 1997)
b. [T/AgrS mili [AgrIO mi [Aux e [PredP ti]]]]
The derivation is problematic in a number of ways. First, nouns or adjectives should be

unable to check the [Tense] or the [Aux] feature, because they are not specified
morphologically for it. Second, as in the case of (76), the head movement of the
predicative element mil violates the Head Movement Constraint, as it crosses two heads
on its way. Finally, the head movement analysis of (77) is not on the right track, because
nouns that precede the clitic cluster can be premodified, as in (61b) repeated below in
(78), so they clearly undergo XP-movement.
(78) a. Ubava žena si

pretty woman be2SG
“You are a pretty woman!” (Mac, Franks & King 2000: 86-87)
4.4.2.4.2 An alternative analysis of cliticization in Macedonian

It has been established that clitic placement in Macedonian hinges on the type of host.
Clitics must procliticize on tensed verbs and l-participles, and encliticize on imperatives
and gerunds. They normally do not cliticize on non-verbal predicative XPs. In what
follows I will present a new analysis of the phenomenon, which is based on the
following two assumptions:
i) Cliticization is obligatory with those hosts that are able to assign case. Since both
finite verbs and l-participles are case assigners, they always trigger proclisis. Likewise,
imperatives and gerunds are also case-assigners. However, for reasons to be explained
in section 4.4.2.4.2.3, they induce enclisis.
ii) Pronominal clitics in Macedonian do not form a uniform class. They comprise weak
and clitic forms, in the sense of Cardinaletti and Starke (1999). Weak forms occur with
the elements that are unable to assign case, such as passive participles, adjectives and
nouns.
4.4.2.4.2.1 Proclisis with case-assigning verbs

Following some insights of Rudin’s (1997) and Tomić’s (2000, forthcoming) proposals,
I submit that clitics must cliticize by raising to Agreement projections. These
projections are located above the l-participle or the finite verb (cf. 79).
(79) [TP [T ste [AgrIO mu [AgrO go [VP dale proektot na Petko]]]]]

be2PL himCL.DAT itCL.ACC givePART.PL project-the to Petko
“You have reportedly given the project to Petko” (Mac)
Since proclisis occurs only in the presence of case assigning hosts, it seems natural to
assume that clitics check case features. This is what is normally proposed in the
literature. However, an alternative might be that clitics check the φ-features of the
objects. Let me briefly explain why the latter option should be rejected.
Berent (1980: 174-175) observes that Macedonian distinguishes a class of neuter
diminutive forms which are derived from feminine nouns referring to female human
beings, such as ženče ‘little woman’ (derived from žena), devojče ‘little girl’ (derived from
devojka), ќerkiče ‘little daughter’ (derived from ќerka). The basic (non-diminutive) variants
are doubled by feminine clitics, as shown in (80) for direct and indirect objects.
(80) a. Ja sakam devojkata

herCL.ACC likePRES.1SG girl-the
“I like this girl”
b. Ì dadov podarok na devojkata
herCL.DAT givePAST.1SG gift to girl-the
“I gave a gift to the girl” (Mac, Berent 1980:175)
If a sentence contains a corresponding neuter diminutive, there are two possibilities.

There are no neuter clitic forms available (cf. the chart in 47), so in Standard
Macedonian neuter nouns are doubled by the masculine clitic go (cf. 81a). However,
some speakers employ the natural gender with the neuter indirect object forms and
double the diminutive indirect object by using the feminine dative clitic ì, as in (81b).
(81) a. Go sakam devojčeto

himCL.ACC likePRES.1SG little-girl-the
“I like the little girl”
b. Ì dadov podarok na devojčeto
herCL.DAT givePAST.1SG gift to little-girl-the
“I gave a gift to the little girl” (Mac, Berent 1980:175)
This shows that the doubled clitic cannot possibly check the φ-features of the object for
two reasons. First, there are no neuter clitics to check the φ-features of neuter objects.
Second, native speakers may apply clitic doubling even when there is a feature
mismatch between the clitic and its associate (cf. 81).
Moreover, notice that doubling of the accusative clitic is impossible with passive
participles, even though it is required in the constructions with the l-participle (cf. 82b).
Both the passive and the l-participle carry the same set of φ-features, but passive
participles are unable to assign accusative case, and thus are incompatible with clitic
doubling (cf. 82a).
(82) a. (*Gi) bea naredni na tezgata

themCL.ACC bePAST.3PL placePASS.PL on counter+the
čašite za vino i rakija
glasses+the for wine and rakija
“The glasses for wine and rakija were placed on the counter”
b. *(Gi) bea naredile čašite
themCL.ACC bePAST.3PL placePART.PL glasses+the
za vino i rakija na tezgata
for wine and rakija on counter+the
“The glasses for wine and rakija had been placed on the counter”
(Mac, L. Grujoska, p.c.)
Hence, the contrast between (82a) and (82b) conclusively shows that clitic doubling is
related to case checking.
With this assumption in mind, let me spell out my analysis in detail. Recall from
section 4.4.2.1 that Macedonian has largely lost morphological case on nouns.
However, case distinctions are uniformly retained on pronominal clitics, which are the
only elements that show a full case inflection paradigm. Following Belletti’s (1999)
seminal analysis of cliticization in Italian, I will assume that this means that clitics bear a
strong case feature, which must be checked syntactically (via movement). The case
checking occurs under the Spec-head relation, so the clitics must move along with their
associates (that is, the direct and indirect objects) to agreement projections.
Furthermore, recall from section 4.4.2.2 that clitic doubling in Macedonian is
contingent on definiteness or specificity of the doubled object. I will take it to mean
that only DPs can be doubled and not NPs. Correspondingly, following Uriagereka
(1995), I will assume that clitics are D-heads, and that they constitute a DP together
with their associates. This means that the base form of the sentence in (79), repeated
for convenience in (83a), is as in (83b).100
(83) a. Ste mu go dale proektot na Petko

be2PL himCL.DAT itCL.ACC givePART.PL project-the to Petko
“You have reportedly given the project to Petko” (Mac)
b.
TP
T’
T AgrOP
AgrO’
AgrO VP
DPdat
V’
D DPdat
V DPacc
D DPacc
muDAT na Petko dale goACC proektot
The finite verb or the l-participle raises to AgrO. The movement of the verb creates
equidistance, and as a result the direct object may move and check accusative case in
Spec, AgrOP.
100 Observe that in contrast to Bulgarian and Serbo-Croatian, in Macedonian the l-participle dale
targets a head position. I will account for the lack of the XP-movement at the end of section
4.4.2.4.2.2.
(84)
TP
T’
T AgrOP
DPacc
AgrO’
clACC DP
V VP
DPdat V’
clDAT DPdat tV tDPacc
Next, the verb continues to move from AgrO to AgrIO by head movement. This
creates equidistance and the indirect object DP may move to Spec, AgrIOP to check
dative case. Finally, the verb moves to T.
(85)
TP
subject T’
T AgrIOP
AgrIO’
DPdat
AgrIO
clDAT DPdat AgrOP
AgrO AgrIO
DPacc AgrO’
clACC V AgrO AgrO VP

clACC DPacc
tDPdat V
tV
However, the clitics must still procliticize onto the verb in T. Why does this happen? A
number of proposals have been made in the literature. According to Belletti (1999:
550), this is due to PF considerations: Agreement projections are not strong heads, so
they may not contain any material that needs a PF interpretation. Therefore, they must
be emptied before Spell-Out. In Nash and Rouveret’s (2002: 177) view, the proclicis on
T occurs because clitics must raise onto a “substantive” (lexical) category endowed with
active φ-features. Independently of these proposals it has been claimed (cf. Bošković
2002) that clitics in Macedonian and Bulgarian cliticize by adjoining to a single head
(such as T in 85). In this way they contrast with Wackernagel clitics in Serbo-Croatian,
which target the specifiers of the relevant Agreement projections, and never cluster in a
single head (cf. section 4.4.1.2). Some evidence for this idea will be given in section
4.4.3.4.3 on the basis of Bulgarian. The subsequent section will present more
supportive arguments, which have not been raised in the literature so far. They will
follow from the fact that Macedonian observes the Person Case Constraint.
4.4.2.4.2.2 Explaining the Person Case Constraint

The derivation in (85) indicates that the l-participle head-adjoins to T, together with the
pronominal clitics. This section will explain the mechanics of this process and will
demonstrate that by adjoining to T, the pronominal clitics check the φ-features of T.
Suitable evidence for this idea comes from the observance of the Person Case
Constraint (PCC) in Macedonian.
The PCC was first described by Perlmutter (1971), who noticed a restriction in the
occurrence of pronominal clitics in ditransitive constructions: if an accusative clitic co-
occurs with a dative clitic, the accusative must carry the 3rd person feature. The
constraint is attested only with weak elements, such as clitics, weak pronouns, and
agreement affixes. Moreover, it does not hold for constructions without an external
argument, such as unaccusatives or passives. The constraint is illustrated by means of
the Swiss German data in (86) that Anagnostopoulou (1999) attributes to Henk van
Riemsdijk.
(86) a. D’Maria zeigt mir en

the Maria shows to me him
“Mary shows him to me”
b. *D’Maria zeigt em mich
the Maria shows to him me
“Mary shows me to him” (Swiss German, Anagnostopoulou 1999: 267)
Example (86a) is grammatical, because the 1st person dative clitic is accompanied by the
3rd person accusative clitic. However, the sentence in (86b) is ill-formed, because the 3rd
person dative co-occurs with the 1st person accusative.
The effects of the constraint have been observed in a number of unrelated
languages, and according to Bonet (1994), the constraint is universal. Bonet’s
assumption is incorrect, though, because the PCC is not operative in Serbo-Croatian
(cf. 87), Czech (cf. 88) and Polish (cf. 89), where non-3rd person accusative clitics may
co-occur with dative clitics.101
(87) Ja im te preporučujem
I themCL.DAT youCL.ACC recommendPRES.1SG
“I am recommending you to them” (S-C, N. Milićević, p.c.)
(88) Jestliže mu vás předám živou a zdravou

if himCL.DAT youCL.ACC.PL bringPRF.PRES.1SG alive and healthy
“If I can bring you to him safe and sound” (Czech , Lenartová 2001)
(89) a. Dał-bym mu cię za żonę

givePART.M.SG+COND.1SG himCL.DAT youCL.ACC for wife
bez wahania
without hesitation
“I would give you to him as a wife without hesitation”
(Pl, cf. Cetnarowska 2003)
Rivero (2005: 1093) notices that the PCC is observed in Bulgarian. Thus, a non-3rd
person accusative clitic is incompatible with a dative clitic, as shown in (90a). The dative
clitic accepts only accusative clitics marked for the 3rd person, as indicated in (90c). Yet,
if a strong form of the dative pronoun is used, the result is grammatical (cf. 90b).
101 See Anagnostopoulou (1999) for examples of other languages where the PCC does not hold.
(90) a. *Az im te preporŭčvam

b. Az im preporŭčvam na tjax
I themCL.DAT recommendPRES.1SG to themDAT
“I am recommending you to them”
c. Az im ja preporŭčvam
I themCL.DAT herCL.ACC recommendPRES.1SG
“I am recommending her to them” (Bg, cf. Hauge 1999)
I would like to point out the PCC is observed in Macedonian as well, as shown in (91),
which correspond to the Bulgarian examples in (90).
(91) a. *Jas im te preporačuvam

b. Jas te preporačuvam na niv
I youCL.ACC recommendPRES.1SG to themCL.ACC
c. Jas im ja preporačuvam
I themCL.DAT herCL.ACC recommendPRES.1SG
“I am recommending her to them” (Mac, L. Grujoska p.c.)
Following Anagnostopoulou (1999: 287ff), I will assume that the constraint is the result
of the incompatibility of person and number feature checking in the syntactic
configuration sketched in (92). Suppose that the head F contains number and person
features, which must be checked. The Person Case Constraint holds when a dative clitic
raises first from its base position within the VP in order to check a person feature of F,
whereas the accusative clitic moves second to check the remaining number feature on
F, tucking in beneath the dative clitic.102 The derivation converges only if the accusative
clitic carries just a number feature, and not a person feature, because the latter has
already been checked by the dative. On the assumption that the 3rd person pronouns
contain only a number feature, they are the only eligible candidates for the movement.
If a non-3rd person pronominal clitic raises, then the derivation will crash, because the
person feature on the accusative will remain unchecked.
102 Hans Broekhuis (p.c.) remarks that Anagnostopoulou’s account presupposes that the dative
clitic is unable to check the number feature on T, which might be problematic. However,
according to many proposals (cf. Taraldsen 1995, Boeckx 1997, Chomsky 2000) datives do not
enter into complete agreement and hence do not check the number feature, but only the person
feature of T. One of the examples that support this claim are quirky subjects constructions in
Icelandic, in which dative subjects do not agree in number with the verb, while nominative
objects induce number agreement. See Anagnostopoulou (1999: 275ff) for a detailed discussion.
Rivero (2005) observes that Bulgarian, which observes the PCC, patterns with Icelandic: the
singular dative subject na Ivan occurs with the verb marked for the plural, which agrees in person
with the nominative object.
(i) a. Na Ivan mu se zeluvaxa devojki

IvanDAT himCL.DAT REFL kissPAST.3PL girlsNOM
“Ivan felt like kissing girls”
b. *Na Ivan mu se zeluvaxme nie
IvanDAT himCL.DAT REFL kissPAST.1PL weNOM
“Ivan felt like kissing us” (Bg, Rivero 2005: 1095)
(92)
FP
F’
VP
clDAT F
clACC F
V clDAT clACC
Let us turn to Macedonian and see how the Person Case Constraint effect can be
derived in this language. I take the T head in (93) to be the equivalent of
Anagnostopoulou’s F head in (92), which contains a person and number feature.
Furthermore, I assume that clitics undergo head movement in Macedonian. This is
justified by the fact that they are D-heads.
Thus, in (93) the dative clitic raises first to T, in which the l-participle is located, in
order to check a person feature there. The accusative clitic will move second and will
tuck in beneath the dative to check the remaining number feature on T. However, as in
(92), the derivation will converge only if the accusative clitic carries just a number
feature, and not a person feature, because the latter has already been checked by the
dative. Consequently, only the 3rd person pronoun may move there, as it contains a
number feature and a null person feature. The derivation will crash if a non-3rd person
pronominal clitic raises, because the person feature on the accusative will remain
unchecked.
(93)
TP
subject T’
clDAT T AgrIOP
clACC T
AgrIO’
DPdat
AgrIO
clDAT DPdat AgrOP
AgrO AgrIO
DPacc
tV AgrO
clACC DPacc
After the clitic adjunction, the constituent in T is specified for 3rd person and either the
singular or plural number. The derivation converges if the feature specification of the
subject is the same. In case it is different than the 3rd person, it is necessary to insert the
1st or the 2nd person auxiliary.103 I suggest that the auxiliary is left-adjoined to the
complex head in T.104 This explains why auxiliaries in Macedonian are overt only in the
1st and the 2nd person (cf. 94a).
103 This implies that the subject by default carries the 3rd person feature, so in fact has a null
person feature. It has a non-3rd person feature only when the subject is realized by the first or
second person pronoun, and it is only then that an auxiliary is needed (H. Broekhuis, p.c.).
104 Additional support for the idea that the auxiliary and the pronominal clitics are left-adjoined to
the l-participle and form a complex head comes from the fact that in Macedonian the l-participle
always moves as a unit together with the clitics, for instance when it is raised to the left of the
question particle li. I assume that the complex head si+mu+gi+dal in (ia) is left adjoined to li.
(i) a. Si mu gi dal li parite?

beCL.2SG himCL.DAT them CL.ACC givePART.M.SG Q money-the
“Did you give him the money?”
b. *Dal si mu gi li parite? (Mac, Rudin et al. 1999: 544)
(94) a. Jas sum mu go dal pismoto

I beAUX.1SG himCL.DAT itCL.ACC givePART.M.SG letter-the
ne edno dete
to a child
“I gave the letter to a child”
b. Jana mu go dala pismoto ne edno dete
Jana himCL.DAT itCL.ACC givePART.F.SG letter-the to a child
“Jana gave the letter to a child” (Mac)
Finally, the subject may raise from Spec, vP to Spec, TP and check the φ-features of T.
The derivation proposed here describes compound tenses constructed with the l-
participle. However, in the case of simple tenses, formed with a finite verb such as
prodava in (95), the derivation will be the same up to the stage represented in (93), at
which the pronominal clitics adjoin to the l-participle in T. In the case of finite verbs,
however, there will be no need to generate an auxiliary in T, because unlike the l-
participle, they always specify tense and person features. Thus, the finite verb will move
directly to T, and the clitics will left-adjoin to it.
(95) (Toj) prodava jabolka

he sellPRES.3SG apples
“He sells apples” (Mac)
Summarizing, I have described the mechanism of proclisis on finite verbs and l-

participles. Before concluding the section, I will briefly explain the relevance of the
proposed derivation for l-participle fronting in Macedonian.
As was mentioned in chapter 2, section 2.1, l-participle movement across the
auxiliary ‘to be’ is impossible in Macedonian and results in strong ungrammaticality.
(96) a. (Jas) sum javil na pregled

(I) beAUX.1SG appearPART.M.SG for examination
“I have appeared for examination”
b. *Javil sum na pregled (cf. Mac, Friedman 1977)
In Migdalski (2005) I claimed that the ill-formedness of (96b) results from the proclitic
status of the auxiliary sum. Since the auxiliary does not need to be supported by overt
material to its left, participle movement is unnecessary, and hence prohibited. However,
if this explanation were correct, placement of any element in front of the auxiliary
should be disallowed. For instance, sentence (96a), in which the auxiliary is preceded by
the subject, should be equally ungrammatical, contrary to fact.
I suggest here that the proclitic character of sum is irrelevant for participle
movement. What is crucial is that Macedonian lacks the 3rd person auxiliary, which
occurs as the last element in the clitic cluster in Bulgarian and Serbo-Croatian, following
the pronominal clitics marked with AgrO and AgrIO in the template in (97).
(97) [TP T[+φ] ... [Aux (BE1/2) [AgrIO(P) [AgrO(P) [Aux (BE3) [vP subject[+φ] v [PartP Part[+φ]
object]]]]]]]
In Bulgarian and Serbo-Croatian the l-participle is attracted by the φ-features of T. The

only way to check them is by XP-movement into Spec, TP. Head movement of the l-
participle is blocked by the 3rd person auxiliary clitic above PartP, which occupies Aux0.
However, this slot is not filled in Macedonian, so the participle may be attracted by the
feature of Aux105 and move there by head movement. From this projection it may raise
to other head positions, such as AgrO (cf. the derivation in 84).
The suggestion that the XP-movement of the l-participle is contingent on the
presence of the 3rd person auxiliary is quite powerful. In chapter 5 (section 5.3.4.2.5) I
will show that it is one of the reasons for the lack of the movement in Polish and in
Czech (in the initial stage of the derivation).
4.4.2.4.2.3 Enclisis with imperatives and gerunds

Recall from section 4.4.2.3.2 that imperatives (cf. 98a) and gerunds (cf. 98b) must be
left-adjacent to clitics in Macedonian.
(98) a. Daj mi ja knigata!

giveIMPV.2SG meCL.DAT herCL.ACC book-the
“Give me the book!”
a’. *Mi ja daj knigata
b. Davajќi mi ja knigata, me bakna
givingGER meCL.DAT herCL.ACC book-the meCL.ACC kissed3SG
“Giving me the book, (s)he kissed me”
b’. *Mi ja davajќi knigata, me bakna (Mac, Tomić 1996a: 824)
These categories induce enclisis in many different languages, such as Greek, Italian, and
Spanish (cf. Rivero & Terzi 1995, Rooryck 1992, Terzi 1999, Zanuttini 1997), so
Macedonian is not exceptional in this respect. On a par with tensed verbs and l-
participles, gerunds and imperatives are case-assigners. However, they have severely
reduced morphology. Gerunds do not show any agreement morphology at all, which
makes them similar to infinitives. Imperatives, though, exhibit an invariant specification
for the 2nd person. They also make a distinction between the singular and the plural, as
shown in (99), where plural morphology is manifested by the suffix te.
(99) a. čita+j
read+IMPV+2SG
b. čita+j+te
read+IMPV+2PL (Mac, Tomić forthcoming)
There have been some attempts to capture the morphological impoverishment of

gerunds and imperatives in syntactic terms. For example, Beukema and Coopmans
(1989) suggest that imperatives have a [-Tense] Infl.106 Correspondingly, Belletti (1999:
569) proposes that if a clause contains an imperative, T is not able to check the
imperative morphology. Likewise, Terzi (1999: 94) argues that imperative clauses have
weak V features of T, which is unable to attract verbs. Still, even if verbs in imperative
clauses may not target T, there must be a way to assign an imperative interpretation to
105 In section 4.4.4.1 I suggest that Aux contains Number feature. Note that if Aux is occupied by
the 3rd person auxiliary clitic in Bulgarian and Serbo-Croatian, this projection does not have the
Specifier, on the assumption due to Bošković (2002) that clitics are non-branching elements.
Hence, the l-participle may not land in Spec, AuxP.
106 This suggests that the feature specification of imperatives is identical to infinitives. Rooryck
(1992) shows that the idea is difficult to maintain crosslinguistically. For instance, Latin has a
future imperative morpheme.
them. Belletti (1999) suggests that this is done by an (Imp)erative Operator, located in
the CP-area. The syntactic function of this Operator is to bind an empty category that
fills the subject position in imperative sentences and to assign it a value that
corresponds to [2nd person, singular/plural]. Since Belletti is not specific about the
location of this operator, I would like to claim that it may be related to the presence of
the functional head Σ, which is responsible for licensing negation, polarity, and in
general, Illocutionary Force (cf. chapter 2, section 2.3.6.2.2, and sections 4.3.3.2 and
4.4.4.2 in the present chapter). Some support for this assumption comes from Spanish,
where imperatives are incompatible with negation (cf. 100a), and infinitives or
imperatives must be used in these contexts instead (cf. 100b and c).
(100) a. *No lee!

NEG readIMPV.2SG
b. No leas!
NEG readSUBJ.2SG
“Don’t read!”
c. No leer!
NEG readINF
“Don’t read!” (Spanish, cf. Laka 1994)
Consequently, Laka (1994) proposes that imperatives are located in Σ, which in her
system is the same projection that hosts negation. Tomić (2001b), however, shows that
in Macedonian negative imperatives are possible, which in her view indicates that
negation and imperatives head different projections, which she terms NegP and ModP,
respectively.
(101) Ne čitaj go!

NEG readIMPV.2SG itACC
“Don’t read it!” (Mac, Tomić 2001b: 160)
I suggest that the postulation of two separate projections is unnecessary. It seems

plausible that there is a single operator related to Force licensing located in Σ, which
specifies both negation and imperative features. Whether both negation and imperative
morphology can be spelt out simultaneously might be subject to a parametric variation.
Negation and imperatives are semantically related, because both of them are
assigned under scope. Moreover, an additional argument for the uniform treatment of
these categories comes from stress shift. I mentioned in section 4.4.2.3.4 that in
Macedonian stress falls on the antepenultimate syllable (cf. 102a). In the context of
imperatives (cf. 102b and c) and negation (cf. 103) stress is shifted rightward onto the
following elements and calculated across word boundaries.
(102) a. DOnesi!
“Bring!”
b. doNEsi goCL.ACC!
“Bring it!”
c. doneSI miCL.DAT goCL.ACC!
“Bring it to me!” (Mac, Rudin et al 1999: 551; Baerman & Billings 1998: 20)
(103) Ne si mu GI dala jaBOLkata

NEGbeAUX.2SG himCL.DAT themCL.ACC givePART.F.SG apples+the
“Reportedly, you haven’t given him the apples” (Mac, Tomić 2001b: 649)
I propose that imperatives and negation are the only elements that are able to shift
lexical stress, because only the elements that target the Σ projection are able to do that.
This is not surprising in view of Cardinaletti & Starke’s (1999: 225 fn 64) observation
that since affirmation and negation always produce special stress patterns, Σ may
contain both polarity and accentuation features.
In contrast to imperatives, gerunds normally retain their fixed lexical stress when
they are accompanied by enclitics (cf. 104).
(104) KupuVAJќi miCL.DAT goCL.ACC, ...

“Buying it for me…” (Mac, cf. Franks 1998)
On the assumptions that Σ contains both the Force and accentuation features, this
property receives a straightforward explanation. The semantics of gerunds is not related
to polarity or Illocutionary Force, which means that they are not licensed by Σ. The
only formal property they share with imperatives is the reduced morphology.
Therefore, gerunds do not target Σ, but a lower projection above TP I will term
GerundP for convenience.
I propose that the enclisis requirement results from the fact that T is “inactive” in
imperative and gerundive clauses, so it does not attract verbs. Correspondingly,
pronominal clitics, which must normally raise to a “substantive” category endowed with
φ-features (cf. Nash & Rouveret 2002) may not check their φ-features against T. Instead
they are attracted directly into the closest projection above T without violating the
Minimal Link Condition (Chomsky 1995): Σ0 in the case of imperatives, and Gerund0 in
the case of gerunds. As an illustration, (105) presents the derivation of (102c).107 It
shows that the pronominal clitics mi and go left-adjoin to Σ0 (cf. 105b). Subsequently,
the verb donesi left-adjoins to the clitics in Σ0, and checks a Force-related imperative
feature there (cf. 105c). As a result, it ends up to the left of the clitics.
(105) a. [ΣP [Σ [AgrIOP [DP mi]k [AgrIO donesii [AgrOP [DP goj] [AgrO ti [VP tk [V ti [DP tj ]]]]]]]]]
b. [ΣP [Σ mik + goj + Σ [AgrIOP tk [AgrIO donesii [AgrOP tj [AgrO ti [VP tk [V ti [DP tj
]]]]]]]]]]
c. [ΣP [Σ donesi i + mik + goj + Σ] [AgrIOP tk [AgrIO ti [AgrOP tj [AgrO ti [VP tk [V ti [DP tj
]]]]]]]]
The derivation of a clause containing a gerund will proceed in largely the same way, the
only difference being that instead of Σ0, the clitics and the verb will adjoin to Gerund0.
Summarizing, the preceding sections have analyzed cliticization in the presence of
case-assigning hosts. The subsequent sections will investigate the ways pronominal
forms behave when there are no case-assigning hosts available.
107Since enclicis proceeds in the same way as proclisis up to the stage in (84), I do not present the
derivation in (105) from the very beginning. Moreover, note that since the pronominal clitics are
moving in (105) by themselves (without the DP associates), they can undergo XP-movement.
4.4.2.4.2.4 Patterns with non-case-assigning categories

This section provides arguments for the proposal that the class of clitics in Macedonian
instantiates two types of deficient elements: clitics and weak pronouns, in the sense of
Cardinaletti and Starke (1999), henceforth C&S.
4.4.2.4.2.4.1 Cardinaletti and Starke’s (1999) division of deficient elements

C&S analyze the standard distinction between clitics and strong pronouns, and on the
basis of crosslinguistic investigations they conclude that a theory of tripartitions is
needed, with a division into strong elements, weak elements, and clitics. These forms
are exemplified by means of the Slovak data in (106/107). Jemu in (106/107a)
instantiates a strong pronoun, which can be coordinated and appear clause initially. Ono
in (106/107b), which is a weak pronoun, may occur at the beginning of a clause, but it
may not be coordinated. Mu in (106/107c) is a clitic, and allows neither coordination
nor the clause-initial occurrence.
(106) a. Jemu to bude pomáhat

himDAT itNOM will helpINF
b. Ono mu to bude pomáhat
itNOM himCL.DAT itNOM will helpINF
c. *Mu to bude pomáhat
himCL.DAT itNOM will helpINF
(107) a. Jemu a Milanovi to bude pomáhat

himDAT and MilanDAT itNOM will helpINF
b. *Ono a to druhe mu bude pomáhat
itNOM and this other himDAT will helpINF
c. *Mu a jej to bude pomáhat
himCL.DAT and herCL.DAT itNOM will helpINF
(Slovak, Cardinaletti & Starke 1999: 166)
Both weak and strong elements are argued to occupy XP positions, while clitics reside
in heads. Clitics must move to case assigning positions in order to recover case.
Moreover, C&S claim that for reasons of economy, an element with the least structure
possible should be realized. This is captured by the “Minimise Structure” (Cardinaletti
& Starke 1999: 198) principle. It states that weak or strong forms are selected only if the
realization of a clitic element is independently ruled out. This means that the realization
of weak pronouns or strong pronouns should be the last resort procedure.
4.4.2.4.2.4.2 Weak and clitic forms in Macedonian

I submit that a similar tripartite division holds for pronominal forms in Macedonian.
Strong forms were presented in the chart in (47). Since they have the same syntactic
distribution as object DPs, they are not discussed here. Weak pronouns are
morphologically the same as clitics, but they have a different distribution. Clitics must
cliticize on case assigning hosts, such as tensed verbs and l-participles and may appear
in clause-initial positions. Conversely, weak pronouns, such as mu in (108), which are
found with nouns, adjectives, and passive participles, do not pro- or encliticize.
However, they are prosodically deficient, so they always occur with other deficient
elements, such as the copula clitic e in (108), and require phonological support to their
left, so they may not be positioned clause-initially.
(108) a. *Mu e tatko (na deteto)

himCL.DAT be3SG father to child-the
b. Tatko mu e na deteto
father himCL.DAT be3SG to child-the
“He is the father of this child (so he has to take care of him)!”
c. Toj mu e tatko
he himCL.DAT be3SG father
“He is his father (and not anyone else)!”
(Mac, cf. Tomić 2000: 295-296; L. Grujoska, p.c.)
I argued in section 4.4.2.4.2 that clitics in Macedonian must move to agreement

projections in order to check case. In the constructions with passive participles,
adjectives, and nouns there are no agreement projections available, because these
categories are not case assigners. Therefore, clitics may not be selected, because they are
unable to check case, and there is no case to be recovered. As a result, in line with the
Minimize Structure principle, weak pronouns must be realized. Notice that weak
pronouns always carry dative (cf. 108 and 109), which is an inherent case. This is
expected, as inherent case is purely semantic, related to thematic structure, so it does
not have to be checked, as it can be interpreted at LF.
Weak pronouns need phonological support to their left. However, as XPs they
have more structure than clitics, so they are more independent than clitics with respect
to their position in the clause. As example (109) indicates, they need not occur in the
second position, nor do they have to be verb-adjacent. They may scramble quite freely
in the clause.108
(109) Na Petreta (mu e) od strana na komisijata

to PeterDAT himCL.DAT be3SG from side of commission-the
(mu e) poveќe pati (mu e) rečeno
himCL.DAT be3SG more times himCL.DAT be3SG tellPASS.N
da bide točen
“Peter was more than once told by the commission to be punctual”
(Mac, cf. Tomić 2000: 299)
It was noted in section 4.4.2.3.3 that the weak forms only have to appear in the second
position when they are preceded by predicative hosts, such as nouns, adjectives, and
passive participles (cf. 110), which agree in φ-features with the subject.
(110) Rečeno mu e poveќe pati (*mu e)

tellPASS.N himCL.DAT be3SG more times himCL.DAT be3SG
da bide točen
“He was told to be punctual more than once” (Mac, O. Tomić p.c.)
108Chapter 5 will demonstrate that deficient pronouns in Polish, which are commonly argued to
be weak forms as well (cf. Franks & King 2000) have the same distribution.
The second position effect stems from the fact that all the weak forms raise to T,
whereas the predicative hosts move to Spec, TP in order to check the φ-features of T.
Consequently, there is only one element that precedes them, which is the occupant of
Spec, TP. As an illustration, (111) provides a derivation of (110).109
(111) [TP Rečenot [T mu+e [PredP ti [XP da bide točen ]
In chapter 2 I showed that in Bulgarian and Serbo-Croatian the l-participle may raise to
Spec, TP and check the φ-features of T. Even though Macedonian does not have this
option because of the lack of an overt 3rd person auxiliary and different cliticization
strategies (cf. section 4.4.2.4.2.2), it exhibits a similar process of predicate (locative)
inversion, exemplified in (111), in which the predicative element rečeno raises to Spec,
TP and checks the φ-features of T.
To sum up, this section has analysed strategies of cliticization in Macedonian. I
have suggested that three types of patterns can be distinguished: proclisis, enclisis, and
postposition. Proclisis and enclisis always involve cliticization of heads in the presence
of hosts that are case assigners. Postposition is found with non-case assigning
categories, such as adjectives, nouns, and passive participles. It affects weak pronouns,
which I have shown are XPs.
4.4.3 Bulgarian
The chapter will conclude with an analysis of the inventory of clitics in Bulgarian. It will
be shown that they form a natural class with Macedonian clitics, as they also have to be
verb-adjacent. However, they have different prosodic properties, and they never
undergo movement together with the l-participle or the finite verb. This part of the
chapter is organized as follows. Section 4.4.3.1 presents the paradigm of the clitics,
while section 4.4.3.2 describes their distribution. Section 4.4.3.3 discusses clitic
doubling. Section 4.4.3.4 briefly overviews previous accounts of cliticization in
Bulgarian. The alternative analysis given in section 4.4.3.4.3 argues that the pronominal
and auxiliary clitics in Bulgarian are adjoined to T. The section concludes with a
comparison between cliticization in a Wackernagel clitic language, exemplified by
Serbo-Croatian, and in a verb-adjacent clitic language, exemplified by Bulgarian.

