University of Cambridge  
Faculty of Modern and Medieval Languages 
Department of Theoretical and Applied Linguistics 
  
 
 
The L1 Acquisition of Clitic Placement in 
Cypriot Greek 
 
 
 
Theoni Neokleous  
Lucy Cavendish College  
 
 
 
Dissertation submitted for the degree of Doctor of Philosophy 
 
Cambridge 
May 2014 
 
!!
""!
DECLARATION 
This dissertation is the result of my own work and includes nothing which is the 
outcome of work done in collaboration except where specifically indicated in the 
text. No part of this dissertation has been submitted for any other qualification at 
any University. This dissertation does not exceed the regulation length, including 
footnotes, references and appendices but excluding the bibliography. 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
!!
"""!
“From the point of view that I adopt here, the fundamental empirical problem of 
linguistics is to explain how a person can acquire knowledge of language” 
Chomsky, N. (1973, “Conditions on Transformations”) 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
!!
"#!
ABSTRACT  
 
This thesis investigates the first language acquisition (L1A) of pronominal object 
clitics in Cypriot Greek (CG) by typically developing (TD) children, focusing on 
an exceptional form of non–adult–like clitic placement attested in early data. The 
aim of the present study is twofold. On the one hand, it aims to sketch the 
developmental stages in the course of L1A of CG in relation to other clitic 
languages. On the other, it investigates whether and to what extent syntactic 
(Agouraki 2001, Terzi 1999a, 1999b), prosodic (Condoravdi and Kiparsky 2001) 
and interface approaches (Mavrogiorgos 2012, Revithiadou 2006) can account for 
early clitic production. 
Research on L1A of clitic pronouns has demonstrated both clitic 
realisation and omission in child languages. However, no instances of clitic 
misplacement have been reported for early European languages, with the 
interesting exceptions of CG (Petinou & Terzi 2002) and European Portuguese 
(Lobo & Costa 2012). 
The present thesis examines the L1A of CG in the age range 2!4 on the 
basis of spontaneous and experimental data, cross!sectional as well as 
longitudinal, with a focus on clitic placement. Spontaneous speech data were 
collected from 8 children, and one of the children was also followed 
longitudinally for a period of 6 months. An elicited production task performed by 
50 children was used to generate 3rd person singular accusative object clitics. The 
results of the study indicate that, for CG: (i) clitic placement in enclisis 
environments is adult–like from the onset in structures involving single clitics and 
clitic clusters, as well as in Clitic Doubling and Clitic Left Dislocation; (ii) clitic 
misplacement is attested in proclisis contexts in a subset of children aged 2;6 to 
3;0; (iii) clitic misplacement does not correlate with early non–finite forms; (iv) 
occasional realisation of two copies of the clitic is attested in some children aged 
2;6 to 3;0; (v) by age 3;6, TD children manifest adult–like clitic placement. 
These findings raise issues regarding the acquisition of clitics in different 
classes of languages (Tobler–Mussafia, finiteness–sensitive languages, languages 
exhibiting second position restrictions), as well as the role of syntax, prosody and 
the syntax–phonology interface in clitic L1A. The current study suggests that only 
Tobler–Mussafia languages display clitic misplacement, as attested in the L1A of 
CG. Clitic misplacement in CG is interpreted within an interface account in line 
with Revithiadou (2006) and, following the spirit of Bo"kovi# (2000), it is 
assumed that the placement requirement imposed on CG clitics “can be captured 
in its entirety through a filtering effect of the phonology on the syntax” 
(2000:105). Clitic placement in CG is an interface phenomenon: the syntax 
provides two copies of clitic pronouns (Franks 1998) and the syntactic outcome is 
filtered through a phonology–controlled procedure. 
  
 
 
 
 