As in the other Balkan Slavic languages, the clitics in Bulgarian must occur in the order
given in (112).
(112) li> Mod > šte > AUX (except 3rd SG e) > DAT > ACC > e
(Bg, cf. Tomić 1996a, Franks&King 2000, Rivero 2005)
Just as Macedonian, Bulgarian distinguishes only between dative and accusative

pronominal clitics. All clitic forms have non-reduced counterparts, but the dative full
109 It is still necessary to explain why the pronominal clitic always raises together with the copula.
I tentatively suggest this happens due to the phonological weakness of these elements. Note that
mu does not procliticize on e the way pronominal clitics do on finite verbs and l-participles,
because it may not appear clause-initially.
forms are perceived as archaic, and are usually replaced by the preposition na ‘to’
followed by an accusative full form.
Singular Plural
Acc (full/clitic) Dat (full/clitic) Acc (full/clitic) Dat (full/clitic)
1 men(e)/me mene/mi nas/ni nam/ni
2 teb(e)/te tebe/ti vas/vi vam/vi
3M nego/go nemu/mu tjax/gi tjam/im
3F neja/ja nej/ì
REFL sebe si/ se sebe si/si
(Bg, Franks and King 2000:52)
The auxiliary verbs in the present tense are also clitics, but the past tense auxiliary verbs
are not.
(114) Auxiliary forms
Present Past
SG PL SG PL
1 sŭm sme bjax bjaxme
2 si ste beše bjaxte
3 e sa beše bjaxa
(Bg, cf. Krapova 1999a)
Bulgarian has two other types of clitics, which do not show any inflectional distinctions
in the paradigm.
(115) the interrrogative complementizer li (cf. section 2.3.6.3.3 in chapter 2)

the future auxiliary proclitic šte (cf. section 2.3.5.3 in chapter 2)
4.4.3.2 Properties and positions of the Bulgarian clitics

The Bulgarian clitics do not have to appear in the second position, but their placement
has to satisfy the divergent phonological and syntactic conditions given in (116).
(116) The Bulgarian clitics are:

a. Phonologically enclitic, so they need a phonological host to their left
b. Syntactically proclitic, so they require a verbal host to their right
Requirement (116b) is suspended if the verb is in the clause initial position. The
workings of the two conditions are exemplified in the outputs in (117).
(117) a. Vera mi go dade včera

Vera meCL.DAT itCL.ACC gave3SG yesterday
b. Včera mi go dade Vera
c. Včera Vera mi go dade
d. *Vera mi go včera dade
e. *Mi go dade Vera včera

f. *Vera dade mi go včera
g. Dade mi go Vera včera (Bg, Franks and King 2000: 63)
Sentence (117c) indicates that the Wackernagel law is not operative in Bulgarian, as the
clitics are preceded by two constituents. In (117d) the clitic cluster is not left-adjacent to
the verb, which violates condition (116b), while in (117e) it is clause initial, which is at
odds with condition (116a) and results in ungrammaticality. The sentence in (117f)
proves that the clitics are syntactically proclitic, so they must precede the verb, if
possible. If the verb is clause-initial, as in (117g), the clitics may follow it, as otherwise
the phonological requirement in (116a) is not met.
The pronominal clitics in Bulgarian clearly opt for the preverbal position, and
appear there if there is any element preceding them. The examples in (118) show that
even the conjunctions i and a suffice to supply pronominal and auxiliary clitics with
necessary phonological support to their left.
(118) a Toj napisa pismo na majka si

he wrote3SG letter to mother selfCL.DAT
i ì go izprati
and herCL.DAT itCL.ACC sent3SG
“He wrote a letter to his mother and sent it to her”
(Bg, Franks and King 2000: 52, 63)
b. Tova myasto e tolkova krasivo, a sŭm bil
this place be3SG so beautiful but beAUX.1SG bePART.M.SG
tam samo vednŭž
there only once
“This place is so beautiful but I have been there only once”
(Bg, S. Marinov, p.c.)
Apart from Polish (cf. chapter 5 section 5.2.2), the cliticization strategy in (118) is not
possible in any other Slavic language, nor was it attested in Old Church Slavonic.
Sławski (1946: 25, 62) claims that the enclisis on conjunctions became available only in
the first half of the 19th century. The property reveals an important characteristic of the
Bulgarian cliticization. Given that conjunctions are outside the syntactic domain of the
clause, and certainly much higher than the TP level, it seems that the requirement of the
enclisis in Bulgarian is purely phonological in nature. In other words, the requirement
of phonological adjacency does not imply syntactic adjacency.
This insight is confirmed by the fact that li, which is also enclitic, may not be
supported by a conjunction.
(119) *I li ti go dade Petko včera?

and Q youCL.DAT itCL.ACC givePAST.3SG Petko yesterday
“And did Petko give it to you yesterday?”
(Bg, Franks and Bošković 2001: 179)
However, this is to be expected, because li can be a focus licenser, which attracts and
puts the constituent to its left in focus (cf. chapter 2, section 2.3.6.3.3). It may not be
preceded by a conjunction, because conjunctions are never focused.
Li is also the only enclitic in Bulgarian that does not require adjacency to the verb
(cf. 120).
(120) a. Včera li Penka ja e dala

yesterday Q Penka herCL.ACC beAUX.3SG givePART.F.SG
knigata na Petko?
book-the to Petko
b. Penka včera li ja e dala
Penka yesterday Q herCL.ACC beAUX.3SG givePART.F.SG
knigata na Petko?
book-the to Petko
“Was it yesterdayFOC that PenkaTOP gave the book to Petko?”
(Bg, Tomić 1996a: 833)
The fact that it can be separated from the other clitics by the subject Penka in (120a)
indicates that it is located higher than the other clitics. It is usually claimed to be hosted
in C (cf. Rivero 1993, 1994a, Bošković 1995, and others), because it is in
complementary distribution with complementizers, such as če ‘that’ (cf. 121a), although
just as the other complementizers, it may appear with the subjunctive marker da (cf.
121b).
(121) a. Mislja, če šte se vŭrne (*li) dovečera

think1SG that will REFL return3SG Q tonight
“I think that he will come back tonight”
b. Da se vŭrne li dovečera?
Da REFL return3SG Q tonight
“Should s/he come back tonight?” (Bg, Rudin 1986: 66)
Section 2.3.6.3.3 in chapter 2 thoroughly analyzed negated questions. They are formed
with the negated particle ne, which attracts the highest clitic in the cluster, and the two
elements become left-adjoined to li. As a reminder, the construction is exemplified in
(122).
(122) Ne go li e viždal?
NEG himCL.ACC Q be3SG seePART.M.SG
“Didn’t he see him?” (Izvorski et al 1997: 191)
4.4.3.3 Clitic doubling in Bulgarian

In Bulgarian clitic doubling is possible with direct and indirect objects when they are
both specific and topicalized. It is not a uniform phenomenon and is subject to a great
deal of dialectal and stylistic variations (see Arnaudova 2003, Franks & Rudin 2005,
Rudin 1997, Schick 2000, Tomić forthcoming for details). It seems to be always dependent
on discourse factors. In this way it differs from clitic doubling in Macedonian (cf.
section 4.4.2.2), which is obligatory with all definite direct objects (cf. 124). As (123)
shows, this requirement does not hold for Bulgarian.
(123) a. Petŭr prodade kolata

Peter sellPAST.3SG car-the
“Peter sold his car”
b. Petŭr ja prodade kolata
Peter itCL.ACC sellPAST.3SG car-the
“As for Peter, he has sold his car” (Bg, Vačkov 1998: 166)
(124) a. Petar *(ja) prodade kolata

Peter itCL.ACC sellPAST.3SG car-the
“Peter sold his car” (Mac, Vačkov 1998: 166)
There are two contexts, though, in which clitic doubling is always required. The first
instance concerns oblique subjects, which are usually Experiencers.
(125) a. Na mene ne *(mi) e studeno

to meDAT NEG meCL.DAT be3SG cold
“I am not cold”
b. Nego *(go) boli stomaxŭt
himDAT himCL.DAT hurt3SG the-stomach
“His stomach hurts” (Bg, Franks & Rudin 2005: 106)
The other case involves left-dislocated topicalized objects, which are specific, but do
not have to be definite (cf. 126b).
(126) a. Pismata Marija vinagi gi prašta s zakasnenie

letters-the Mary always themCL.ACC send3SG with delay
“Mary is always sending letters with a delay”
b. Edin paket go izgubixa po pogreška
one package itCL.ACC losePAST.3PL by mistake
“They lost a package by mistake” (Bg, Arnaudova 2003: 163)
Since clitic doubling in Bulgarian does not seem to be a uniform syntactic

phenomenon, which largely depends on specific discourse structure requirements that
need to be defined in future research, I will not discuss it here any further.
4.4.3.4 Towards an analysis of cliticization in Bulgarian

It has been shown that unlike in Serbo-Croatian, Bulgarian clitics need not appear in
the Wackernagel position. As in Macedonian, they must be verb-adjacent.110 The main
difference between Bulgarian and Macedonian clitics is related to their prosodic
requirement. In Bulgarian, they are always enclitic, so they may never appear clause-
initially. In Macedonian, clitics may be proclitic or enclitic, depending on the type of
host onto which they cliticize. The minimal pair in (127) illustrates the contrast.
(127) a. Mi go dade Vera včera (*Bg, √Mac)

meDAT itACC gave Vera yesterday
b. Dade mi go Vera včera (√Bg, *Mac)
gave meDAT itACC Vera yesterday (Franks 1998)
In the subsequent sections I will show that these prosodic differences have syntactic
motivations; for example, I will demonstrate that unlike in Macedonian, finite verbs and
participles in Bulgarian do not undergo movement together with the pronominal clitics.
Before I develop a theory of clitic placement in Bulgarian, I will briefly overview some
previous accounts. Section 4.4.3.4.1 will discuss the analyses which postulate that clitics
110Recall from section 2.3.5.1.2 in chapter 2 though, that there is a small class of aspectual
adverbs that may intervene between the auxiliary clitic and the l-participle.
occupy head positions. Section 4.4.3.4.2 will outline Bošković’s (2001) account, which
proposes that clitics target specifiers.
4.4.3.4.1 Clitics are in head positions

Most of the analyses of cliticization in Bulgarian (cf. Franks 1998, Franks & King 2000,
Rudin 1997, Tomić 1996a, 1997, 2000 and others) claim that clitics reside in separate
functional heads, as schematized in (128).
(128)
TP
T’
T AgrIO
Dat.Cl. AgrO
non-3rd SG Aux
Acc.Cl. Aux
e PartP
These analyses assume that the clitics must cluster with each other. The cluster is
formed by formation of a complex head through a series of successive cyclic rightward
head adjunctions. For instance, the clitic cluster in the sentence in (129a) is formed as
shown in (129b-c).
(129) a. Petko mi go dade

Petko meCL.DAT itCL.ACC givePAST.3SG
“Petko gave it to me”
b. [AgrIOP mi [AgrOP go+dadei [VP ti]]]
c. [AgrIOP mi+[go+dadei ]j [AgrOP tj [VP ti ]]] (Bg, cf. Bošković 2002)111
The finite verb dade raises as a head from VP and right-adjoins to the accusative clitic in
AgrO (cf. 129b), forming a complex head. The complex head raises to AgrIO, and right
adjoins to the dative clitic mi.
The construction of a clitic cluster in a clause that contains the l-participle
proceeds as sketched in (130).
111Bošković (2002) does not assume the derivation presented here. I only quote his overview of
the previous accounts.
(130) a. Ti si mu gi dal
you be2SG himCL.DAT themCL.ACC givePART.M.SG
“You have given them to him”
b. [TP si [AgrIOP mu [AgrOP gi+dali [VP ti ]]]]
c. [TP si [AgrIOP mu+[gi+dali ]j [AgrOP tj [VP ti ]]]]
d. [TP si+[mu+[gi+dali ]j]k [AgrIOP tk [AgrOP tj [VP ti ]]]] (Bg, cf. Bošković 2002)
The l-participle is claimed to undergo head movement from V to AgrO, where it right-
adjoins to the accusative clitic gi. Next, the complex head formed in this way raises to
AgrIO, and right-adjoins to the dative clitic mu. Finally, the complex of pronominal
clitics and the l-participle right-adjoins to the auxiliary in T.
If Kayne’s (1994) LCA is adopted, the analysis is problematic from a theoretical
point of view, because it makes use of rightward adjunction. Moreover, the assumption
that l-participle fronting occurs via head movement has the drawbacks mentioned in
chapter 2, section 2.2.
4.4.3.4.2 Clitics are in specifier positions

An alternative view on cliticization is taken by Bošković (2002), who claims that both
auxiliary and pronominal clitics in Bulgarian and Macedonian are located in specifiers of
null heads. Along with Chomsky (1995: 249), he argues that clitics are both maximal
and minimal elements; that is, they are generated in non-head positions, but may move
as heads.
I will consider his approach on the basis of the example in (131a). On the
assumption that all clitics are in specifiers, the string in (131) will have the structure as
in (131b) before the clitic cluster is formed.
(131) a. Ti si mi go dal
you beAUX.2SG meCL.DAT itCL.ACC givePART.M.SG
“You have given them to me” (Bg, Bošković 2002)
b. [TP si [T’ [AgrIOP mi [AgrIO’[AgrOP go [AgrO’[VP dal]]]]]]]
Bošković proposes that the participle dal moves cyclically as a head past each of the
clitics to the heads immediately above them. Subsequently, the clitics themselves raise
from a specifier and adjoin to the left of the l-participle as a head. Thus, the participle
first adjoins to AgrO (cf. 132a). Next, AgrO together with the participle adjoins to
AgrIO, and dal lands above the specifier that hosts the accusative clitic go. Go moves as
a head, and adjoins to AgrIO (cf. 132b). Assuming with Kayne (1994) that adjunction is
always to the left, go is placed to the left of the participle dal. At this point AgrIO
contains the string go dal and moves on past the dative clitic mi to T. This enables mi to
adjoin to go dal as a head (cf. 132c). Now T contains mi go dal and moves past si to a
higher functional head called F. Finally, the auxiliary clitic si raises from Spec, TP and
adjoins to the left of F (cf. 132d).
(132) a. [TP si [T’ [AgrIOP mi [AgrIO’[AgrOP go [AgrO’ dali [VP ti]]]]]]]

b. [TP si [T’ [AgrIOP mi [AgrIO’ [goj + dali] [AgrOP tj [AgrO’ti [VP ti]]]]]]]
c. [TP si [T’ mil + [goj + dali]k [AgrIOP tl [AgrIO’ tk [AgrOP tj [AgrO’ti [VP ti]]]]]]]
d. [F sin+[mil+[goj+dali]k]m][TPtn[T’tm[AgrIOP tl [AgrIO’ tk [AgrDOP tj [AgrDO’ ti [VP ti ]]]]]]]
Since the clitic cluster is created by left-adjoining the verb to the clitics, Bošković’s
analysis is in line with Kayne’s (1994) LCA and avoids the problematic idea of right
adjunction of some previous approaches. The positioning of AgrIOP higher than

AgrOP in the structure secures the desired dative-accusative-verb order.
However, Franks & Rudin (2005) and Franks (forthcoming) point out that
Bošković’s account faces empirical problems. Namely, it is possible to split the auxiliary
clitic from the pronominal clitics in Bulgarian with a small class of aspectual adverbs
(cf. chapter 2, section 2.3.5.1.2). This is wrongly predicted not to be possible in his
approach.
(133) Az sŭm veče ti ja dala

I beAUX.1SG already youCL.DAT itCL.ACC givePART.F.SG
“I have already given it to you” (Bg, Franks forthcoming)
Moreover, Bošković argues that the auxiliary clitic originates in Spec, TP, which is a a
strange suggestion, because auxiliaries are never analyzed as elements that undergo XP-
movement. Furthermore, he claims that his proposal is valid for all verb-adjacent clitic
languages, which includes Macedonian. However, it seems to me that his approach may
not predict the variable behaviour of the Macedonian clitics, which procliticize only
when the clitic host is a finite verb or an l-participle. In fact, Bošković (to appear, fn 42)
acknowledges disregarding these diverging contexts.
4.4.3.4.3 An alternative analysis

As in the case of Macedonian, the analysis of cliticization in Bulgarian that I am
developing relates to the Person-Case Constraint (PCC), which is observed in Bulgarian
as well. This is indicated by the examples in (134), which correspond to the Macedonian
outputs in (90).
(134) a. *Az im te preporŭčvam

I themCL.DAT youCL.ACC recommend1SG
b. Az te preporŭčvam na tjax
I youCL.ACC recommend1SG to themACC
c. Az im ja preporŭčvam
I themCL.DAT herCL.ACC recommend1SG
“I am recommending her to them” (Bg, cf. Hauge 1999)
Sentence (134a) shows that a non-3rd person accusative clitic is incompatible with a
dative clitic. However, if a strong form of the accusative pronoun tjax preceded by the
preposition na is used, the result is grammatical (cf. 134b). The dative clitic is
compatible only with accusative clitics marked for the 3rd person, as indicated in (134c).
I suggest that the Bulgarian clitic cluster is derived along the lines of
Anagnostopoulou’s (1999: 287) account for the PCC languages. Since I have already
presented Anagnostopoulou’s approach in section 4.4.2.4.2.2, I will outline the analysis
in a somewhat simplified form.
I propose that the auxiliary clitics are merged in two positions in Bulgarian: the 3rd
person singular e originates in the head of AuxP. I will provide some motivations for
this idea in section 4.4.4.1. All the other auxiliary forms are merged in T. The
pronominal clitics are generated as phrasal arguments within the VP, but they land in
head positions. The dative clitic raises first, lands in T adjoining to the auxiliary or the
finite verb if they are present there, and checks the person feature of T. The accusative
clitic moves next and checks the remaining number feature on T, adjoining to the
dative clitic. However, the derivation converges only if the accusative clitic carries just a
number feature, and not a person feature, because the latter has already been checked
by the dative. On the assumption that the 3rd person pronouns represent only a number
feature, they are the only eligible candidates for the movement. If a non-3rd person
pronominal clitic raises, the derivation will crash, because the person feature on the
accusative will remain unchecked. Finally, if the construction contains the 3rd person
auxiliary e instead of some other finite verb merged in T, it remains in its base generated
position in Aux if there are pronominal clitics present, or raises to T, if it is the only
clitic in the cluster.
(135) [TP [T <clDAT>i + <clACC>j + T] … [Aux (e) [VP V ti tj]]]
The derivation given in (135) is considerably simpler than in Bošković’s (2002)

proposal. It also avoids the potentially problematic idea of right-adjunction of the
previous head movement accounts discussed in section 4.4.3.4.1. Finally, it handles the
PCC effect straightforwardly.
In order to provide more support for the analysis developed here, let me juxtapose
some properties of cliticization in Bulgarian and Macedonian, which represent
languages with verb-adjacent clitics, with Serbo-Croatian, which is a Wackernagel
position clitic language.
I mentioned in section 4.4.2.4.2.2 that Serbo-Croatian does not observe the Person
Case Constraint. It seems that in general the constraint does not hold in Wackernagel
position clitic languages (apart from Serbo-Croatian, also in Czech, Slovak, and
Slovene) as well as in the languages which have weak pronouns, rather than clitics (e.g.
in Polish and Russian; cf. the discussion in section 5.2.2, chapter 5). In section 4.4.1.2 I
claimed, following the insights of Bošković’s (2001, to appear) and Stjepanović’s (1999)
analyses, that pronominal clitics do not cluster in a single head in Serbo-Croatian.
Rather, the dative clitic resides in Spec, AgrIOP, and the accusative clitic is hosted in
Spec, AgrOP. Since the PCC holds only when pronominal clitics are adjoined to a
single head, the fact that it is not operative in Serbo-Croatian follows straightforwardly.
The assumption that clitics do not cluster in a single head in the Wackernagel clitic
languages allows us to account for a few more differences between Serbo-Croatian on
the one hand, and Bulgarian and Macedonian on the other hand. For instance, I
mentioned in section 4.4.1.2 that a higher part of the pronominal clitic cluster may be
deleted in VP-ellipsis in Serbo-Croatian (cf. 136b).
(136) a. Mi smo mu ga dali,

we beAUX.1PL himCL.DAT itCL.ACC givePART.M.PL
a i vi ste mu ga dali, (takodje)
and also you beAUX.2PL himCL.DAT itCL.ACC givePART.M.PL too
“We gave it to him, and you did, too”
b. Mi smo mu ga dali, a i vi ste mu ga dali, (takodje)
c. *Mi smo mu ga dali, a i vi ste mu ga dali, (takodje)
(S-C, cf. Stjepanović 1998, 1999)
Bošković (2002) points out that a similar operation in Bulgarian or Macedonian gives
rise to strong ungrammaticality.
(137) a. *Nie sme mu go dali, i vie ste

we beAUX.1PL himCL.DAT itCL.ACC givePART.PL and you beAUX.2PL
mu go dali (sŭšto)
himCL.DAT himCL.ACC givePART.M.PL too
“We gave it to him, and you did too”
b. *Nie sme mu go dali, i vie ste mu go dali (sŭšto)
c. *Nie sme mu go dali, i vie ste go mu dali (sŭšto) (Bg, Bošković 2002: 331)
This is to be expected if the pronominal clitic cluster in Bulgarian is located in the same
head, and forms a single constituent as a whole.
Likewise, I reported an observation due to Progovac (1993), who claims that clitics
may climb from an embedded subjunctive clause (cf. 138) but not out of an embedded
indicative clause (cf. 139).
(138) a. Milan želi da ga vidi

Milan wish3SG that himCL.ACC see3SG
“Milan wishes to see him”
b. ?Milan ga želi da vidi (S-C, Progovac 1993)
(139) a. Milan kaže da ga vidi

Milan say3SG that himCL.ACC see3SG
“Milan says that he can see him”
b. *Milan ga kaže da vidi (S-C, Progovac 1993)
In Bulgarian clitics may never raise from an embedded clause to the main clause,
whether they move out of a subjunctive clause introduced by the subjunctive
complementizer da (cf. 140) or an indicative clause headed by the indicative
complementizer če (cf. 141).
(140) a. Manol iska da go vidi

Manol wish3SG that himCL.ACC see3SG
“Manol wishes to see him”
b. * Manol go iska da vidi (Bg, S. Marinov, p.c.)
(141) a. Manol kazva če go vižda

Manol say3SG that himCL.ACC see3SG
“Manol says that he can see him”
b. *Manol go kazva če vižda (Bg, S. Marinov, p.c.)
The contrast receives a straightforward explanation once it is assumed that clitics in

Bulgarian target a single head position. Their movement across the complementizer
would lead to a Head Movement Constraint violation, so it is prohibited. The climbing
is unproblematic in Serbo-Croatian, in which clitics target specifiers and undergo XP-
movement.
Finally, more evidence for the claim made in this section comes from the way
clitics pattern with negation. As shown in (142) and (143), negation attracts pronominal
clitics in Bulgarian, but not in Serbo-Croatian, where it attracts the verb instead. It was
claimed in chapter 2 (section 2.3.6.3.2) that these elements incorporate into negation.
Since negation is a head, it may incorporate into other heads. Therefore, it may
incorporate into the finite verb in Serbo-Croatian, but not into the pronominal clitics,
which occupy XP-positions.
(142) a. Ne me boli
NEGmeCL.ACC hurt3SG
a’. *Ne boli me
b. Ne mi se struva, če ...
NEG meCL.DAT REFL seem3SG that
b’. *Ne struva mi se, če... (Bg, S. Marnov, p.c.)
(143) a. Ne boli me
NEG hurt3SG meCL.ACC
a’. *Ne me boli
b. Ne cini mi se da...
NEG seem3SG meCL.DAT REFL that
b. * Ne mi se cini da... (S-C, N. Milićević, p.c.)
4.4.4 Concluding remarks

The following two sections will give some general remarks concerning cliticization in all
the South Slavic languages analyzed in this chapter. Section 4.4.4.1 will try to provide an
account for the ordering of the 3rd person auxiliary, which always follows the other
clitics. Section 4.4.4.2 will discuss potential motivations for clitic climbing.
4.4.4.1 Ordering of the clitics

As has been mentioned throughout this chapter, clitics in South Slavic follow the
ordering given in (144).112
(144) li> Mod > AUX (apart from (j)e) > DAT > ACC > (j)e
It has been demonstrated that the clitics are syntactically active. For example, the
highest clitic may be attracted by negation in Bulgarian, whereas in Serbo-Croatian the
dative clitic may be raised from an embedded clause to the main clause. This clearly
indicates that the sequence in which they appear is the result of a syntactic operation.
The previous sections have investigated the ways pronominal clitics are inserted in
the syntactic structure. However, I still have not explained why the 3rd person auxiliary
occurs in a different position than the other auxiliary forms. This will be done in the
subsequent sections. Section 4.4.4.1.1 will briefly review an account due to Bošković
(2001; to appear). Section 4.4.4.1.2 will discuss Tomić’s (1996a) view on the topic.
Section 4.4.4.1.3 will give some alternative suggestions.
4.4.4.1.1 Bošković (2001, in press)

According to Bošković (2001: 125ff), the final position of je in the clitic cluster is
phonologically conditioned. One of his main arguments for this claim comes from the
112I disregard the fact that in Macedonian both singular and plural variants of the 3rd person
auxiliary appear last in the cluster. Their distribution was accounted for in section 4.4.2.4.2.1.
behaviour of different auxiliary forms, such as je in (145a) and su in (145b) in sentences

interrupted by intonation pauses.
(145) a. ?#On je, # kao šte sam vam rekla#,

he beAUX.3SG as be1SG youCL.DAT sayPART.F.SG
predstavio se Petru#
introducePART.M.SG selfACC PeterDAT
“He, as I told you, introduced himself to Peter”
a’. *#On se, # kao šte sam vam rekla#, predstavio je Petru
b. ?#Oni su, # kao šte sam vam rekla#,
they beAUX.3PL as be1SG youCL.DAT sayPART.F.SG
predstavili se Petru#
introducePART.M.PL selfACC PeterDAT
“They, as I told you, introduced themselves to Peter”
b’. *#Oni se, # kao šte sam vam rekla#, predstavili su Petru#
(S-C, Bošković 2001: 126)
Even though the two pairs in (145) contain different auxiliaries, they have the same
syntactic distribution, and both je and su must precede the pronominal clitics.
According to Bošković, this means that all the auxiliary clitics occupy the same position
in syntax, and it is only at PF that je is spelt out at the end of the cluster.
Furthermore, Bošković observes that both su and je may precede sentential
adverbs. However, when je co-occurs with pronominal clitics, it may not do that. This is
indicated by the meaning of the ambiguous adverb pravilno ‘cleverly’. It permits the
sentential reading only when it is preceded by the auxiliaries alone.
(146) a. On je pravilno odgovorio Mileni

he beAUX.3SG correctly answerPART.M.SG MilenaDAT
“He did the right thing in answering Milena”
“He gave Milena a correct answer”
a’. On joj je pravilno odgovorio
he herCL.DAT beAUX.3SG correctly answerPART.M.SG
“*He did the right thing in answering Milena”
“He gave Milena a correct answer”
b. Oni su pravilno odgovorili Mileni
they beAUX.3PL correctly answerPART.M.PL MilenaDAT
(S-C, Bošković 2001: 126-127)
This shows that both je and su are located higher than the pronominal clitics. The fact
that je is normally spelt out lower is, in Bošković’s view, due to a PF filter, which forces
the clitic to be pronounced in this position It is not entirely clear what kind of PF
requirement prohibits the pronunciation of je in the higher position. Bošković refers to
the “process of losing clitichood” by je as the responsible factor, which prevents it from
occurring outside the edges of the cluster. I find the explanation inconclusive. It is not
entirely clear why the loss of clitichood should involve the pronunciation at the end of
a cluster, rather than, say, in the middle of it. Moreover, the loss of clitic properties is
often intermediated by the reanalysis of a clitic as an affix (cf. the discussion of Polish
in chapter 1, section 1.3.4.2.2.1), which in fact implies that the clitic je should occur as
close to the verb as possible, rather than at the end of the cluster, following the
pronominal clitics.
4.4.4.1.2 Tomić (1996a)

According to Tomić (1996a: 839-841), the exceptional placement of the 3rd person
auxiliary is due to its mixed clitic and root-like properties. The clitic function is assumed
when je is placed at the end of a cluster (cf. 147a). The root-like behaviour is observed
when je adjoins into another auxiliary clitic, such as sam (cf. 147b). This results in the
creation of a strong auxiliary form, with je being the root, and sam a person/number
affix.113
(147) a. On mu ih je dao
he himCL.DAT themCL.ACC beCL.3SG givePART.M.SG
“He gave them to him”
b. Ja jesam mu ih dao
I be1SG himCL.DAT themCL.ACC givePART.M.SG
“I gave them to him indeed” (S-C, cf. Tomić 1996a)
However, this proposal does not explain the idiosyncratic behaviour of je in front of the
question particle li. As demonstrated in (148a), je is the only auxiliary clitic which may
appear clause-initially. It is actually preferred over the strong form in this position (cf.
148a’). The other clitic variants, such as si in (148b) are prohibited in this context, and
may only occur as strong forms (cf. 148b’).
(148) a. Je li ga našao?
beCL.3SG Q himCL.ACC findPART.M.SG
“Did he find him?”
a’. *Jest(e) li ga našao?
be3SG Q himCL.ACC findPART.M.SG
b. *Si li ga našao?
beCL.2SG Q himCL.ACC findPART.M.SG
b’. Jesi li ga našao?
be2SG Q himCL.ACC findPART.M.SG
“Did you find him?” (S-C, Tomić 1996a)
4.4.4.1.3 Some alternative suggestions

I agree with Tomić (1996a) that there are two instances of je. However, I would like to
claim that while one of them is the 3rd person auxiliary, the other one is the
morphological realization of the Σ-head, which contains features related to polarity,
focus, and Illocutionary Force. This morphological realization is present not only in the
verbal domain, as it is also found on strong pronominal forms in some West Slavic
languages. For example, the clitic variant of the masculine accusative pronoun ‘him’ in
Polish go has the strong counterpart jego (cf. chapter 5, section 5.2.1 for a discussion of
Polish pronouns, and Cardinaletti & Starke 1999 for an analysis of related Slovak data).
As for the auxiliary function of je, let me point out that the situations in which
some person clitic forms occur in different positions than other clitics in the cluster
have also been attested outside the South Slavic languages. For example, Poletto (2000:
30ff) observes similar distribution of subject clitics in Northern Italian dialects, which
show disjoint occurrences in the structure according to their person marking. She
113In chapter 2, section 2.3.6.2.2 this form was claimed to be created via incorporation of the
auxiliary clitic into je, which was argued to be the head of the Polarity Phrase.
suggests capturing their distribution by assuming that they encode [±speaker] and
[±hearer] distinctions syntactically, with the former corresponding to the 1st person, and
the latter to the 2nd person. I am not sure whether a similar division could be made in
Slavic. I leave the issue for future research, and for the time being I assume that je
specifies only the [number] feature, while the other auxiliaries additionally carry
[person] distinctions. I will provide more arguments for this idea in section 5.3.4.2.4.1
in chapter 5, where I discuss the position of auxiliaries in Polish.
4.4.4.2 Why do clitics climb?

It has been shown that in each of the Slavic languages studied in this chapter clitics
must raise out of their base positions within the VP. Subsequently, they move to
different projections, such as Spec, AgrOP and Spec, AgrIOP in Serbo-Croatian, and T
in Bulgarian and Macedonian. They may undergo head or phrasal movement, and check
either case or φ-features. The movement affects all types of clitics, which eventually end
up as a cluster, usually in a designated position. The important question is what drives
the movement. Certainly, their phonological deficiency cannot be the reason, because
the clitics can receive phonological support in their base positions as well. The
movement cannot be triggered by a single syntactic feature, either, given that it affects a
number of categorially unrelated elements with a different morphological content.
As far as climbing of the pronominal clitics is concerned, it has sometimes been
argued that this is motivated by discourse structure requirements (cf. Halpern 1992;
Uriagereka 1995). Pronominal forms always carry old information, and are specific and
referential; therefore they may be forced to appear outside the VP, in line with
Diesing’s (1990) Mapping Hypothesis. The way the movement actually proceeds varies
across languages. In a Wackernagel clitic language, such as Serbo-Croatian, pronominal
clitics land in XP positions. In verb-adjacent clitic languages, such as Bulgarian, they
raise as XPs from the argument positions within the VP, but become adjoined to T as
heads.
The clitics in Serbo-Croatian must appear in the second position. It was noted in
section 4.4.1.3 that the requirement is reminiscent of the verb second effect, but the
exact motivation for this phenomenon is undetermined. In Bulgarian and Macedonian
the clitics are always verb-adjacent, which may be interpreted as a gradual shift towards
affixhood. Chapter 5 will provide more evidence for this assumption, which will be
based on properties of cliticization in Polish. In each of these languages clitics may be
attracted by the functional head Σ, which encodes polarity features (assertion and
negation), features related to Illocutionary Force marking, as well as focus.
By means of summary, the chart in (149) presents properties of the pronominal
clitics in the languages discussed in this chapter.
(149) Properties of pronominal clitics in South Slavic
Bulgarian/ Macedonian Serbo-Croatian

position verb-adjacent, enclitic verb-adjacent, proclitic Wackernagel
or enclitic depending enclitic
on the host
ordering Dat-Acc Dat-Acc Dat-Acc
pronominal yes yes no
clustering
type of X0 X0 XP
movement
the Person Case yes yes no
Constraint
Chapter 5 The status of clitics and
compound tenses in Polish
5.1 Introduction
It was mentioned in chapter 1 that the syntax of compound tenses displays a
considerable typological variation across Slavic. Diachronically, the main reason for this
diversity was the imbalanced and uneconomical system of expressing tense and aspect
distinctions inherited from Proto-Slavic, which was simplified in different ways in each
language group. For instance, the South Slavic languages enhanced the system mainly
through a semantic reanalysis of superfluous constructions, but their compound tenses
largely retained their morphological and syntactic make-up.
The present chapter is devoted to Polish, which is a West Slavic language. In
comparison to South Slavic, the inventory of tenses in Polish is quite reduced: the
language has lost the aspectual tenses and the pluperfect, and the former present
perfect is used as the default past tense. Importantly, the impoverishment of the
temporal distinctions has been accompanied by a reduction of the auxiliary forms into
affixes. It will be demonstrated that this morphological process has direct consequences
for the syntax of the ‘l-participle+auxiliary’ constructions.
The chapter will begin, however, with a brief overview of pronominal clitics in
section 5.2. Their properties will be compared to the characteristics of clitics in South
Slavic. It will be shown that the pronominal forms in Polish enjoy a greater autonomy
in the clause. A review of the auxiliary clitics in section 5.3 will demonstrate that they,
conversely, are gradually turning into affixes.
5.2 The pronominal forms

This section will describe pronominal forms in Polish. It will be shown that Polish has
strong and weak pronouns, but it does not have pronominal clitics. The pronominal
system in Polish will be compared to those found in the South Slavic languages.
5.2.1 The paradigm of pronouns

The paradigm of the pronominal clitics in Polish is presented in (1). The chart indicates
that both pronouns and clitics may appear in three different cases. In comparison with
nouns, which distinguish seven morphological cases, the case declension on clitics is
quite reduced.
224 The status of clitics and compound tenses in Polish
(1) Pronominal forms in Polish
Accusative Genitive Dative

strong clitic strong clitic full clitic
1SG mnie mię mnie mię (dated) mnie mi
(dated)
2SG ciebie cię ciebie cię tobie ci
3SG.M jego go jego go jemu mu
(niego) (niego) (niemu)
3SG.F ją (nią) ją jej (niej) jej jej (niej) jej
3SG.N je (nie) je jego go jemu mu
(niego) (niemu)
1PL nas nas nas nas nam nam
2PL was was was was wam wam
3PL.VIR ich (nich) ich ich (nich) ich im (nim) im
3PL.NV je (nie) je ich (nich) ich im (nim) im
REFL siebie się siebie się sobie (se)
(dialectal)
(Pl, cf. Spencer 1991: 367, Franks&King 2000: 150)
Some of the pronouns have an additional variant which begins with the letter n. These
forms are used if the pronoun is the object of a preposition.
(2) Jan pisze [do niego/*jego]

Jan writePRES.3SG to himCL.ACC/himACC
“Jan is writing to him” (Pl)
It has been noticed in the literature that the pronominal forms in Polish are losing their
clitic status and that they increasingly tend to pattern like strong pronouns (cf.
Cetnarowska 2003, Franks & King 2000, and Witkoś 1998). This observation is most
often drawn from the positions occupied by the pronouns in the clause structure (cf.
section 5.2.2 for details). However, the decline of the clitics is also evident in the
pronominal paradigm, in which the morphological forms of some clitic and non-clitic
variants are very similar. For instance, the clitic and full forms of the 1st and 2nd person
plural nas and nam are orthographically the same; they only differ in their prosody: the
clitics are deaccented, whereas the strong forms receive sentence stress. Moreover,
some clitic forms are falling out of use and are replaced by strong variants. For
example, the 1st person accusative and genitive clitic mię is nowadays considered archaic
and the form mnie is normally selected instead.
5.2.2 The positions occupied by the

pronominal forms
The distribution of pronominal forms in Polish differs from the pattern found in the
South Slavic languages. In chapter 4 I noted that in Serbo-Croatian clitics must always
appear in the Wackernagel position, whereas in Bulgarian and Macedonian they are
verb-adjacent. In Polish pronominal clitics do not have to appear in the second (cf. 3a)
or in any other specific position. Their placement is largely conditioned by information
structure and prosodic factors. They are enclitic, so they need a phonological host to
The pronominal forms 225
their left. Therefore, in contrast to strong pronouns, they normally may not appear at
the beginning of a clause (cf. 3b), and are also avoided in sentence-final contexts (cf. 3c
and d). According to Franks (1998: 83), the latter restriction is due to a constraint
against ending a prosodic phrase with a clitic. However, the pronominal clitics may
occur at the end of a clause if there is only one other constituent in the clause available
apart from them. Then the only option of avoiding the initial placement is to appear
clause-finally (cf. 3e).
(3) a. Często spotykam go na ulicy

often meetPRES.1SG himCL.ACC on street
“I often meet him in the street”
b. Jego/*go spotykam, ale z nim nie rozmawiam
himACC/himCL.ACC meetPRES.1SG but with himINSTR NEG conversePRES.1SG
“HIM, I meet, but I never talk to him”
c. ??Często spotykam go
often meetPRES.1SG himCL.ACC
d. Spotykam go często
meetPRES.1SG himCL.ACC often
“I meet him often”
e. Spotykam go
“I meet him (regularly)” (Pl, cf. Spencer 1991: 367-368)
Pronominal clitics accept a wide range of phonological hosts. They may even be hosted
by extra-clausal elements, such as unaccented coordinating conjuncts i and a (cf. 4).
This is a very rare pattern, which in South Slavic is found only in Bulgarian.114 It
indicates that the pronominal enclisis in Polish is a purely phonological requirement. It
is sufficient for the pronominal clitics to be supported by any phonologically overt
material, regardless of its grammatical category or feature specification.
(4) Janek poszedł do tego pana

Janek goPART.M.SG to this man
i mu wszystko powiedział
and himCL.DAT everything tellPART.M.SG
“Janek went to that man and told him everything” (Pl, Rappaport 1988: 321)
As far as the ordering of the pronominal clitics with respect to each other is concerned,
Polish differs from the South Slavic languages as well. In chapter 4 I showed that clitics
in South Slavic cluster and always appear in a specific order, which may never be
changed. For instance, the dative pronominal clitic must precede the accusative
pronominal clitic, and it is impossible to reverse their ordering.
(5) li Mod > AUX (except 3rd SG) > DAT > ACC > AUX 3rd SG
In Polish both the “dative-accusative” and the “accusative-dative” patterns are possible.
Cetnarowska (2003) observes that the choice of a particular ordering may reflect the
requirements of theme-rheme articulation. This is exemplified in (6b), which lists
answers to the question in (6a).
114I will show in section 5.3.2 that in contrast to Bulgarian, auxiliary clitics may not be supported
by extraclausal elements in Polish, though.
(6) a. Czy Tomek pożyczyl Joannie swój samochód?

if Tomek lendPART.M.SG JoannaDAT his-own carACC
“Did Tomek lend his car to Joanna?”
b. Tak, w końcu jej go pożyczył, mimo że
yes, in end herCL.DAT itCL.ACC lendPART.M.SG although
Basi pożyczyć go nie chciał
BasiaTOP.DAT lendINF itCL.ACC NEG wantINF
“Yes, eventually he lent it to her, even though he didn’t want to lend it to
Basia”
b.’ Tak, w końcu go jej pożyczył, mimo że
yes, in end itCL.ACC herCL.DAT lendPART.M.SG although
roweru pożyczyć jej nie chciał115
bicycleTOP.GEN lendINF herCL.DAT NEG wantINF
“Yes, eventually he lent it to her, even though he didn’t want to lend her a
bicycle” (Pl)
The two variants in (6b and b’) are synonymous. However, (6b) is concerned with
Joanna (referred to as jej ‘herCL.DAT’), who is interpreted as the topic and contrasted with
Basia. The sentence in (6b’) is about Tomek’s car (referred to as go ‘itCL.ACC’), which is
contrasted with his bicycle. In both cases, the topics occur first in the pronominal clitic
cluster.
The contrast between South Slavic and Polish is related not only to the ordering of
clitics, but also to their clustering. In Bulgarian and Macedonian pronominal clitics
always cluster with each other. In Polish they tend to cluster as well, but a sequence of
pronominal forms can be disrupted when it is necessary to prepose one of the clitics
for topic or focus reasons (cf. 7b).
(7) a. Jan chce mu go wypożyczyć

Jan want3SG himCL.DAT itCL.ACC lendINF
“Jan wants to lend it to him”
b. Jan mu chce go wypożyczyć a nie sprzedać
Jan himCL.DAT want3SG itCL.ACC lendINF and not sellINF
“Jan wants to lend it to him rather than sell it” (Pl)
Moreover, in the South Slavic languages pronominal clitics are adjacent not only to
each other, but also to the auxiliary clitics. In Polish pronominal clitics may be split
from auxiliaries by some other constituents, such as the l-participle (cf. 8a) or an adverb
(cf. 8b).
(8) a. Kiedy-ś widział go ostatnim razem?

when+AUX.2.SG seePART.M.PL himCL.ACC last time
“When did you see him last time?”
b. Kiedy-ś naprawdę go kupił?
when+AUX.2.SG really himCL.ACC buyPART.M.PL
“When did you really buy it? (Pl, cf. Rappaport 1988: 320-321)
The data in (6) through (8) demonstrate that pronominal clitics enjoy a remarkable
freedom of placement in the clause structure in Polish. This indicates that they undergo
115 The object roweru occurs in genitive, rather than accusative, because it is affected by the rule of
‘Genitive of Negation’, which is discussed in section 5.3.4.2.2.3.

The pronominal forms 227
phrasal movement and target XP positions (cf. Franks 1998), on a par with weak
pronouns in Macedonian.
The proposal receives additional support from the fact that pronominal clitics in
Polish do not observe the Person Case Constraint.
(9) Dała-by-m mu cię

givePART.F.SG+COND+AUX.1.SG himCL.DAT youCL.ACC
za żonę bez wahania
as wife without hesitation
“I would give you to him as a wife without hesitation”
(Pl, cf. Cetnarowska 2003)
The languages in which the Person Case Constraint is active require that the accusative
clitic following the dative appear in the 3rd person. In (9) the accusative clitic is specified
for the 2nd person, which means that the constraint does not hold in Polish. In chapter
4 I suggested that the constraint is observed only if the clitics adjoin to a single head.
This is what happens in Bulgarian and Macedonian, but not in Serbo-Croatian, where
each of the pronominal clitics is located in the specifier of separate agreement
projections. The well-formedness of (9) therefore strengthens the proposal developed
on the basis of the examples in (6) through (8) that pronominal clitics target different
XP positions in Polish, and that they may not become adjoined to a single head.
The only restriction concerning the placement of pronominal clitics mentioned so
far concerns their occurrence at the beginning of a clause. However, this does not mean
that their position in the clause structure is determined solely by prosodic requirements.
Witkoś (1998: 159) specifies the following sites that can be occupied by pronominal
clitics in Polish.
(10) A an adjacent postverbal position

B. an (adjacent) preverbal position
In other words, they have to be adjacent to the verb when they follow it (cf. 11a). They
do not have to be verb-adjacent when they precede it (cf. 11b).
(11) a. Widział-em go wczoraj na koncercie

seePART.M.SG+AUX.1SG himCL.ACC yesterday on concert
“I saw him yesterday at a concert”
a.’ ??Widziałem wczoraj go na koncercie116
b. Wczoraj go wszyscy widzieli na koncercie
yesterday himCL.ACC everyone seePART.VIR.PL on concert
“Everyone saw him at the concert yesterday” (Pl)
I propose to capture the adjacency condition in (11a’) by arguing that the clitic go must
reach Spec, AgrO, which is located immediately below the the position occupied by the
verb, in order to check accusative case. A detailed phrase structure of the Polish clause
will be provided in section 5.3.4.2.4.2, after the system of auxiliary forms is examined.
To conclude, the chart in (12) presents the most important differences between the
pronominal clitics in Polish and South Slavic.
116The acceptability of the sentence improves when the adverbial na koncercie is contrastively
focused.
(12) Properties of pronominal clitics in Polish and South Slavic
Bulgarian Macedonian Serbo-Croatian Polish

position verb-adjacent, verb-adjacent, Wackernagel enclitic
enclitic usually proclitic enclitic
ordering Dat-Acc Dat-Acc Dat-Acc Dat-Acc and
Acc-Dat
pronominal yes yes no no
clustering
adjacent to yes yes no only in the
the verb postverbal
position
type of X0 X0 XP XP
movement
the Person yes yes no no
Case
Constraint
The chart indicates that the pronominal forms in Polish enjoy a greater autonomy in
the clause structure than their South Slavic counterparts. This makes them seem to be
more like weak pronouns than clitics. They do not need to cluster or be adjacent to the
verb. They undergo phrasal movement and may surface in different orderings with
respect to each other.
5.3 The auxiliary forms and the syntax of

compound tenses
The present section will discuss the syntax of compound tenses in Polish and is
organized as follows. Section 5.3.1 will analyse the status of auxiliaries in Polish. Section
5.3.2.1 and 5.3.2.1 will show that the perfect and conditional auxiliaries can be either
clitics or affixes, and that the latter option is a more recent diachronic development.
Section 5.3.3 will discuss some previous accounts of the syntax of compound tenses in
Polish, while section 5.3.4 will present an alternative analysis.
5.3.1 The paradigm of the auxiliaries

Polish has two types of auxiliary clitics: perfect and conditional. The paradigm of the
perfect auxiliary forms is presented in (13). Just as in other West Slavic languages and
Macedonian, the singular and plural variants of the 3rd person are morphologically null.
(13) Perfect auxiliaries in Polish
SG PL
1 -m -śmy
2 -ś -ście
3 - -
Apart from the forms in (13), Polish has a conditional auxiliary by. This auxiliary carries
the same inflectional endings as the perfect auxiliaries in (13).
The auxiliary forms and the syntax of compound tenses 229
(14) Conditional auxiliaries in Polish
SG PL
1 by-m by-śmy
2 by-ś by-ście
3 by by
Polish has also a future auxiliary będzie. It is a perfective form of the verb ‘be’, with the
same person/number inflection as finite verbs in Polish (cf. section 5.3.4.2.4.1 for an
explanation of differences between finite verb inflection and the person/number
marking on the auxiliary clitics).
(15) Future auxiliaries in Polish
SG PL
1 będę będziemy
2 będziesz będziecie
3 będzie będą
The future auxiliary is not a clitic. It is accompanied by an infinitive or an l-participle as

the main verb. The choice of either form of verb does not influence the meaning of the
sentence.117
(16) a. Janek będzie parzył kawę

Janek bePRF.1SG brewPART.IMPF.M.SG coffeeACC
“Janek will brew coffee” / “Janek will be brewing coffee”
b. Janek będzie parzyć kawę
Janek bePRF.1SG brewINF.IMPF coffeeACC (Pl)
Będzie imposes aspectual restrictions on the main verb, and it may appear only with
imperfective forms. Hence, the variants of (16) constructed with the l-participle or the
infinitive marked for perfective aspect are ungrammatical (cf. 17a). In order to render a
future meaning with a perfective verb, the verb must be used in the present tense (cf.
17b).
(17) a. *Janek będzie zaparzał/zaparzać kawę

Janek bePRF.1SG brewPART.PRF.M.SG/brewINF.PRF coffeeACC
b. Janek zaparzy kawę
Janek brewPRF.PRES.3SG coffeeACC
“Janek will brew coffee” (Pl)
The fact that będzie is followed by the infinitive might give an impression that this is a
bi-clausal construction. However, Grenoble (1995) argues that this is not the case,
117Nitsch (1956) and Mikoś (1985: 454) observe that the use of the l-participle following będzie is
the most common when the subject of the clause is masculine. The infinitive is found more often
when the subject is plural or feminine. Both of them claim that the choice of the variant of the
main verb is dictated by stylistic considerations. See Whaley (2000a: 53ff) for more discussion.
because the future auxiliary cannot form double negatives (cf. 18a). In this way it differs
from modal verbs, such as móc ‘be able to’, which admit two negatives.118
(18) a. Janek nie będzie (*nie) chodzić do szkoły

Janek NEG bePRF.1SG NEG goINF to school
b. Janek nie może nie chodzić do szkoły
Janek NEG can1SG NEG goINF to school
“It is not the case that Janek may not go to school” (Pl)
Since będzie is not a clitic, it patterns like other finite verbs in Polish. Therefore, I
postpone discussion of its syntactic behaviour to section 5.3.4.2.4.
5.3.2 Positions of the auxiliaries

As far as the position of the auxiliary clitics is concerned, two strategies can be
distinguished. First, the auxiliaries can appear attached to the l-participle as affixes. This
is the pattern they follow in contemporary Polish most often. It is exemplified for the
perfect auxiliary in (19a) and for the conditional auxiliary in (19b).
(19) a. Rano wyjechali-śmy z Wrocławia119

morning leavePART.VIR.PL+AUX.2PL from Wrocław
“We left Wrocław in the morning”
b. Rano wyjechali-by-śmy z Wrocławia
morning leavePART.VIR.PL+COND+AUX.2PL from Wrocław
“We would leave Wrocław in the morning” (Pl)
Second, the auxiliary may appear after the first constituent in main clauses, and function
as a second position clitic. This variant dates back to the period when Polish was a
Wackernagel clitic language (cf. chapter 1, section 1.3.4.2.2.1), but is still productive,
particularly in the Southern dialects (cf. Topolińska 1961). However, in Modern Polish
this is not the default strategy of the auxiliary clitic placement, and the element
preceding it is always interpreted as topicalized or focused. The auxiliary most often
follows the subject, especially when it is realized as a pronoun. Constructions with the
auxiliary in the 2nd singular, 1st and 2nd plural are quite standard (cf. 20a); the ones with
the 1st singular form are archaic (cf. 20b),120 and the 3rd singular and plural forms do not
exist, because the auxiliary is null in the 3rd person (cf. the chart in 13).
118 The sentence in (18b) does not exemplify negative concord, because the two negations are
interpreted compositionally. See Błaszczak (2001) for an extensive discussion.
119 In the Polish orthographic convention the auxiliary affix is written together with the l-
participle. However, for the sake of clarity of the presentation, the affix is italicized and preceded
by a hyphen.
120 In chapter 1, section 1.3.4.2.2.2 I report that the 1st person singular form of the auxiliary was
the first one to be reanalyzed as an affix on the l-participle. The reason for this seems to be
morphological ambiguity of the auxiliary clitic, which is the same as an instrumental case
morpheme in one of the case paradigms.
(20) a. My-śmy wyjechali z Wrocławia rano

we+AUX.1PL leavePART.VIR.PL from Wrocław morning
“We left Wrocław in the morning”
a’. My-by-śmy wyjechali z Wrocławia rano
we+COND+AUX.1PL leavePART.VIR.PL from Wrocław morning
“We would leave Wrocław in the morning”
b. %Ja-m wyjechał z Wrocławia rano (dated)
I+AUX.1SG leavePART.M.SG from Wrocław morning
“I left Wrocław in the morning”
b’. Ja-by-m wyjechał z Wrocławia rano
I+COND+AUX.1SG leavePART.M.SG from Wrocław morning (Pl)
The auxiliary clitic may also be preceded by a topicalised prepositional phrase (cf. 21a),
an object (cf. 21b and c), an adverbial, an AP (cf. 21d), a complementizer (cf. 21e), or a
wh-phrase (cf. 21f).
(21) a. Z Warszawy-śmy wyjechali rano

from Warsaw+AUX.1PL leavePART.VIR.PL morning
“We left Warsaw in the morning” (… but we left Kraków in the evening)
b. Książkę-ś o Polsce napisał
book+AUX.2SG about Poland writePART.M.SG
“You wrote a book about Poland” (… and not an article)
c. Książkę o Polsce-ś napisał
book about Poland+AUX.2SG writePART.M.SG
“You wrote a book about Poland” (… and not about the Czech Republic)
d. Bliski-ś mi jak brat
close +AUX.2SG meCL.DAT as brother
“You are as close to me as my brother”
e. Zauważył-em, że-ś uciekł
noticePART.M.SG+AUX.1SG that+AUX.2SG run-awayPART.M.SG
“I noticed you’ve run away”
f. Kiedy-ście przyjechali?
when+AUX.2PL arrivePART.VIR.PL
“When did you arrive?” (Pl, cf. Mikoś & Moravcsik 1986)
However, the auxiliary clitic may not encliticize on all types of lexical elements. For
instance, coordinating conjunctions i and a may not lend support to auxiliary clitics (cf.
22) even though they may serve as hosts for pronominal clitics (cf. 4 above).
(22) a. *Poszedłem do tego pana

goPART.M.SG+AUX.1SG to this man
i-m mu wszystko powiedział
and+AUX.1SG himCL.DAT everything tellPART.M.SG
“I went to this man and told him everything”
b. Poszedłem do tego pana i mu wszystko powiedział-em
(Pl, cf. Rappaport 1988: 321)
There is some disagreement in the literature concerning the range of positions that can
be occupied by the auxiliary clitic. Franks (1998) and Franks and Bański (1999) argue
that when the auxiliary is not an affix on the l-participle (cf. 23a), it may appear
anywhere in the clause as long as it precedes the l-participle (cf. 23e).
(23) a. My znowu wczoraj poszli-śmy do parku

we again yesterday goPART.VIR.PL +AUX.1PL to park
“We went to the park yesterday again”
b. My-śmy znowu wczoraj poszli do parku
c. My znowu-śmy wczoraj poszli do parku
d. My znowu wczoraj-śmy poszli do parku
e. *My znowu wczoraj poszli do parku-śmy (Pl, cf. Franks 1998)
These examples have been repeatedly quoted by different authors, but their
acceptability varies among native speakers. For example, neither I nor any other native
speakers I have consulted find (23d) to be acceptable. In fact, (23c) is not completely
well-formed, either; it is only saved by the fact that the auxiliary encliticizes on the
adverbial znowu ‘again’, which here must be necessarily interpreted as focused.121 The
most felicitous position for the auxiliary clitic is (23b), in which it immediately follows
the first element in the sentence.
Thus, the claim that the auxiliary clitics may appear “anywhere” in the clause
preceding the l-participle is definitely too strong. It seems that the clitics show a strong
tendency to appear in the second position (cf. Mikoś & Moravcsik 1986 for a similar
view).
Summarizing, it has been shown that there are two patterns of the auxiliary
placement in Polish: affixation on the l-participle and encliticization on non-verbal
hosts in the clause initial positions. The next two sections will examine these two
patterns in detail and demonstrate how they differ.
5.3.2.1 Auxiliaries as clitics

The following sections will discuss the properties of the auxiliary clitics in Polish. I have
mentioned earlier that whenever the auxiliary is not affixed on the l-participle, it must
encliticize on the clause-initial element. However, it is still necessary to determine
whether the process is only triggered by phonological requirements of the auxiliary
enclitic, or whether it has syntactic effects as well.
5.3.2.1.1 Placement of the auxiliary clitic – syntax or phonology?

The auxiliary clitics tend to occupy the second position in Polish. In this way they
resemble clitics in Serbo-Croatian. In chapter 4, section 4.4.1.3.1 I reported that
according to some analyses, cliticization in Serbo-Croatian is determined by rules of
phonology, because clitics seem capable of splitting syntactic constituents. As indicated
in (24), at first blush the auxiliaries may appear inside constituents in Polish as well.
121 In this way Polish resembles Czech, which is a very closely related language. Clitics must
appear in the Wackernagel position in Czech. However, there are exceptional contexts in which
clitics may occur embedded further in the clause than in the second position, and encliticize on
the element that is focused (e.g. Petr in ib). This pattern is often referred to as “clitic slippage” in
the literature (cf. Short 2002: 495).
(i) a. Helena říkala, že se Petr odstĕhoval

Helena sayPART.F.SG that CL.REFL Petr move-awayPART.M.SG
“Helena said that Peter had moved”
b. Helena říkala, že PETR se odstĕhoval
“Peter, Helena said, had moved” (Czech, Franks 1998)
(24) a. Którego-ście studenta oblali?

which+AUX.2PL student failPART.VIR.PL
“Which student did you fail, after all?”
b. Tego-m piosenkarza znał
thisACC+AUX.1SG singerACC knowPART.M.SG
“I knew this singer!”
c. Bardzo-śmy świadomi tych problemów byli
very+AUX.1PL awareVIR.PL these problems bePART.VIR.PL
“We were aware of these problems VERY well”
d. Bardzo świadomy-ś tych problemów był
very awareVIR.SG+AUX.2SG these problems bePART.M.SG
“You were aware of these problems very well”
(Pl, Borsley & Rivero 1994: 395ff; Rappaport 1988: 322)
Thus, in (24a) the auxiliary clitic encliticizes on the wh-word którego, which is separated
from its complement student. Likewise, in (24b) the auxiliary intervenes between the
demonstrative tego and the DP piosenkarza. In (24c) the auxiliary splits the adverb bardzo
from the adjective it premodifies. Example (24d) indicates that the clitic does not have
to appear after the first head. Here it follows the first phrase (AP), bardzo świadomy. The
AP is separated from the copula był by the scrambled DP tych problemów.
However, as in the case of Serbo-Croatian, placement of these elements in front of
the auxiliary clitics can be shown to be a result of left-branch extraction, which is widely
available in Polish (cf. Corver 1992; Rappaport 2000, and Bošković 2005). Moreover,
the examples in (25) indicate that the left branch extraction may also occur across non-
clitic elements, so it is not conditioned by the clitic status of the auxiliary verb.
(25) a. Którego oblejesz studenta?

which failPRES.PRF.2SG student
“Which student are you going to fail?”
b. Tego chcę zobaczyć piosenkarza, który jest najlepszy
thisACC want1SG seePRF.INF singerACC who be3SG best
“I want to see the singer who is the best”
c. Bardzo jesteśmy świadomi tych problemów
very be1PL awareVIR.PL these problems
“We are aware of these problems very well”
d. Bardzo świadomy jestem tych problemów
very awareM.SG be1SG these problems
“I am aware of these problems very well” (Pl)
Furthermore, it is possible to show the clitic may only follow an element that has
independently undergone syntactic movement. For example, neither the auxiliary clitic
nor a lexical verb may intervene between a preposition and its complement (cf. 26).
From a phonological point of view, this position should be accessible for a clitic,
because the preposition is an independent phonological word that can be stressed.
(26) a. do Krakowa-ś pojechał

to Cracow+AUX.2SG goPART.M.SG
“You went to Cracow”
a’. *do-ś Krakowa pojechał
b. do Krakowa jedziesz
to Cracow goPRES.2SG
“You are going to Cracow”
b’. *do jedziesz Krakowa (Pl, cf. Borsley & Rivero 1994: 406)
However, from a syntactic point of view the restriction is to be expected, because

prepositions can never be extracted in Polish.
(27) a. O czym Jan rozmawia?

about what Jan talk3SG
“What is Jan talking about?”
b. *Czym Jan rozmawia o? (Pl, cf. Borsley & Rivero 1994: 406)
Thus, I conclude that in Polish, just as in Serbo-Croatian, placement of the auxiliary

clitics is determined by syntax, rather than phonology.
5.3.2.1.2 Prosodic and morphological interactions between the auxiliary and

its host
We have seen that the auxiliary clitic may only appear after the first syntactic position in
the clause. This section will provide more evidence for the idea that even though the
auxiliary clitic is phonologically impoverished, it is an independent syntactic unit.
As an enclitic, the auxiliary needs phonological support. However, this does not
mean that it forms a morphological or prosodic word with its host. Example (28)
shows that the second auxiliary clitic may be optionally deleted in a coordinated clause
consisting of two adjectives if the other clitic is identical.
(28) Zmęczony+m i głodny(+m)

tiredADJ.M.SG+AUX.1SG and hungry(+AUX.1SG)
“I am tired and hungry” (Pl, cf. Franks & Bański 1999)
The deletion of the second clitic should be prohibited if the auxiliary were a part of the
inflection of the host, because this would involve deletion of a part of a word. The fact
that this is possible indicates a non-morphological relation between the auxiliary and its
adjectival host.
Furthermore, encliticization of the auxiliary on non-verbal elements does not
influence word stress placement, which would be expected if the process were a result
of a phonological operation. Word stress is very regular in Polish and almost without
exception falls on the penultimate syllable, which is capitalized in (29).
(29) a. któREgo
which
b. BArdzo
very (Pl)
As demonstrated in (30), the penultimate stress of the word followed by the auxiliary
clitic is retained. This means that the clitics does not enter the prosodic word of the
non-verbal host.
(30) a. któREgo-ście
whichACC+AUX.2PL
b. BArdzo-śmy
very+AUX.1PL (Pl, cf. Bański 2000: 65)
However, even though the encliticization of the auxiliary is a syntactic process, its actual
realization might be influenced by prosodic factors. Bański (2000: 96) points out that
the encliticization is possible when the final syllable of the clitic host does not have a
coda; that is, when it ends in a vowel.122
(31) a. do Krakowa-ś pojechał

to CracowGEN+AUX.2SG goPART.M.SG
“You went to Cracow”
b. *do Katowic-ś pojechał
to KatowiceGEN+AUX.2SG goPART.M.SG
b’. do Katowic pojechał-eś
to KatowiceGEN goPART.M.SG+AUX.2SG (Pl)
Otherwise, when the auxiliary encliticization is blocked for phonological reasons, a

focus particle że might be inserted to provide appropriate support for the clitic (cf.
section 5.3.4.1.3 for a detailed description of że support).
(32) do Katowic-że-eś pojechał

to KatowiceGEN+FOC+AUX.2SG goPART.M.SG
“You went to Katowice?” (Pl)
The next section will investigate the properties of the auxiliary when it is affixed on the
l-participle.
5.3.2.2 Auxiliaries as affixes

The auxiliaries in Polish may also appear as affixes on the l-participle. This is a more
common strategy than the encliticization of the auxiliary on the clause-initial element. It
also represents a more recent stage in the diachrony of the syntax of compound tenses.
When the auxiliary is an affix on the l-participle, the relationship between the two
elements is closer than in the case of auxiliary clitics and their non-verbal hosts. Suitable
evidence for it comes from morphological and prosodic patterns.
First of all, as an affix, the auxiliary may not be separated from the participle by any
overt material (cf. 33a), not even by pronominal clitics (cf. 33b).
(33) a1. Kupili-śmy lustro

buyPART.VIR.PL+AUX.1PL mirror
“We bought a mirror”
a’. *Kupili lustro-śmy
122This is a descriptive statement and certainly not a rule. The actual realization of this tendency
varies among speakers.
b 1. Dałe-ś mi go
givePART.M.SG+AUX.2SG meCL.DAT itCL.ACC
“You gave it to me”
b’. *Dał mi go -ś (Pl, cf. Borsley & Rivero 1994: 386; 394)
In this way Polish differs from the South Slavic languages, in which the pronominal
clitics always intervene between the l-participle and the 3rd person auxiliary.
(34) a. Dal mu go e
givePART.M.SG himCL.DAT itCL.ACC beAUX.3SG
“He gave it to him”
b. Dal sŭm mu go
givePART.M.SG beAUX.1SG himCL.DAT itCL.ACC
“I gave it to him” (Bg)
In Polish the 3rd person singular and plural forms of the auxiliary disappeared from the
paradigm between the 15th and the 17th. This was an important step in the evolution of
the compound tenses in Polish, because it meant that the l-participle and the auxiliary
clitics to the right of it became unexceptionally adjacent. Thus, the loss of the 3rd
person form certainly speeded up the reanalysis of auxiliary clitics as affixes. However,
as will be demonstrated by using a variety of prosodic and morphological tests, the
process has still not been completed.
The first test is related to the phonological rule of Vowel Raising and has been
described by Booij & Rubach (1987), among others. The rule applies to stems that have
/o/ before a voiced consonant in the last syllable. In some specifically defined contexts,
when the vowel is not followed by any other vowel within the same word, it raises to
[u], which is orthographically represented as ó. The rule is informally described in (35).
(35) /o/ → [u] / _ +voice C0 # (B. Hermans, p.c.)
In this way the following outputs are obtained.
(36) a. Anna pomogła

Anna helpPART.F.SG
“Anna helped”
b. Jan pomógł
Jan helpPART.M.SG
“Jan helped” (Pl)
As indicated in (37), the rule does not operate when the perfect auxiliaries are added.
This suggests that they are not perceived by speakers as separate words (that is, as
clitics), but rather as parts of the verb (that is, as affixes). Therefore, /o/ does not
count as the last vowel in the word.
(37) a. ja mogł-em b. my mogl-iśmy

I canPART.M.SG+AUX.1SG we canPART.VIR.PL+AUX.1PL
“I could” “We could” (Pl)
By contrast, the rule is at work with the conditional auxiliary by followed by the perfect
auxiliary. This indicates that by is not an affix.
(38) Ja mógł-by-m
I canPART.M.SG+COND+AUX.1SG
“I could have” (Pl)
However, there are some other prosodic tests, which indicate that the reanalysis of the
perfect auxiliary clitics as affixes has not been completed, because the auxiliaries are not
uniformly interpreted as affixes throughout the whole paradigm (cf. also section
1.3.4.2.2.2 in Chapter 1). One of them is related to stress assignment, which in Polish
almost without exception falls on the penultimate syllable. The paradigm of the
perfective auxiliaries shows a split between the singular and the plural forms, because
the stress assignment rule applies only in the singular. The whole paradigm is presented
in (39) for the l-participle czytał ‘read’, with the stressed syllable capitalized and the
auxiliaries in italics.
(39)
SG PL
1 czyTAł-em czyTAli-śmy
2 czyTAł-eś czyTAli-ście
3 CZYtałM czyTAliVIR
czyTAłaF czyTAłyNON-VIR
czyTAłoN
Thus, the stress patterns show that the reanalysis of the auxiliary as an affix is more
advanced among the singular forms. The perfective auxiliaries in the plural are still
treated as clitics, given that the part+aux complexes receive the antepenultimate stress.
Still, an increasing number of speakers applies the rule to the plural forms as well, and
stress the penultimate syllable, even though this is still considered substandard by
normative linguists.
Example (38) indicates that the conditional auxiliary by is not an affix, because its
attachment to the l-participle does not trigger the rule of Vowel Raising. This fact is
further verified by stress assignment in the ‘participle+by+perfect auxiliary’ formations.
As the chart in (40) demonstrates, the conditional auxiliary does not enter the prosodic
word of the host, on a par with the plural forms in (39).
(40)
SG PL
1 CZYtał-by-m czyTAli-by-śmy
2 CZYtał-by-ś czyTAli-by-ście
3 CZYtał-byM czyTAli-byVIR
czyTAła-byF czyTAły-byNON-VIR
czyTAło-byN
There is also syntactic evidence for the divergent status of the perfect auxiliaries in the
singular on the one hand, and the conditional auxiliaries and perfect auxiliaries in the
plural on the other hand. Mikoś & Moravcsik (1986) and Franks & Bański (1999) point
out that the singular forms of identical perfect auxiliaries cannot be deleted in structures
with coordinated l-participles.
(41) a. Czytał-em i ziewał-em

readPART.M.SG+AUX.1SG and yawnPART.M.SG+AUX.1SG
“I was reading and yawning”
b. *Czytał-em i ziewał-em (Pl)
In this way the auxiliary -em resembles the person/number inflection on finite verbs,
which cannot be deleted under identity, either.
(42) a. Daj-esz i bierz-esz

give2SG and take2SG
“You give and take”
b. *Dajesz i bierz-esz (Pl, cf. Bański 2000: 77)
However, it is possible to elide the perfect auxiliary in the plural, but only when the
normative, antepenultimate stress pattern is applied (cf. 43a). When the participle is
stressed in the “substandard” way, the coordination with a deleted perfect auxiliary is
ruled out (cf. 43b).
(43) a. CzytAli-śmy i piSAli

readPART.VIR.PL+AUX.1PL and writePART.VIR.PL
“We read and wrote”
b. *CzytaLI+śmy i pisaLI (Pl)
This shows that the auxiliary deletion is impossible for the speakers that treat the
participle-auxiliary complex as a single word, with the auxiliary being an inflection affix
on the l-participle.
To my knowledge, all speakers stress the penultimate syllable of the l-participle
when it is followed by the conditional auxiliary by. Therefore, the conditional auxiliaries
can be always deleted in structures with coordinated l-participles under identity.
(44) CZYtał-by-m i PIsał-by-m

readPART.M.SG+COND+AUX.1SG and writePART.M.SG (Pl)
To summarize, it has been shown that the auxiliaries in Polish exhibit two different
morphological patterns: they can be either clitics or affixes on the l-participle. The
intriguing question is whether these two patterns can receive a single syntactic
treatment. Four previous approaches to this problem will be presented in the next
section.
5.3.3 Some previous accounts of the syntax of

compound tenses in Polish
A number of analyses of the structure of compound tenses in Polish have been
suggested in the literature (cf. Witkoś 1998 and Błaszczak 2001 for an overview). Here
I selected only four, which represent different ways of accounting for the divergent
properties of the auxiliary clitics and affixes. The proposals make various statements as
to the position of the auxiliary clitics and affixes in the clause structure, the projections
in which the perfect and conditional auxiliaries originate, as well as to whether
affixation and encliticization of the auxiliary takes place derivationally (i.e., in syntax) or
in the lexicon.
5.3.3.1 Borsley & Rivero (1994)

Borsley & Rivero (1994), henceforth B&R, offer the earliest syntactic account of the
compound tenses in Polish. Many of the later analyses were put forward in response to
B&R’s approach; therefore it is necessary to present their assumptions in some detail.
B&R argue that both affixation of the auxiliary on the l-participle (cf. 45a) and the
encliticization of the auxiliary on non-verbal forms (cf. 45b) occur via head movement.
In the first instance the l-participle incorporates into the auxiliary. In the second
instance the l-participle stays in situ, and the auxiliary encliticizes on a non-verbal
element, such as the subject pronoun ty at PF
(45) a. Ty czytał-eś książkę

you readPART.M.SG+AUX.2SG bookACC
“You have read the book”
IP
Ty I’
I VP
V NP
czytałi eś
ti książkę
b. Ty-ś czytał książkę

you+AUX.2SG readPART.M.SG bookACC
IP
Ty I’
I VP
V NP
-ś
czytał książkę
As far as the position of the auxiliaries is concerned, B&R assume that both the perfect
and the conditional forms occupy I0, because they are inseparable from each other (cf.
46b). Thus, the sequence of the l-participle chciał and the auxiliary complex by-ś in (46a)
receives the analysis given in (46a’).
(46) a. Chciał-by-ś
wantPART.M.SG+COND+AUX.2SG
“You would like to”
a’. [I Vi [Aux COND]] ti
b. *By-chciał-eś (Pl, cf. Borsley & Rivero 1994: 389)
Let me briefly evaluate B&R’s account on the basis of the arguments they put forward
themselves, as well as the data presented in section 5.3.2.
B&R’s idea of the participle-to-auxiliary incorporation gains support from the
penultimate stress assignment, which is frequently recalculated when the auxiliary is
affixed to the l-participle (cf. the chart in 39 and the examples in 47). This suggests that
the participle-auxiliary complex is interpreted as a single word.
(47) a. Ty-ś CZYtał książkę

you+AUX.2.SG readPART.M.SG bookACC
“You’ve read the book”
b. Ty czyTAł-eś książkę
you readPART.M.SG+AUX.2.SG bookACC (Pl)
Moreover, some evidence for B&R’s hypothesis comes from divergent requirements
for adjacency between the l-participle and the auxiliary, which are related to the
ordering between these two constituents. As demonstrated in (48), when the auxiliary
occurs to the right of the l-participle, it must be adjacent to it. It may never be separated
from the l-participle by any other constituent (cf. 48a), not even by pronominal clitics
(cf. 48b).
(48) a. Dał-eś mi książkę

givePART.M.SG+AUX.2SG meCL.DAT bookACC
“You gave me a book”
a’. *Dał książkę-eś mi
b. Dał-eś mi go
givePART.M.SG+AUX.2SG meCL.DAT itCL.ACC
“You’ve given it to me”
b’. *Dał-mi-go-ś (Pl, Borsley & Rivero 1994: 389)
By contrast, the auxiliary need not be left-adjacent to the l-participle. Whenever it

encliticizes on non-verbal clause-initial elements, it can be separated from the participle
by other constituents, such as the direct object in (49a and b) or the adverbial in (49c).
(49) a. Ewy-ś książkę czytał

EwaGEN+AUX.2SG bookACC readPART.M.SG
“It’s Ewa’s book you have read”
b. Ty-ś ten list napisał
you+AUX.2SG this letterACC writePART.M.SG
“You’ve written this letter” (… and not read it)
c. My-śmy szybko opróżnili lodówkę

we+AUX.1PL quickly emptyPART.VIR.PL fridgeACC
“We have quickly emptied the fridge” (Pl)
Thus, it is possible to make a generalization that whenever the l-participle raises, it must
incorporate (or left-adjoin) to the auxiliary. When the l-participle remains in situ (cf. 49),
the auxiliary encliticizes on the clause-initial element.
As far as the position occupied by the auxiliary is concerned, B&R observe that it
is equally accessible for present tense verbs (cf. 50b) and for the l-participle (cf. 50c).
(50) a. Ewy-ś książkę czytał

EwaGEN+AUX.2SG bookACC readPART.M.SG
“It’s Ewa’s book you have read”
b. Ewy czytasz książkę
EwaGEN read2SG bookACC
“It’s Ewa’s book you’re reading”
c. Ewy czytałe-ś książkę
EwaGEN readPART.M.SG+AUX.2SG bookACC
“It’s Ewa’s book you have read” (Pl, Borsley and Rivero 1994 397-398)
In B&R’s view this indicates that finite verbs, auxiliary clitics, and the l-
participle+auxiliary complexes have the same distribution and that they may all occupy
I0. The direct object Ewy is claimed to undergo scrambling and move to Spec, IP, as
exemplified for (50a) in (51; cf. Borsley & Rivero 1994: 402)
(51)
IP
NP I’
I VP
Ewyk
NPi
-ś
tk N VP
V ti
książkę
czytał
However, B & R acknowledge that there are types of clauses in which certain positions
are restricted for movement of the l-participle and finite verbs, but not for the
auxiliaries. For instance, the auxiliaries may appear between two wh-words (52a and b),
but this position is inaccessible for the l-participle (cf. 52c).
(52) a. Co komu-ś dał?

what whom+AUX.2SG givePART.M.SG
“What did you give to whom?”
a’. Co-ś komu dał?
b. Gdzie kto by poszedł?
where who COND goPART.M.SG
“Who would go where?
b’. Gdzie by kto poszedł?
c. Co komu dał-eś?
what whom givePART.M.SG+AUX.2SG
“What did you give to whom?”
c’. *Co dał-eś komu? (Pl, Borsley & Rivero 1998: 410)
The well-formedness of (52a and b) is unproblematic for B&R, as they assume,

following Rudin (1988), that only the first wh-word fills the Spec, CP position in Polish,
whereas the other wh-phrases are adjoined to IP. On this approach the two wh-words
precede I0, which allows them to maintain the idea that the auxiliary is always located in
the same position. In order to account for the acceptability of (52a’ and b’), in which
the auxiliary appears after the first wh-word, they suggest that the auxiliary can be
adjoined to IP in the same way as a wh-word. According to B&R, this proposal is
supported by the fact that in clauses with three wh-elements, the auxiliary may intervene
between each of them.123
(53) a. Co by-ś komu kiedy dał?

what COND+AUX.2SG whom when givePART.M.SG
“What would you give to whom when?”
b. Co komu by-ś kiedy dał?
c. Co komu kiedy by-ś dał? (Pl, cf. Borsley & Rivero 1998: 409)
In my view the suggestion that the auxiliary, which is a head, may adjoin to IP has a
dubious value on the assumption that heads may adjoin only to other heads. Apart
from that, the proposal that the auxiliary may optionally adjoin to other categories
seems quite ad hoc and weakens the initial hypothesis that it is always located in a single
syntactic site.
B&R do not explain why the positions between the wh-words cannot be accessed
by the l-participle.124 Clearly, the clitic status of the auxiliary does not play a role here;
because the non-clitic future auxiliary będzie may intervene between wh-phrases as well,
as B&R observe themselves.
123 B&R do not mention that the auxiliary may also follow all the three wh-words (cf. 53c). I add
this example for consistency of the presentation.
124 In fact, B&R’s statement is too strong, because the l-participle may be sandwiched between
two wh-elements when the first wh-word is not D-linked. See Dornisch (1995) for details and
Lubańska (2005) for an in-depth analysis of wh-movement in Polish.
(54) a. Co komu będziesz rozdawał?