 
!!
#!
ACKNOWLEDGEMENTS 
 
This thesis could not have been completed without the contribution of different 
people whom I would like to thank. 
Firstly, I would like to thank my supervisor Teresa Parodi for her help, her 
guidance, her constant support and valuable feedback throughout my PhD 
research and, especially, in the final stages. I also want to thank my advisor 
Napoleon Katsos for his help and his insightful comments at crucial stages of this 
research. This thesis was greatly benefited by discussions with Ianthi Tsimpli, 
Anthi Revithiadou, Vassilios Spyropoulos, Arhonto Terzi, Dora Alexopoulou, 
and, especially, Marios Mavrogiorgos, to whom I owe a special thank you for 
willingly discussing with me various aspects of his analysis. 
I also owe a special thank you to Dimitris Michelioudakis for long 
discussions on syntactic aspects of my work throughout my doctoral research, for 
his enthusiasm, constructive criticism and valuable feedback, as well as for 
reviewing chapters 1 and 3.  
 Earlier versions of this work were presented at GALA, ISB8, MGDLT5, 
ISTAL20, 10th ICGL, CamLING, LoT and LangUE, as well as in colloquia and 
workshops in Cambridge and Nicosia; I thank all the audiences for their 
comments and suggestions. I also thank Chris Cummins for proof–reading this 
thesis. 
This project could not have been completed without the group of adorable 
toddlers that participated in my study and their families. I want to thank all and 
each one of them, especially $%&'() and little A., *+µ,-'. and little /. Special 
thanks go to 0&1-. and 2345. and their little angel *. 
I thank Cyprus Research Promotion Foundation for funding this project 
(67870/0609/42) and Leventis Foundation for supporting me financially, as 
well as the Research Centre for English and Applied Linguistics and Lucy 
Cavendish College for their financial assistance. Special thanks go to Kleanthes 
K. Grohmann, my project coordinator, for his help throughout our collaboration 
and his initiative in various academic–related aspects of our project. 
Finally, I would like to thank my linguist friends Norbert Vanek and Helen 
Engemann. And, the best housemates ever: Eleni Lampaki, Rana Bilbeisi, Eleni 
Kapogianni, Lidia Napiorkowska and Enass Abo–Hamed for their help when it 
was most needed and for making Cambridge feel like a home away from home. I 
also thank 2µ%3,, 95:'%.5+, ;3<,1-,, =.'>. and =?'%) for making me feel that 
I never left home. 
A big THANK YOU goes to my parents, 0&1-. and 233,, and to /-.@') 
for their love, support and belief in me all these years. 
!!
1 
TABLE OF CONTENTS 
DECLARATION……...……………………………………………………………..ii 
ABSTRACT………………………………………………………………………...iv 
ACKNOWLEDGEMENTS…………………………………………………………....v 
TABLE OF CONTENTS..……………………………………………………………1 
LIST OF TABLES…...……………………………………………………………...5 
LIST OF FIGURES………………………………………………………………...8 
INTRODUCTION………..……………………………………………...………..9 
CHAPTER 1: THE CYPRIOT GREEK CP 
1.1. Introduction….…………………………..…….…………………………12 
1.2. The articulated C–structure in CG..…………………..……………….....13 
1.3. Topic and COp positions…………………………………………………21 
1.4. Wh–question formation……...…………………………………………...26 
1.5. Clefts…………………………………………………………………......30 
1.6. Summary…………………………………………………………..……..34 
CHAPTER 2: CYPRIOT GREEK CLITIC PRONOUNS 
2.1. Introduction……...…..…………..…………..……………….…………..35 
2.2. Basic properties of CG clitics.…...............................................................35 
2.2.1. The morphological paradigm….…...……….….…….……..…....36 
2.2.2. Distribution…...………………….………………………………43 
2.2.3. Coordination and Modification……...……...……………………46 
2.2.4. Semantic properties….…………………………………………...46 
2.2.5. Phonological properties….....……………………….…….….......50 
2.2.6. Phrase structure status……….………………………….………..54 
2.3. Clitic placement in CG…………………………………………………...56 
2.3.1. Enclisis and proclisis contexts....…...……………………….…...57 
2.3.1.1. Modal particles..……………...………………………….60 
2.3.1.2. Wh–elements……...……………………………………..60 
2.3.1.3. Negative particles.…………………………………….....61 
2.3.1.4. The factive complementizer pu and complementizers 
an/otan/ama/afu………………………………………………………….62 
2.3.1.5. Stressed constituents in the left periphery of the CG CP..63 
!!
2 
2.3.2. Free ordering of pronominal clitics in CG………...…………..…63 
2.3.3. Clitic clusters in CG..……..…………………..………………….65 
2.3.4. CG clitics and second position restrictions……………….....…...65 
2.3.5. CG, European Portuguese and Galician.…………………………67 
2.4. Summary………………........…………...…….…………………………70 
CHAPTER 3: ACCOUNTS OF CLITIC PLACEMENT IN CYPRIOT 
GREEK 
3.1. Introduction...………………….…………..……………………………..71 
3.2. Syntactic accounts..…….………….……………………………………..72 
3.2.1. The V–to–M analyses for finite enclisis…………………..…..…72 
3.2.2. The filled–C requirement …………………………………..........78 
3.3. Prosodic inversion account………..…...………………………………...80 
3.4. Interface accounts……….………………………..……………………...83 
3.4.1. The syntax–PF interface analysis for Tobler–Mussafia 
languages…………………………………………………………………83 
3.4.2. The PF–controlled spell–out of copies…………………………...84 
3.5. A dynamic account……………………………………………………….93 
3.6. Interim conclusions for clitic placement in CG………………………….94 
3.6.1. Verb adjacency and structural and/or prosodic dependency…..…94 
3.6.2. The ban from clause initial position………………..…...…..……94 
3.6.3. CG clitics are not inherently enclitics…..………………………..95 
3.6.4. The role of syntax…………………………………………..........95 
3.6.5. Enclisis–proclisis alternation and finiteness……...………...........96 
3.6.6. Enclisis–proclisis alternation: neither pure syntax nor pure PF….97 
3.7. Summary………………………...…………………………………….…98 
CHAPTER 4: THE L1 ACQUISITION OF CLITICS 
4.1. Introduction.…………………………………………………....………...99 
4.2. Finiteness–sensitive languages………………………………………....100 
4.2.1. Italian…………………………………………………………...100 
4.2.2. Romanian……………………………………………………….104 
4.2.3. Spanish/Catalan………………………………………………....107 
4.2.4. Standard Modern Greek………………………………………...109 
!!
3 
4.3. Second position clitic languages………………………………………..113 
4.4. Clitic scrambling languages…..……………………………………...…115 
4.5. Tobler–Mussafia languages.……………………………………………116 
4.5.1. European Portuguese.…………………………………………...116 
4.5.2. Galician……………………………………………………...….122 
4.5.3. Cypriot Greek…………………………………………………...124 
4.6. Interim conclusions..……………………………………………............131 
CHAPTER 5: METHODOLOGY 
5.1. Introduction.………………………………………………..……...........133 
5.2. Hypotheses of the present study….……………………………………..133 
5.3. Spontaneous data……….………………….…………...……………….137 
5.3.1. Participants……………………………………………………...137 
5.3.2. Method………………………………………………………….138 
5.3.3. MLT analysis….………………………………………………..141 
5.3.4. Data analysis……………………………………………………143 
5.3.5. Clitic misplacement and finiteness……………………………..144 
5.4. Elicited production experiment………...……………………………….148  
5.4.1. Participants……………………………………………………...148 
5.4.2. Method.…………………………………………………………149 
5.4.3. Double realisation of the clitic pronoun………………………...153 
CHAPTER 6: RESULTS 
6.1. Introduction.………………………………………………..………...…155 
6.2. Spontaneous data.………………………….…………...……………….155 
6.2.1. Clitic production and clitic placement……….…………………156 
6.2.2. Clitic ordering within clitic clusters…………………………….172 
6.2.3. Production of Clitic Doubling and Clitic Left 
Dislocation……………………………………………………………...173 
6.2.4. Clitic misplacement and finiteness………….………………….176 