what whom bePRF.2SG give-awayPART.M.SG
“What will you give away to whom?”
b. Co będziesz komu rozdawał? (Pl, cf. Borsley & Rivero 1998: 409)
Moreover, B&R’s idea that both auxiliary encliticization and incorporation of the l-
participle into the auxiliary affix occur in the same structure does not pay attention to
different interpretations triggered by these two processes. As was noted in section
5.3.2.2, the affixation of the auxiliary onto the l-participle is the default strategy of
forming compound tenses and conditional structures in Modern Polish. Conversely,
encliticization of the auxiliaries into the clause-initial element always gives rise to a
focused or topicalized reading. B&R argue that this elelement targets Spec, IP, on a par
with subjects, but crosslinguistically the occupants of the subject position are not
normally interpreted as focused. In fact, it will be shown in section 5.3.4.1 that the
constituents that encliticize onto the auxiliary undergo A’-movement, so they cannot
target Spec, IP (TP).
Furthermore, Dornisch (1997) and Witkoś (1998 ch. 1) point out empirical
shortcomings of B&R’s proposal. For instance, they observe that when the l-participle
is affixed to the auxiliary, it can be either preceded or followed by a pronominal clitic
(cf. 55a). However, when the auxiliary encliticizes to an element at the beginning of the
clause, the pronominal clitic can occur only between the auxiliary and the l-participle
(cf. 55b), but it may not precede the auxiliary (cf. 55b’).
(55) a. My (go) spotkali-śmy (go) wczoraj

we himCL.ACC meetPART.VIR.PL+AUX.1PL himCL.ACC yesterday
“We met him yesterday”
b. My-śmy (go) spotkali (go) wczoraj
we+AUX.1PL himCL.ACC meetPART.VIR.PL himCL.ACC yesterday
b.’ *My go-śmy spotkali wczoraj (Pl, cf. Dornisch 1997: 191)
The acceptability difference between (55a) and (55b’) poses a problem for B&R’s idea
that the auxiliary in Polish is always generated in the same position. If this were the
case, the clitic go should always be able to appear on both sides of the l-participle.125
In spite of the shortcomings mentioned above, B&R’s hypothesis successfully
captures the fact that the perfect and conditional auxiliaries in Polish can be both
enclitics and auxiliaries in syntactic terms. In the latter case they stay in a much closer
relationship with the l-participle than in the corresponding constructions in South
Slavic. Therefore, in an alternative account of the syntax of compound tenses in Polish
developed in section 5.3.4 I will adopt B&R’s idea that the l-participle incorporates into
the auxiliary in I0 (T0).
5.3.3.2 Bošković (1997)

Boskovic (1997) extends his analysis of participle fronting in Serbo-Croatian (cf.
chapter 2, section 2.2.2) to Dutch and Polish. In his brief overview of Polish data he
claims that in Polish, just as in Serbo-Croatian, the participle head-adjoins to the
auxiliary located in Aux0. The process is triggered by an [+aux] feature. The idea that
125In my opinion, the sentence in (55b’) is excluded because the auxiliary clitic does not appear in
the second position.
the two elements are adjoined gains support from the fact that no lexical material may
intervene between the l-participle and the conditional auxiliary (cf. 56c).
(56) a. [AuxP Chciałi-eś [VP ti tę książkę]]

wantPART.M.SG+AUX.2SG this book
“You wanted this book”
b. Chciałi+by+śj tj ti tę książkę
wantPART.M.SG+COND+AUX.2SG this book
“You would want this book”
c. *Chciał tę książkę by-ś
wantPART.M.SG this book COND+AUX.2SG (Pl, Bošković 1997: 173-174)
Unlike B&R, who suggest that both the perfect auxiliary and the conditional auxiliary by
are generated in the same head position, Bošković argues that by is higher in the
structure and takes the AuxP headed by the perfect auxiliary as its complement. In this
way he accounts for the fact that the participle is always adjoined to the highest verbal
head in the sequence. If both the perfect auxiliary and the conditional auxiliary are
present, the participle may adjoin only to the latter. This is evidenced by the example in
(57), in which the presence of a higher verbal head above the l-participle precludes its
adjunction to the perfect auxiliary -(e)ś.
(57) *By chciał-eś tę książkę

COND wantPART.M.SG+AUX.2SG thisACC bookACC
“You would want this book” (Pl, Bošković 1997: 174)
Bośković concludes the l-participle in Polish adjoins to the highest verbal head
available. However, if there is more than one auxiliary in the clause, by must adjoin to
the highest verb as well (cf. 56b).
Bośković’s analysis of the compound tense formation in Polish is not extensive
and is used by him mainly to support his theories of participle movement in Dutch and
Serbo-Croatian. However, his idea that the conditional auxiliary is generated in a
different position than the perfect auxiliary will be used in an alternative analysis of
participle movement in Polish presented in section 5.3.4.
5.3.3.3 Szczegielniak (1997)

Szczegielniak tries to recast B&R’s proposal in a Minimalist theory of feature checking
(Chomsky 1995). Moreover, drawing on the ideas of Cardinaletti and Starke (1999: 187)
concerning the architecture of clitic pronouns (cf. chapter 4, section 4.4.2.4.2.4.1), he
argues that clitic auxiliaries have a similar impoverished morphological structure. The
poverty can be compensated either by establishing a Spec-head relationship with Σ0,
which is the locus of polarity, focus, and prosody-related features, or by incorporation
into Σ0. According to Szczegielniak, the auxiliary clitic in Polish resorts to the latter
option. Since Σ0 may encode topic and focus features, movement of an element to
Spec, ΣP gives rise to its focused or topicalized interpretation. This is what happens in
(58), where the subject ty is interpreted as topicalized or focused. The auxiliary -ś raises
as a head and adjoins to Σ0.
(58) a. Ty-ś zabił Janka

you+AUX.2SG killPART.M.SG JanekACC
“You killed Janek”
b. [ΣP Ty [Σ -śi] [T zabiłj] [Aux ti] [ VP [V tj] Janka ]] (Pl, cf. Szczegielniak 1997)
When Σ is not present in the derivation, the auxiliary circumvents its structural
impoverishment by incorporation into the l-participle, which moves to T0. The
incorporation usually leads to creation of a prosodic unit, which consists of the
participle and the clitic. The derivation is presented in (59b).
(59) a. Ty zabił-eś Janka

you killPART.M.SG+AUX.2SG JanekACC
“You killed Janek”
b. [TP Ty [T zabiłj+eśi] [AuxP ti [VP [V tj ] Janka ]]]] (Pl, cf. Szczegielniak 1997)
In contrast to (58), the subject in (59) is hosted in Spec, TP, rather than Spec, ΣP. This
correlates with the fact that it does not have to be interpreted as focused or topicalized.
If the conditional auxiliary by is present in the Numeration, the derivation proceeds
as sketched in (60b)
(60) a. Ty zabił-by-ś Janka

you killPART.M.SG+COND+AUX.2SG JanekACC
“You would kill Janek”
b. [TP Ty4 [T zabił3+by2+ś1] [MP t2 [AuxP t1 [vP[DP t4] [ v t3] [VP[V t’3] [DP Janka ]]]]]]
The VP is created by merging the participle zabił and the object Janka. The subject ty is
generated in Spec, vP. Next, the perfect auxiliary ś merges with vP to form AuxP,
which is subsequently merged with by and gives MP (Mood Phrase). Alternatively, MP
might be located below AuxP, and the conditional auxiliary picks up the perfect
auxiliary on its way up the structure. Then T is selected and merged with MP, which
gives rise to TP. Finally, the subject ty moves to Spec, TP and checks the EPP feature
of T0.
If Σ0 is present in the derivation, it attracts the ‘by+ the perfect auxiliary’ clitic
complex, which circumvents its structural deficiency in this position. Σ0 also encodes a
Topic/Focus feature, which forces movement of the subject to Spec, ΣP (cf. 61b).
(61) Ty by-ś zabił Janka

you COND+AUX.2SG killPART.M.SG JanekACC
“You would kill Mark”
b. [ΣP Ty4 [Σ by2+ś1 [TP t’4 [T zabił3+t’2+t’1] [MP t2 [AuxP t1 [VP[DP t4] [ V t3] [VP[V t’3]
[DP Marka ]]]]]]]]
Szczegielniak submits that the l-participle moves to T0 in order to check Tense. This
forces him to assume that the l-participle has overt Tense morphology. As he
acknowledges himself, this is problematic, because the l-participle occurs also in future
tense constructions (cf. 62), so it is a stipulation to suggest that it has a past
interpretation.
(62) Ja będę malował stół

I bePRF.1SG paintPART.M.SG tableACC
“I will be painting the table” (Pl)
Consequently, he proposes that there are two distinct types of the l-participle in Polish.
I consider this suggestion a weakness in his analysis.
Szczegielniak argues that the movement of the l-participle to the auxiliary is
optional, because it does not occur if the auxiliary can encliticize into an element that
precedes it, such as the subject or a focused/topicalized element. Hence, he concludes
that the operation cannot be regarded as feature driven as in Chomsky (1995).126
Following Chomsky’s 1995 autumn lectures, Szczegielniak proposes that head
movement takes place at PF. He states that PF movements are not feature driven, but
exemplify “category conflation”, which results from a tendency to fuse all verbal or
nominal projections by head movement. Thus, he claims that the movement of the
auxiliary to Σ0 is triggered by the need to eliminate morphological deficiencies of the
raised element. However, this proposal is inconsistent with his idea that the l-participle,
which in his view also raises as a head at PF, checks the Tense feature of T0. In other
words, it is not clear why only certain types of head movement may lead to feature
checking if they all occur at PF.
Crucially, Szczegielniak’s claim that participle movement occurs at PF is not on the
right track. It was shown in chapter 2 that the operation may have semantic effects, so
it does not take place only to provide support for the phonologically impoverished
auxiliary. Constituents of various types may be preposed to the position preceding the
auxiliary, and depending on the element that undergoes movement, the sentence
receives a different interpretation. Thus, the movement is not optional, because it is
related to the intended meaning of the clause. For example, encliticization of the
auxiliary into the subject pronoun in (61) gives rise to its topicalized or focused
interpretation. This type of reading does not need to occur when the l-participle is
affixed to the auxiliary in T0.
In spite of these shortcomings, Szczegielniak offers a very interesting proposal
concerning the structure of compound tenses in Polish. His ideas, and especially the
proposal of a ΣP in the clause structure were a major source of inspiration for the
theory of compound tenses in Polish developed in section 5.3.4.
5.3.3.4 Witkoś (1998)

The theories of the participle-auxiliary orders in Polish presented so far have been
largely derivational in nature. The participle has been argued to undergo head
movement or head incorporation by raising from some lower projection to I0 (Borsley
& Rivero 1994), Aux0 (Bošković 1997), and T0 or Σ0 (Szczegielniak 1996). The analysis
due to Witkoś (1998), which will be reviewed in this section, represents a lexicalist
approach. Since Witkoś’s proposal is a challenge both to the previous analyses as well
as to the alternative derivational account I develop in section 5.3.4, his ideas will be
scrutinized in detail.
126 Note that the “optionality” of movement does not exclude feature checking. For instance, a
feature in a functional head may make movement possible, but it does not force it to apply
overtly. The overt displacement might be enforced by some other factors, such as discourse
structure requirements.
Much in the spirit of the early version of the Minimalist Program (Chomsky 1995),
Witkoś assumes that verbs enter the computational component of grammar as fully
inflected forms. In his view this means that the “l-participle-perfect auxiliary”
complexes are always generated as single words. Moreover, he suggests that since
movement is dispreferred and regarded as a costly operation in the Mininalist
framework, there should be no overt raising of the l-participle to I0 in Polish unless
substantial evidence to the contrary is found.
In order to substantiate his claim concerning the lack of verb movement in Polish,
Witkoś (1998: 27ff) refers to an analysis of verb and adverb placement by Śpiewak &
Szymańska (1995); henceforth Ś&Sz. Ś&Sz demonstrate that adverbs may follow and
precede VP, as well as separate a verb from its nominal complement.
(63) a. Tomek ostrożnie podniósł dziewczynkę

Tomek carefully liftPART.M.SG little-girl
“Tomek lifted the girl carefully”
b. Tomek podniósł ostrożnie dziewczynkę
c. Tomek podniósł dziewczynkę ostrożnie
(Pl, Śpiewak & Szymańska 1995: 130)
Hence, the position of adverbs seems to be quite free and apparently reflects
information structure requirements. However, there is a group of adverbs, such as
prawie ‘nearly’, naprawdę ‘really’, and po prostu ‘simply’, which may only precede the verb.
(64) a. Ja prawie skończył-em swoją pracę

I nearly finishPART.M.SG+AUX.1SG REFL work
“I nearly finished my work”
b. *Ja skończył-em prawie swoją pracę
c. *Ja skończył-em swoją pracę prawie (Pl, cf. Śpiewak & Szymańska 1995: 131)
The distribution of the adverb prawie exhibits the same restrictions when the auxiliary
encliticizes to it (cf. 65a) or some other non-verbal element, such as the pronominal
subject wy (cf. 65b).
(65) a. Prawie-ście skończyli swoją pracę

almost+AUX.2PL finishPART.VIR.PL REFL work
“You have almost finished your work”
b. Wy-ście prawie skończyli swoją pracę
you+AUX.2PL almost finishPART.VIR.PL REFL work
c. *Wy-ście skończyli prawie swoją pracę
d. *Wy-ście skończyli swoją pracę prawie (Pl, cf. Witkoś 1998: 28-29)
Ś&Sz remark that the restriction in the position of prawie might be due to the fact that it
is a scopal element, which requires the verb to be in its scope. Movement of the verb
across it is prohibited, because it results in an illegitimate reversal of scope (cf. 64b and
65c).
Witkoś seems to ignore the special scopal semantics of the adverb prawie ‘almost’
and states that the lack of verb movement across it is unexpected on the assumption
that the verb leaves the VP, as claimed by Borsley & Rivero (1994). Moreover, he
suggests that the position of the auxiliary clitic in (65a) following the adverb prawie
indicates that the adverb has been topicalized across the auxiliary.127 Thus, the adverb
data lead him to posit that adverb placement does not favour the verb movement
hypothesis.
I consider Witkoś’s interpretation of the adverb positions to be misguided. He
disregards the special scopal effects triggered by the scopal adjuncts and attempts to
generalize the prohibition on verb movement across them to all contexts. This line of
reasoning is not on the right track, which can be shown on the basis of interactions
between prawie and another adjunct, such as the temporal adverbial dzisiaj ‘today’.
(66) a. Wy-ście dzisiaj prawie skończyli swoją pracę

you+AUX.2PL today almost finishPART.VIR.PL your work
“You have almost finished your work today”
b. Wy-ście prawie skończyli dzisiaj swoją pracę
c. *Wy-ście prawie dzisiaj skończyli swoją pracę
d. *Wy-ście dzisiaj skończyli prawie swoją pracę (Pl)
The examples in (66 a and b) show that the temporal adverb dzisiaj may follow or
precede the l-participle, which indicates that the participle may move across it.
However, the temporal adverb dzisiaj may not separate the scopal adjunct prawie from
the l-participle, because then prawie would scope over the temporal adverb rather than
the VP. Witkoś claims that the lack of participle movement across prawie manifests lack
of participle movement out of the VP. If this were the case, sentence (66b) should be
unacceptable, because here the l-participle has raised across the temporal adverb.
In spite of his initial claims that there is little evidence for overt verb movement in
Polish, Witkoś (1998) points out that that the position of subjects indicates that the l-
participle may leave the VP. In (67a) the subject receives an existential reading, and
remains in its thematic base-position. The subject in (67b) has a generic interpretation,
as it characterizes a general property of children in the past. It has raised from its base
position and presumably occupies Spec, TP.
(67) a. Na ulicy bawiły się dzieci

on street playPART.NV.PL REFL children
“Children were playing in the street”
b. Dzieci bawiły się na ulicy
children playPART.NV.PL REFL on street
“Children used to play in the street” (cf. Pl, Witkoś 1988: 87)
Since the existential subject in (67a) occurs to the right of the l-participle, the l-
participle must have moved from its base position within the VP. However, contrary to
B&R’s claims, Witkoś argues that the l-participle never moves very high, and that it
does not reach the I/T projection. Given that it does not carry any tense specification,
it cannot be attracted by the Tense feature in T.128 He proposes that it only targets
127This is an unusual claim, because adverbs are normally taken to be immobile elements and can
be used indicators of verb movement (cf. Cinque 1999; Pollock 1989).
128 On similar grounds Lavine (2000 ch. 3) argues against derivation of the incorporated
participle-auxiliary forms via participle movement to I0. He states that the idea is problematic
from a feature-checking perspective, because the feature that drives this movement of the
incorporated form would remain unchecked when the auxiliary encliticizes into a non-verbal
element and the participle-to-auxiliary incorporation does not apply. In my view, the argument
does not hold. T0 (or I0) contains both a Tense feature, which is checked by the auxiliary, as well
AspP, where aspectual morphology is licensed. The proposal is supported by the fact
that the l-participle always carries aspectual specification.
(68) [AgrSP subject [TP T [AspP [Asp l-participle] [AgrOP object [VP V ... DP]]]]]
(cf. Witkoś 1998: 99)
Unlike the affixes, both the perfect and conditional auxiliaries are generated in T0, and
are therefore considerably higher than the l-participle. They may move still further to
AgrS. According to Witkoś (1998: 97), in clauses with the conditional auxiliary by, T0
has either a strong or a weak verbal [+v] feature. If it is strong, the l-participle raises
from Asp0 to T0 and incorporates into the conditional (cf. 69a). If the feature is weak,
the incorporation does not take place (cf. 69b).
(69) a. Ty zjadł-by-ś ciastko

you eatPART.M.SG+COND+AUX.2SG cakeACC
“You would eat the cake”
a’. [AgrSP tyi [AgrS zjadłj-byśk [TP [T tk] [AspP [Asp tj] [AgrOP ciastkol [AgrO tj][VP ti tj tl ]]]]]]
b. Ty-by-ś zjadł ciastko
you+COND+AUX.2SG eatPART.M.SG cakeACC
b’. [AgrSP tyi [AgrS byśk [TP [T tk] [AspP [Asp zjadłj] [AgrOP ciastkol [AgrO tj][VP ti tj tl ]]]]]]
By contrast, there is no optionality in the feature strength in the case of perfect

auxiliaries, and the [+v] feature of T is always weak, so it never triggers movement of
the l-participle. The non-incorporated (cf. 70a) and incorporated (cf. 70b) variants
result from two different numerations.
(70) a. Ty-ś zjadł ciastko

you+AUX.2SG eatPART.M.SG cakeACC
“You ate the cake”
a’. [AgrSP tyi [AgrS śk [TP [T tk ] [AspP [Asp zjadłj] [AgrOP ciastkol [AgrO tj][VP ti tj tl ]]]]]]
b. Ty zjadł-eś ciastko
you eatPART.M.SG+AUX.2SG cakeACC
“You ate the cake”
b’. [AgrSP tyi [AgrS śk [TP [T tk ] [AspP [Asp zjadł-eś j] [AgrOP ciastkol [AgrO tj][VP ti tj tl ]]]]]]
Thus, on Witkoś’s approach there are two ways in which the l-participle-auxiliary
structures are generated in Polish: the one with the conditional auxiliary by is created
derivationally, whereas the structure with the perfect auxiliary enters syntax as a fully-
inflected form.
Witkoś’s approach is certainly an interesting attempt at a lexicalist analysis of the
structure of Polish compound tenses. However, the problem is that the author confines
himself to strict Minimalist restrictions on overt movement, and tries to rigidly follow
them even when data speak to the contrary. Furthermore, he stipulates that the
incorporation of the l-participle into the conditional auxiliary hinges on the strength or
weakness of the [+v] feature. Yet, it is unclear what kind of feature is represented by
[+v], and how it is decided whether it is weak or strong in a given context. The
proposal that the [+v] feature of T0 is always weak is equally unjustified. According to
as φ-features, which are checked by the l-participle. The φ-features are interpretable, so they do
not need to be checked.
Chomsky (1995), feature strength may correspond to overt presence of a morphological

element that realizes this feature. In Polish both the conditional morpheme and the
perfect auxiliary are lexically overt, so there is no principled reason to assume that
either of them differs in strength.
Witkoś’s analysis also faces empirical problems. If the participle-auxiliary
complexes enter the derivation as single words, it is unexpected that one of the
auxiliaries in the plural may be deleted under identity in a sentence with two
coordinated participles.
(71) Czytali-śmy i pisali(-śmy)

readPART.VIR.PL+AUX.1PL and writePART.VIR.PL+AUX.1PL
“We read and wrote” (Pl, Mikoś & Moravcsik 1986)
Admittedly, it was shown in section 5.3.2.2 that the auxiliaries are being reanalyzed as
affixes on the l-participle, so when the change is completed in the future, it might be
correct to argue that the l-participle enters the derivation fully inflected. Such a claim,
however, is still unjustified for Modern Polish.
By means of summary, I present the main claims of the approaches presented in
this section in the chart in (72).
(72)
B&R (1994) Bošković Szczegielniak Witkoś (1998)

(1997) (1997)
position of the I0/I0 Aux0/Aux 0 Aux0/ comes from the
auxiliary affix Mood0, lexicon together
(perfect/conditional) moves to T0 with the verb
position of the I0/I0 not discussed Σ0 T0/T0,
auxiliary enclitic moves to AgrS0,
(perfect/conditional) optionally with
the l-participle
type of movement X0 X0 X0 X0
5.3.4 An alternative account

We have seen that the auxiliaries in Polish appear in two variants: synthetically as affixes
on the l-participle, or analytically as second position clitics. The former option is a more
recent diachronic development and prevails in Modern Polish; the latter is reminiscent
of the Wackernagel law, which was observed in Old Polish. In chapter 1, section
1.3.4.2.2.1 I pointed out that the reanalysis of the auxiliary as an affix started already in
the 16th century and has been proceeding very slowly over time. According to
Kowalska’s (1976: 63) calculations, the auxiliary appears as the second position clitic
only in 3% of compound tense formations in her corpus of Polish texts from the first
half of the 20th century. However, in spite of the rarity of these cases, the process seems
to be far from completion.
Furthermore, it has been reported that the reinterpretation of the auxiliary status
does not proceed uniformly within the paradigm, and that it is completed only among
the singular forms. Thus, it seems that the auxiliaries are subject to two diachronic
changes, rather than just one: a shift from the Wackernagel clitic to an affix on the l-
participle, and the spread of the affix form within the singular and the plural paradigms.
The important question is whether these two processes are really related, or perhaps
one of them occurs independently of the other.
Franks & Bański (1999) analyze the change in the position of the auxiliary by using
Kroch’s (1989) theory of “Grammar Competition”. According to this theory, a
diachronic change occurs in a language when there are two competing analyses of a
single linguistic phenomenon in the individual grammars of native speakers. Language
change is achieved when one of the analyses wins over the other. Franks & Bański
suggest that in the case of Polish the competition seems to be between the two
interpretations of the auxiliary: as a clitic or as an affix on the l-participle. In Modern
Polish there is a strong preference for the latter. However, if the affixed variants enjoy
such a predominant position in the grammar, and the process of the reanaysis has been
taking place for so many centuries, it is surprising that the auxiliary clitics still exist.
Bański (2000: 195) argues that the slow rate of the process is due to the fact that the
competition between grammars involves three options, rather than just two. The third
option covers the present tense copula ‘be’, which is constructed with an old strong
form of the verb ‘be’ jest and person-number affixes (cf. 73a; see section 5.3.4.2.4 for
details). They are morphologically the same as the affixes on the l-participle (cf. 73b; see
also section 1.3.4.2.2.1 in chapter 1, and section 5.3.4.2.4 in the present chapter), but
different from the ones found on finite verbs (cf. 73c).
(73) a. Jest-eśmy bardzo zmęczeni

be+1PL very tired
“We are very tired”
b. Kupili-śmy ciekawą książkę
buyPART.VIR.PL+AUX.1PL interesting book
“We bought an interesting book”
c. Jedzi-emy na wycieczkę
go+1PL on trip
“We are going on a trip” (Pl)
It is not clear to me why Bański singles out the copula ‘be’ as a special verbal case that
hinders the reinterpretation of the auxiliary clitics in Polish as affixes.
In my view, the diachronic change is far from the end because in fact there is no
competition between the two forms of the auxiliary in Modern Polish at all. I would like
to argue that auxiliary affixation and the auxiliary enclitization correspond to two
different syntactic processes, which make use of a morphologically identical form of the
auxiliary, but which are syntactically and semantically independent of each other.
The auxiliary always encliticizes onto the elements that have been moved to the
clause initial position for reasons of focus or topicalization. It may also move by itself
to a functional head located in the left periphery of the clause in which a special
grammatical function, such as subjunctive mood, polarity or focus is encoded. Thus,
the operation is performed in order to express special semantic concepts which are
found less often in language, and therefore auxiliary encliticization is statistically a less
common option than auxiliary affixation, which is used in order to render the temporal
relations in the clause.
When the auxiliary is right-adjacent to the l-participle, most speakers treat it as an
affix in the singular, and as an enclitic in the plural. Some speakers have already
reinterpreted the auxiliary as an affix throughout the whole paradigm. However, all of
them are able to perform auxiliary encliticization in order to express focus or modality.
Hence, the only process of language change that is taking place at the moment consists
in the extension of the auxiliary affixation to the whole paradigm of the “l-
participle+auxiliary” complexes. It is entirely independent of the auxiliary enclitization
on clause-initial elements.
The situation in which the auxiliary enclitics and the auxiliary affixes performed the
same grammatical function in Polish would be very uneconomical. It would imply
presence of morphological doublets, that is the forms which are morphologically the
same, but which are not functionally differentiated. Since both variants of the auxiliary
still exist in the language, this means that they are not in competition, because they are
used for different reasons.
The analysis developed here presupposes that since the auxiliary clitic and the
auxiliary affix are formally and functionally unrelated, they should occur in different
syntactic positions. This implies that it should also be possible to find contexts in which
only one of the options is available. This is indeed the case, as is exemplified in (74).
The sentence contains a verb of volition chcieć, which requires that the auxiliary is
encliticized on the complementizer, rather than affixed on the l-participle in the
embedded clause (see section 5.3.4.1.1 for details).
(74) a. Chcę, że-by-ś przestał mi przeszkadzać

want1SG that+COND+AUX.2SG stopPART.M.SG meDAT disturb
“I want you to stop disturbing me”
b. *Chcę, że przestał-byś mi przeszkadzać (Pl, cf. Borsley & Rivero 1994)
Moreover, texts written in older variants of Polish contain clauses in which the two
positions of the auxiliary are filled simultaneously. Thus, in the examples in (75) the
auxiliary is doubled and appears simultaneously on the l-participle and the clause-initial
element, which is interpreted as topicalized or focused.129
(75) a. Tedy-m ja owszeki stracił-em miasto

then+AUX.1SG I irretrievably losePART.M.SG+AUX.1SG city
“Then I irretrievably lost the city”
b. Iż-eś ty pobił-eś wszytki sprzeciwiające się mnie
that+AUX.2SG you beatPART.M.SG+AUX.2SG all opposing REFL me
“That you defeated all those who opposed me”
(16th c. Pl, Decaux 1955: 34; cf. Bański 2000: 125)
The following sections will discuss these two types of the auxiliary placement and will
provide support for the claim that the auxiliary participates in divergent syntactic
operations, depending on whether it is a clitic (cf. section 5.3.4.1) or an affix (cf. section
5.3.4.2). I will begin with cases of enclitization, which comprise conditional and
subjunctive structures formed with the enclitic by (cf. section 5.3.4.1.1), topicalization
(cf. section 5.3.4.1.2), and że-support (cf. section 5.3.4.1.3). Subsequently, I will address
the syntax of the incorporated l-participle-auxiliary forms in Polish, which will be
contrasted with the corresponding structures in South Slavic (cf. section 5.3.4.2). The
conclusions that I reach there will allow me to make definite claims about the position
of different auxiliary verbs in the structure of Polish (cf. section 5.3.4.2.4) and finally,
about the syntax of the ‘l-participle-auxiliary’ constructions (cf. section 5.3.4.2.5).
129Note that this phenomenon resembles complementizer agreement in West Flemish (cf.
Haegeman 1992) and Bavarian (cf. Bayer 1984).
5.3.4.1 Auxiliary encliticization

The following sections will examine instances of encliticization of the auxiliary onto
clause initial elements. It will be shown that these elements are always located higher
than TP, and occupy projections that license different types of modality, focus, and
topicalization.
5.3.4.1.1 The conditional / subjunctive auxiliary by

This section analyzes the auxiliary by. It will be demonstrated that it occurs in two types:
as a conditional and as a subjunctive marker, depending on whether it encliticizes on
the l-participle or a clause-initial element. The two types of by will be argued to target
different syntactic positions.
5.3.4.1.1.1 The position of by

By is an invariant conditional auxiliary, which is always immediately followed by the
perfect auxiliary. Since these two elements are inseparable, it is often argued (cf. Borsley
& Rivero 389-390; Witkoś 1998) that they enter the derivation as a single inflected
‘conditional auxiliary+perfect auxiliary’ complex. This view receives support from the
fact that when the perfect auxiliary is attached to by, it cannot be elided in coordinated
clauses under identity (cf. 76b). By contrast, it is possible to elide both of the auxiliaries
simultaneously (cf. 76a; compare the data in section 5.3.2.2).
(76) a. Włączyli-by-śmy radio i

turn-onPART.VIR.PL+COND+AUX.1PL radio and
posłuchali-by-śmy muzyki
listenPART.VIR.PL+COND+AUX.1PL music
“We would turn on the radio and listen to music”
b. *Włączyli-by-śmy radio i posłuchali-by-śmy muzyki (Pl, Bański 2000: 80)
However, I would like to argue for a derivational treatment of the ‘by + perfective
auxiliary’ complexes (cf. Dogil 1987, Booij and Rubach 1987; Bošković 1997, and
Dornisch 1997 for a similar view). If they were to emerge from the lexicon together
with the perfect auxiliary as a single unit, they should be always able to appear together.
Still, only by can be attached to impersonal participles.130 Encliticization of the perfect
auxiliary, either alone, or together with by on the impersonal participle results in
ungrammaticality.
(77) a. Włączono-by radio i posłuchano-by muzyki

turn-onIPT+COND radio and listenIPT+COND music
“One would switch on the radio and listened to music”
b. *Włączono-by-m radio i posłuchano-by-m muzyki
turn-onIPT+COND+AUX.1SG radio and listenIPT+COND+AUX.1SG music
c. *Włączono-m radio i posłuchano-m muzyki
turn-onIPT+AUX.1SG radio and listenIPT+AUX.1SG music (Pl)
130Impersonal participles were introduced in chapter 3, section 3.5. They never show subject
agreement and were argued not to be able to raise to T.
More evidence for the claim that by and the perfect auxiliary do not form a single lexical
word comes from language acquisition. Aguado and Dogil (1989) notice that Polish
children often place the conditional auxiliary independently of the perfect auxiliary,
which they choose to attach to the l-participle (cf. 78a; 78b represents standard Polish).
This demonstrates that it is possible to generate by independently of the perfect
auxiliary, so the two elements must occupy different positions in the clause structure.
(78) a. Basia powiedziała, że-by ją za godzinę (child language)

Basia sayPART.F.SG that+COND herACC after hour
obudził-eś
wakePART.M.SG+AUX.2SG
b. Basia powiedziała, że-by-ś ją za godzinę obudził
Basia sayPART.F.SG that+COND+AUX.2SG herACC after hour wakePART.M.SG
“Basia said that you should wake her up in an hour”
(Pl, Aguado and Dogil 1989: 106)
Hence, I would like to claim that by is merged as the head of MoodP, and left-adjoins to
the perfect auxiliary in T0, as is exemplified in (79) for the conditional auxiliary in (76a).
(79) …[T byi+śmy [Mood ti ]
5.3.4.1.1.2 The meanings of by

It has been observed (cf. Borsley and Rivero 1994; Mikoś and Moravcsik 1986) that the
‘by+perfect auxiliary’ complex is obligatorily attracted by some conjunctions, such as
gdy+by ‘if’ jak+by ‘as if’ o+by ‘I wish’/‘may’, and że+by ‘so that’, which express various
types of modal meanings, such as condition and potentiality in (80) or desire in (81).131
If the auxiliary is affixed on the l-participle rather than the complementizer, the result is
ungrammatical (cf. 80b and 81b).
(80) a. Gdy-by-m miał czas,

if+beCOND+AUX.1SG havePART.M.SG timeACC
poszedł-by-m do kina
goPART.M.SG+COND+AUX.1SG to cinema
“If I had time, I would go to the cinema”
b. *Gdy miał-by-m czas, poszedłbym do kina (Pl, Borsley & Rivero 1994: 418)
131It is a matter of debate whether these forms are formed in syntax via movement of by into the
position occupied by these conjunctions (as in Mikoś and Moravcsik 1986 and in the analysis
developed in this section), or whether they emerge from the lexicon as fully inflected items, as in
Bański (2000: 110). One of Bański’s arguments for a lexical approach to by-conjunctions is related
to the meaning of some of these forms, which according to him is not compositional. For
instance, aby consists of by and a coordinating conjunction a ‘and’, but the meaning of the whole
structure is ‘in order to’, which does not correspond to the meaning of a plus by. However, the
problem is that by may express a variety of modal meanings, such as potentiality, non-factuality,
or prediction, so it is difficult to provide a compositional meaning of the whole ‘by-conjunction’
complex without specifying the exact semantic contribution of by first.
(81) a. Że-by-ś tylko tego nie robił!

that+COND+AUX.2SG only this NEG doPART.M.SG
“May you never do that!”
b. *Że tylko tego nie robił-by-ś! (Pl, cf. Bański 2000: 113)
Correspondingly, the same requirement holds for the position of the auxiliary in
purpose clauses. It must occur right-adjacent to the complementizer że and may not be
affixed on the l-participle.
(82) a. Basia prosiła, że-by-ś ją za godzinę obudził

Basia askPART.F.SG that+COND+AUX.2SG herACC after hour wakePART.M.SG
“Basia asked you to wake her up in an hour”
a’. *Basia prosiła, że ją za godzinę obudził-by-ś (Pl, Dogil 1987: 40)
b. Muszę się dużo uczyć że-by-m zdał ten egzamin
must1SG REFL much studyINF that+COND+1SG passPART.M.SG this exam
“I have to study hard so that I pass this exam”
b’. *Muszę się dużo uczyć że zdał-by-m ten egzamin (Pl)
Likewise, by must encliticize on the complementizer że when it introduces a subjunctive

complement of volition verbs (cf. 83a and b). Że is the most common subjunctive
complementizer, but as demonstrated in (83c and d), it does not need to be present, or
it can be replaced by the complementizer a, which carries a similar meaning.
(83) a. Chcę, że-by-ś to zrobił

want1SG that+COND+AUX.2SG it doPART.M.SG
“I want you to do it”
b. *Chcę, że zrobił-by-ś to
c. Chcę, by-ś to zrobił
d. Chcę, a-by-ś to zrobił
(Pl, cf. Aguado & Dogil 1989: 105; Borsley & Rivero 1994: 418; Bański 2000:
82)
All the examples which require encliticization of by onto the complementizer contain
subjunctive complements, so it can be assumed that the auxiliary is attracted by a modal
feature related to subjunctivity located in a functional head in the left periphery of the
clause. The assumption is supported by the fact that if the auxiliary is not adjoined to a
complementizer and instead is affixed on the l-participle, only the indicative meaning is
normally possible (cf. 84a).
(84) a. Powiedział, że to zrobili-by-śmy indicative clause

sayPART.M.SG that it doPART.VIR.PL+COND+AUX.1PL
“He said we would do it”
b. *Powiedział, że to zrobili-by-śmy purpose clause
sayPART.M.SG that it doPART.VIR.PL+COND+AUX.1PL
“He told/asked us to do it”
(Pl, cf. Aguado & Dogil 1989: 105; Bański 2000: 84)
The indicative meaning may also obtain when the conditional auxiliary raises to a
position to the right of the complementizer że, as in (85). This leads to a focused
reading of the auxiliary complex byśmy.
(85) Powiedział, że by-śmy to zrobili indicative clause

sayPART.M.SG that COND+AUX.1PL it doPART.VIR.PL
“He said we would do it” (Pl, cf. Bański 2000: 85)
Nevertheless, there is some evidence that the movement of byśmy in (85) does not
involve adjunction of by into the complementizer. For instance, in contrast to the
subjunctive purpose clause in (86a), the complementizer and the conditional auxiliary
may be split by a covert or overt subject, such as my in the indicative clause in (86b).
Moreover, the two auxiliaries are written separately. According to Bański (2000: 84ff),
the orthographic convention reflects different positions of pro or overt subjects in these
two types of clauses.132
(86) a. Powiedział, że-by-śmy [pro]/ my to zrobili purpose clause

sayPART.M.SG that+COND+AUX.1PL we it doPART.VIR.PL
“He told us to do it”
b. Powiedział, że [pro]/my by-śmy to zrobili indicative clause
sayPART.M.SG that we COND+AUX.1PL it doPART.VIR.PL
“He said we would do it” (Pl, cf. Bański 2000: 86)
Furthermore, Bański (2000: 81ff) observes that prosodic requirements of by differ

depending on its occurrence in an indicative clause or in a subjunctive clause. In
general, by is an enclitic, so it may not appear at the beginning of a sentence. However,
it has been reported (cf. Booij & Rubach 1987: 40; Borsley & Rivero 1994: 389) that
some speakers accept by clause-initially. According to Bański, the speakers who permit
by in this position do so only when the sentence has a subjunctive interpretation.
(87) By teraz to zrobił

COND now it doPART.M.SG
“(For him) to do it now”
“*He would do it now” (Pl, Bański 2000: 83)
To summarize, it has been shown that that there are two types of conditional auxiliaries
in Polish. The auxiliary must encliticize onto a complementizer that introduces
subjunctive clauses, but it does not need to do so when the complementizer selects an
indicative clause. I suggest that this means that by may occupy two different positions in
Polish. The proposal is specific not only to Polish, because there is some evidence for
this idea in other Slavic languages, too. For example, in chapter 4, section 4.3.3.2 I
pointed out, following Willis (2000), that in Old Church Slavonic certain
complementizers such as a ‘if’ always attracted the conditional auxiliary, irrespectively
of the order of the constituents that followed it.
(88) a. A by bylъ sьde

if COND.3SG bePART.M.SG here
“If he had been here”
b. A by sьde bylъ
c. A by bylъ prorokъ
if COND.3SG bePART.M.SG prophet
“If he had been the prophet” (OCS, Vaillant 1977: 219)
132 It is more usual not to insert the subject, unless it needs to be focused or topicalized.
Some other auxiliaries, such as da, attracted the conditional auxiliary only optionally, as
in the two different translations of Luke 4:42.
(89) a. Drъžaaxõ i da ne bi otъšelъ otъ nixъ

held3PL him that NEG COND.3SG leavePART.M.SG from them
(OCS, Codex Marianus, Willis 2000: 330)
b. Drъžaaxõ i da bi ne otъšlъ otъ nixъ
held3PL him that COND.3SG NEG leavePART.M.SG from them
“And they held him, so that he would not leave them”
(OCS, Codex Zographensis, Willis 2000: 330)
It was argued in chapter 4 that the placement of the auxiliary was contingent on the
semantics of the complementizer. The complementizer that selected conditional
sentences, such as a, always attracted the auxiliary. By contrast, the complementizer da,
which usually introduced indicative clauses, did not have to be adjacent to the
conditional auxiliary.
Likewise, Tomić (2005) shows that in some contemporary Balkan Slavic languages
conditional auxiliaries also appear in different positions in indicative and subjunctive
clauses. For example, in Macedonian the auxiliary clitic ќe occurs to the left of the
subjunctive complementizer da (cf. 90a), but to the right of the indicative
complementizer deka (cf. 90b).
(90) a. Ќe da dojde
CL.MOD thatSUBJ comeSUBJ.PAST.3SG
“S(he) seems to have come”
b. Reče deka ќe dojde
sayPAST.3SG thatIND CL.MOD comePRES.1SG
“(S)he said that (s)he would come” (Mac, Tomić 2005: 365)
Given the crosslinguistic data from various Slavic languages I suggest that there are two
positions in which the conditional auxiliary by can be hosted in Polish. Following
Tomić’s (1995) terminology, I will dub the projection where the subjunctive mood is
licensed ModP, and I will refer to the lower modal projection licensing non-subjunctive
meanings as MoodP.
To begin with an analysis of indicative (non-subjunctive) clauses, I propose that
MoodP is located immediately below TP. The conditional auxiliary by obligatorily raises
from Mood to T via head movement (cf. 91a). Subsequently, the l-participle may raise
via head movement from within the VP and left-adjoin to the auxiliary complex in T
(cf. 91b). The movement of the l-participle is not obligatory. If it does not take place,
the auxiliary complex receives a focused interpretation (cf. 85 above).
(91) a. (Powiedział) że by-śmy im tę książkę pożyczyli

sayPART.M.SG that COND+AUX.1PL themCL.DAT this book lendPART.VIR.PL
“He said that we would lend this book to them”
a’. [C że [T byi+śmy] [AgrIOP im [AgrOP tę książkę [VP pożyczyli]]]]
b. (Powiedział) że pożyczyli-by-śmy im tę książkę
sayPART.M.SG that lendPART.VIR.PL+COND+AUX.1PL themCL.DAT this book
b’. [C że [T pożyczylij+byi+śmy] [AgrIOP im [AgrOP tę książkę [VP tj ]]]] (Pl)
It is evident that the l-participle adjoins to the auxiliary complex in T0, because it may
not be separated from it by any lexical material, whether it is a full DP (cf. 92a; cf.
Bošković 1997: 173) or a pronominal clitic (cf. 92b).
(92) a. *że pożyczyli tę książkę by-śmy

that lendPART.VIR.PL this book COND+AUX.1PL
b. *że pożyczyli ją by-śmy
that lendPART.VIR.PL it/herCL.ACC COND+AUX.1PL (Pl)
Moreover, it is possible to insert the subject to the left of the fronted l-participle. This
shows that the ‘by + perfect auxiliary’ complex is not incorporated into the
complementizer. This also indicates that the l-participle does not land in Spec, TP, the
way it does in Bulgarian and Serbo-Croatian. The subject my ‘we’ in (93) is topicalized.
(93) że my pożyczyli-by-śmy im tę książkę

that we lendPART.VIR.PL+COND+AUX.1PL themCL.DAT this book (Pl)
Apart from the subject, other elements may intervene between the complementizer and
the auxiliary (or the l-participle that has adjoined to the auxiliary), such as adverbs (cf.
94a) and topicalized objects (cf. 94b). Both of them receive a focused interpretation.
(94) a. (Powiedział) że jutro by-śmy im

sayPART.M.SG that tomorrow COND+AUX.1PL themCL.DAT
tę książkę pożyczyli
this book lendPART.VIR.PL
“He said that tomorrow we would lend this book to them”
b. (Powiedział) że tę książkę by-śmy im
sayPART.M.SG that this book COND+AUX.1PL themCL.DAT
wtedy pożyczyli
then lendPART.VIR.PL
“He said that we would lend this book to them then” (Pl)
By contrast, in the case of subjunctive clauses, the conditional auxiliary must

obligatorily be right-adjacent to the complimentizer. It cannot be separated from it by
any overt material. This is illustrated for an adverbial in (95a) and a subject in (95b).
(95) a. (Powiedział) że (*jutro) -by-śmy im jutro

sayPART.M.SG that tomorrow +COND+AUX.1PL themCL.DAT tomorrow
this book lendPART.VIR.PL
“He said that we should lend this book to them tomorrow”
b. (Powiedział) że (*wy)-by-ście wy im
sayPART.M.SG that youPL+COND+AUX.2PL youPL themCL.DAT
this book lendPART.VIR.PL (Pl)
As an illustration, the derivation of the embedded clause in (95b) is presented in (96b).

First by moves as a head from Mood and adjoins to the auxiliary in T. Next, the
complex is attracted by a subjunctive feature in Mod, which is located immediately
below the complementizer.
(96) a. że-by-ście wy im tę książkę pożyczyli

that+COND+AUX.2PL youPL themCL.DAT this book lendPART.VIR.PL
b. [C że [ModP byi+ściej] [TP wy [T <ti’ + tj>] [MoodP [Mood ti] [AgrIOP im [AgrOP tę
książkę [VP pożyczyli]]]] (Pl)
To summarize, this section has shown that the auxiliary enclitic by in Polish may occupy
two different positions in the clause, depending on its semantics. When it functions as
the subjunctive marker, it moves to the head of ModP above TP. When it is the
conditional auxiliary, it stays in its base position in Mood below TP.
(97) [CP [ModP BYSUBJ [TP [MoodP BYCOND ]]]]
5.3.4.1.2 Topicalization
Another context in which the auxiliary is not affixed on the l-participle involves
auxiliary encliticization on constituents that have been preposed for reasons of focus or
topicalization. It was shown in section 5.3.2.1.1 that the fronted elements comprise a
variety of different categories. The movement prevents the auxiliary enclitic from
occurring clause initially, so Borsley & Rivero claim that it happens at PF. This does not
seem to be the correct suggestion, because the process displays syntactic locality
restrictions. For instance, the auxiliary may not encliticize onto extra-sentential
elements, such as the vocative in (98a) or the conjunction (99a). Rather, it must
encliticize on the initial element that is part of the same clause, such as a wh-word (cf.
98b) or an adverbial (99b), or else it must be affixed on the l-participle (cf. 98c and 99c).
(98) a. *Jurku-ś, czemu otworzył drzwi?

JurekVOC+AUX.2SG why openPART.M.SG doorACC
“Jurek, why did you open the door?”
b. Jurku, czemu-ś otworzył drzwi?
c. Jurku, czemu otworzył-eś drzwi? (Pl, Mikoś & Moravcsik 1986: 330)
(99) a. *I-ś wtedy otworzył drzwi?

and+AUX.2SG then openPART.M.SG door
b. I wtedy-ś otworzył drzwi?
“And was it then that you opened the door”
c. I wtedy otworzył-eś drzwi? (Pl)
In section 5.2.2 I mentioned that pronominal clitics may be supported by conjuncts and
other extra-clausal elements, because this process occurs for purely phonological
reasons, so they accept any type of overt lexical material as prosodic support.
(100) a. I mu wtedy otworzył-eś drzwi?

and himCL.DAT then openPART.M.SG+AUX.2SG doorACC
“And you open opened the door for him then?”
b. *Mu wtedy otworzył-eś drzwi? (Pl)
This is not the case with the auxiliary enclitization on the focused/topicalized elements,
which leads me to conclude that this is a syntactically-constrained phenomenon.
According to Borsley & Rivero (1994), the perfect auxiliary is always generated in
I0, whether it is incorporated into the l-participle or encliticized onto the word
appearing in the first position. The clause-initial elements reach their ultimate landing
site by scrambling (cf. 101a), or in the case of wh-elements, via wh-movement (cf. 101b).
(101) a. Tego-m piosenkarza znał

thisACC+AUX.1SG singerACC knowPART.M.SG
“I knew this singer!”
b. Którego-ście studenta oblali?
which+AUX.2PL student failPART.VIR.PL
“Which student did you fail, after all?” (Pl)
All these clause-initial elements create an inseparable unit together with the auxiliary.
However, even though the unit is inseparable, the two elements do not form a
phonological word, because they resist penultimate stress assignment, as shown in (102)
for the example in (101b).
(102) a. *KtóreGO-ście studenta oblali?

b. KtóREgo-ście studenta oblali? (Pl)
Consequently, I conclude that the unit is formed in syntax. Since the auxiliary can be
preceded by phrasal elements, such as wh-words and pronouns, I suggest that they
occupy the specifier of the projection which is headed by the auxiliary. Following
Borsley and Rivero (1994: 399), I assume that these elements reach this position by
scrambling. Now, in order to determine whether the auxiliary really resides in I/T, as
Borsley and Rivero (1994: 399) suggest, and whether the elements that precede the
auxiliary target Spec, TP it is necessary to examine the properties of scrambling.133
It has been observed that scrambling in Polish shares a number of properties with
wh-movement (cf. Willim 1989; Borsley & Rivero 1994; Lubańska 2000). For instance,
both types of displacement are constrained by the same restrictions, such as the Wh-
island Constraint, the Complex NP Constraint, and the Left Branch Condition. As an
example, the sentence in (103a) illustrates wh-movement, whereas the ones in (103b)
demonstrate scrambling of an object to the clause-initial position in the context of the
Wh-island Constraint. See Willim (1989) for a discussion of the other constraints.
(103) a. *[CP Coj [TP pro zastanawiał-eś się

whatj wonderPART.M.SG+AUX.2SG REFL
[CP komui [TP Jan da ti tj ?]]]]
(to) whoi Jan givePRES.3SG ti tj
“*What did you think that John to whom gives?”
b. *Janaj [oni zastanawiali się [kto ti odwiedza tj ]]
Johnj [they wonderPART.VIR.PL REFL [whoi ti visitPRES.3SG tj ]]
“*Johni, they wondered who visited himi” (Pl, Willim 1989: 110; 124)
Wh-movement represents A’-movement. Since scrambling is subject to the same

conditions and restrictions as wh-movement, it must instantiate A’-movement as well.
This assumption is further verified by the fact that scrambling in Polish licences
parasitic gaps.
133 Borsley & Rivero (1994: 400-401) claim that scrambling in Polish “adjoins a phrase to a Xmax
in the clause and creates a chain with the characteristics of wh-chains”. My interpretation of this
statement is that scrambled elements adjoin to TP/IP by creating an A’-bar chain. Topicalization
seems to be the more appropriate term here.
Parasitic gaps, which are marked with e in (104), are compatible only with A’-
movement. Their existence depends on the presence of a real gap in the clause, which is
identified with t in (104). Bondaruk (1995) observes that parasitic gaps are attested in
clauses in which the real gap that licenses the parasitic gap results from scrambling (cf.
104a and b). The clauses in which scrambling does not occur are ill-formed due to the
lack of the gap (cf. 104a’ and b’). Thus, scrambling is the only way to create the required
gap. Since the traces left by scrambled constituents behave like A’-bound variables,
scrambling must take place via A’-movement.
(104) a. Ten artykuł Piotr odłożył t nie analizując e

this article Piotr put-awayPART.M.SG not analysing
“Piotr put away this paper without analysing”
a.’ *Piotr odłożył ten artykuł nie analizując e
Piotr put-awayPART.M.SG this article not analysing
b. Tę spódnicę trzeba dobrze wyprasować t,
this skirt one-should well ironINF
żeby móc nałożyć e
that+COND canINF put-onINF
“This skirt should be well-ironed before it is worn”
b.’ *Trzeba dobrze wyprasować tę spódnicę,
one-should well ironINF this skirt
żeby móc nalożyć e
that+COND canINF put-onINF (Pl, cf. Bondaruk 1995)
Moreover, Borsley & Rivero (1994: 399-401) postulate that scrambling in Polish must
be movement to an A’-position, because it has reconstruction properties, which is a
characteristic of A’-movement (cf. Déprez 1989). As shown in (105b), the anaphor
swojego can be bound by the object Jana even when it is topicalized to the clause-initial
position.
(105) a. Jan spotkał swojego brata

Jan meetPART.M.SG his-ownREFL brother
“Jani met hisi own brother”
b. Swojego brata Jan spotkał
his-ownREFL brother Jan meetPART.M.SG (Pl, Borsley & Rivero 1994: 400)
To conclude, the tests that have been carried out so far uniformly indicate that Polish
has scrambling of the A’-type. However, the sentences examined above did not include
auxiliary clitics. What remains to be shown is that the examples with auxiliary clitics
exemplify the same type of scrambling possibilities. Thus, the sentence in (106), which
corresponds to (104), indicates that scrambling of the object across an auxiliary clitic
licenses parasitic gaps.
(106) Tę książkę-ś odłożył t nie analizując e?

this book+AUX.2SG put-awayPART.M.SG not analysing
“Was it this book that you put away without analysing?” (Pl)
Likewise, the pairs in (107) demonstrate that movement of the direct object in front of
the auxiliary clitic exhibit reconstruction properties, as well.
(107) a. Ty-ś spotkał swojego brata

you+AUX.2SG meetPART.M.SG your-ownREFL brother
“Youi met youri own brother”
b. Swojego brata-ś spotkał
your-ownREFL brother+AUX.2SG meetPART.M.SG (Pl)
To summarize, the test I have carried out prove that the displacement of various
constituents across the auxiliary clitic exemplify A’-movement. This indicates that they
may not target Spec, TP (or Spec, IP), because it is an A-position. Following the
insights due to Szczegielniak (1996; cf. section 5.3.3.3), but contra Borsley & Rivero
(1994), I assume that in these cases the auxiliary clitic raises higher than I/T and that it
lands in Σ, which is a projection that encodes topic and focus features. The scrambled
elements land in Spec, ΣP, as exemplified in (108) for (107b).
(108) [ΣP <swojego brata>i [Σ -śj [TP [T tj ... [PartP spotkał ti]]]]]
5.3.4.1.3 Że-support
Section 5.3.2.1.2 showed that there are circumstances in which encliticization of the
auxiliary onto the preceding element is blocked for phonological reasons. This happens
most often when the last syllable of the clitic host has a highly sonorous coda (cf.
Bański 2000: 96). The only way then to save the derivation is to attach the auxiliary to
the l-participle as an affix (cf. 109b), or to insert the focus particle że, which will provide
appropriate phonological support for the clitic (cf. 109c).
(109) a. *do Katowic-ś pojechał

to KatowiceGEN+AUX.2SG goPART.M.SG
b. do Katowic pojechał-eś
to KatowiceGEN goPART.M.SG+AUX.2SG
“You went to Katowice”
c. do Katowic-że-eś pojechał
“You went to Katowice?!” (Pl)
When supported by że, the auxiliary clitic may appear in the clause initial position. As
shown in (110), the auxiliary-że complex may be followed by the subject (cf. 110a) or a
scrambled direct object (110b).
(110) a. Że-ś ty czytał tę ksiażke

FOC+AUX.2SG you readPART.M.SG this book
nie zaskoczyło nikogo
NEG surprisePART.N.SG nobody
“The fact that you read this book did not surprise any one”
b. Że-ś tych meili jeszcze nie wysłał
FOC+AUX.2SG theseGEN e-mailsGEN still NEG sendPART.M.SG
to my wiemy
this we know
“The fact that you haven’t sent these e-mails yet is known to us” (Pl)
This indicates once again that Borsley and Rivero’s (1994) idea that the auxiliary is
located in I0 in all contexts cannot be on the right track. If this were the case, it should
be impossible for the subject ty in (110a) to follow the auxiliary clitic.
As in the other types of auxiliary encliticization, że does not enter the prosodic
word of the preposed non-verbal element. This is shown in (111), where że does not
change the normal penultimate stress pattern.
(111) a. do KaTOwic-że-eś pojechał

“You went to Katowice?!”
b. KIEdy-że-ś tam poszedł
when+FOC+AUX.2SG there goPART.M.SG
“When did you go there?!” (Pl)
Że is homophonous with the indicative complementizer. However, the complementizer

że has a different distribution from the supportive particle że, and there is strong
evidence that the two elements constitute different categories in spite of their
morphological identity. For instance, insertion of the supportive że is obligatory only
when there is no suitable phonological host for the auxiliary clitic (cf. 112a). If the
auxiliary is affixed on the l-participle, że is superfluous and hence may not appear in the
structure (cf. 112b). If the auxiliary can encliticize on the clause initial element, że may
be optionally inserted to render additional emphasis to this element (cf. 112c).
(112) a. We wtorek-*(że)-śmy go nie widzieli

on Tuesday+FOC+AUX.1PL himCL.ACC NEG seePART.VIR.PL
“On Tuesday we didn’t see him”
b. We wtorek (*że) go nie widzieli-śmy
on Tuesday+FOC himCL.GEN NEG seePART.VIR.PL+AUX.1PL
c. Wtedy-(że)-śmy go nie widzieli
then FOC+AUX.1PL himCL.ACC NEG seePART.VIR.PL
“We didn’t see him then” (Pl, cf. Bański 2001)
By contrast, the indicative complementizer must always appear in the clause and is
never dropped.
(113) Powiedział, *(że) już zjedli-ście obiad

sayPART.M.SG that already eatPART.VIR.PL+AUX.2PL dinner
“He says that you have already had dinner” (Pl)
Furthermore, the two types of że may co-occur. The complementizer is always located
higher than the focus particle. It is also possible to place some lexical material between
the two elements (cf. 114b), which is then interpreted as topicalized or focused.
(114) a. Powiedział, że że-ście już zjedli obiad

sayPART.M.SG that FOC+AUX.2PL already eatPART.VIR.PL dinner
“He says that you have already had dinner”
b. Powiedział, że ten park że-ście już widzieli
sayPART.M.SG that this park FOC+AUX.2PL already seePART.VIR.PL
“He says that you’ve seen this park already” (Pl, cf. Bański 2001)
Moreover, the focus particle że may be accompanied by the interrogative

complementizer czy. As in the case of the indicative complementizer, że is located lower
than czy.
(115) Zastanawiam się, czy że-ście już to widzieli

wonderPRES.1SG REFL if FOC+AUX.2PL already it seePART.VIR.PL
“I wonder whether you have seen it yet” (Pl, cf. Bański: 2001)
The position of the supportive że with respect to other constituents suggest that even
though its occurrence is motivated prosodically, it is located in a designated syntactic
position in the left periphery of the clause. To my knowledge, there have been no
detailed accounts of że-placement specifying its syntactic position proposed so far. The
use of że-support seems to be increasing in spoken Polish (cf. Szpiczakowska 1988),134
which led Franks & Bański (1999) to suggest that the ‘że+AUX’ complexes might be new
auxiliaries, with że being reanalyzed as a verbal stem. In his later work, Bański (2000,
2001) proposes that że is a ubiquitous dummy head which may be inserted at PF in
order to supply a host for stranded auxiliary clitics, which are unable to form a prosodic
word otherwise. He argues that in this respect że is similar to do-support in English.
However, as Bański (2001) points out himself, there are problems with the idea
that że occurs just for phonological reasons. For example, że-support may apply when
there is no prosodic need for it, as in (116a), where the auxiliary can be affixed on the l-
participle (cf. 116b). The only reason to insert że in (116a) is to put an additional focus
on poszli.
(116) a. POSZLI że-ście już tam?

goPART.VIR.PL FOC+AUX.2PL already there
“Have you gone there yet?”
b. Poszli-ście już tam? (Pl, cf. Bański 2000: 211)
Likewise, że is commonly affixed on the imperative. As (117b) shows, its form may be
morphologically reduced to ż.
(117) a. Idź(-że) już!

goIMPV.2SG+FOC already
“Go there now!”
b. Zróbcie(-ż) to!
doIMPV.2PL+FOC it
“Do it” (Pl, cf. Bański 2001)
The current status of że in Polish becomes clarified once the diachronic development of
this element is taken into account. Both types of że have a common ancestor. Że was
used as an enclitic focus marker in Old Church Slavonic (cf. 118) and in Old Polish (cf.
119).
134It will be shown presently that this statement does not seem to be justified diachronically. It is
more likely that the use of supportive że tends to be more accepted by normative linguists, and
hence may appear to them to be more widespread than before.
(118) a. Rečeno že bystъ

sayPASS.N.SG FOC beAOR.3SG
“But it was said” (OCS, Willis 2000: 325)
(119) Ty, jen-że-ś jest przez początku

you who+FOC+AUX.2SG beFOC since beginning
“Thou, who hast been since the beginning…”
(Old Polish, Andersen 1987: 37)
In Old Polish texts the focus particle was frequently attached to demonstratives in
order to add emphasis. These forms have been lexicalized in Modern Polish into tenże
‘thisGEN’; tegoż ‘thisGEN’, temuż ‘thisDAT’, and tymże ‘thisINSTR’. Moreover, it formed a
complex conjunction together with the 3rd person copula jest and the question particle li:
je(st)+że+li, which has been lexicalized as the complementizer jeżeli ‘if/whether’135 (cf.
Decaux 1955: 205-206).
According to Decaux (1955: 208-209), the function of że as a complementizer,
which predominates in Modern Polish, was extremely uncommon in Old Church
Slavonic and Old Polish texts. The usual complementizer was iże, and że emerged as a
complementizer only in the 16th century, when the initial vowel i was lost. The
categorial shift from the focus particle to the complementizer was accompanied by a
change in the direction of cliticization: whereas the focus particle was an enclitic, the
complementizer was a proclitic.
I would like to suggest that the reinterpretation of że has syntactic repercussions as
well. As a declarative complementizer, it is the head of CP. When że functions as a
focus particle, it is one of the possible spell-outs of the Σ0-related focus feature. When
the auxiliary clitic is unable to encliticize onto a host, it must compensate for its
prosodic deficiency by incorporating into że in Σ0. This is exemplified for the 1st person
plural auxiliary -śmy in (120).
(120) [ΣP [Σ że+-śmyi ] [TP [T ti]]]
However, the Σ-head does not have to be spelt out overtly. For example, the auxiliary
clitic may encliticize into the topicalized lexical element occupying Spec, ΣP, if the
prosodic structure of this element is appropriate for encliticization (in the case at hand,
if it does not end in an obstruent). This option is illustrated for the 1st person plural
subject pronoun my, which is the host for the enclitic auxiliary -śmy.
(121) [ΣP Myj [Σ -śmyi ] [TP tj [T ti]]]
Summarizing, the preceding sections investigated cases of auxiliary enclitization. It was

demonstrated that the enclitics move from their base positions to functional heads
which lexicalize meanings related to focus, topic, and subjunctivity. The elements that
give support to the enclitics undergo A’-movement.
135 In contrast to the South Slavic languages, the question particle li is obsolete in Modern Polish.
Jeżeli has an alternative variant jeśli, which presumably derives from jest+li. Jeżeli seems to render
additional focus meaning, which is to be expected given that it contains the focus particle że.
5.3.4.2 Auxiliary affixation on the l-participle

The present section will analyze the position of auxiliary affixes in the clause structure.
Unlike the auxiliary clitics, which can encliticize on any fronted category, the auxiliary
affixes are always adjacent to the verb. Therefore, I will be only concerned with their
placement with respect to the l-participle.
5.3.4.2.1 Participle movement in Polish and in South Slavic

Chapter 2 extensively analysed fronting of the l-participle across the auxiliary ‘be’ in
South Slavic. The movement was argued to be an instance of predicate/locative
inversion, and the l-participle was claimed to raise to Spec, TP. A number of arguments
were raised in support of this approach. For instance, it was shown that since the l-
participle is marked for gender and number, it is a suitable candidate for movement into
Spec, TP in order to check the φ-features of T0. The hypothesis is confirmed by the fact
that the fronted participle may never co-occur with the subject, which means that the
two constituents compete for the same position.
In Polish the l-participle has exactly the same morphology as in South Slavic. It
carries φ-features, so it is potentially eligible for movement to Spec, TP. However, it
may be preceded by the subject.
(122) My pojechali-śmy do Bukaresztu pociągiem

we goPART.VIR.PL+AUX.1PL to BucharestGEN trainINSTR
“We went to Bucharest by train” (Pl)
Admittedly, the presence of the pronoun in (122) adds emphasis to the interpretation of
the clause, but this is due to the fact that Polish is a pro-drop language, so subject
pronouns are usually elided. However, in South Slavic the subject may not be placed in
front of the l-participle, not even to render extra emphasis. Whenever the subject is
clause-initial, it must precede the auxiliary, while the l-participle remains in situ. If the
same word order is applied in Polish, the auxiliary encliticizes on the subject. The
subject is then normally interpreted as topicalized or focused, which is not necessarily
the case in South Slavic.
(123) My-śmy pojechali do Bukaresztu pociągiem

we+AUX.1PL goPART.VIR.PL to BucharestGEN trainINSTR (Pl)
There are a number of ways to explain this variation across the Slavic languages. For
instance, there exists a linguistic division between East Slavic on the one hand and the
West and South Slavic on the other hand concerning the availability of declinable
auxiliary clitics. The East Slavic languages, such as Russian or Ukrainian, have lost their
perfect auxiliaries completely (cf. chapter 1, section 1.3.4.2.1). At the same time they are
non-pro-drop languages, and this property is often attributed to the lack of the
auxiliaries in these languages (cf. Franks 1995 ch. 7). It could be hypothesized, then,
that due to the weakening of the auxiliaries, Polish is losing its pro-drop properties.
Presumably, some of the person features are being lost on the auxiliary because of its
morphological impoverishment, and the overt realization of the subject is a way to
recover them. In order to maintain uniformity with the analysis of the related
constructions in South Slavic, it could be argued that when the subject precedes the
participle-auxiliary sequences (cf. 122), it moves higher than the l-participle and lands in
a high verbal position (e.g. in Spec, Focus/Topic Phrase). This idea might be potentially
supported by the position occupied by sentential adverbs. Watanabe (1993) and
Bošković (1997) assume that they are adjoined to TP. As long as the position of
sentential adverbs is a reliable criterion for the structure of Polish, the data in (124)
show that the subject may occur higher than sentential adverbs in the structure, and
target positions inaccessible for the l-participle.
(124) a. Ty na pewno zaspał-eś

you certainly oversleepPART.M.SG+AUX.2SG
“You’ve certainly overslept”
b. *Ty zaspał-eś na pewno
you oversleepPART.M.SG+AUX.2SG certainly (Pl)
However, it was pointed out in chapter 2 that the same distribution is found in the
South Slavic languages, where the subject may raise across sentential adverbs as well.
Another option may be to suggest that as a result of the morphological weakening
of the auxiliaries, the structure of the compound tenses in Polish is different than in
South Slavic. As was pointed out in section 5.3.2.2, not all native speakers of Polish
treat the perfect auxiliaries uniformly. All of them analyze the forms in the singular
paradigm as affixes, but many of them still interpret the variants in the plural as clitics. I
will assume, though, that regardless of the clitic/affix status of the auxiliary, all speakers
treat the l-participle as a head. It undergoes head movement to T, where it adjoins to
the auxiliary. As a result, it may be freely preceded by the subject, which occupies Spec,
TP.
In the subsequent sections I will carry out different tests, which will allow me to
determine that the l-participle does indeed move as a head in Polish. The most
important one will be related to its interaction with sentential negation. Section 5.3.4.2.2
will differentiate between sentential and constituent negation. Since only the former is
relevant for l-participle movement, it is necessary to distinguish between the two types,
using criteria related to prosody (cf. section 5.3.4.2.2.2) and certain morphological
relations between negation and the object, such as the genitive of negation (cf. section
5.3.4.2.2.3).
5.3.4.2.2 Position and types of negation in Polish

The Slavic languages have a very lax word order, with many constituents being moved
in order to establish specific discourse structure relations. Frequently the only reliable
way of deciding about the placement of a constituent in the clause structure is to
investigate the elements that are assumed not to undergo movement, such as adverbs
and negation, or the words that must appear in fixed positions, such as clitics.
However, it was shown in section 5.2.2 that pronominal clitics do not have to target a
designated position in Polish, so they cannot be used to determine verb placement.
Therefore, it seems more useful to look at the position of negation instead.
It was demonstrated in chapter 2 that negation is a useful device not only for
establishing the position of the verb, but also for deciding whether the verb moves as a
head or as an XP. The negative marker is a head, which projects NegP. It obligatorily
attracts other X0 elements, either only verbs, as in Serbo-Croatian, or the closest head
of any category, as in Bulgarian. Since the l-participle undergoes XP-movement in
Bulgarian and Serbo-Croatian, it may not be attracted by negation.
Furthermore, it was pointed out in chapter 2, section 2.3.6.1 that it is important to

distinguish between sentential (cf. 125a) and constituent (contrastive) negation (cf.
125b).
(125) a. Jan nie pije piwa

Jan NEG drink3SG beerGEN
“Jan doesn’t drink beer”
b. Jan pije nie piwo a wino
Jan drink3SG NEG beerACC but wineACC
“Jan doesn’t drink beer but wine” (Pl)
The two types of negation differ in their syntactic licensing in Slavic. Sentential
negation always involves incorporation of a verbal head into the negative marker.
Constituent negation is assigned by placing the negative marker in the position
immediately preceding the element that is to be negated. It does not involve syntactic
incorporation, because the negated element can be either a head or an XP, while
incorporation is possible only between two heads.
Hence, the X0/XP nature of the element that is attracted by the negative particle
can be uniformly determined only in the case of sentential negation. Luckily, unlike
South Slavic, Polish has very explicit means of distinguishing between sentential and
constituent negation, which are related to stress shift and the assignment of genitive of
negation. They will be overviewed in sections 5.3.4.2.2.2 and 5.3.4.2.2.3, respectively.
On the basis of these findings I will determine the position of NegP that licenses
sentential negation in section 5.3.4.2.3. This will allow me to make definite claims about
the location of the l-participle and the type of movement it undergoes in section
5.3.4.2.3, as well as about the position of the auxiliaries in the clause structure in section
5.3.4.2.4.
5.3.4.2.2.1 Position of sentential and constituent negation

It has been observed in the literature (cf. Willim 1990: 212ff, Witkoś 1998: 214ff, and
Błaszczak 2001: 55) that the negative particle nie immediately precedes finite and non-
finite verbal heads in Polish. Insertion of any overt material between negation and the
verb results in ungrammaticality.
(126) a. Jan nie czyta gazet

Jan NEG read3SG newspapersGEN
“Jan doesn’t read newspapers”
b. *Jan nie gazet czyta (Pl)
Negation may not be separated from the verb even by reflexive or pronominal clitics.
In this respect Polish patterns with Serb-Croatian, where negation is obligatorily pre-
verbal as well.
(127) a. Jan się nie nudzi

Jan REFL NEG bore3SG
“Jan is not bored”
a’. *Jan nie się nudzi
b. Jan jej nie słyszy

Jan herCL.GEN NEG hear3SG
“Jan can’t hear her”
b’. *Jan nie jej słyszy (Pl, cf. Willim 1990: 212)
Błaszczak (2001: 125-127) observes that there is some evidence for overt attraction of
verbs in Polish by negation, which comes from their ordering with respect to frequency
and manner adverbs. The default pattern is ‘adverb-verb’. However, when the verb is
negated, it tends to precede the adverb.
(128) a. Ewa ładnie się ubiera

Ewa prettily REFL dress3SG
“Ewa dresses well”
a’. Ewa nie ubiera się ładnie
Ewa NEG dress3SG REFL prettily
“Ewa doesn’t dress well”
b. Jan często pisze listy
Jan often write3SG lettersACC
“Jan writes letters frequently”
b’ Jan nie pisze często listów
Jan NEG write3SG often lettersGEN
“Jan doesn’t write letters frequently” (Pl, Błaszczak 2001: 125-126)
The sentences listed in (126) through (128) exemplify sentential negation. However, just
as Bulgarian and Serbo-Croatian, Polish also has an option of expressing constituent
negation by placing the negative particle in front of the constituent that negation scopes
over. Unlike in the case of sentential negation, the negated constituent does not have to
be a verbal element or a head. It can be of any category that needs to be negated, such
as the noun phrase in (129a) or a prepositional phrase in (129b). The sentence is
normally followed by a “correction phrase” introduced by the complementizer ale or
lecz ‘but’.
(129) a. Jan czytał nie gazetę ale list

Jan readPART.M.SG NEG newspaperACC but letterACC
”Jan didn’t read a newspaper but a letter”
b. Ania przyszła nie w poniedziałek, lecz we wtorek
Ania comePART.F.SG NEG on Monday but on Tuesday
“Ania didn’t come on Monday, but on Tuesday” (Pl, Błaszczak 2001: 57)
To summarize, sentential negation attracts only verbal heads, whereas constituent

negation may scope over different categories, which can be phrasal. This means that
only sentential negation involves the syntactic operation of incorporating a negated
element into the head of NegP, because incorporation is possible only between two
heads. Constituent negation consists in placement of a negative adverbial in front of the
negated constituent, and does not involve incorporation. In the following sections I will
show that in Polish the two types of negation differ in prosodic and syntactic
properties. By examining these differences it will be possible to determine the exact
position of NegP in the clause structure, and its relation to the l-participle.
5.3.4.2.2.2 Prosodic properties of negation

Apart from the selectional restrictions, sentential negation exhibits different
phonological properties than constituent negation. Namely, the former imposes stress
shift on the verbal element that is negated. Thus, since stress falls on the penultimate
syllable in Polish, it is moved from a one-syllable verb to the negative particle when the
verb is negated. This shows that negation forms a prosodic word with the verb it
precedes.
(130) a. WIEM
know1SG
“I know”
b. NIE wiem
NEG know1SG
“I don’t know”
c. *Nie WIEM (Pl, cf. Ozga 1976: 137)
Conversely, constituent negation does not give rise to the stress shift, which means that
the negative particle nie does not form a prosodic word with non-verbal elements.
(131) a. Nie JA
NEG I
“Not me”
b. *NIE ja (Pl, Błaszczak 2001: 120)
The contrast is to be expected on the assumption that a syntactic process may have
prosodic effects. Only sentential negation involves syntactic incorporation, so only this
type of negation changes stress assignment.
5.3.4.2.2.3 Genitive of negation

It has been demonstrated that only sentential negation involves syntactic incorporation
of a negated verbal head into the negative marker in Neg0. The present section will
discuss the phenomenon of genitive of negation, which is a very reliable criterion of
sentential negation. I will examine the elements that are affected by this rule, and this
will allow me to locate NegP in the clause structure.
Polish, along with Slovene, is the only Slavic language that obligatorily exhibits
genitive of negation (Błaszczak 2001: 60).136 The rule requires that the direct object that
appears in accusative case in affirmative clauses switch into genitive case when the
sentence is negated (cf. 132b).
(132) a. Ewa karmi ptaki/*ptaków

Ewa feed3SG birdsACC/birdsGEN
“Ewa feeds birds”
136The other Slavic languages that have genitive of negation limit its application to some specific
constructions. See Babby (1980), Pesetsky (1982), Franks (1995), Brown (1999), Błaszczak (2001),
and the references cited therein for a detailed discussion.
b. Ewa nie karmi *ptaki/ptaków

Ewa NEG feed3SG birdsACC/birdsGEN
“Ewa doesn’t feed birds” (Pl, cf. Błaszczak 2001: 61)
The examples in (133) demonstrate that the direct objects selected by the transitive verb
may also appear in dative (133a) or instrumental (133b), but they remain unaffected by
negation. Thus, the generalization is that negation does not have any influence on
inherent case.
(133) a. Jan pomaga Ewie