6.3. Elicited production experiment……….………………………………...180 
6.3.1. Clitic production..………………………………………………180 
6.3.2. Clitic placement………………………………………………...183 
6.3.3. Between–group analysis for clitic placement………...………...185 
!!
4 
6.4. Double realization of the clitic pronoun…..…….…………...…………185 
6.5. Summary………………………………………………………………..187 
CHAPTER 7: THE L1 ACQUISITION OF CLITIC PLACEMENT IN 
CYPRIOT GREEK  
7.1. Introduction…………………………………………………………..…188 
7.2. Recapitulation: the L1 acquisition of clitic placement in CG…………..188 
7.3. Clitic misplacement in CG ……………………………………………..191 
7.4. Clitic misplacement in early grammars ………………………………..193 
7.5. Double realization of the clitic pronoun in early grammars…………....199 
7.6. Developmental patterns and formal accounts of clitic placement in 
CG………………………………………………………………………………200 
7.7. Developmental patterns and the PF!controlled spell–out of copies…....202 
CONCLUDING REMARKS…………………………………………………215 
LIMITATIONS OF THE CURRENT STUDY – IMPLICATIONS FOR 
FUTURE RESEARCH……………………………………………………….217 
REFERENCES……………………………………………….……………………....218 
APPENDICES………………………………………………………………..232 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
!!
5 
LIST OF TABLES 
Table 2.1: The morphological paradigm of CG clitic pronouns…………………36 
Table 2.2: The morphological paradigm of CG strong pronouns………………..37 
Table 2.3: The morphological paradigm of definite articles in CG…………...…41 
Table 2.4: Proclisis contexts in CG, EP and Galician…………………………....69 
Table 4.1: Production of pre–verbal clitic pronouns in early Italian (Guasti 
1993/94)………………………………………………………………………...101 
Table 4.2: Overall proportion of overt and omitted direct object clitics and 
production of full NPs (in obligatory clitic contexts) (Schaeffer 2000)………..103 
Table 4.3: Clitic production in early SMG: the Stephany Corpus (Marinis 2000, 
Tsakali 2006)…………………………………………………………………....110 
Table 4.4: Clitic production in early SMG: the Christofidou Corpus (Marinis 
2000) and the Doukas Corpus (Tsakali 2006)………………………………….111 
Tables 4.5: Clitic placement in child Serbo/Croatian (Naturalistic data) (Ilic & Ud 
Deen 2004)……………………………………………………………………...114 
Tables 4.6: Clitic placement in child Serbo/Croatian (Experimental data) (Ilic & 
Ud Deen 2004)………………………………………………………………….114 
Table 4.7: Proportion of full objects, demonstrative pronouns, strong personal 
pronouns and clitics in Dutch clitic scenarios (Schaeffer 2000)………………..115 
Table 4.8: Proportion of clitic misplacement in children with SLI (Petinou & 
Terzi 2002)…………………………………………………………………...…125 
Table 4.9: Proportion of clitic misplacement in TD children (Petinou & Terzi 
2002)…………………………………………………………………………....125 
Table 4.10: Proportion of clitic misplacement in 3 younger TD children (Petinou 
& Terzi 2002).......................................................................................................126 
Table 4.11: Clitic production in TD children aged 3 to 6 for clitics–in–islands test 
(Grohmann 2011)……………………………………………………………….129 
Table 4.12: Clitic placement in TD children aged 3 to 6 for clitics–in–islands test 
(Grohmann 2011)……………………………………………………………….129 
Table 4.13: Clitic production and clitic placement for all age groups for clitics–in–
islands test (Grohmann et al. 2012)…………………………………………….129 
Table 5.1: Responses elicited in the spontaneous speech study………………..141 
!!
6 
Table 5.2: Responses elicited in the longitudinal study………………………...141 
Table 5.3: Participants: elicited production experiment………..….…………...148 
Table 5.4: The morphological paradigm of CG clitics…………………………151 
Table 5.5: Subset of participants: elicited production experiment……..……….154 
Table 6.1: Clitic production and clitic/object omission (Spontaneous data)…...158 
Table 6.2: Production of single clitics and clitic clusters in enclisis and proclisis 
contexts (Spontaneous data)…………………………………………………….161 
Table 6.3: Clitic production in enclisis contexts: imperatives and root clauses 
(Spontaneous data)……………………………………………………………...162 
Table 6.4: Clitic production in proclisis contexts: clauses headed by modality 
markers and other proclisis–triggering elements, negatives and wh–questions 
(Spontaneous data)……………………………………………………………...163 
Table 6.5: Correct and incorrect clitic placement in enclisis and proclisis contexts 
(Spontaneous data)……………………………………………………………...164 
Table 6.6: Correct and incorrect clitic placement in enclisis contexts: imperatives 
and root clauses (Spontaneous data)……………………………………………165 
Table 6.7: Correct and incorrect clitic placement in proclisis contexts: negatives 
and clauses headed by modality markers (Spontaneous data)………………….166 
Table 6.8: Correct and incorrect clitic placement in proclisis contexts: wh–
questions and other proclisis–triggering elements (Spontaneous Data)………..166 
Table 6.9: Clitic production and clitic/object omission (Longitudinal data)…...168 
Table 6.10: Clitic production (Longitudinal data)……………………………...169 
Table 6.11: Clitic production in enclisis and proclisis contexts (Longitudinal 
data)……………………………………………………………………………..170 
Table 6.12: Production of Single Clitic (SC), Clitic Doubling (CD) and Clitic Left 
Dislocation (CLLD) (Spontaneous data)……………………………………….174 
Table 6.13: Production of strong pronouns and DPs in Clitic Doubling (CD) and 
Clitic Left Dislocation (CLLD) (Spontaneous data)……………………………174 
Table 6.14: Production of Clitic Doubling (CD) per syntactic context 
(Spontaneous data)……………………………………………………………...174 
Table 6.15: Production of Clitic Left Dislocation (CLLD) per syntactic context 
(Spontaneous data)……………………………………………………………...175 
!!
7 
Table 6.16: Production of Single Clitic (SC), Clitic Doubling (CD) and Clitic Left 
Dislocation (CLLD) (Longitudinal data)……………………………………….175 
Table 6.17: Production of strong pronouns and DPs in Clitic Doubling (CD) and 
Clitic Left Dislocation (CLLD) (Longitudinal data)……………………………175 
Table 6.18: Production of Clitic Doubling (CD) per syntactic context 
(Longitudinal data)……………………………………………………………...176 
Table 6.19: Production of Clitic Left Dislocation (CLLD) per syntactic context 
(Longitudinal data)……………………………………………………………...176 
Table 6.20: Use of finite and early non–finite forms in constructions involving 
misplaced clitics………………………………………………………………...178 
Table 6.21: Finite forms in constructions involving misplaced clitics…………179 
Table 6.22: The distribution of the non–3SG verb forms (Varlokosta et al. 
1998)……………………………........................................................................179 
Table 6.23: Production of target responses, other answers and no answers per age 
group per experimental condition (Elicited production experiment).…………..181 
Table 6.24: Clitic production and clitic omission per age group per syntactic 
context (Elicited production experiment)……………………………………....183 
Table 6.25: Clitic placement per experimental condition (Elicited production 
experiment)……………………………………………………………………..184 
Table 7.1: Pre– and post–verbal ADOCs in Romanian (Antonio corpus) (Avram 
& Coene 2007)………………………………………………………………….204 
Table 7.2: Pre– and post–verbal ADOCs in Romanian (Bianca corpus) (Avram & 
Coene 2007)…………………………………………………………………….204 
Table 7.3: Production and omission of single enclitics and single proclitics 
(Tzakosta 2003)…………………………………………………………………206 
Table 7.4: Mean clitic production per group in the present and past condition 
(Mastropavlou et al. 2014)…………………………………………………...…207 
 