Jan help3SG EwaDAT
“Jan helps Ewa”
a’. Jan nie pomaga Ewie/*Ewy
Jan NEG help3SG EwaDAT/EwaGEN
“Jan does not help Ewa”
b. Jan kieruje samochodem
Jan drive3SG carINSTR
“Jan drives a car”
b’. Jan nie kieruje samochodem/*samochodu
Jan NEG drive3SG carINSTR/carGEN
“Jan doesn’t drive a car” (Pl, cf. Błaszczak 2001: 61-62)
Correspondingly, negation has no effect on the case of nominal adjuncts (cf. 134). It is
only nominal complements that appear in genitive case (cf. 135).
(134) a. Deszcz padał trzy godziny

rainNOM fallPART.M.SG threeACC hoursACC
“It was raining for three hours”
b. Deszcz nie padał trzy godziny/*trzech godzin
rainNOM NEG fallPART.M.SG threeACC hoursACC/threeGEN hoursGEN
“It wasn’t raining for three hours” (Pl, Willim 1990: 211)
(135) a. Jan przeszedł trzy mile

Jan walk-throughPART.M.SG threeACC milesACC
“Jan walked three miles”
b. Jan nie przeszedł *[trzy mile]/ [trzech mili]137
Jan NEG walk-throughPART.M.SG threeACC milesACC/ threeGEN milesGEN
(Pl, Willim 1990: 211)
Genitive of negation has a different distribution with internal arguments of negated

unaccusative verbs. Błaszczak (2001: 62) observes that their case is not affected by
negation, so they must appear in the nominative form.
137 It is not entirely clear whether trzech mili is a complement or actually an adjunct. Most analyses
(e.g. Franks & Dziwirek 1993, Franks 1995, Witkoś 1998) argue for the latter interpretation of the
constituent, which is regarded as a member of an exceptional class of temporal, distance, and
measurement adjuncts that take genitive of negation. Witkoś (1998: 273 fn 32) points out that
both variants of (135b) are acceptable under the appropriate interpretation. Genitive on the DP
adverb implies that Jan covered a distance shorter than three miles. When the accusative form of
the DP is used, the adverb of duration has negation within its scope and thus involves
constituent negation on the adverb. It may trigger contrastive reading, such as “John did not
cover three miles, but four”.
(136) Studenci/*studentów nie przyszli na wykład

studentsNOM /studentsGEN NEG comePART.VIR.PL on lecture
“The students didn’t come to the lecture” (Pl, Błaszczak 2001: 62-63)
Correspondingly, the underlying objects of passive participles must also always occur in
nominative.
(137) Gazety/*gazet nie były dostarczone

newspapersNOM.NV/newspapersGEN.NV NEG bePART.NV.PL deliverPASS.NV.PL
“The newspapers have not been delivered” (Pl)
Moreover, in contrast to sentential negation, constituent negation has no bearing on the

case of the nominals that are negated, whether they are subjects (cf. 138a) or objects (cf.
138b).
(138) a. Nie Marek, a Tomek przyniósł chipsy i wino

NEG MarekNOM but TomekNOM bringPART.M.SG chipsACC and wineACC
“Chips and wine were brought not by Marek but by Tomek”
b. Jan kupił nie książkę a kwiaty
Jan buyPART.M.SG NEG bookACC but flowersACC
“Jan didn’t buy a book but flowers” (Pl)
The exact mechanism of the accusative-to-genitive shift in the presence of sentential

negation has been a matter of debate in the literature for a long time (cf. the references
listed in footnote 136, and especially Błaszczak 2001, Przepiórkowski 2000, and Witkoś
1998 for an analysis of the genitive of negation in Polish in particular). The details of
these accounts or the reason for the case change are not relevant here and exceed the
scope of this dissertation. What is important for the analysis developed in this chapter
is that genitive of negation never operates on the external argument of transitive and
unergative verbs, or internal arguments of unaccusative verbs. It is related to sentential
negation and only affects the objects marked for accusative case. Therefore, NegP that
licenses sentential negation cannot be higher than TP in Polish. I propose that it is
located immediately above AgrOP, which means that it scopes only over the projection
in which accusative case is checked (cf. Błaszczak 2001 for related claims). As an
illustration, I provide a derivation of a clause with a negated verb in (139).
(139) a. Nie kupuję roweru

NEG buy1SG bicycleGEN
“I’m not buying a bicycle”
b. [TP [T <nie+kupuję>i] ... [NegP [Neg nie + ti’] [AgrOP roweruj [AgrO t i’] ... [VP t i tj ]]]
The finite verb kupuję undergoes head movement from V to AgrO. The object DP
rowerACC raises out of VP to Spec, AgrOP in order to check case. Under the influence of
negation immediately above AgrOP, the case on the object switches into genitive. The
verb is attracted by the negative particle and incorporates into it. Finally, the verb
moves to T together with the negative particle, where it checks Tense.
5.3.4.2.3 Negation and types of verbs

The properties of genitive of negation discussed in the previous section indicate that
NegP is located above AgrOP in Polish. Moreover, it has been established that Neg
attracts verbal heads, which incorporate into the negative particle nie, and that the two
elements form a prosodic word. The present section will investigate the types of verbs
that are attracted by negation.
In section 5.3.4.2.2.1 I mentioned that in Polish as in Serbo-Croatian the negative
particle must immediately precede finite and non-finite verbs. Insertion of any overt
material between negation and the verb results in ungrammaticality.
(140) a. Jan nie jedzie jutro na wycieczkę

Jan NEG go3SG tomorrow on trip
“Jan is not going on a trip tomorrow”
b. *Jan nie jutro jedzie na wycieczkę (Pl)
Correspondingly, negation precedes the copula ‘be’ in Polish. This is the pattern found
in Bulgarian and Serbo-Croatian as well, but the important difference between Polish
and the other languages is that the present tense copula in South Slavic is a clitic and
must occur in designated clitic positions. In Polish the present tense copula ‘be’ is not a
clitic and has the same distribution as other finite verbs (cf. section 5.3.4.2.4.1).
(141) To jabłko nie jest wystarczająco dojrzałe

this apple NEG be3SG enough ripeN.PL
“This apple is not ripe enough” (Pl)
There are a few other differences with respect to verb movement and negation between
Polish and South Slavic. In Serbo-Croatian and Bulgarian negation may attract all types
of auxiliaries. In Polish, negation may precede only the non-clitic future auxiliary.
(142) a. Ty nie będziesz się śmiał

you NEG bePRF.2SG REFL laughPART.M.SG
“You won’t laugh”
b. *Ty będziesz się nie śmiał (Pl, cf. Błaszczak 2001: 81)
When the auxiliary is a clitic or an affix (cf. 143 for the conditional auxiliary and 144 for
the perfect auxiliary), the l-participle must be attracted instead.
(143) a. Ty by-ś nie zgubił kluczy

you COND+AUX.2SG NEG losePART.M.SG keysGEN
“You wouldn’t lose the keys”
b *Ty nie by-ś zgubił kluczy (Pl)
(144) a. My-śmy nie widzieli zorzy polarnej

today+AUX.1PL NEG seePART.VIR.PL light polar
“We didn’t see the northern lights”
b. *My nie-śmy widzieli zorzy polarnej (Pl)
The contrast in (143) and (144) is crucial for the analysis developed in this chapter,
because it shows that the l-participle incorporates into negation in Polish. Since the
negative marker is a head, and incorporation is possible only between heads, the
acceptability of (144a) proves that the l-participle moves as a head in Polish. The
variants of (144a) are ungrammatical in Bulgarian and Serbo-Croatian, because the l-
participle undergoes XP movement in these languages.
To summarize, the preceding sections have demonstrated that there is only one
NegP in Polish, which dominates AgrOP. As in South Slavic, the head of NegP
obligatorily attracts verbal heads, which incorporate into the negative marker nie.
However, in contrast to the South Slavic languages, negation may not attract verbal
clitics, such as the conditional auxiliary by and the perfect auxiliaries. A striking property
of Polish is the fact that it allows incorporation of the l-participle into negation. This
indicates that the l-participle undergoes head movement in Polish, and has the same
syntactic distribution as finite verbs.
In order to make more definite claims about the position and the movement of the
l-participle in Polish, it is necessary to establish the syntactic distribution of the
auxiliaries that it appears with. This will be done in the subsequent section.
5.3.4.2.4 Position and types of the auxiliaries

The following sections will determine the positions occupied by different auxiliaries in
the clause structure in Polish. It will be shown that it is possible to distinguish two types
of auxiliaries on the basis of their interactions with negation and the l-participle.
It has been established that negation always immediately precedes the future
auxiliary będzie, finite verbs including the copula jest, and the l-participle. Conversely, the
perfect and conditional auxiliaries may never be preceded by negation. I would like to
argue that the contrast is not superficial and that it corresponds to substantial structural
and semantic differences between the two groups of verbs. For ease of reference, I will
term the members of the former class “lexical”, while the latter class will be dubbed
“functional”. In this way I follow the distinction introduced by Lema and Rivero (1989)
and Rivero (1991, 1994, 2000) to differentiate between the auxiliaries that license
participle preposing in South Slavic.
5.3.4.2.4.1 Syntactic properties of functional and lexical auxiliaries

I propose that the lexical auxiliaries are generated below AgrOP, which is the projection
immediately dominated by NegP. They undergo head movement and incorporate into
negation. The functional auxiliaries that may not be preceded by the negative particle
are merged above NegP: by originates as the head of MoodP, while the perfect
auxiliaries are generated directly in T. The proposal receives support not only from the
interaction of functional auxiliaries with the negative particle, but also from their
distribution in the presence of the main verb. As demonstrated in (145) and (146), by
and the perfect auxiliary may never appear lower in the clause structure than
immediately right-adjacent (affixed) to the l-participle.
(145) a. Płynęli-śmy szybko na drugi brzeg

swimPART.VIR.PL+AUX.1PL quickly on second shore
“We were swimming quickly to the other shore”
b. *Płynęli szybko-śmy na drugi brzeg
c. *Płynęli szybko na drugi brzeg-śmy (Pl)
(146) a. Płynęli-by-śmy szybko na drugi brzeg

swimPART.VIR.PL+COND+AUX.1PL quickly on second shore
“We would swim quickly to the other shore”
b. *Płynęli szybko-by-śmy na drugi brzeg
c. *Płynęli szybko na drugi brzeg-by-śmy (Pl)
However, the future auxiliary będzie, which is a lexical auxiliary, can follow the main
verb and be separated from it by some overt lexical material, such as the adverb szybko
‘quickly’ and negation.
(147) Płynąć/płyneli szybko nie będziemy,

swimINF/swimPART.VIR.PL quickly NEG bePRF.1PL
bo jest silny wiatr
because be3SG strong wind
“We won’t swim quickly because the wind is strong” (Pl)
The contrast is expected on the assumption that the functional elements are generated
higher in the clause than the lexical ones. The perfect auxiliary in (145) is located in T,
and the l-participle raises to become left-adjoined to it (148a). Similarly, the conditional
auxiliary is generated in Mood. The l-participle raises to Mood0 and left-adjoins to by.
Finally, they raise together to T in order to incorporate into the perfect auxiliary -śmy
(cf. 148b).
(148) a. [TP [T płynęlii + śmy]... [VP [V ti ]] perfect auxiliary

b. [TP [T płynęlii + byj śmy] [MoodP [Mood tj + ti’ ] ... [VP [V ti ]]]
conditional auxiliary
Będzie originates in a projection that is positioned lower than T, as it can be separated

from the l-participle by some other constituents. It must also be generated lower than
AgrOP, because it is preceded by negation. For ease of reference, I term the projection
where będzie is merged Aux. I will provide justification for the choice of this position
later in this section.
Turning to the functional auxiliaries, I have claimed that the perfect auxiliary and
the conditional auxiliary are generated in T0 and Mood0, respectively. The l-participle,
finite verbs, and infinitives originate as heads of the VP. These are lexical verbs, capable
of assigning case and theta roles. I would like to argue, though, that the future auxiliary
będzie, as well as the present tense copula jest/są ‘be3SG/be3PL’ are generated in neither of
these positions. I propose that they are merged in Aux0, which is located above VP, and
below NegP. This is the same head which hosts the 3rd person auxiliary je in Serbo-
Croatian and e in Bulgarian (cf. chapter 4, section 4.4.4.1.3), and which is endowed with
the number feature.
(149) [TP [T Perfect Aux [MoodP [Mood by [AgrIOP [NegP [AgrOP [AuxP będzie/jest/są
[VP/PartP]]]]]]]]]
The feature content of Aux0 in Polish is motivated by the fact that the only suppletive
elements in the copula paradigm are the 3rd person forms: jest in the singular and są in
the plural. Since the 3rd person is a null person form, jest and są do not carry any person
feature by themselves. Consequently, they can be used as stems for the other members
of the copula paradigm. In standard Polish jest is used as the stem both in the singular
and the plural, to which person/number morphemes are added (cf. 150).138
138In some modern dialects of Polish jest is the stem only for the singular forms, whereas są is the
stem for the plural: są-śmy, (be1PL), są-śće (be2PL), są (be3PL), (cf. Andersen 1987: 37, and chapter 1,
section 1.3.4.2.2.1).
(150) Paradigm of the copula ‘be’ in standard Polish
SG PL
1 jest-em jest-eśmy
2 jest-eś jest-ście
3 jest są
The chart in (150) presents the most common strategy of copula formation. The
alternative way is to encliticize the person/number morphemes onto the clause initial
element. The two variants are compared in (151). The striking property is that when the
copula is encliticized onto the pronominal subject, jest and są are copula stems, which
remain the same irrespectively of the person specification of the copula enclitic (cf.
Bański 2000: 169).
(151) Two ways of copula formation in Polish
copula affix copula enclitic

1SG ja jest-em ja-m jest
I be1SG I+AUX.1SG be3SG
2SG ty jest-eś ty-ś jest
you be2SG you+AUX.2SG be3SG
3SG on/ona/ono jest
he/she/it be3SG
1PL my jest-eśmy my-śmy są
we be1PL we+AUX.1PL be3PL
2PL wy jest-eście wy-ście są
you be2PL we+AUX.2PL be3PL
3PL oni/one są
theyVIR/theyNV be3PL
The pattern observed in (151) confirms that są and jest do not specify any person
feature, as they are compatible with any person variant of the copula enclitic. Moreover,
they must be located lower than the other members of the paradigm in this context. If
they were to move, they would compete for the same syntactic position with the enclitic
following the clause initial element. As an illustration, the contrast between jest/są and
the other copulas is exemplified in (152). In (152a) the copula jest raises to T0 in order
to left-adjoin into the auxiliary affix -śmy, which carries the 1st person plural feature. In
(152b) -śmy is the 1st person plural clitic, which originates in T0, and becomes
incorporated into the subject pronoun my. The 3rd person (or non-person) plural copula
są may not raise to T0, as this position is occupied by the the auxiliary affix -śmy. I
propose that są remains in its base position, that is in Aux0. The ungrammaticality of
(152c) proves that only the non-person copula forms remain in situ. The ones that carry
a person feature must raise to T0. They may not do so if this position is already
occupied by an auxiliary clitic, such as -śmy.
(152) a. My jest-eśmy zadowoleni

we be+AUX.1PL gladVIR.PL
“We are glad”
b. My-śmy są zadowoleni
we+AUX.1PL be3PL gladVIR.PL
“We are glad”
c. *My-śmy jest-eśmy zadowoleni (Pl)
Thus, the generalization is that only the 1st and the 2nd person copula raise to T0, while
the 3rd person variants remain in Aux0. In this way I am also able to show that the
syntactic structures of the auxiliaries in Polish and in the South Slavic languages are in
fact quite similar. In each of these languages the 3rd person auxiliary remains lower than
the other forms. In Bulgarian, Macedonian, and Serbo-Croatian this is overtly
demonstrated in the arrangement of the clitics in the cluster, as the 1st and 2nd person
auxiliary forms are higher and are separated from the 3rd person by the pronominal
forms. This pattern is more difficult to discern in Polish, because the language is lacking
pronominal clitics.
Even though the auxiliaries in Polish and South Slavic target the same syntactic
positions, an important difference between these languages concerns the fact that in
South Slavic the copula has the same morphological form as the auxiliary (cf. 153 for
Bulgarian).
(153) a. Dovolen sŭm ‘be’ as a copula

gladM.SG bePRES.1SG
“I am glad”
b. Pročel sŭm stixove ‘be’ as an auxiliary
readPART.M.SG bePRES.1SG poems
“I have read poems” (Bg)
In Polish the two forms have divergent properties: the copula has properties of a lexical
element, and can incorporate into negation (cf. 154a), while the auxiliary is a functional
word, because it cannot be preceded by nie (cf. 154b).
(154) a. Jestem zadowolony ‘be’ as a copula

bePRES.1SG gladM.SG
“I am glad”
a’. Nie jestem zadowolony
NEG bePRES.1SG gladM.SG
“I am not glad”
b. Czytał-em wiersze ‘be’ as an auxiliary
readPART.M.SG+AUX.1SG poemsACC
“I have read poems”
b.’ Nie czytal-em wierszy
NEG readPART.M.SG+AUX.1SG poemsGEN
“I haven’t read poems”
b’’. *Nie-em czytał wierszy (Pl)
Despite the morphological contrast between the auxiliary and the copula in Polish, I
will demonstrate that the differences are more superficial than they seem on the
surface. The main evidence for this claim will come from their inflectional paradigms.
I have argued that Aux0 is the position where both the present tense copula jest and
the future auxiliary będzie are generated. However, the two elements have different
inflection. Będzie exhibits the same paradigm as lexical verbs in the present tense,
exemplified in (155) for jechać ‘to ride’.
(155) Paradigm of the copula ‘be’ and the finite verb jechać
SG PL
future auxiliary jechać future auxiliary jechać
1 będ-ę jad-ę będzi-emy jedzi-emy
2 będzi-esz jedzi-esz będzi-ecie jedzi-ecie
3 będzi-e jedzi-e będ-ą jad-ą
Conversely, jest has inflection affixes which are the same as the auxiliary affixes attached
to the l-participle. In (156) the paradigm of jest is contrasted with the auxiliary forms on
the l-participle jechał ‘ride’.139
(156) Paradigm of the copula and auxiliary affixes
copula affix auxiliary affixes

1SG ja jest-em jechał-em
I be1SG ridePART.M.SG+AUX.1SG
2SG ty jest-eś jechał-eś
you be2SG ridePART.M.SG+AUX.2SG
3SG on/ona/ono jest jechał
he/she/it be3SG ridePART.M.SG
1PL my jest-eśmy jechali-śmy
we be1PL ridePART.VIR.PL+AUX.1PL
2PL wy jest-eście jechali-ście
you be2PL ridePART.VIR.PL+AUX.2PL
3PL oni/one są jechali
theyVIR/theyNV be3PL ridePART.VIR.PL
The principal difference between the two paradigms in (155) and (156) is the lack of the
morpheme ś in the finite verb forms in (155), as in jest-eśmy ‘be1PL’ and jechali-śmy
‘ridePART.VIR.PL+AUX.1PL’, as well as jest-eście ‘be2PL’ and jechali-ście ‘ridePART.VIR.PL+AUX.2PL’
versus będzi-emy ‘bePRF.1PL’ and będzi-ecie ‘bePRF.2PL’. Following the ideas developed in
chapter 4, section 4.4.4.1.3, it could be hypothesized that this morpheme represents the
“speaker/hearer” feature, as it appears only on the 1st and 2nd person forms (cf. Poletto
2000). The morpheme is absent on the 3rd person auxiliary, which only carries the
number feature.140
At first blush the contrast between the two paradigms is surprising, because the
copula ‘be’ does not follow the same pattern as będzie, which is the form of ‘be’ used as
the future auxiliary. Quite unexpectedly, the copula matches the paradigm of the
auxiliary that is affixed to the l-participle. However, the contrast receives a
straightforward explanation once a diachronic development of the copula is taken into
account. As was argued in chapter 1, section 1.3.4.2.2.1, the form of the copula in
Modern Polish were created via morphological merge of the emphatic (orthotonic)
139 For reasons of clarity, I present only the masculine/virile forms of the l-participle. The
feminine variant of ‘ride’ is jechała, and the neuter one is jechało. The feminine and the neuter
forms in the singular have the non-virile form in the plural jechały.
140 Andersen (1987: 34) defines the meaning of ś as “subject other than speaker”, because the
morpheme is missing in the 1st singular form in Polish. However, this specification is misguided,
because the equivalent of ś in South Slavic occurs in the first person singular, too.
variant of ‘be’ jest with the affixes that originated from an enclitic variant of ‘be’ (cf.
Andersen 1987). This is shown for the 1st person plural form in (157).141
(157) jest+(e)smy → jesteśmy

beFOC.3SG+beCL.1PL be1PL (Pl)
The l-participle is accompanied by the same type of affixes as auxiliaries, because it also
occurred with an enclitic variant of ‘be’ functioning as the auxiliary (cf. Długosz-
Kurczabowa & Dubisz 2001: 307).
(158) przyszli smy → przyszli-śmy

comePART.VIR.PL beAUX.1PL comePART.VIR.PL+AUX.1PL (Pl)
In spite of the similarity in the forms of affixes, the l-participle and the copula have
different verbal properties. For instance, the former is non-finite, exhibits adjectival
morphology, and is able to assign case and theta roles. The latter is finite, functions as a
linking verb and is unable to assign case or project theta roles. Therefore, I suggest that
the l-participle is generated as the head of V0, while the copula is merged in the
extended projection of the verb, as the head of AuxP.
The future auxiliary będzie has a different paradigm than the present tense copula
jest. However, these two elements are members of the same aspectual pair. Native
speakers of Polish do not perceive jest and będzie as marked for any aspectual value, but
there are diachronic reasons to assume that this is the case. In chapter 1, section
1.3.3.5.2.2 I mentioned that the equivalent of będzie in Old Church Slavonic represented
the perfective form of ‘be’ (cf. Dostál 1954: 146). The assumption that będzie is
perfective is further supported by the facts that it contains the nasal vowel ę in the root,
which is a descendant of a Proto-Indo-European perfectivizing infix (cf. chapter 1,
section 1.3.3.5.2.2). Nasality signals perfectivity in some other Polish verbs as well, for
instance in usiąść ‘to sitPRF’, which contains the nasal vowel ą, in contrast to siedzieć ‘to
sitIMP’, which is imperfective. Therefore, given that jest and będzie represent the same
verb, I submit that they are both generated in Aux0.
To summarize, it has been demonstrated that Polish has two types of auxiliaries,
which differ in their syntactic behaviour. The subsequent section will show that the two
groups of auxiliaries differ in their semantics as well.
5.3.4.2.4.2 Semantic properties of functional and lexical auxiliaries

Apart from the syntactic ordering with respect to each other, the functional and lexical
forms have different semantic content. The auxiliary clitics carry only formal features,
which are number and person in the case of perfect auxiliaries, and the mood feature in
the case of by. The lexical elements are richer in their semantics. For instance, they are
able to appear in aspectual pairs (cf. 159a), which means that they must have raised
across AspP. The functional elements are unable to show aspectual distinctions. This is
demonstrated in the compound tense constructions in (159b), in which the aspectual
meaning is carried by the l-participle, rather than by the perfect auxiliary.
141 E preceding smy is a linking vowel occurring for phonological reasons.

(159) a. piszę/napiszę
writeIMP.1SG/writePRF.1SG
b. pisał-em/napisał-em
writeIMP.PART.M.SG+AUX.1SG/ writePRF.PART.M.SG+AUX.1SG (Pl)
Likewise, the future auxiliary będzie is also a member of an aspectual pair. As was noted
in chapter 1, section 1.3.3.5.2.2, it is a perfective variant of the verb ‘be’ and forms an
aspectual pair with the present tense copula jest, which is imperfective. Moreover, będzie
imposes aspectual restrictions on the main verbs that it appears with. Thus, as shown in
(160), only imperfective forms of verbs are permitted in compound future
constructions in Polish.
(160) Będziemy kąpali/*wykąpali się w morzu

bePRF.1PL bathePART.IMP.VIR.PL/bathePART.PRF.VIR.PL REFL in sea
“We will be bathing in the sea” (Pl)
The perfect (functional) auxiliary is not specified for any aspectual distinctions, because
it originates above AspP. Consequently, it may occur with both perfective and
imperfective variants of the l-participle.142
(161) a. Kąpali-śmy się w morzu

bathePART.IMP.VIR.PL+AUX.1PL REFL in sea
“We were bathing in the sea”
b. Wykąpali-śmy się w morzu
bathePART.PRF.VIR.PL+AUX.1PL REFL in sea
“We bathed in the sea” (Pl)
To summarize, I present the positions of the two types of auxiliaries in the clause
structure in (162). It shows that the perfect and the conditional auxiliary forms are
located in T and Mood, respectively. Thus, they are higher than the lexical future
auxiliary będzie, which occupies the same position as the copula jest.
(162) [TP [T Perfect Aux [MoodP [Mood by [AspP [AgrIOP [NegP [AgrOP [AuxP będzie/jest
[VP/PartP]]]]]]]]]]
142 In chapter 2 I mentioned that in Bulgarian the past auxiliary beše, which diachronically
corresponds to the form of the verb ‘be’ in the imperfectum in Old Church Slavonic, cannot
combine with the l-participles that carry imperfectum morphology, whether they are specified for
perfective or imperfective aspect (cf. ib). It may only appear with aorist participles (cf. ia; see
Krapova 1999a: 61).
(i) a. Ivana beše pisala stixove

Ivana bePAST.3SG writePART.IMPF.AOR. F.SG poems
“I had written poems”
b. *Ivana beše napišela/pišela stixove
Ivana bePAST.3SG writePART.PRF.IMP.F.SG/writePART.IMPF.IMP.F.SG poems
(Bg, Krapova 1999a: 60-61)
Thus, the generalization is that the participle should carry the opposite aspectual value from the
one specified by the auxiliary. I leave the explanation of this semantic relation for future research.
5.3.4.2.5 Why is XP-movement of the l-participle impossible in Polish?

I will conclude this chapter by trying to respond to the question posed in section
5.3.4.2, which was concerned with the impossibility of l-participle fronting via predicate
inversion.
It has been shown that the l-participle moves as a head in Polish. The main
motivation for this assumption comes from the fact that the l-participle is able to
become incorporated into the negative particle, which is the head of NegP. However,
this still does not exclude XP movement of the l-participle to Spec, TP. In principle, it
should be equally possible to raise NegP into Spec, TP after the l-participle has
incorporated into negation. For instance, NegP in (139), repeated in a modified version
in (163), should be eligible for XP-movement into Spec, TP, that is to the position in
front of the auxiliary affix -em.
(163) a. Nie kupił-em roweru

NEG buyPART.M.SG+AUX.1SG bicycleGEN
“I didn’t buy a bicycle”
b. [TP [T -em ... [NegP [Neg nie + kupiłi’ [AgrOP roweruj [AgrO ti’ ... [VP t i tj ]]]]]]]
In fact, I suggest that this is the way the derivation proceeds in Czech, as will be
demonstrated briefly in the next section for comparison.
5.3.4.2.5.1 An excursus on participle fronting in Czech

Czech has been scarcely discussed in this dissertation (although see some remarks on
the copula/auxiliary distinction in chapter 1, section 1.3.4.2.2.3). However, even though
it is a Western Slavic language closely related to Polish, it displays participle fronting via
predicate inversion, on a par with the South Slavic languages. Thus, the examples in
(164) indicate that the l-participle in Czech may be preposed to the sentence-initial
position.
(164) a. Ja jsem koupil knihy

I beAUX.1SG buyPART.M.SG booksACC
“I have bought the book”
b. Koupil jsem knihy (Czech, cf. Rivero 1991; 2000: 315)
As in South Slavic, the l-participle may not be fronted across the auxiliary clitic if the
position at the beginning of the clause is occupied by the subject (cf. 165a).
Correspondingly, the l-participle may only raise entirely on its own, so full VP-
movement, together with an object DP, is barred (cf. 165b)
(165) a. *Ja koupil jsem knihy

I buyPART.M.SG beAUX.1SG booksACC
b. *Koupil knihy jsem
buyPART.M.SG booksACC beAUX.1SG (Czech, cf. Rivero 2000: 314)
As in Polish, negation in Czech exhibits different patterns depending on whether it

negates the copula or the auxiliary ‘be’. In the former case, it must precede the copula
(cf. 166). In the latter, it must precede the l-participle (cf. 167).
(166) a. Jsem zdráv

be1SG healthy
“I am healthy”
b. Nejsem zdráv
NEG+be1SG healthy
“I am not healthy”
c. *Jsem ne-zdráv (Czech, J. Dotlačil p.c.)
(167) a. Nekoupil jsem knihy

NEG+buyPART.M.SG beAUX.1SG booksACC
“I haven’t bought books”
b. *Nejsem koupil knihy (Czech, cf. Rivero 1991)
I assume that this means that the verb ‘be’ is generated in two different positions
depending on its function: in T0 when it is an auxiliary verb, and in Aux0 when it is a
copula (cf. also Veselovská 2004 for an analysis of the position of the verb ‘be’ in
Czech).
In contrast to Polish, Czech does not have the rule of Genitive of Negation, so the
objects in negative clauses retain their structural case. I propose that this means that
NegP does not dominate AgrOP in Czech. In other words, it is located lower in the
structure than in Polish. As an illustration, I provide a derivation of (167a) in (168b).
(168) a. Nekoupil jsem knihy (Czech)

NEG+buyPART.M.SG beAUX.1SG booksACC
b.
TP
[ne+koupili] T’
T AgrOP
knihyj NegP
jsem
Neg
[ne+koupili] VP
V DP
ti tj
The derivation shows that the l-participle koupil incorporates into negation, while the
direct object knihy moves to Spec, AgrOP to check case. Next, the ne+koupil complex
raises as NegP to Spec, TP, and checks the φ-features of T.
5.3.4.2.5.2 The solution

There are two crucial differences between Czech and Polish with respect to the
structure of compound tenses. First, in Czech the auxiliary is homophonous with the
copula. In Polish, these two elements are morphologicaly different (cf. section
5.3.4.2.4.1 above). I would like to claim that this fact has direct repercussions for the
nature of l-participle movement in Polish. It was mentioned in chapter 2, section 2.3.2
that participle fronting via locative/predicate inversion is possible only across the
auxiliary ‘to be’. Due to its morphological impoverishment the auxiliary does not
resemble the verb ‘to be’ in Polish any more. I suggest that this precludes participle
movement via locative inversion in Modern Polish.
Another contrast between the two languages is the fact that in Czech the auxiliary
is a clitic which must receive phonological support, but does not need to become
affixed to the l-participle. In Polish the perfect auxiliary is more impoverished than in
Czech, and it must be affixed to the l-participle. The affix status of the auxiliary makes
the l-participle fronting via XP-movement in Polish impossible for two reasons. First,
this is an instance of remnant movement, so although the PartP moves without its
complement, it contains the trace of the object that has been evacuatated out of it for
case checking. It is a common property of traces that they block encliticization or
affixation. The most well-known case is the “wanna-contraction”. Wanna is a contracted
form of want and the infinitival particle to.
(169) a. Which wine do you want to drink

b. Which wine do you wanna drink
It has been observed (cf. Lasnik and Saito 1984; Van Riemsdijk and Williams 1986: 150
and the references cited therein) that the contraction is blocked when there is a wh-trace
present between want and to.
(170) a. Who do you want to get the wine

b. *Who do you wanna get the wine (Van Riemsdijk and Williams 1986: 150)
The derivations for sentence (169a) is presented in (171a), and for (170a) in (171b).
(171) a. [which wine]i do youj want [TP PROj to drink [e]i ]

b. [who]i do you want [TP [e]i to get the wine]
(Van Riemsdijk and Williams 1986: 150)
Likewise, the contraction is blocked by the intervention of an overt lexical element

between want and to.
(172) a. I want the glass to be emptied

b. *I wanna the glass be emptied
The examples of wanna contraction closely correspond to the auxiliary affixation in

Polish, because both processes are blocked by the presence of an overt element, such as
the direct object wino in (173a) as well as the object trace, if the object moves to Spec,
AgrOP before the l-participle lands in Spec, TP (cf. 173b).
(173) a. *Wypił wino-em

drinkPART.M.SG wine+AUX.1SG
a’. [TP <wypił wino>i [T -em ... [AgrOP ... [VP t i ]
b. *[XPWypił ti]-em winoi
drinkPART.M.SG+AUX.1SG wine
b’. [TP <wypiłi tj> [T -em ... [AgrOP winoj ... [VP t i tj ]]]]
Another property of l-participle fronting via locative inversion is the fact that it always
occurs as XP-movement, even if an intransitive verb is fronted. It is standardly assumed
that affixes may attach to heads, rather than phrases. Therefore, the intransitive l-
participle skończyłem ‘finish’ in (174a) may not undergo XP-movement to Spec, TP,
because this will preclude the affixation of the auxiliary into the l-participle.
Consequently, the derivation presented in (174b) will crash.
(174) a. Skończył-em
finishPART.M.SG+AUX.1SG
“I have finished”
b.’ [TP [VP skończył] [T -em ... [tVP]]]
Summarizing, I have shown the l-participle may not undergo XP movement in Polish
due to the affixal nature of the auxiliary. The l-participle must always raise as a head in
order to permit the auxiliary affixation. Moreover, the movement via locative inversion
is impossible, because the auxiliary does not correspond to the verb ‘to be’ any more.
The structure in (175) represents a converging derivation, in which the negated l-
participle undergoes head raising.
Conclusions 285
(175) a. Nie wypił-em wina

NEG drinkPART.M.SG+AUX.1SG wineGEN
b.
TP
T’
T NegP
Neg
[nie+wypiłi]-em
[nie+wypiłi] AgrOP
winaj VP
V DP
ti tj
In contrast to the related derivation of the Czech example in (168), AgrOP in

dominated by NegP, and hence it is affected by the rule of genitive of negation. The l-
participle incorporates into the negative particle nie in Neg0, and the two elements raise
as a complex head to T0, where they left-adjoin to the auxiliary affix -em.
5.4 Conclusions
This chapter has described the properties of pronominal clitics and the syntax of
compound tenses in Polish. The pronominal clitics have been shown to be steadily
becoming more independent with respect to their position in the clause. By contrast,
the auxiliary clitics are in the process of being reanalyzed as affixes on the l-participle.
Thus, the pronominal and the auxiliary clitics are developing in opposite directions: the
former are being reinterpreted as strong, whereas the latter are becoming
morphologically impoverished.
Summarizing the ideas developed in this chapter, I propose the syntactic template
for the structure of compound tenses in Polish in (176). The perfect auxiliaries are
functional elements, so they are generated directly in the functional projection that
specifies their feature content. Since they carry person and number features, I suggest
they are merged in T0. Right above them I place Mod0, which is the site where the
subjunctive mood is lexicalized. To simplify the template, I generate the Σ0 in the same
position. This is the site where polarity and Force features are located. This projection
hosts the emphatic particle że, and it may attract perfect and conditional auxiliaries,
while A’-scrambled constituents land in Spec, ΣP. The perfect auxiliaries, to which the
l-participle may adjoin, are hosted in T0. This is also the head against which finite verbs
check Tense. The modal auxiliary by originates in Mood0, from which it moves in order
to adjoin to the perfect auxiliary in T0 either by itself or as a free rider on the l-
participle. The future auxiliary is merged as the head of AuxP, together with the 3rd
person copulas jest and są. Będzie raises to T0 in order to check Tense. Finite verbs, the
present tense copula jest, and the l-participle, which are merged in V, move to T0 as
well, where they check Tense or φ-features.
(176)
Mod/Σ
subjunctive by T
emphatic że
perfect auxiliaries Mood
modal by Asp
AgrIO
Neg
AgrO
Aux
jest/są/będzie Part/V
lexical verbs/
thel-participle
Summary
This dissertation has overviewed the diachrony and synchrony of the structure of
compound tenses in Slavic. The underlying assumption was the idea that the tense
system of the contemporary Slavic languages has been simplified due to an excess of
aspect marking via both tense and aspect morphology in Proto Slavic and Old Church
Slavonic. The simplification proceeded in two ways: morphosyntactically, through a
decline of the aspectual tenses, and semantically, through a reinterpretation of the
present perfect. The former strategy occurred in West and East Slavic languages, which
have lost the aorist and the imperfectum, and reduced the auxiliary in the present
perfect. The latter strategy was applied in South Slavic. These languages have largely
retained the aspectual tenses, but have reanalyzed the present perfect as a marker of
non-evidentiality.
Bulgarian and Serbo-Croatian have preserved the morphological make-up of the
Old Church Slavonic present perfect. It is constructed with the auxiliary ‘to be’, which
has the same form as the copula, and appears in different aspectual variants in order to
render temporal distinctions. The auxiliary is accompanied by the l-participle, which is a
descendant of a class of Proto-Indo-European *lo-adjectives. Throughout the history of
Slavic these adjectives have become completely verbalized, and now are case and theta
role assigners. The l-participle always shows agreement in gender and number with the
subject. I argued in chapter 2 that this property makes it eligible for movement to Spec,
TP and for checking the φ-features of T. I also suggested that the operation proceeds
via XP-movement, on a par with locative inversion in English. The main motivation for
this claim came from the complementary distribution of the raised subject and the
fronted participle, and the requirement of strict adjacency between the clause-initial
participle and the auxiliary. More evidence was drawn from properties of double
participle constructions, short participle movement, participle fronting across the future
auxiliary šte in Bulgarian, and the patterns of sentential negation. In this way I
challenged the previous accounts of l-participle fronting, which postulate that the
movement consists in head raising or head adjunction.
Chapter 3 studied the compound tense formed with the auxiliary ‘have’ and the
past participle. It has received scarce attention in linguistics so far, so this thesis offers
the first detailed analysis of the construction in generative literature. Unlike the ‘be’-
perfect, which has a uniform distribution across Slavic, the ‘have’-perfect has developed
a complete paradigm only in Kashubian and Macedonian. In these two languages the
participle does not agree with the subject in φ-features, and always appears in the
invariant singular neuter form. It is morphologically the same as the passive participle,
and hence was claimed to be unable to assign structural case to the object and project
an external theta role. These functions are performed by the auxiliary ‘to have’. A few
other Slavic languages, which have not grammaticalized the ‘have’-perfect yet, use a
similar construction, termed the ‘stative perfect’. Its distribution is quite limited,
because it is found only with animate subjects and transitive participles, which always
appear in the perfective form and agree with the direct object. On the basis of the
development of related structures in Germanic and Romance languages I concluded
that the stative perfect is the underlying, non-grammaticalized variant of the ‘have’-
perfect. The stative perfect and the ‘have’-perfect were contrasted with the impersonal
participle construction, which is found in Polish, Byelorussian, Ukrainian, and some
dialects in Russian. This form represents the most verbalized type of the passive
participle.
288 Summary
Chapter 4 investigated clitics in South Slavic. Their distribution is important for the
structure of the compound tenses for two reasons. First, Slavic languages have a very
unconstrained word order, but the arrangement of clitics in the clause is exceptionally
rigid. Therefore, clitics can be used as indicators of syntactic movement in order to
determine the placement of other elements in the sentence. Second, the forms of the
auxiliary ‘to be’ in the present tense are clitics themselves.
The South Slavic languages exhibit a considerable diversity in their clitic systems.
Serbo-Croatian is one of the few world’s languages which still have Wackernagel clitics.
In this way it represents the pattern of the early Proto-Indo-European dialects, which
was lost in most languages. Clitics in Bulgarian display divergent syntactic and
phonological requirements: phonologically they are enclitic, so they need a
morphologically overt host to their left, but syntactically they are proclitic, hence they
opt for a preverbal position. Macedonian clitics show a very intricate distribution, with
divergent requirements depending on the categorial status of their host. In the
alternative analysis developed in chapter 4 I proposed that clitics in Macedonian do not
form a uniform category, but rather comprise clitics proper and weak pronouns. The
former occur with case-assigning verbs, while the latter are found in constructions with
non-verbal predicative XPs, such as APs, DPs, and passive participles.
The status of the auxiliary clitics was argued to determine the pattern of l-participle
fronting. It was demonstrated that the overt presence of the 3rd person auxiliary je/e,
which always occupies the final slot in the clitic cluster, is the condition for the XP-
movement. In the languages which have lost the 3rd person auxiliary, such as
Macedonian, Czech, and Polish, the auxiliary raises out of the PartP as a head.
The final part of the dissertation analyzed the syntax of Polish. In contrast to the
South Slavic languages it simplified the structure of the compound tenses through a
reduction of the tense inventory and their morphological composition, rather than via a
semantic reinterpretation. The morphological reduction consists in the reanalysis of the
auxiliary enclitic as an inflectional affix on the l-participle, and has been proceeding
slowly since the 16th century. The phenomenon coincided with a change in the auxiliary
placement, but I have argued that the two processes have been independent of each
other. In the older variants of Polish the auxiliary uniformly targeted the Wackernagel
position. In Modern Polish the auxiliary is attached to the l-participle in most contexts,
but it encliticizes onto the sentence-initial element in the clauses that express meanings
related to subjunctivity, focus, and topicalization. This is best observed in the case of
the auxiliary by, which can be a marker of subjunctive mood or conditionality,
depending on its enclitization on the l-participle or the clause-initial constituent.
Finally, it has been shown that in contrast to the auxiliary clitics, which are being
morphologically reduced, the pronominal clitics in Polish are steadily becoming more
independent with respect to their position in the clause. They have few properties in
common with the South Slavic pronominal clitics: they do not need to cluster or appear
in designated positions, and they can scramble across the clause quite freely.
I have argued that the ways the compound tenses have evolved in different Slavic
groups have been influenced by external language contacts. For instance, the aspectual
tenses have been retained in Bulgarian and Macedonian because they also exist in some
other Balkan languages. Likewise, the ‘have’-perfect has been fully grammaticalized in
Kashubian and Macedonian due to extensive crosslinguistic exposure of their speakers
to German and Arumanian, respectively. It might be interesting to determine whether
the internal properties of other syntactic subsystems of Slavic, such as the impoverished
DP structure in Balkan Slavic, may have an impact on the structure of compound
tenses as well. This issue, however, has to await further research.
References
Abney, Steven. 1987. The English Noun Phrase in its Sentential Aspect. PhD
Dissertation. MIT.
Acquaviva, Paolo. 1995. Operator Composition and Negative Concord. GenGenP
(Geneva Generative Papers) 3:72-105.
Allen, Margaret. 1978. Morphological Investigations. Ph.D. Thesis. University of
Connecticut.
Anagnostopoulou, Elena. 1999. On Double Object Alternations and Clitics. Ph.D.
dissertation. Tilburg University, Models of Grammar Group.
Anagnostopoulou, Elena. 2003. Participles and Voice. In Perfect Explorations, eds.
Artemis Alexiadou, Monika Rathert and Arnim von Stechow, 1-36. Berlin:
Mouton de Gruyter.
Andersen, Henning. 1987. From Auxiliary to Desinence. In Historical Development of
Auxiliaries, eds. Martin Harris and Paolo Ramat. Berlin: Mouton de Gruyter.
Anderson, Stephen. 1993. Wackernagel's Revenge: Clitics, Morphology, and the Syntax
of Verb-Second Position. Language 69:68-98.
Andrejczin, Lubomir. 1938. Kategorie znaczeniowe koniugacji bułgarskiej. Kraków: Polska
Akademia Umiejętności.
Arnaudova, Olga. 2003. Focus and Bulgarian Clause Structure. Ph.D. dissertation.
University of Ottawa.
Asbury, Anna. 2005. Adpositions as Case Realisations - Structures and Consequences.
Leiden Papers in Linguistics 2:69-92.
Babby, Leonard. 1980. Existential Sentences and Negation in Russian. Ann Arbor: Karoma
Publishers.
Baerman, Matthew, and Loren Billings. 1998. Macedonian Clitics and the Trisyllabic
Stress Window. University of Trondheim Working Papers in Linguistics 31:13-32.
Bajerowa, Irena. 1964. Kształtowanie się systemu polskiego języka literackiego w XVIII wieku.
Wrocław: Ossolineum.
Baker, Mark. 1988. Incorporation: A Theory of Grammatical Function Changing. Chicago:
University of Chicago Press.
Baker, Mark, Kyle Johnson, and Ian Roberts. 1989. Passive Arguments Raised. Linguistic
Inquiry 20:219-251.
Bański, Piotr. 2000. Morphological and Prosodic Analysis of Auxiliary Clitics in Polish
and English, Institute of English. Ph.D. dissertation. Warsaw University.
Bański, Piotr. 2001. Last Resort Prosodic Support in Polish. In Current Issues in Formal
Slavic Linguistics, eds. Gerhild Zybatow, Uwe Junghanns, Grit Mehlhorn and
Luka Szucsich, 179-186. Frankfurt am Main: Peter Lang.
Barbiers, Sjef. 1995. The Syntax of Interpretation. PhD dissertation. University of
Leiden.
Barbiers, Sjef, and Sybesma, Rint. 2004. On the Different Verbal Behavior of
Auxiliaries. Lingua 114:389-398.
Bartula, Czesław. 1981. Podstawowe wiadomości z gramatyki staro-cerkiewno-słowiańskiej na tle
porównawczym. Warszawa: Państwowe Wydawnictwo Naukowe.
Bašić, Monika. 2003. Participle Movement in Serbo-Croatian. Ms. University of
Tromsø.
Bayer, Josef. 1984. COMP in Bavarian. The Linguistic Review 3:209-274.
Behaghel, Otto. 1928. Deutsche Syntax. Vol. 3: Die Satzgebilde. Heidelberg: Winter.
290 References
Belletti, Adriana. 1999. Italian/Romance Clitics: Structure and Derivation. In Clitics in

the Languages of Europe, ed. Henk van Riemsdijk, 543-580. Berlin: Mouton de
Gruyter.
Bentley, Delia, and Eythórsson, Thórhallur. 2004. Auxiliary Selection and the Semantics
of Unaccusativity. Lingua 114:447-471.
Berent, Gerald. 1980. On the Realisation of Trace: Macedonian Clitic Pronouns. In
Morphosyntax in Slavic, eds. Catherine Chvany and Richard D. Brecht, 150-186.
Columbus: Slavica.
Beukema, Frits, and Peter Coopmans. 1989. A Government-Binding Perspective on the
Imperative in English. Journal of Linguistics 25:417-436.
Billings, Loren. 2002. Why Clitics Cluster Together in Balkan Slavic: Non-templatic
Morphology. In Formal Approaches to Slavic Linguistics 10, ed. Jindrich Toman,
75-96. Ann Arbor: Michigan Slavic Publications.
Blevins, James P. 2003. Passives and Impersonals. Journal of Linguistics 39:473-520.
Błaszczak, Joanna. 2001. Investigation into the Interaction between the Indefinites and Negation.
Berlin: Akademie Verlag.
Boeckx, Cédric. 1997. Agreement Constraints in Icelandic and Elsewhere. Ms.
University of Connecticut.
Boeckx, Cédric. 1998. A Minimalist View on the Passive. UConn Working Papers in
Linguistics 2.
Bok-Bennema, Rieneke. 2001. Evidence for an Aspectual Functional Head in French
and Spanish. In Progress in Grammar: Articles at the 20th Anniversary of the
Comparison of Grammatical Models Group in Tilburg, eds. Marc van Oostendorp
and Elena Anagnostopoulou. Utrecht: Roccade.
Bondaruk, Anna. 1995. Are Polish Parasitic Gaps Truly Parasitic? Paper presented at
The Fourth PASE Conference, Poznań.
Bondaruk, Anna, and Magdalena Charzyńska-Wójcik. 2002. Is There Expletive pro in
Impersonal Passives: A Reappraisal. In PASE Papers in Language Studies:
Proceedings of the Ninth Annual Conference of the Polish Association for the Study of
English. Gdańsk, 26-28 April 2000, ed. Danuta Stanulewicz, 85-93. Gdańsk:
Wydawnictwa Uniwersytetu Gdańskiego.
Bonet, Eulalia. 1994. The Person-Case Constraint: A Morphological Approach. In The
Morphology-Syntax Connection, eds. H. Harley and C. Phillips. Cambridge, MA:
MIT Press.
Booij, Geert, and Jerzy Rubach. 1987. Postcyclic versus Postlexical Rules in Lexical
Phonology. Linguistic Inquiry 18:1-44.
Borsley, Robert, and María-Luisa Rivero. 1994. Clitic Auxiliaries and Incorporation in
Polish. Natural Language and Linguistic Theory 12:373-422.
Borsley, Robert, María-Luisa Rivero, and Janig Stephens. 1996. Long Head Movement
in Breton. In The Syntax of the Celtic Languages, eds. Robert Borsley and Ian
Roberts. Cambridge: Cambridge University Press.
Bošković, Željko. 1995. Participle Movement and Second Position Cliticization in
Serbo-Croatian. Lingua 96:245-266.
Bošković, Željko. 1997. The Syntax of Nonfinite Complementation: An Economy Approach.
Cambridge, MA: The MIT Press.
Bošković, Željko. 2000. Second Position Cliticization: Syntax and/or Phonology? In
Clitic Phenomena in European Languages, eds. Frits Beukema and Marcel Den
Dikken, 71-119. Amsterdam: Benjamins.
Bošković, Željko. 2001. On the Nature of the Syntax-Phonology Interface. Cliticization and
Related Phenomena. Amsterdam: Elsevier.
References 291
Bošković, Željko. 2002. Clitics as Non-branching Elements and the Linear

Correspondence Axiom. Linguistic Inquiry 33:329-340.
Bošković, Željko. 2005. Left Branch Extraction, Structure of NP, and Scrambling.
Studia Linguistica 59:1-45.
Bošković, Željko. to appear. Clitic Placement in South Slavic. Journal of Slavic Linguistics.
Bräuer, Herbert. 1957. Untersuchungen zum Konjunktiv im Altkirchenslavischen un im
Altrussischen: Slavistische Veröffentlichungen II. Wiesbaden: Osteuropa-
Institut an der Freien Universität Berlin.
Breitbarth, Anne. 2005. Live Fast, Die Young - the Short Life of the Early Modern
German Auxiliary Ellipsis. Ph.D. dissertation. Tilburg University, Models of
Grammar Group.
Bresnan, Joan. 1994. Locative Inversion and the Architecture of Universal Grammar.
Language 70:72-131.
Breza, Edward, and Jerzy Treder. 1981. Gramatyka kaszubska: zarys popularny. Gdańsk:
Zrzeszenie Kaszubsko-Pomorskie.
Broekhuis, Hans. 2005. Predicate Movement. Ms. Tilburg University.
Broekhuis, Hans, and Kees van Dijk. 1995. The Syntactic Function of the Auxiliaries of
Time. In Linguistics in The Netherlands 1995, eds. Marcel Den Dikken and Kees
Hengeveld, 37-48. Amsterdam: Benjamins.
Broekhuis, Hans, and Krzysztof Migdalski. 2003. Participle Fronting in Bulgarian. In
Linguistics in the Netherlands 2003, eds. Paula Fikkert and Leonie Cornips, 1-12.
Amsterdam: Benjamins.
Brown, Sue. 1999. The Syntax of Negation in Russian. A Minimalist Approach. Stanford:
CSLI Publications.
Browne, Wayles. 2002. Serbo-Croat. In The Slavonic Languages, eds. Bernard Comrie and
Greville G. Corbett, 125-187. London: Routledge.
Bubenik, Vit. 2001. Morphological and Syntactic Change in Medieval Greek and South Slavic
Languages. Muenchen: Lincom Europa.
Caink, Andrew. 1999. Against ‘Long Head Movement’: Lexical Insertion and the
Bulgarian Auxiliary ‘Be’. In Topics in South Slavic Syntax and Semantics, ed. Lars
Hellan, 91-123. Amsterdam: Benjamins.
Caink, Andrew. 2000. Full Form Auxiliaries in Serbian/Croatian/Bosnian. In Formal
Approaches to Slavic Linguistics 8: The Philadelphia Meeting, eds. Tracy Holloway
King and Irina Sekerina, 61-77. Ann Arbor: Michigan Slavic Publications.
Cardinaletti, Anna, and Michal Starke. 1999. The Typology of Structural Deficiency: A
Case Study of The Three Classes of Pronouns. In Clitics in the Languages of
Europe, ed. Henk van Riemsdijk, 185-234. Berlin: Mouton de Gruyter.
Cédric, Boeckx. 1998. A Minimalist View on the Passive. UConn Working Papers in
Linguistics 2.
Cetnarowska, Bożena. 2003. On Pronominal Clusters in Polish. In Investigations into
Formal Slavic Linguistics: Contributions of the Fourth European Conference on Formal
Description of Slavic Languages - FDSL IV held at Potsdam University, November 28-
30, 2001, eds. Peter Kosta, Joanna Błaszczak, Jens Frasek, Ljudmila Geist and
Marzena Zygis, 13-30. Frankfurt am Main: Peter Lang.
Ceynowa, Florian. 1879. Zarés do grammatikj kasebsko-slovjnskjé mové (An Outline of the
Grammar of the Kashubian-Slovincian language). Poznań.
Chomsky, Noam. 1965. Aspects of the theory of syntax. Cambridge, MA: MIT Press.
Chomsky, Noam. 1981. Lectures on Government and Binding: Dordrecht: Foris.
Chomsky, Noam. 1986. Barriers: Cambridge, Mass.: MIT Press.
292 References
Chomsky, Noam. 1993. A Minimalist Program for Linguistic Theory. In The View from
Building 20: Essays in Linguistics in Honour of Sylvian Bromberger, eds. Ken Hale
and Samuel Keyser. Cambridge, MA: MIT Press.
Chomsky, Noam. 1995. The Minimalist Program. Cambridge, MA: The MIT Press.
Chomsky, Noam. 2000. Minimalist Inquiries: the Framework. In Step by Step. Essays on
Minimalist Syntax in Honor of Howard Lasnik, eds. Roger Martin, David Michaels
and Juan Uriagereka, 89-155. Cambridge, MA: MIT Press.
Chomsky, Noam. 2001. Derivation by Phase. In Ken Hale. A Life in Language, ed.
Michael Kenstowicz, 1-52. Cambridge, MA: MIT Press.
Cinque, Guglielmo. 1993. A Null Theory of Phrase and Compound Stress. Linguistic
Inquiry 24:239-298.
Cinque, Guglielmo. 1999. Adverbs and Functional Heads. A Cross-Linguistic Perspective.
Oxford: Oxford University Press.
Comrie, Bernard. 1977. In Defence of Spontaneous Demotion: the Impersonal Passive.
In Syntax and Semantics: Grammatical Relations, eds. Peter Cole and Jarrold M.
Sadock, 47-58. New York: Academic Press.
Comrie, Bernard. 1985. Tense. Cambridge: Cambridge University Press.
Comrie, Bernard. 1988. On the Semantics of Tense and Aspect in Bulgarian. Jouko
Lindstedt, (Slavica Helsingiensia, 4.) Helsinki. University Press, 1985, 319 pp.
74 FIM. Review article. Lingua 74:72-74.
Comrie, Bernard, and Greville G. Corbett. 2002. The Slavonic Languages. London:
Routledge.
Coopmans, Peter. 1989. Where Stylistic and Syntactic Processes Meet: Locative
Inversion in English. Language 65:728-751.
Corver, Norbert. 1992. On Deriving Certain Left Branch Extraction Asymmetries: A
Case Study in Parametric Syntax. In Proceedings of the Northeast Linguistic Society
22, 67-84. University of Delaware.
Čamdžić, Amela. 1999. Toward the Analysis of Topic and Focus in Serbo-Croat. SOAS
Working Papers in Linguistics 9:3-19.
Čamdžić, Amela. 2004. On the Syntax of Long Verb Movement. Ph.D. dissertation.
University College London.
Ćavar, Damir, and Chris Wilder. 1997. Auxiliaries in Serbian/Croatian and English. In
Formale Slavistik, eds. Uwe Junghanns and Gerhild Zybatow, 3-12. Frankfurt
am Main: Vervuert Verlag.
Ćavar, Damir, and Chris Wilder. 1999. “Clitic Third” in Croatian. In Clitics in the
Languages of Europe, ed. Henk van Riemsdijk, 429-467. Berlin: Mouton de
Gruyter.
Dalewska-Greń, Hanna. 1997. Języki słowiańskie. Warszawa: Wydawnictwo Naukowe
PWN.
Damborský, Jiří. 1967. Participium l-ove ve slovanštine. Warszawa: Państwowe
Wydawnictwo Naukowe.
Decaux, Étienne. 1955. Morphologie des enclitiques polonaise. Travaux de l'Institut d'Études
Slaves, 23. Paris: l'Institut d'Études Slaves.
Den Besten, Hans. 1981. A Case Filter for Passives. In Theory of Markedness in Generative
Grammar, eds. Adriana Belletti, Luciana Brandi and Luigi Rizzi, 65-122. Pisa:
Scuola Normale Superiore.
Den Besten, Hans. 1989. Studies in West Germanic Syntax. PhD dissertation. Tilburg
University, Models of Grammar Group.
References 293
Den Besten, Hans, and Gert Webelhuth. 1987. Remnant Topicalization and the
Constituent Structure of VP in the Germanic SOV Languages. GLOW
Newsletter 18:15-16.
Den Besten, Hans, and Gert Webelhuth. 1990. Stranding. In Scrambling and Barriers, eds.
Günther Grewendorf and Wolfgang Sternefeld, 77-92. Amsterdam:
Benjamins.
Déprez, Viviane. 1989. On the Typology of Syntactic Positions and the Nature of
Chains. Ph.D dissertation. MIT.
Diesing, Molly. 1990. The Syntactic Roots of Semantic Partitions. Ph.D. dissertation.
University of Massachusetts, Amherst.
Dimitrova-Vulchanova, Mila. 1995. Clitics in Slavic. Studia Linguistica 49:54-92.
Dimitrova-Vulchanova, Mila. 1998. Are Bulgarian Pronominal Clitics in the
Wackernagel Position? In Norwegian Contributions to the Twelfth International
Congress of Slavists. Cracow 1998, 7-32: Meddelelser 80, Universitetet i Oslo,
Slavisk-Baltisk Avdeling.
Dimitrova-Vulchanova, Mila, and Lars Hellan. 1999. Clitics and Bulgarian Clause
Structure. In Clitics in the Languages of Europe, ed. Henk van Riemsdijk, 469-514.
Berlin: Mouton de Gruyter.
Dimitrovski, Todor. 1957. Revija na Izvestija na Instituta za Bulgarski Ezik. Makedonski
Jazik 8:245-248.
Długosz-Kurczabowa, Krystyna, and Stanisław Dubisz. 2001. Gramatyka historyczna
języka polskiego. Warszawa: Wydawnictwa Uniwersytetu Warszawskiego.
Dobrovie-Sorin, Carmen. 1994. The Syntax of Romanian. Berlin: Mouton de Gruyter.
Dogil, Grzegorz. 1987. Lexical Phonology and Floating Inflection in Polish. In
Phonologica 1984, ed. Wolfgang U. Dressler, 39-47. Cambridge: Cambridge
University Press.
Dogil, Grzegorz, and Miquel Aguado. 1989. Clitics in Lexical Phonology: Alleged
Counterevidence? Linguistische Berichte 120:99-116.
Dornisch, Ewa. 1995. Discourse-Linking and Multiple Wh-questions in Polish. In
Licensing in Syntax and Phonology. Volume 1 of PASE Studies and Monographs, ed.
Edmund Gussmann, 71-87. Lublin: Folium.
Dornisch, Ewa. 1997. Auxiliaries and Functional Projections in Polish. In Formal
Approaches to Slavic Linguistics (FASL) IV: The Cornell Meeting 1995, eds. Wayles
Browne, Ewa Dornisch, Natasha Kondrashova and Draga Zec, 183-209. Ann
Arbor: Michigan Slavic Publications.
Doros, Aleksander. 1975. Werbalne konstrukcje bezosobowe w języku rosyjskim i polskim na tle
innych języków słowiańskich. Wrocław: Ossolineum.
Dostál, Antonín. 1954. Studie o vidovém systému v staroslovnštine. Prague: Státní pedagogicke
nakladatelství.
Dziwirek, Katarzyna. 1994. Polish Subjects. New York: Garland Press.
Elliott, Elisabeth M. 2001. The Sum (‘Be’) and Imam (‘Have’) Resultative Constructions
in Macedonian and Bulgarian within a Typology of Resultative Constructions
in Slavic. PhD dissertation. University of Toronto.
Embick, David. 2003. Locality, Listedness, and Morphological Identity. Studia Linguistica
57:143-170.
Embick, David. 2004. On the Structure of Resultative Participles in English. Linguistic
Inquiry 35:355-392.
Embick, David, and Roumyana Izvorski. 1995. On Long Head Movement. In
Proceedings of ESCOL '94, eds. Janet Fuller, Ho Han and David Parkinson, 104-
115. DMLL Publications, Cornell University.
294 References
Embick, David, and Roumyana Izvorski. 1997. Participle-Auxiliary Orders in Slavic. In

Formal Approaches to Slavic Linguistics: The Cornell Meeting, eds. Wayles Browne,
Ewa Dornisch, Natasha Kondrashova and Draga Zec, 210-239. Ann Arbor:
Michigan Slavic Publications.
Emonds, Joseph. 2000. Lexicon and Grammar: the English Syntacticon. Berlin: Mouton de
Gruyter.
Emonds, Joseph. 2006. Adjectival Passives: The Construction in the Iron Mask. In The
Syntax Companion (SynCom), eds. Martin Everaert and Henk van Riemsdijk.
Oxford: Blackwell.
Erb, Marie Christine. 2001. Finite Auxiliaries in German, Tilburg University, Models of
Grammar Group: Ph.D. dissertation.
Fontana, Josep M. 1997. On the Integration of Second Position Phenomena. In
Parameters of Morphosyntactic Change, eds. Ans van Kemenade and Nigel Vincent,
207-250. Cambridge: Cambridge University Press.
Fox, Danny. 1999. Reconstruction, Binding Theory and the Interpretation of Chains.
Linguistic Inquiry 30:157-196.
Franks, Steven. 1995. Parameters of Slavic Morphosyntax. New York: Oxford University
Press.
Franks, Steven. 1998. Clitics in Slavic. Paper presented at Comparative Slavic Morphosyntax
Workshop, Bloomington.
Franks, Steven. 2000. Clitics at the Interface: An Introduction to Clitic Phenomena in
European Languages. In Clitic Phenomena in European Languages, eds. Frits
Beukema and Marcel Den Dikken, 1-46. Amsterdam: John Benjamins.
Franks, Steven. forthcoming. Adverb Interpolation in the Bulgarian Clitic Cluster. In A
festschrift for Charles Gribble, eds. Robert Rothstein, Ernest Scatton and Charles
Townsend. Bloomington: Slavica.
Franks, Steven, and Katarzyna Dziwirek. 1993. Negated Adjunct Phrases Are Really
Partitive. Journal of Slavic Linguistics 1:280-305.
Franks, Steven, and Piotr Bański. 1999. Approaches to ‘Schizophrenic’ Polish Person
Agreement. In Formal Approaches to Slavic Linguistics: the Seattle Meeting, eds.
Katarzyna Dziwirek, Herbert Coats and Cynthia M. Vakareliyska. Ann Arbor:
Franks, Steven, and Tracy Holloway King. 2000. A Handbook of Slavic Clitics: New York:
Oxford University Press.
Franks, Steven, and Željko Bošković. 2001. An Argument for Multiple Spell-out.
Franks, Steven, and Catherine Rudin. 2005. Bulgarian Clitics as K0 Heads. In Formal
Approaches to Slavic Linguistics: The South Carolina Meeting, 2004, eds. Steven
Franks, Frank Gladney and Mila Tasseva-Kurktchieva, 104-116. Ann Arbor:
Friedman, Victor, A. 1977. The Grammatical Categories of the Macedonian Indicative.
Columbus: Slavica.
Friedman, Victor, A. 1978. On the Semantic and Morphological Influence of Turkish
on Balkan Slavic. In Papers from the 14th Regional Meeting Chicago Linguistic Society,
eds. Donka Farkas, Wesley M. Jacobsen and Karol W. Todrys. Chicago:
Chicago Linguistic Society.
Friedman, Victor, A. 2002. Macedonian. In The Slavonic Languages, eds. Bernard Comrie
and Greville G. Corbett, 249-305. London: Routledge.
Gallis, Arne. 1960. Die neuen slavischen Perfekte vom Typus factum habeo und *casus
sum, *casum habeo. Scando-Slavica 6:176-188.
References 295
Giorgi, Alessandra, and Fabio Pianesi. 1997. Tense and Aspect: From Semantics to
Morphosyntax. New York: Oxford University Press.
Gogolewski, Stanisław. 1963. Wpływy niemieckie na kaszubski system czasów
przeszłych. Rozprawy Komisji Językowej Łódzkiego Towarzystwa Naukowego IX:69-
75.
Gołab, Zbigniew. 1959. Some Arumanian-Macedonian Isogrammatisms and the Social
Background of their Development. Word 15:415-435.
Goodall, Grant. 1993. On Case and the Passive Morpheme. Natural Language and
Linguistic Theory 11:31-44.
Górecka, Janina, and Witold Śmiech. 1972. Czas przyszły złożony w języku polskim.
Rozprawy Komisji Językowej Łódzkiego Towarzystwa Naukowego 18:11-38.
Graves, Nina. 2000. Macedonian - a Language with Three Perfects? In Tense and Aspect
in the Languages of Europe, ed. Östen Dahl, 479-493. Berlin: Mouton de Gruyter.
Grenoble, Lenore. 1995. The Imperfective Future Tense in Russian. Word 46:183-205.
Haegeman, Liliane. 1992. Theory and Description in Generative Syntax: A Case Study in West
Flemish: Cambridge: Cambridge University Press.
Hale, Ken, and Samuel Keyser. 1993. On Argument Structure and the Lexical
Expression of Syntactic Relations. In The View from Building 20: Essays in
Linguistics in Honor of Sylvain Bromberger, eds. Ken Hale and Samuel Keyser.
Cambridge, MA/London: MIT Press.
Halpern, Aaron L. 1992. Topics in the Placement and Morphology of Clitics. Ph.D.
dissertation. Stanford University.
Hauge, Kjetil Rå. 1999. The Word Order of Predicate Clitics in Bulgarian. Journal of
Slavic Linguistics 7:91-139.
Hendriks, Peter. 1976. The Radožda-Vevčani Dialect of Macedonian. Ph.D.
dissertation. University of Amsterdam.
Hewson, John, and Vit Bubenik. 1997. Tense and Aspect in Indo-European Languages:
Theory, Typology, Diachrony. Amsterdam: Benjamins.
Hoekstra, Eric. 1991. Licensing Conditions on Phrase Structure. PhD dissertation.
University of Groningen.
Hoekstra, Teun. 1984. Transitivity. Grammatical Relations in Government-Binding Theory.
Dordrecht/Cinnaminson: Foris.
Hoekstra, Teun. 1986. Passives and Participles. In Linguistics in the Netherlands 1986, eds.
Frits Beukema and Aafke Hulk, 95-105. Dordrecht: Foris.
Hoekstra, Teun. 2004. Small Clauses Everywhere. In Arguments and Structure: Studies on
the Architecture of the Sentence, eds. Rint Sybesma, Sjef Barbiers, Marcel Den
Dikken, Jenny Doetjes and Guido Vanden Wyngaerd. Berlin: Mouton de
Gruyter.
Hoekstra, Teun, and René Mulder. 1990. Unergatives as Copular Verbs; Locational and
Existential Predication. The Linguistic Review 7:1-79.
Hoffmann, Gerhard. 1934. Die Entwicklung des umschriebenen Perfektums im
Altenglischen und Frühmittelenglischen. Ph.D. dissertation. University of
Breslau (Wrocław).
Holmberg, Anders. 2000. Scandinavian Stylistic Fronting: How Any Category Can
Become an Expletive. Linguistic Inquiry 31:445-484.
Holmberg, Anders. 2006. Stylistic Fronting. In The Syntax Companion (SynCom), eds.
Martin Everaert and Henk van Riemsdijk. Oxford: Blackwell.
Holmberg, Anders, and Christer Platzack. 1995. The Role of Inflection in Scandinavian
Syntax. Oxford: Oxford University Press.
296 References
Huntley, David. 2002. Old Church Slavonic. In The Slavonic Languages, eds. Bernard
Comrie and Greville G. Corbett, 125-187. London: Routledge.
Iatridou, Sabine. 1990. About AgrS(P). Linguistic Inquiry 21:551-577.
Izvorski, Roumyana, Tracy Holloway King, and Catherine Rudin. 1997. Against Li
lowering in Bulgarian. Lingua 102:187-194.
Jabłońska, Patrycja. 2006. Radical Decomposition and Argument Structure. Ph.D
dissertation. University of Tromsø.
Jackendoff, Ray. 1977. X-bar syntax. Cambridge MA: MIT Press.
Jaegli, Oswaldo. 1986. Passive. Linguistic Inquiry 17:587-622.
Joseph, Brian D. 1983. The Synchrony and Diachrony of the Balkan Infinitive. Cambridge:
Cambridge University Press.
Julien, Marit. 2001. The Syntax of Complex Tenses. The Linguistic Review 18:125-167.
Kayne, Richard. 1993. Towards a Modular Theory of Auxiliary Selection. Studia
Linguistica 47:3-31.
Kayne, Richard. 1994. The Antisymmetry of Syntax: Cambridge, Ma: MIT Press.
Kayne, Richard. 2004. Prepositions as Probes. In Structures and Beyond: Cartography of
Syntactic Structures, vol. 3, ed. Adriana Belletti, 192-212. Oxford: Oxford
University Press.
Kayne, Richard, and Jean-Yves Pollock. 2001. New Thoughts on Stylistic Inversion. In
Subject Inversion in Romance and the Theory of Universal Grammar, eds. Aafke Hulk
and Jean-Yves Pollock, 107-162. New York: Oxford University Press.
Kern, Johan Hendrik. 1912. De met het participium praeteriti omschreven werkwoordsvormen in 't
Nederlands, in Verhandelingen der Koninklijke Akademie van Wetenschappen, afdeeling
Letterkunde, Nieuwe Reeks, deel XII no. 2. Amsterdam: Johannes Müller.
Kibort, Anna. 2004. Passive and Passive-like Constructions in English and Polish.
Ph.D. dissertation. Pembroke College, University of Cambridge.
Klemensiewicz, Zenon, Tadeusz Lehr-Spławiński, and Stanisław Urbańczyk. 1964.
Gramatyka historyczna języka polskiego. Warszawa: Państwowe Wydawnictwo
Naukowe.
Klima, Edward. 1964. Negation in English. In The Structure of Language, eds. Jerry Fodor
and Jerrold J. Katz. Englewood Cliffs NJ: Prentice-Hall.
Koneski, Blaže. 1987. Gramatika na makedonskiot literaturen jazik. Skopje: Kultura.
Koster, Jan. 1987. Domains and Dynasties. The Radical Autonomy of Syntax.
Dordrecht/Providence: Foris.
Kowalska, Alina. 1976. Ewolucja analitycznych form czasownikowych z imiesłowem na -ł w
języku polskim. Katowice: Uniwersytet Śląski.
Krapova, Iliyana. 1999a. The System of Auxiliaries in Bulgarian. In Topics in South Slavic
Syntax and Semantics, eds. Mila Dimitrova-Vulchanova and Lars Hellan, 59-89.
Krapova, Iliyana. 1999b. Auxiliaries and Complex Tenses in Bulgarian. In Formal
Approaches to Slavic Linguistics, eds. Katarzyna Dziwirek, Herbert Coats and
Cynthia M. Vakareliyska, 320-344. Ann Arbor: Michigan Slavic Publications.
Kratzer, Angelika. 1994. The Event Argument and the Semantics of Voice. Ms.
University of Massachusetts, Amherst.
Kratzer, Angelika. 2000. Building Statives. In Proceedings of the 26th Annual Meeting of the
Berkeley Linguistics Society, eds. Lisa Conathan, Jeff Good, Darya Kavitskaya,
Alyssa Wulf and Alan C.L. Yu, 385-399. Berkeley: University of California,
Berkeley Linguistics Society.
Krause, Wolfgang, and Thomas Werner. 1960. Tocharisches Elementarbuch. Heidelberg:
Winter.
References 297
Křižková, Helena. 1960. Vývoj opisného futura v jazycích slovanských, zvlášte v ruštine. Prague:
Státní pedagogické nakl.
Kroch, Anthony. 1989. Function and Grammar in the History of English Periphrastic
do. In Language Change and Variation, eds. Ralph Fasold and Deborah Shiffrin.
Philadelphia: Benjamins.
Kupść, Anna, and Małgorzata Marciniak. 1997. Some Notes on HPSG Binding Theory
for Polish. In Formale Slavistik, eds. Uwe Junghanns and Gerhild Zybatow,
193-203. Frankfurt am Main: Vervuert Verlag.
Kuryłowicz, Jerzy. 1964. The Inflectional Categories of Indo-European. Heidelberg: Carl
Winter.
Ladusaw, William A. 1980. Polarity Sensitivity as Inherent Scope Relations. New York:
Garland Press.
Laka, Itziar. 1994. On the Syntax of Negation. New York: Garland Press.
Lambova, Mariana. 2001. On A-bar Movements in Bulgarian and their Interaction. The
Linguistic Review 18:327-374.
Lambova, Mariana. 2003. On Information Structure and Clausal Architecture: Evidence
from Bulgarian. Ph.D. dissertation. University of Connecticut.
Lambova, Mariana. 2004. On Triggers of Movement and Effects at the Interfaces. In
Triggers, eds. Anne Breitbarth and Henk van Riemsdijk, 231-258. Berlin:
Mouton de Gruyter.
Lasnik, Howard. 1995. A Note on Pseudogapping. In Papers on Minimalist Syntax, eds.
Robert Pensalfini and Hiroyuki Ura, 143-163. Cambridge, Mass.: MIT,
Department of Linguistics and Philosophy, MITWPL.
Lasnik, Howard, and Mamoru Saito. 1984. On the Nature of Proper Government.
Lavine, James. 2000. Topics in the Syntax of Nonagreeing Predicates in Slavic. Ph.D.
dissertation. Princeton University.
Lehmann, Winfred P. 1974. Proto-Indo-European Syntax. Austin: University of Texas
Press.
Lehmann, Winfred P. 1993. Theoretical Bases of Indo-European Linguistics. London:
Routledge.
Leko, Nedžad. 1996. Notes on Negation and Related Topics in Bosnian. GenGenP
(Geneva Generative Papers) 4:15-34.
Lema, José, and María-Luisa Rivero. 1989. Long Head Movement: ECP vs. HMC. In
Proceedings of NELS 20, ed. Tim Sherer, 333-347. Amherst: GLSA, University
of Massachusetts.
Lenartová, Denisa. 2001. Czech Pronominal Clitics. Paper presented at The workshop on
Slavic pronominal clitics, Berlin, 8-9 February 2001.
Levin, Beth, and Malka Rappaport. 1986. The Formation of Adjectival Passives.
Lindstedt, Jouko. 1985. On the Semantics of Tense and Aspect in Bulgarian. Ph.D.
Dissertation. University of Helsinki.
Lindstedt, Jouko. 1994. On the Development of the South Slavonic Perfect. Three Papers
on the Perfect. EUROTYP Working Papers, Series VI, July 1994 5:32-53.
Lindstedt, Jouko. 2000a. Is There a Balkan Verb System? In Papers from Third Conference
on Formal Approaches to South Slavic and Balkan Languages, Plovdiv September 1999,
eds. Mila Dimitrova-Vulchanova, Iliyana Krapova and Lars Hellan, 1-8.
Trondheim: University of Trondheim.
298 References
Lindstedt, Jouko. 2000b. The Perfect - Aspectual, Temporal and Evidential. In Tense and
Aspect in the Languages of Europe, ed. Östen Dahl, 365-383. Berlin: Mouton de
Gruyter.
Linebarger, Marcia C. 1981. The Grammar of Negative Polarity. Bloomington: Indiana
University Linguistic Club.
Lorentz, Friedrich. 1919. Kaschubische Grammatik. Danzig: Gedania.
Lubańska, Maja. 2000. Contrastive Scrambling in Polish. In PASE Papers in Language
Studies. Proceedings of the 8th Annual Conference of the Polish Association for the Study of
English, ed. Bożena Rozwadowska. Wrocław: Aksel.
Lubańska, Maja. 2005. Focus on Wh-Questions. Frankfurt am Main: Peter Lang.
Lunt, Horace, G.1952. A Grammar of the Macedonian Literary Language. Skopje: Drzavno
knigoizd. na NR Makedonija.
Lunt, Horace G. 1974. Old Church Slavonic Grammar. The Hague: Mouton.
Lysaght, Thomas E. 1982. Old Church Slavonic (Old Bulgarian)-Middle Greek-Modern English
Dictionary. Wien: Hollinek.
Maslov, Jurij S. 1988. Resultative, Perfect, and Aspect. In Typology of Resultative
Constructions, eds. Vladimir P. Nedjalkov and Bernard Comrie. Amsterdam:
Benjamins.
Massam, Diane. 2001a. Pseudo Noun Incorporation in Niuean. Natural Language and
Massam, Diane. 2001b. On Predication and the Status of Subjects in Niuean. In Objects
and Other Subjects, eds. William Davies and Stanley Dubinsky, 225-246.
Amsterdam: Kluwer.
Mathesius, Vilém. 1975. A Functional Analysis of Present Day English on a General Linguistic
Basis. The Hague: Mouton.
Meillet, Antoine. 1958. Wstęp do językoznawstwa indoeuropejskiego. Warszawa: Państwowe
Wydawnictwo Naukowe.
Migdalski, Krzysztof. 2005. To Have or to Be: On the Syntax of Perfect Tenses in
South Slavic. In Formal Approaches to Slavic Linguistics: The South Carolina Meeting,
2004, eds. Steven Franks, Frank Gladney and Mila Tasseva-Kurktchieva, 230-
240. Ann Arbor: Michigan Slavic Publications.
Mikoś, Michael J. 1985. Alternative Forms of the Future Imperfective Tense in Polish.
Slavic and East European Journal 29:448-460.
Mikoś, Michael J., and Edith A. Moravcsik. 1986. Moving Clitics in Polish and Some
Crosslinguistic Generalizations. Studia Slavica 32:327-336.
Mitchell, Bruce. 1985. Old English Syntax.vol. 1. Oxford: Clarendon Press.
Młynarczyk, Anna. 2004. Aspectual Pairing in Polish. Ph.D. dissertation. Institute of
Linguistics OTS, Utrecht University.
Müller, Gereon. 1998. Incomplete Category Fronting. Dordrecht: Kluwer.
Nash, Léa, and Alain Rouveret. 2002. Cliticization as Unselective Attract. Catalan Journal
of Linguistics 1:157-199.
Nitsch, Kazimierz. 1956. Tajemnice polskiego czasu przyszłego złożonego. Język polski
36:190-196.
Oesterreicher, Henryk. 1926. Imiesłów bierny w języku polskim. Kraków: Polska Akademia
Umiejętności.
Ordóńez, Francisco. 1998. Post-Verbal Asymmetries in Spanish. Natural Language and
Linguistic Theory 16:313 - 345.
Osawa, Fuyo. 1999. The Relation between Tense and Aspect: The Emergence of the T
system. UCL Working Papers in Linguistics 11:521-543.
References 299
Oubouzar, Erika. 1974. Über die Ausbildung der zusammengesetzten Verbformen im