 
 
 
 
 
!!
8 
LIST OF FIGURES 
Figure 6.1: Clitic placement in proclisis contexts (Spontaneous data)…………165 
Figure 6.2: Proportion of children exhibiting adult–like and non–adult–like clitic 
placement in clauses headed by modality markers per age group (Elicited 
production experiment)…………………………………………..……………..185 
 
 
 
 
 
 
 
 
 9 
INTRODUCTION 
 
Cardinaletti and Starke describe clitic pronouns as severely deficient pronominal 
elements (1999:168). This deficiency has, however, proven inversely proportional 
to the attention they have received in the literature. The past two decades have 
seen a proliferation of scholarly work on cliticisation from a theoretical as well as 
from an acquisition perspective.  
The present thesis investigates the first language acquisition (L1A) of 
pronominal object clitics in Cypriot Greek (CG) by typically developing (TD) 
children, focusing on an exceptional form of non–adult–like clitic placement 
attested in early data. The aim of the present study is twofold. On the one hand, it 
aims to sketch the developmental stages in the course of L1A of Cypriot Greek in 
relation to other clitic languages. On the other, it investigates whether and to what 
extent syntactic (Agouraki 2001, Terzi 1999a, 1999b), prosodic (Condoravdi and 
Kiparsky 2001) and interface approaches (Mavrogiorgos 2012, Revithiadou 2006) 
can account for early clitic production. 
Research on L1A of clitic pronouns has demonstrated both clitic 
realisation and omission in child languages. However, no instances of clitic 
misplacement have been reported for early European languages, with the 
interesting exceptions of Cypriot Greek (Petinou & Terzi 2002) and European 
Portuguese (Lobo & Costa 2012). Following the tri–partition of clitic languages 
proposed in Mavrogiorgos (2012), most Romance languages, including Catalan, 
Italian, Spanish and Romanian, as well as Standard Modern Greek, are 
categorised as finiteness–sensitive; languages such as Serbo–Croatian adhere to 
second position restrictions; while Cypriot Greek, Galician and European 
Portuguese pattern alike in obeying the Tobler–Mussafia law. 
The present thesis aims at showing that a correlation holds between clitic 
misplacement in child language and the Tobler–Mussafia properties of the 
language being acquired. This is implemented by systematically examining the 
L1A of Cypriot Greek (CG) in the age range 2!4 on the basis of spontaneous and 
experimental data, cross!sectional as well as longitudinal, with a focus on clitic 
 10 
placement. Spontaneous speech data were collected from 8 children, and one of 
the children was also followed longitudinally for a period of 6 months. An elicited 
production task performed by 50 children was used to generate 3rd person 
singular object clitics. The results of the study indicate that, for CG: 
(i) clitic placement in enclisis environments is adult–like from the 
onset in structures involving single clitics and clitic clusters, as well as in Clitic 
Doubling and Clitic Left Dislocation; 
(ii) clitic misplacement is attested in proclisis contexts in a subset of 
children aged 2;6 to 3;0; 
(iii) clitic misplacement does not correlate with early non–finite forms; 
(iv) occasional realisation of two copies of the clitic is attested in some 
children aged 2;6 to 3;0; 
(v)  by age 3;6, typically developing children manifest adult–like clitic 
placement. 
These findings raise issues regarding the acquisition of clitics in different 
classes of languages (Tobler–Mussafia, finiteness–sensitive languages, languages 
exhibiting second position restrictions), as well as the role of syntax, prosody and 
the syntax–phonology interface in clitic L1A. The current study suggests that only 
Tobler–Mussafia languages display clitic misplacement, as attested in the L1A of 
CG. 
Clitic misplacement in CG is interpreted within an interface account in 
line with Revithiadou (2006) and, following the spirit of Bo!kovi" (2000), it is 
assumed that the placement requirement imposed on CG clitics “can be captured 
in its entirety through a filtering effect of the phonology on the syntax” 
(2000:105). Clitic placement in CG is an interface phenomenon: the syntax 
provides two copies of clitic pronouns (Franks 1998) and the syntactic outcome is 
filtered through a phonology–controlled procedure. 
The thesis is organised as follows. Chapter 1 provides an overview of the 
language examined and lays out the basic assumptions with regard to the CG 
clause structure. Chapter 2 focuses on the morpho–phonological properties, as 
well as the distribution of object clitic pronouns in CG, and chapter 3 reviews the 
syntactic, prosodic and interface accounts that have been put forward for their 
 11 
placement. Chapter 4 reviews a number of studies conducted on the L1 
acquisition of clitics cross–linguistically. Chapter 5 presents the methodology 
adopted in the current study and chapter 6 presents the results obtained. The thesis 
concludes with chapter 7, which discusses the theoretical implications of the 
findings of the present study for the developmental patterns attested in the course 
of clitic L1A, and evaluates the formal accounts proposed for cliticisation in CG. 
! 12 
CHAPTER 1: THE CYPRIOT GREEK CP 
 
1.1. Introduction 
 
Cypriot Greek (henceforth CG) is a variety of Modern Greek spoken by 
approximately 700,000 people1 residing in the Republic of Cyprus2. I deliberately 
use the term variety instead of dialect or language. Whether CG constitutes a 
dialect of Modern Greek or a separate language is a highly debatable issue. 
Arvaniti (2010, and a host of references therein) offers a thorough presentation of 
the socio–linguistic situation in Cyprus and a review of the scholarly work 
accomplished on various aspects of it, including the status of CG as compared to 
Standard Modern Greek, the development of a Cypriot Koiné and of dialectal 
levelling on the island, the manifestation of code–switching and/or code–mixing, 
the speakers’ awareness of the linguistic varieties spoken on the island, the 
language attitudes amongst Cypriot speakers, and the relationship between 
education and language. This thesis examines the acquisition of CG, the variety 
that young Greek Cypriot children acquire effortlessly and in a native–like way, 
whatever its status, and hence leaves aside sociolinguistic aspects of the linguistic 
landscape in Cyprus. 
The Greek–speaking population of the island uses CG in everyday oral 
communication and Standard Modern Greek (henceforth SMG)3, one of the 
official languages of the Republic of Cyprus (the other two being Turkish and 
English), in written texts, as well as in formal speech acts. From a diachronic !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
1 According to the most recent census carried out by the Statistical Service of Cyprus, in 2011 
667,398 Cypriot citizens were residing on the island (source: 
http://www.mof.gov.cy/mof/cystat/statistics.nsf/populationcondition_22main_en/populationcondit
ion_22main_en?OpenForm&sub=2&sel=2, Accessed in September 2012, File: POPULATION 
ENUMERATED BY CITIZENSHIP, SEX AND POSTAL CODE, 2011), the vast majority of 
whom are native speakers of CG. Apart from the residents of the island, the dialect is spoken by 
some tens of thousands of Greek Cypriot immigrants who reside with their families in foreign 
countries, including Greece, the United Kingdom, South Africa and Australia. 
2 The island’s population is comprised of Greek Cypriots and Turkish Cypriots; there are also 
other ethnic and religious groups, including Armenians and Maronites.  
3 Cf. Arvaniti (2010) for a description of what she calls Cypriot Standard Greek. Arvaniti (2010) 
argues that SMG as used in Cyprus has been increasingly diverging from the Standard variety as 
spoken in Greece and the two are recognizably different. !
! 13 
perspective, CG evolved from the Koiné standard of the Hellenistic period. The 
purpose of the current chapter is to offer a brief description of the Cypriot variety 
from a synchronic perspective, as well as to highlight the points of convergence 
and divergence from the Standard variety. This chapter is an overview of the main 
characteristics of the morpho–syntax of CG. This overview will constitute the 
background for the specific topic of the current thesis, namely the acquisition of 
object clitic placement by young Greek Cypriot children. The various properties 
of clitic constructions in CG can only be discussed within a clearly defined 
framework regarding the syntax of the variety. Therefore, I first formulate the 
clause structure of CG in the current chapter before then proceeding, in the 
following chapter, to a discussion of the structures under investigation. Moreover, 
it is crucial to highlight the similarities and differences of the CG clause as 
compared to the SMG clause that allow for a direct comparison of acquisition data 
from CG– and SMG–speaking children. 
The first section (1.2) discusses the left periphery of the CG clause, 
namely the functional heads projected above IP, and, in particular, their relative 
order and feature specification. Roussou’s (2000) proposal for an articulated CP 
in SMG is adopted and adapted for CG. Hence, a preliminary formulation of the 
CG CP is offered in the first section, which will then be refined in the subsequent 
sections. The rest of the chapter is devoted to the following topics, addressed in 
separate sections: the Topic and COp(erator) heads (1.3), wh–question formation 
(1.4) and the syntax of it–clefts (1.5).  
 
1.2. The articulated C–structure in CG 
 
The phenomena observed in child language with regard to clitic placement are 
directly related to inflection, as well as to the functional projections above IP, and, 
in particular, the heads where negative particles, modal particles and 
complementizers are realised. It is therefore clear that the structure of the CP 
needs particular attention. 
I adopt Roussou’s (2000) proposal for the SMG left periphery, as outlined 
in (1). Roussou (2000) suggests that the SMG clause structure involves three C 
! 14 
heads, each carrying different features: the lowest one (CM) is specified for 
modality, the middle one (COp) has a clause–typing feature and the highest one 
(C) has a feature for subordination.  
 
(1) [C pu [Topic/Focus [COp oti/an/na/as [Neg dhen/min [CM tha/na/as [I cl + 
V...]]]]]]                                                                           (Roussou 2000:79) 
 
CG behaves similarly to SMG in many respects. Similarities in the 
properties and distribution of different modal particles and negators, in the use of 
topicalised phrases, and in wh–question formation are the points of convergence 
between the two varieties. Building on these common properties and the similar 
distribution of modal and negative particles in CG and SMG, I will attempt to 
offer a first formulation of the CG CP. 
CG, like SMG, makes use of the particles as, na and tha. The particle na 
has been traditionally analysed as the subjunctive marker (Veloudis & Philippaki–
Warburton 1983), while it has also been proposed that na is a complementizer 
(Agouraki 1991, Tsoulas 1993). The particle tha has been traditionally treated as 
the future particle, while as is a hortative particle found in root contexts alone 
(Roussou 2000:65). Revithiadou and Spyropoulos (2008) indicate that major 
grammaticalization processes that affected the morpho–syntax of the verb group 
in Greek (see Horrocks 1997, Philippaki–Warburton & Spyropoulos 2004, among 
others) have created the particles as, na and tha as follows: afes “let” ! as, ina 
(COMP) ! na, and thelo ina “want to” ! tha. 
As and na are only compatible with the negator min in SMG and its CG 
equivalent men4, and they obligatorily precede it. As shown in examples (2) and 
(3), in CG negative clauses, na and as immediately precede men, patterning like 
their SMG counterparts.  
 
(2) Na  (men (CG) / min (SMG))  to                 feri.  
M   NEG                                  it–CL.ACC  bring–3S 
“S/he should (not) bring it” !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
4 The resemblance of the two forms is obvious and their properties and distribution do not differ.!
! 15 
(3) As  (men (CG) / min (SMG))  to                 feri.  
M   NEG                                  it–CL.ACC  bring–3S 
“Let him/her (not) bring it” 
 
The particle tha is used in both CG and SMG for future tense formation. 
This is the reason why it has been traditionally analysed as the future particle. Yet 
while tha in SMG is used in both affirmative and negative future clauses, in CG it 
is only used in negated future clauses ((4) vs. (5)). The particle used in CG 
affirmative future clauses is enna5 (4). Apart from the difference in the specific 
form of the particle used, CG resembles SMG in all other respects of future tense 
formation. To be precise, the CG particles enna in affirmatives and tha in 
negatives combine with the verb that bears future morphology and give rise to a 
periphrastic future tense. 
 