deutschen Verbalsystem. Beiträge zur Geschichte der deutschen Sprache und Literatur
(Halle) 95:5-96.
Ouhalla, Jamal. 1991. Functional Categories and Parametric Variation. London: Routledge.
Ozga, Janina. 1976. Clitics in English and Polish. Papers and Studies in Contrastive
Linguistics 4:127-140.
Pancheva, Roumyana. 2005. The Rise and Fall of Second-Position Clitics. Natural
Language and Linguistic Theory 23:103-167.
Pavlov, Ivan. 2000. The Non-Argument Field of Bulgarian Clause Structure. Ph.D.
dissertation. University of Trondheim.
Perlmutter, David. 1971. Deep and Surface Structure Constraints in Syntax. New York:
Rinehart & Winston Inc.
Pesetsky, David. 1982. Paths and Categories. Ph.D. dissertation. MIT.
Pesetsky, David. 1989. The Earliness Principle. Ms. MIT, Cambridge, Mass.
Phillips, Colin. 1996. Order and Structure. Ph.D. dissertation. MIT.
Piotrowski, Jan. 1981. Składnia słowińska wobec wpływów języka niemieckiego. Wrocław:
Ossolineum.
Pisarkowa, Krystyna. 1984. Historia składni języka polskiego. Wrocław: Ossolineum.
Poletto, Cecilia. 2000. The Higher Functional Field: Evidence from Northern Italian Dialects.
New York: Oxford University Press.
Pollock, Jean-Yves. 1989. Verb Movement, Universal Grammar and the Structure of
IP. Linguistic Inquiry 20:365-424.
Progovac, Ljiljana. 1993. Locality and Subjunctive-like Complements in Serbo-Croatian.
Journal of Slavic Linguistics 1:116-144.
Progovac, Ljiljana. 1996. Clitics in Serbian/Croatian: Comp as the Second Position. In
Approaching Second: Second Position Clitics and Related Phenomena, eds. Aaron
Halpernand and Arnold Zwicky, 411-428. Stanford: CSLI Publications.
Progovac, Ljiljana. 1999. Eventive To and the Placement of Clitics in Serbo-Croatian. In
Crossing Boundaries: Advances in the Theory of Central and Eastern European
Languages, ed. István Kenesei, 33-44. Amsterdam: John Benjamins.
Progovac, Ljiljana. 2000. Where do Clitics Cluster? In Clitic Phenomena in European
Languages, eds. Frits Beukema and Marcel Den Dikken, 249-258. Amsterdam:
Benjamins.
Progovac, Ljiljana. 2001. Clausal Functional Projections in Serbian. In Formal Approaches
to Slavic Linguistics: the Bloomington Meeting 2000, eds. Steven Franks, Tracy
Holloway King and Misha Yadroff, 229-256. Ann Arbor: Michigan Slavic
Publications.
Progovac, Ljiljana. 2005a. Syntax of Serbian: Clausal Architecture. Bloomington: Slavica.
Progovac, Ljiljana. 2005b. Negative and Positive Feature Checking and the Distribution
of Polarity Items. In Negation in Slavic, eds. Sue Brown and Adam
Przepiórkowski. Bloomington: Slavica.
Przepiórkowski, Adam. 2000. Long Distance Genitive of Negation in Polish. Journal of
Slavic Linguistics 8:151-189.
Radanović-Kocić, Vesna. 1988. The Grammar of Serbo-Croatian Clitics: A Synchronic
and Diachronic Perspective. Ph.D. dissertation. University of Illinois, Urbana.
Rappaport, Gilbert. 1988. On the Relationship Between Prosodic and Syntactic
Properties of Pronouns in the Slavic Languages. In American Contributions to the
Tenth International Congress of Slavists, ed. Alexander M. Schenker, 301-327.
Bloomington: Slavica.
300 References
Rappaport, Gilbert. 2000. Extraction from Nominal Phrases in Polish and the Theory
of Determiners. Journal of Slavic Linguistics 8:159-198.
Reinhart, Tanya. 1983. Anaphora and Semantic Interpretation. London/Sydney: Croom
Helm.
Riemsdijk, Henk van. 1999. Clitics: A State of the Art Report In Clitics in the Languages of
Europe, ed. Henk van Riemsdijk, 1-30. Berlin: Mouton de Gruyter.
Riemsdijk, Henk van. 2002. The Unbearable Lightness of GOing. Journal of Comparative
Germanic Linguistics 5:143-196.
Riemsdijk, Henk van, and Edwin Williams. 1986. Introduction to the Theory of Grammar.
Cambridge: The MIT Press.
Rittel, Teodozja. 1975. Szyk członów w obrębie form czasu przeszłego i trybu przypuszczającego.
Wrocław: Ossolineum.
Rivero, María-Luisa. 1991. Long Head Movement and Negation: Serbo-Croatian vs.
Slovak and Czech. The Linguistic Review 8:319-351.
Rivero, María-Luisa. 1993. Bulgarian and Serbo-Croatian yes-no Questions: V0-raising
to li vs. li-hopping. Linguistic Inquiry 24:567-575.
Rivero, María-Luisa. 1994a. Clause Structure and V-movement in the Languages of the
Balkans. Natural Language and Linguistic Theory 12:63-120.
Rivero, María-Luisa. 1994b. Negation, Imperatives, and Wackernagel effects. Rivista di
Linguistica 6:91-118.
Rivero, María-Luisa. 1997. On two Locations for Complement Clitic Pronouns: Serbo-
Croatian, Bulgarian, and Old Spanish. In Parameters of Morphosyntactic Change,
eds. Nigel Vincent and Ans van Kemenade, 170-206. Cambridge: Cambridge
University Press.
Rivero, María-Luisa. 2000. Finetness and Second Position in Long Verb Movement
Languages: Breton and Slavic. In The Nature and Function of Syntactic Categories,
ed. Borsley Robert. San Diego: Academic Press.
Rivero, María-Luisa. 2005. Topics in Bulgarian Morphology and Syntax: A Minimalist
Perspective. Lingua 115:1083-1128.
Rivero, María-Luisa, and Arhonto Terzi. 1995. Imperatives, V-Movement, and Logical
Mood. Journal of Linguistics 31:301-332.
Roberts, Ian. 1987. The representation of Implicit and Dethematized Subjects. Dordrecht: Foris.
Roberts, Ian. 1992. Verbs and Diachronic Syntax. Dordrecht: Kluwer.
Roberts, Ian. 1994. Two Types of Verb Movement in Romance. In Verb Movement, eds.
Norbert Hornstein and David Lightfoot. Cambridge: Cambridge University
Press.
Rochemont, Michael. 1978. A Theory of Stylistic Rules in English. New York: Garland
Press.
Rooryck, Johan. 1992. Romance Enclitic Ordering and Universal Grammar. The
Linguistic Review 9:219-250.
Ross, John. 1969. Auxiliaries as Main Verbs. In Studies in Philosophical Linguistics, ed.
William Todd, 77-102. Evanston, IL: Great Expectations Press.
Rozwadowska, Bożena. 1992. Thematic Constraints on Selected Constructions in English and
Polish. Wrocław: Wydawnictwo Uniwersytetu Wrocławskiego.
Rozwadowska, Bożena. 1997. Towards a Unified Theory of Nominalizations. External and
Internal Eventualities. Wrocław: Wydawnictwo Uniwersytetu Wrocławskiego.
Rudin, Catherine. 1986. Aspects of Bulgarian Syntax: Complementizers and Wh-Constructions.
Columbus: Slavica.
Rudin, Catherine. 1988. On Multiple Questions and Multiple Wh-Fronting. Natural
References 301
Rudin, Catherine. 1997. AgrO and Bulgarian Pronominal Clitics. In Formal Approaches to
Slavic Linguistics: The Indiana Meeting, eds. Martina Lindseth and Steven Franks,
224-252. Ann Arbor: Michigan Slavic Publications.
Rudin, Catherine, Christina Kramer, Loren Billings, and Matthew Baerman. 1999.
Macedonian and Bulgarian li Questions: Beyond Syntax. Natural Language and
Růžička, Rudolf. 1963. Das syntaktische System der altslavischen Partizipien und sein Verhältnis
zum griechischen. Berlin: Veröffentlichungen des Instituts für
Slawistik/Deutsche Akademie der Wissenschaften zu Berlin.
Salvi, Giampaolo. 1987. Syntactic Restructuring in the Evolution of Romance
Auxiliaries. In Historical Development of Auxiliaries, eds. Martin Harris and Paolo
Ramat, 225-236. Berlin: Mouton de Gruyter.
Scatton, Ernest, A. 1984. A Reference Grammar of Modern Bulgarian. Columbus: Slavica.
Schenker, Alexander M. 2002. Proto-Slavonic. In The Slavonic Languages, eds. Bernard
Comrie and Greville G. Corbett, 60-124. London: Routledge.
Schick, Ivanka. 2000. Clitic Doubling Constructions in Balkan-Slavic Languages. In
Clitic Phenomena in European Languages, eds. Frits Beukema and Marcel Den
Dikken, 260-292. Amsterdam: John Benjamins.
Schields, Kenneth C. 1992. A History of Indo-European Verb Morphology. Amsterdam:
Benjamins.
Schmalstieg, William, R. 1983. An Introduction to Old Church Slavic. Columbus: Slavica.
Schooneveld, Cornelis H. van 1951. The Aspectual System of the Old Church Slavonic
and Old Russian verbum finitum byti. Word 7:93-103.
Schooneveld, Cornelis H. van 1959. A Semantic Analysis of the Old Russian Finite Preterite
System: Slavic Printings and Reprintings 7. The Hague: Mouton.
Schütze, Carson. 1994. Serbo-Croatian Second Position Clitic Placement and the
Phonology-Syntax Interface. MIT Working Papers in Linguistics 21:373-473.
Short, David. 2002. Czech. In The Slavonic Languages, eds. Bernard Comrie and Greville
G. Corbett, 455-532. London: Routledge.
Siegel, Dorothy. 1973. Non-sources for Un-passives. In Syntax and Semantics II, ed. John
Kimball. New York: Seminar Press.
Siewierska, Anna. 1988. The Passive in Slavic. In Passive and Voice, ed. Masayoshi
Shibatani, 243-289. Amsterdam: Benjamins.
Sławski, Franciszek. 1946. Miejsce enklityki odmiennej w dziejach języka bułgarskiego. Kraków:
Polska Akademia Umiejętności.
Słoński, Stanisław. 1926. Tak zwane perfectum w językach słowiańskich. Prace
Filologiczne 10.
Smith, John Charles. 1995. Agreement between Past Participle and Direct Object in
Catalan: The Hypothesis of Castilian Influence Revisited. In Linguistic Change
under Contact Conditions, ed. Jacek Fisiak, 271-289. Berlin: Mouton de Gruyter.
Smith, Neil V. 1989. The Twitter Machine: Reflections on Language. Oxford: Blackwell.
Spencer, Andrew. 1991. Morphological Theory: An Introduction to Word Structure in Generative
Grammar. Oxford: Blackwell.
Sportiche, Dominique. 1988. A Theory of Floating Quantifiers and its Corollaries for
Constituent Structure. Linguistic Inquiry 19:425-449.
Stieber, Zdzisław. 1973. Zarys gramatyki porównawczej języków słowiańskich. Fleksja werbalna.
Warszawa: Państwowe Wydawnictwo Naukowe.
Stjepanović, Sandra. 1998. On the Placement of Serbo-Croatian Clitics: Evidence from
VP Ellipsis. Linguistic Inquiry 29:527-537.
302 References
Stjepanović, Sandra. 1999. What do Second Position Cliticization, Scrambling, and

Multiple wh-fronting have in Common? Ph.D. dissertation. University of
Connecticut.
Stone, Gerald. 2002. Cassubian. In The Slavonic Languages, eds. Bernard Comrie and
Greville G. Corbett, 759-794. London: Routledge.
Stowell, Tim. 1981. Origins of Phrase Structure. Ph.D. dissertation. MIT.
Svenonius, Peter. 2004. Russian Prefixes are Phrasal. Ms. University of Tromsø.
Szczegielniak, Adam. 1997. Deficient Heads and Long Head Movement in Slovak. In
Formal Approaches to Slavic Linguistics. The Indiana Meeting, eds. Martina Lindseth
and Steven Franks, 312-333. Ann Arbor: Michigan Slavic Publications.
Szpiczakowska, Monika. 1988. Ruchomość końcówek czasu przeszłego we
współczesnym języku polskim. Zeszyty Naukowe Uniwersytetu Jagiellońskiego. Prace
Językoznawcze 97:41-53.
Śpiewak, Grzegorz, and Izabela Szymańska. 1995. A Poll on Pollock - Functional
Categories in Polish. In Licensing in Syntax and Phonology, ed. Edmund
Gussmann, 125-145. Lublin: Folium.
Świderska-Koneczna, Halina. 1930. Użycie słowa mieć jako słowa posiłkowego w języku
polskim. Prace Filologiczne 15:263-272.
Taraldsen, Knut Tarald. 1995. On Agreement and Nominative Objects in Icelandic. In
Studies in Comparative Germanic Syntax, eds. Hubert Haider, Susan Olsen and
Sten Vikner, 307-327. Dordrecht: Kluwer.
Terzi, Arhonto. 1999. Clitic Combinations, Their Hosts, and Their Ordering. Natural
Thiersch, Craig. 1985. VP and Scrambling in the German Mittelfeld. Ms. Tilburg
University, Models of Grammar Group.
Toman, Jindrich. 1980. Weak and Strong: Notes on be in Czech. In Wege zur
Universalienforschung: sprachwissenschaftliche Beiträge zum 60. Geburtstag von
Hansjakob Seiler, eds. Gunter Brettschneider and Christian Lehmann.
Tübingen: Gunter Narr.
Tomić, Olga. 1983. Linguistic Features and Grammatical Categories. Folia Linguistica
17:387-401.
Tomić, Olga. 1984. Grammatical Categories across Cultures. In Language across Cultures,
eds. Liam Mac Mathúna and David Singleton, 31-39. Dublin: Irish Association
for Applied Linguistics.
Tomić, Olga. 1989. Language Contact, Language System and Language Code. In
Yugoslav General Linguistics, ed. Milorad Radovanović, 361-381. Amsterdam:
Benjamins.
Tomić, Olga. 1996a. The Balkan Slavic Clausal Clitics. Natural Language and Linguistic
Theory 14:811-872.
Tomić, Olga. 1996b. Focus on Focus. In Papers from the First Conference on Formal
Approaches to South Slavic Languages (Plovdiv, October 1995), eds. Mila Dimitrova-
Vulchanova and Lars Hellan, 309-318: University of Trondheim.
Tomić, Olga. 1997. Non-first as a Default Clitic Position. Journal of Slavic Linguistics
5:301-323.
Tomić, Olga. 1999. Cliticization and Clitichood. In Crossing Boundaries: Advances in the
Theory of Central and Eastern European Languages, ed. István Kenesei, 9-32.
Tomić, Olga. 2000. On Clitic Sites. In Clitic Phenomena in European Languages, eds. Frits
Beukema and Marcel Den Dikken, 293-316. Amsterdam: Benjamins.
References 303
Tomić, Olga. 2001a. The Macedonian Negation Operator and Cliticization. Natural
Tomić, Olga. 2001b. Negation and Imperatives. In Current Issues in Formal Slavic
Linguistics, eds. Gerhild Zybatow, Uwe Junghanns, Grit Mehlhorn and Luka
Szucsich, 159-168. Frankfurt am Main: Peter Lang.
Tomić, Olga. 2004. The Syntax of the Balkan Slavic Future Tenses. Lingua 114:517-542.
Tomić, Olga. 2005. The Syntax of Negative Imperatives in Balkan Slavic. In Formal
Approaches to Slavic Linguistics: The South Carolina Meeting, 2004, eds. Steven
Franks, Frank Gladney and Mila Tasseva-Kurktchieva, 365-375. Ann Arbor:
Tomić, Olga. 2006. The Perfect and the Evidential. In The Grammar of Macedonian
(working title), 333-411.
Tomić, Olga. forthcoming. The South Slavic Pronominal Clitics. Journal of Slavic
Linguistics 12: 215-250.
Tommola, Hannu. 2000. On the Perfect in North Slavic. In Tense and Aspect in the
Languages of Europe, ed. Östen Dahl, 655-692. Berlin: Mouton de Gruyter.
Topolińska, Zuzanna. 1961. Z historii akcentu polskiego od wieku XVI do dziś. Wrocław:
Ossolineum.
Travis, Lisa. 1984. Parameters and Effects of Word Order Variation. Ph.D. dissertation.
MIT.
Uriagereka, Juan. 1995. Aspects of the Syntax of Clitic Placement in Western Romance.
Vačkov, Veselin. 1998. Reduplicated Objects and Informational Articulation in
Bulgarian and Macedonian. University of Trondheim Working Papers in Linguistics
31:165-179.
Vaillant, André. 1948. Manuel de vieux slave. Vol 1. Paris: Institut d'études slaves.
Vaillant, André. 1966. Grammaire comparée des langues slaves. Tome III: Le verbe. Paris:
Éditions Klincksieck.
Vaillant, André. 1977. Grammaire comparée des langues slaves. Tome V: La syntaxe. Paris:
Éditions Klincksieck.
Veselovská, Ludmila. 2004. The Extended Verbal Projection in Czech: Three Variants
of the Verb be. Paper presented at Formal Description of Slavic Languages-5,
University of Leipzig.
Wackernagel, Jacob. 1892. Über ein Gesetz der indogermanischen Wortstellung.
Indogermanische Forschungen 1:333-436.
Wasow, Thomas. 1977. Transformation and the Lexicon. In Formal Syntax, eds. Peter
Culicover, Thomas Wasow and Adrian Akmajian, 327-360. New York:
Academic Press.
Watanabe, Akira. 1993. Agr-based Case Theory and its Interaction with the A-bar
System. Ph.D. dissertation. MIT.
Watkins, Calvert. 1962. Indo-European Origins of the Celtic Verb. Dublin: The Dublin
Institute for Advanced Studies.
Whaley, Marika. 2000a. The Evolution of the Slavic ‘Be(come)’-type Compound
Future. Ph.D dissertation. The Ohio State University.
Whaley, Marika. 2000b. A Semantic Analysis of the Slavic Future Perfect. Paper
presented at AATSEEL 2000, Washington, D.C.
Wilder, Chris, and Damir Ćavar. 1994. Long Head Movement? Verb Movement and
Cliticization in Croatian. Lingua 93:1-58.
Willim, Ewa. 1989. On Word Order: A Government-Binding Study of English and Polish.
Kraków: Zeszyty Naukowe Uniwersytetu Jagiellońskiego.
304 References
Willim, Ewa. 1990. On Case Marking in Polish. Papers and Studies in Contrastive Linguistics
25:204-220.
Willis, David. 2000. Verb Movement in Slavonic Conditionals. In Diachronic Syntax:
Models and mechanisms, ed. Anthony Warner, 322-348. Oxford: Oxford
University Press.
Wischer, Ilse. 2004. The HAVE-'Perfect' in Old English. In New Perspectives On English
Historical Linguistics, eds. Christian Kay, Simon Horobin and Jeremy Smith,
243-255. Amsterdam: John Benjamins.
Witkoś, Jacek. 1998. The Syntax of Clitics. Steps Towards a Minimalist Account. Poznań:
Motivex.
Wojtyła-Świerzowska, Maria. 1974. Protosłowiańskie Nomen Agentis. Wrocław:
Ossolineum.
Zanuttini, Raffaella. 1997. Negation and Clausal Structure. A Comparative Study of Romance
Languages. New York: Oxford University Press.
Zec, Draga, and Sharon Inkelas. 1990. Prosodically Constrained Syntax. In The
Phonology-Syntax Connection, eds. Sharon Inkelas and Draga Zec, 365-378.
Stanford: CSLI Publications.
Zeijlstra, Hedde. 2004. Sentential Negation and Negative Concord. Ph.D. dissertation.
University of Amsterdam.
Zwart, Jan Wouter. 1997. Morphosyntax of Verb Movement. A Minimalist Approach to the
Syntax of Dutch. Dordrecht: Kluwer.
Zwicky, Arnold. 1977. On Clitics. Bloomington: Indiana University Linguistics Club.
Samenvatting in het Nederlands
Dit proefschrift behandelt de diachronie en synchronie van de structuur van
samengestelde tijden in de Slavische talen. De onderliggende aanname is dat het
temporele systeem in de tegenwoordige Slavische talen is vereenvoudigd als gevolg van
een overvloed aan aspectuele markering door zowel temporele als aspectuele morfemen
in het Protoslavisch en het Kerkslavisch. Deze vereenvoudiging vond plaats op twee
manieren: morfosyntactisch, door het verdwijnen van de aspectuele tijdsvormen, en
semantisch, door een herinterpretatie van de voltooid tegenwoordige tijd. De
eerstgenoemde strategie vinden wij in de West- en Oost-Slavische talen, die de aorist en
het imperfectum verloren en het hulpwerkwoord in het perfectum gereduceerd hebben.
De tweede strategie vinden we in het Zuid-Slavische talen, die in de regel de aspectuele
tijden behouden hebben, maar de voltooid tegenwoordige tijd geherinterpreteerd
hebben als een markeerder van niet-evidentialiteit.
Het Bulgaars en het Servo-Kroatisch hebben de morfologische opmaak van de
voltooid tegenwoordige tijd in het Kerkslavisch behouden. Het wordt gevormd met het
hulpwerkwoord ‘zijn’, dat dezelfde vorm heeft als het koppelwerkwoord en dat in
verschillende aspectuele varianten optreedt om temporele verschillen uit te drukken.
Het hulpwerkwoord is vergezeld van het zogenaamde l-participium, dat zich ontwikkeld
heeft uit de klasse van Proto-Indo-Europeese adjectieven die beginnen met *lo-. Deze
adjectieven zijn nu volledig geverbaliseerd en kunnen naamval en thematische rollen
toekennen. Het l-participium congrueert altijd in geslacht en getal met het onderwerp.
In hoofdstuk 2 betoog ik dat het l-participium dankzij deze eigenschap naar Spec, TP
verplaatst kan worden om de φ-kenmerken van T te checken. Ik stel verder voor dat dit
door middel van XP-verplaatsing gebeurt, op vergelijkbare wijze als locatief inversie in
het Engels plaatsvindt. De belangrijkste argumenten voor deze stelling zijn de
complementaire distributie van onderwerp en het vooropgeplaatste participium en het
feit dat het vooropgeplaatste participium direct aan het hulpwerkwoord moet
voorafgaan. Meer evidentie wordt gevormd door eigenschappen van dubbele
deelwoordconstructies, korte participiumverplaatsing, verplaatsing van het participium
over het hulpwerkwoord voor de toekomende tijd šte in het Bulgaars, en
zinsontkenning. Met dit voorstel geef ik een alternatief voor analyses die stellen dat de
verplaatsing van het l-participium het gevolg van hoofdverplaatsing/adjunctie is.
In hoofdstuk 3 onderzoek ik de samengestelde tijden die gevormd worden met het
hulpwerkwoord ‘hebben’ en het voltooid deelwoord. Dit is tot nu toe nauwelijks
onderzocht en dit proefschrift biedt dan ook de eerste gedetailleerde analyse van deze
constructie binnen het generatieve kader. In tegenstelling tot de voltooide tijd met het
hulpwerkwoord ‘zijn’, die men in alle Slavische talen aantreft, hebben alleen het
Kasjoebisch en het Macedonisch een volledig paradigma van de voltooide tijd met
‘hebben’ ontwikkeld. In deze twee talen congrueert het participium niet met het
onderwerp in φ-kenmerken, maar verschijnt het altijd in onveranderlijke (enkelvoudige
en onzijdige) vorm. Het is morfologisch gezien identiek aan het passieve participium en
daarom wordt wel beweerd dat het niet in staat is om een externe theta-rol te nemen of
structurele naamval aan het voorwerp toe te kennen. Deze functies worden vervuld
door het hulpwerkwoord ‘hebben’. Een aantal andere Slavische talen, waarin de
voltooide tijd met ‘hebben’ niet gegrammaticaliseerd is, gebruiken een vergelijkbare
constructie die wel de ‘statieve voltooide tijd’ genoemd wordt. De distributie daarvan is
nogal beperkt: het komt alleen voor met levende onderwerpen en met transitieve
voltooide deelwoorden die congrueren met het lijdend voorwerp. Op basis van de
306 Samenvatting in het Nederlands
ontwikkeling van verwante structuren in de Germaanse en Romaanse talen concludeer

ik dat de statieve voltooide tijd de onderliggende, niet-gegrammaticaliseerde variant van
de voltooide tijd met ‘hebben’ is. Deze twee constructies vergelijk ik met constructies
met een onpersoonlijk participium die men aantreft in het Pools, het Witrussisch, het
Oekraïens en sommige Russische dialecten. Deze vorm representeert het meest
gegrammaticaliseerde type van het passieve participium.
In hoofdstuk 4 houd ik me bezig met clitica in het Zuid-Slavisch. Hun distributie is
om twee redenen belangrijk voor de structuur van de samengestelde tijden. In de eerste
plaats hebben de Slavische talen een nogal vrije woordvolgorde, maar is de plaatsing
van de clitica bijzonder rigide. Hierdoor kan de plaatsing van de clitica gebruikt worden
als indicator voor syntactische verplaatsing. In de tweede plaats zijn de vormen van het
hulpwerkwoord ‘zijn’ in de tegenwoordige tijd zelf clitica.
De Zuid-Slavische talen vertonen aanzienlijke verschillen in hun cliticsystemen.
Het Servo-Kroatisch is een van de weinige talen in de wereld die nog Wackernagel-
clitica hebben. Deze taal vertoont dus nog het patroon van de vroege Indo-Europese
dialecten, dat uit de meeste talen verdwenen is. Clitica in het Bulgaars moeten voldoen
aan uiteenlopende syntactische en fonologische eisen: fonologisch zijn ze enclitisch en
vereisen dus een morfologisch gerealiseerde gastheer aan hun linkerzijde: syntactische
gezien, daarentegen, zijn ze proclitisch, omdat ze bij voorkeur in preverbale positie
staan. Macedonische clitica tonen een zeer ingewikkelde distributie, die afhankelijk is
van de categoriale status van hun gastheer. In mijn analyse stel ik dat de clitica in het
Macedonisch geen homogene categorie vormen, maar een mengeling zijn van echte
clitica en zwakke voornaamwoorden. De eerste treden op met werkwoorden die
naamval toekennen, de laatste in constructies met niet-verbale predicatieve XP’s zoals
AP’s, DP’s, en passieve participia.
De status van het hulpwerkwoord in de derde persoon enkelvoud bepaalt hoe het
l-participium vooropgeplaatst wordt. De overte aanwezigheid van het hulpwerkwoord
je/e, dat altijd de laatste positie in het clitic cluster inneemt, is noodzakelijk voor XP-
verplaatsing. In de talen die het hulpwerkwoord in de derde persoon enkelvoud
verloren hebben, zoals het Macedonisch, het Tsjechisch en het Pools, verlaat het
participium de VP als hoofd.
Het laatste gedeelte van het proefschrift geeft een analyse van het Pools. In
tegenstelling tot de Zuid-Slavische talen heeft deze taal de structuur van de
samengestelde tijden niet vereenvoudigd door een semantische herinterpretatie, maar
door een reductie van het aantal tijden en hun morfologische opmaak. Deze
morfologische vereenvoudiging bestaat uit de heranalyse van de enclitische
hulpwerkwoorden als een verbuigingsuitgang op het l-participium. Dit reductieproces is
begonnen in de 16e eeuw en voltrekt zich zeer langzaam. De reductie van het
hulpwerkwoord vond tegelijkertijd plaats met een verandering in de plaatsing van het
hulpwerkwoord, maar ik beargumenteer dat we hier met twee onafhankelijke
verschijnselen te maken hebben. In de oudere stadia van het Pools werd het
hulpwerkwoord onveranderlijk in de Wackernagelpositie geplaatst. In het Moderne
Pools, daarentegen, hecht het hulpwerkwoord zich in de meeste gevallen aan het
l-participium, hoewel het zich ook kan hechten aan gefocusde of getopicaliseerde
constituenten op de eerste zinspositie. Tenslotte laat ik zien dat, in tegenstelling tot de
clitische hulpwerkwoorden, de plaatsing van pronominale clitica in het Pools steeds
Samenvatting in het Nederlands 307
vrijer wordt; in tegenstelling tot de Zuid-Slavische pronominale clitica hoeven zij niet
langer te clusteren of op vaste posities in de zin te verschijnen.

130 Fulltext

Uploaded by

Copyright:

Available Formats

130 Fulltext

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

130 Fulltext

Uploaded by

Copyright:

Available Formats

The Syntax of Compound Tenses in Slavic

Copyright © 2006: Krzysztof Migdalski. All rights reserved.

ter verkrijging van de graad van doctor

Krzysztof Marek Migdalski

geboren op 4 maart 1975 te Bielsko-Biała, Polen

CHAPTER 1 THE DIACHRONY OF COMPOUND TENSES IN SLAVIC7

1.2 The division of the Slavic languages ..........................................................7

1.3 The Tense and Aspect system ...................................................................11

1.4 Conclusions ............................................................................................... 59

CHAPTER 2 THE SYNTAX OF 'BE'-PERFECTS AND THE L-

2.2 Head-movement accounts of participle fronting in Bulgarian and Serbo-

2.3 Towards an alternative analysis ................................................................ 67

2.4 Conclusions ..............................................................................................123

CHAPTER 3 THE SYNTAX OF ‘HAVE’-PERFECTS AND PASSIVE

3.2 A typology of passive participles..............................................................125

3.3 Generalizations concerning ‘have’-perfects.............................................128

3.4 Properties of ‘have’-perfects in Slavic ......................................................129

3.5 Impersonal participles in Polish ..............................................................142

3.6 Grammaticalization of ‘have’-perfects .....................................................147

3.7 Conclusions ..............................................................................................160

CHAPTER 4 CLITICS IN SOUTH SLAVIC ................................................163

4.2 What are clitics?........................................................................................163

4.3 Diachrony of clitics in Slavic....................................................................164

4.4 Clitic positions in contemporary South Slavic languages .......................170

CHAPTER 5 THE STATUS OF CLITICS AND COMPOUND TENSES IN

5.1 Introduction............................................................................................. 223

5.2 The pronominal forms ............................................................................ 223

5.4 Conclusions ............................................................................................. 285

SUMMARY ......................................................................................................... 287

REFERENCES .................................................................................................. 289

SAMENVATTING IN HET NEDERLANDS................................................. 305

Moreover, I thank Ad Backus, Bożena Rozwadowska, Maaike Schoorlemmer, Craig

I received from Bożena Cetnarowska, Steven Franks, Gilbert C. Rappaport, Giuliana

Q the question/focus particle li

The analysis is couched in the framework of generative linguistics; in particular it

(1) If I’d known you’re coming, I’d have baked a cake

participle that is used in compound tenses as an “active” past participle appears in

2. The compound tenses in Slavic

(2) a. Ivan e čel knigata

(3) a. Jan będzie pisał list

b. Kad budemo govorili s Marijom…

(4) a. Ivan e pročel knigata

(5) a. Mъnoзi že ot ijudei běaxõ prišьlo kъ Martě

Diachronically, the l-participle is a Slavic innovation. It derives from a class of Proto-

(6) a. Ja imam skinato mojata kosula

(7) Mam już wszystkie ciasta upieczone

morphological reduction of certain compound tense structures. It will be demonstrated

3. Organization of the thesis

1.2 The division of the Slavic languages

1.2.1 The common ancestor

the attainment of statehood by Bulgaria, Carantania, Croatia, Serbia, Bohemia, Moravia,

1.2.2 Sources of the division

(2) a. Čel sŭm knigata

(3) a. Czytał-em książkę

(4) Ja čital knigu

1.3 The Tense and Aspect system

1.3.1 The Tense System in Proto-Indo-

(5) I was in Longyearbyen last Monday

(6) a. Thatcher treats her Cabinet colleagues like children