(4) Enna (CG) / Tha (SMG)  to                  feri.  
M                                      it–CL.ACC  bring–3S 
“S/he will bring it” 
(5) (Dh)e(n)  tha (CG & SMG)  to                 feri.  
NEG        M                          it–CL.ACC  bring–3S 
“S/he will not bring it” 
 
As shown in (5) the particle tha in negated future clauses, in both CG and 
SMG, is only compatible with the negator (dh)e(n). There are, however, two 
points of divergence in the formation of negated future clauses in the two 
varieties. Firstly, while in SMG two particles are used, with the negative particle 
dhe(n) immediately preceding the future particle tha, in CG, negated future 
clauses are usually headed by the particle etha alone. The latter constitutes an 
amalgam form which combines (dh)e(n) and tha. However, when the CG clause is 
headed by the complementizer an, negated future clauses are formed like their 
counterpart SMG structures with the realisation of both (dh)e(n) and tha (6). With !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
5 Ralli (2006:136) observes that tha originates from the verb thelo (“want”), which takes a na–
complement (tha<thelo+na). Notice the morphological similarity with the CG future particle enna.!
! 16 
respect to feature specification, each of the two particles, dhen and tha, bears a 
different feature: the former denotes negation and the latter futurity. Hence each 
spells out a distinct functional head: dhen realises NegP and tha CM. Thus, the 
CG particle etha expresses both negation and futurity and spells out the two 
corresponding functional heads. 
 
(6) An        dhe(n)  tha  mu                  to                 feris,  
COMP  NEG    M   me–CL.GEN  it–CL.ACC  bring–2S  
thelo        na  to                 ksero. 
want–1S  M  it–CL.ACC  know–1S  
“If you will not bring it to me, I want to know it” 
 
Secondly, negated future clauses in CG may also be headed by enna 
immediately followed by the negator men (7), a structure absent from SMG. In 
this respect, the distribution of enna is reminiscent of the distribution of na (8) in 
that they both occur with the negator men (CG); specifically, they obligatorily 
precede it. 
 
(7) Enna  (men)  to                  feri?  
M       NEG   it–CL.ACC  bring–3S 
“Will s/he (not) bring it?” 
(8) Na  (men (CG) / min (SMG))  to                 feri. 
M   NEG                                  it–CL.ACC  bring–3S 
“S/he should (not) bring it” 
 
The preceding presentation of the properties and distribution of the 
particles as, na and tha in CG shows that these particles behave like their SMG 
counterparts. I adopt Roussou’s (2000) proposal for the SMG CP in (1) for the CG 
CP as well, and I assume that na/tha/as realise CM6, the C head encoding !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
6 Observe the similarity of Roussou’s (2000) proposal to Philippaki–Warburton’s (1987, 1998) 
with respect to the functional projection in which na and as are realised. In Philippaki–Warburton 
(1987, 1998) these particles appear under the Mood Phrase, a head that encodes modality and 
hosts na, as, the indicative marker and the affix marking the imperative.  
! 17 
modality, in both varieties. Yet, for CG, the CP proposal in (1) has to be extended 
to include the CG particle enna as well, since enna has the same distribution as 
na. It is thus assumed that enna is also merged under CM, together with na/tha/as.  
Roussou’s (2000) argumentation as to why na/tha/as are modal particles 
merged under CM rather than inflectional particles merged within the inflectional 
domain is based on two points. The first refers to the distribution of na/tha/as, and 
their position in relation to pronominal clitics. The second is related to the status 
of these particles and the various interpretations assigned to clauses that they head 
depending on the temporal and aspectual specification on the verb involved. 
As for the first argument, na/tha/as in SMG precede pronominal clitics 
and can take “dependent” (–past, +perfective) verbal forms (Roussou 2000:66). 
Both observations also apply to the behaviour of na/enna/tha/as in CG, as shown 
in (9).  
 
(9) As  /  Na  /  Enna  /  Etha  to                  feri.  
M         it–CL.ACC  bring–3S 
“Let him / He should / will / will not bring it” 
 
Current syntactic analyses of cliticization in SMG (Mavrogiorgos 2009, 
2010) and Romance languages (Roberts 2010) assume that clitics adjoin to a 
functional projection within the inflectional domain. Taking into account the 
placement of CG na/enna/tha/as in relation to pronominal clitics, I suggest that 
these particles are realised in some functional head higher than I. The lowest C 
head in (1), namely CM, seems to be a legitimate candidate for na/enna/tha/as to 
adjoin. 
As for the second argument, Roussou (2000:71–72) focuses on the status 
and the properties of na and tha in SMG to show that they are not inflectional but 
rather modal particles. Following Roussou, I assume that although na and tha 
have been traditionally analysed as the subjunctive (Veloudis & Philippaki 
Warburton 1983) and the future particle respectively, they cannot be considered 
inflectional particles. In particular, it seems that the contexts of use of na and tha 
are not restricted to clauses bearing subjunctive and future interpretation 
! 18 
correspondingly, but the particles in discussion also participate in clauses with 
various modal readings. 
Consider, for example, clauses headed by the so–called subjunctive 
particle na. Even though na is used for the formation of subjunctive clauses in 
both CG and SMG, it is not an inflectional element of the mood paradigm, as it 
does not bear morphological inflection; the subjunctive morphology is present on 
the verbal host. The same argument applies to as, which lacks inflection in both 
CG and SMG, as well as to the CG future particles enna and tha, as the 
morphological realisation of the future tense in clauses headed by enna and tha is 
only present on the verb. 
In addition, enna and tha do not necessarily mark futurity, and thus they 
do not behave like future particles alone. Instead, they occur in a number of 
modalised, non–future contexts. Roussou (2000) indicates that the particles 
na/tha/as in SMG “sub–categorise for all possible inflected verbal forms along the 
±past, ±perfective specifications” (2000:72) and the interpretation of the clauses 
headed by these particles is not only based on the individual particle involved, but 
also depends on the temporal and aspectual specification of the verbal host. Both 
points apply to the particles na/tha/as as used in CG, as well as to enna. Examples 
(10–11) illustrate how the interpretation of clauses headed by enna and etha is 
modified when the tense and/or the aspect of the verb is changed.7  
 
(10) Enna / Etha         to                 feri  /  ferni [+/–PERF].  
M     /  NEG  M  it–CL.ACC  bring–3S 
“S/he will / won’t bring it” 
(11) Enna / Etha         to                 efere  /  eferne [+/–PERF].  
M     /  NEG  M  it–CL.ACC  brought–3S 
“S/he would / wouldn’t have brought it” 
 
The previous discussion shows that there are no indications that 
na/enna/tha/as are located within IP in the CG clause. Recall that the particles in !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
7 Note that even though the examples in (10–11) involve clauses headed by enna and etha alone, 
all the particles in discussion behave similarly in this respect; hence CG clauses headed by na, tha 
and as may also have a number of different modal readings depending on the verbal specification. 
! 19 
question do not form inflectional paradigms, and also that clauses headed by them 
can have various modal interpretations. Therefore I follow proposals in Roussou 
(2000) in assuming that these particles are modal rather than inflectional elements. 
The former term better captures their status, their properties and their overall 
behaviour. Moreover, it is well known that modality, unlike mood, is not tied to 
verbal forms and can be expressed by different means, the use of particles being 
one of them. Consequently, na/enna/tha/as could be accommodated under a C 
head with modal specification. Roussou’s (2000) CM (see (1)), a C head specified 
for modality, seems to be an appropriate host. 
However, there is an important difference between na/enna/as and tha in 
CG with respect to their interaction with negators, which needs to be captured as 
well. In particular, na/enna/as and tha differ in the choice of negative particle, as 
well as in the order of the negator and the modal particle. As previously 
discussed, enna/na/as are only compatible with men and must obligatorily precede 
it, while tha can only occur with (dh)e(n) and must immediately follow it. 
Roussou (2000) has proposed that na and as in SMG undergo movement from 
CM to COp, an Operator position between the highest and the lowest C in her 
tripartite C system. This proposal is adopted for their CG counterparts and is 
extended to enna as well. In this way, the difference between na/enna/as and tha 
in their choice of negators and their ordering with respect to them is adequately 
accommodated. 
Turning now to the Neg head8, the discussion on the interaction of modal 
particles with negators has shown that the clause structure proposed by Roussou 
(2000) in (1) can accommodate the CG facts as well, merely substituting the SMG 
negator min by its CG counterpart men. As for the negator dhen, I assume that it 
appears under NegP in both SMG and CG. In this way, given that Neg is the 
functional head immediately preceding CM, where tha is realised, the order 
dhe(n) tha as well as the formation of the CG negator etha easily follow.  
The clausal structure in (12) represents the left periphery of the CG clause, 
as has been formulated so far on the basis of the above discussion. Na/enna/tha/as !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
8 Cf. Alexiadou (1994) for a discussion on whether dhen and min adjoin to a single Neg head or 
each adjoins to a distinct one.!
! 20 
are located under CM, the lower C head. The CG particle etha, composed of a 
negative and a modal particle, spells out the two corresponding C heads, namely 
Neg and CM. Furthermore, tha can only appear under CM, and, therefore, spells 
out only this head; na/enna/as, on the other hand, can realise both CM and COp 
and, correspondingly, spell out both a modal and a clause–typing feature 
(Roussou 2000:74). The postulation of an additional C projection above CM in 
the CG CP (12) is in line with proposals in Agouraki (2010)9 as well, who 
postulates the Force position introduced by Rizzi (1997). This position is identical 
with Roussou’s (2000) COp position for SMG (1)10, which is the analysis I adopt. 
 
(12) [COp na/enna/as [Neg (dh)e(n)/men [CM na/enna/as/tha [I...]]]]] 
 
Roussou’s (2000) proposal for the left periphery differs from Rizzi’s 
(1997) representation in a crucial respect: it involves a tripartite C–structure, 
whereas Rizzi’s involves only two C heads, Fin and Force11. It is important to 
note that Roussou’s CM projection corresponds to Rizzi’s Fin head (Roussou 
2000:73). Rizzi’s Fin head carries information about finiteness including mood 
specification, tense, aspect and agreement. Both Roussou’s CM and Rizzi’s 
(1997) Fin may well serve as adjunction sites for the verb.  
The discussion so far has shown that the functional heads CM, NEG and 
COp are projected in this order above IP in the CG clause structure. The 
articulated CG CP includes the highest C head as well, a head specified for 
subordination (Roussou 2000), in which the complementizer pu is realised in both 
SMG and CG. The slightly modified CG CP, extended by comparison to the one 
in (12), is outlined in (13). The structure in (13) features a tripartite C–structure 
for the CG clause, along the lines of Roussou’s (2000) proposal for SMG, with 
only minor differences from the SMG structure, as outlined in (1).!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
9 Agouraki points out that in terms of Rizzi’s (1997) discussion of the clause left periphery, CG 
provides evidence for the presence of another functional projection, namely of Neg/Aff, the head 
hosting Negation and Affirmation (2010:543). Nevertheless, I consider the postulation of Neg/Aff 
in the CG CP not to be justified. 
10 The only point of divergence concerns the terminology adopted: Agouraki (2010) follows Rizzi 
(1997) and refers to a Force position, while Roussou (2000:73) adopts the terminology of Manzini 
and Savoia (1999) and refers to a COp head. 
11 Force in Roussou’s proposal (2000:80) is split into two heads: COp and the highest C head.!
! 21 
(13) [C pu [COp oti/an/na/as [Neg dhen/min [CM tha/na/as [I...]]]]] 
 
1.3. Topic and COp positions 
 
According to Alexiadou and Anagnostopoulou (1998), subjects in SMG and other 
pro–drop languages, like CG, occupy the Topic position. Objects in OVS order 
also occupy the Topic position, and the DP that constitutes the object of the clause 
must be doubled by a pronominal clitic forming a CL(itic) –L(eft)-D(islocation)12  
structure (14), unless the object of the clause is an XP bearing contrastive focus 
(15). Objects in CG may also appear in R(ight)–D(islocation) (16), as well as in 
Clitic Doubling and CL(itic)–R(ight)–D(islocation) constructions. CLRD and CD 
structures are both represented in example (17) with the presence or the absence 
of a pause immediately before the doubled object DP differentiating the two types 
of constructions. Specifically, (17) is considered a CLRD construction if there is a 
pause immediately before the RDed constituent to moro, but a CD construction if 
there is no pause between the clitic and the doubled object phrase. 
 
(14) To    koritsaki                  endisa          to.  
The  girl–ACC.DIMUN  dressed–1S  it–CL.ACC 
 “The little girl I dressed”  
(15) To    koritsaki                  endisa,         oi       to    aghoraki. 
The  girl–ACC.DIMUN  dressed–1S  NEG  the  boy–ACC 
 “The little girl I dressed, not the little boy” 
(16) Endithike            mesa se  pende  lepta               [PAUSE]  i    Maria. 
Got dressed–3S  within     five     minutes–ACC                the  Mary–NOM 
 “She got dressed within five minutes, Mary”  
 !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
12 “As shown in Agouraki (1993) for SMG, CLLDed objects have both A– and A!–properties. The 
claim extends to the CG data as well. The evidence that CLLDed objects have A–properties relates 
to binding, the existence of subject+verb idioms and discontinuous idioms, quirky subjects, 
pseudo–relatives, subject–oriented adverbs and control verbs. In all these cases it is shown that 
there is no asymmetry between subjects and CLLDed objects. In Agouraki (1993) the pieces of 
evidence for the A!–properties of CLLDed objects were taken to be, firstly, the fact that this is not 
a thematic position and, secondly, the interpretation of Topicalization” (Agouraki 2010:528). !
! 22 
(17) Endise          to                  [PAUSE]  to               moro. 
            Dressed–3S  it–CL.ACC                    the–ACC  baby–ACC 
 “S/he dressed him/her, the baby”  
 
Anagnostopoulou (2006:546–547) offers a very good diagnostic for 
differentiating between CD and CLRD constructions that can be applied in 
constructions like (17) which involve an object phrase doubled by a pronominal 
clitic at the left of the object DP. Such a clause should be considered a Clitic 
Doubling construction if it allows a subject that is not pre–supposed and bears 
main sentence stress to appear at the end of the clause, forming a construction of 
the type CLi–V–XPi–S. And, conversely, if a non–presupposed subject bearing 
main stress cannot appear at the end of such a clause, then it should be considered 
a CLRD construction. If we want to use Anagnostopoulou’s diagnostic to decide 
whether (17) is a CD or a CLRD construction, we should test whether a stressed 
DP can appear at the end of the clause, as in (18). If it can, it is a CD structure, 
otherwise it is a CLRD structure.  
 
(18) Endisen         to                  to   moro           [i     mama            tu].     
            Dressed–3S  it–CL.ACC  the  baby–ACC  the  mum–NOM  his–POSS 
 “She dressed the baby, the baby’s mum”  
 
The topic position in CG is occupied by DPs that appear as topicalised 
subjects or objects (19), as well as preverbal universal quantifiers. These are 
occupying the [Spec,TopicP] position (Agouraki 2010). 
 
(19) Ulus  tus   mathites  edhokamen  tus               vivlia. 
 All     the  students   gave–1PL    them–GEN  books–ACC 
 “To all the students we gave books” 
 
As shown in the previous section, the left periphery of the CG clause 
involves a COp position as well, above which all the other pre–verbal stressed 
constituents, including wh–phrases, existential quantifiers (20), negative 
! 23 
quantifiers (21), negative polarity items (22) and anaphoric/deictic proforms (23) 
appear. For Agouraki (2010) all these constituents adjoin to a Force position. Yet, 
as has already been noted, Agouraki’s (2010) Force is identical to Roussou’s 
(2000) COp, which I have adopted instead (see (13)). Thus, terminology aside, 
my proposal concerning the adjunction site of the elements in question coincides 
with both Roussou (2000) and Agouraki (2010). 
 
(20) O     kathenas              enna  mboruse   na  to                  kami. 
 The  everyone–NOM  M      could–3S  M  it–CL.ACC  do–3S 
 “Everyone could have done it” 
(21) Kanenas        etha         mboruse   na  to                 kami. 
 None–NOM  NEG  M  could–3S  M  it–CL.ACC  do–3S 
 “None could have done it” 
(22) Pote     etha        mborusa   na  to                  kamo. 
 Never NEG  M  could–1S  M   it–CL.ACC  do–1S 
 “I could never have done it” 
(23) Tutos           mbori     na  to                  kami. 
 This–NOM  can–3S  M   it–CL.ACC  do–3S 
 “He can do it”  
 
Notably, within Agouraki’s (2010) approach, these pre–verbal stressed 
constituents function as operators that check an [Emphasis] specification on C. 
This point is directly related to the account she has proposed for the syntax of CG 
clitics which is built on the idea that CG has a filled C–domain requirement13. The 
filled–C requirement is also incorporated into her proposal on focus strategies in 
CG (Agouraki 2010), where she discusses a number of constituents appearing in 
the left periphery of CG. In particular, she argues that a filled–C is necessary so 
that the specification of the sentential force on C is checked in overt syntax !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
13 Agouraki’s (2001) proposal for clitic placement in CG is discussed in detail in chapter 3. 
Briefly, she proposes that pre– and post–verbal placement of CG clitics depend on the filled C–
domain requirement of the variety. She claims that, if the C position is occupied by some element, 
the verb does not move to C, but rather remains in IP. Thus, given that in her account clitics adjoin 
to a functional projection higher than I, proclisis follows. If, however, the C position is not filled, 
the verb moves to C to fulfill the filled–C requirement; in this case, the clitic appears post–verbally 
and enclisis follows.!
! 24 
(2010:542). Therefore,  [Spec,CP] stressed constituents, in complementary 
distribution with stressed V–in–C and stressed Neg–in–C, check an [Emphasis] 
specification on C. Emphasis is assumed to be one of the possible specifications 
of sentential force on C, denoting emphasis on the event14. The validity of the 
aforementioned proposal depends on the behaviour of CG clitics and, in 
particular, on whether the verbal host in clitic constructions undergoes V–to–C 
movement, as suggested in Agouraki (2001). This thesis is intended to shed light 
on issues related to the syntax of CG clitics, and, hence, check the validity of 
Agouraki’s (2010) filled–C requirement / requirement for the specification of the 
sentential force on C. 
With respect to the relative order of the COp and the Topic projections in 
the CG CP, it should be noted that both the Force head in Agouraki (2010) and the 
corresponding COp head in Roussou (2000) appear lower than TopicP. I adopt 
this approach on the basis of the following fact (Agouraki 2010:530–31): in 
clauses involving both a topicalised phrase and an element in the Spec Force/COp 
head, such as a wh–element, the topicalised constituent always precedes the other 
constituent; compare grammatical (24) with ungrammatical (25). Interestingly, 
(25) is equally bad with a pre– and a post–verbal clitic, indicating that irrespective 
of where the clitic appears, the relative order of the constituents appearing under 
TopicP and Spec Force/COp is not free15. The right order in shown in (24). 
 
(24) To    vivlio           pu        to                 evales? 
The  book–ACC  where  it–CL.ACC  put–2S 
(25) *Pu      to    vivlio           to                 evales  /  evales to ? 
 Where  the  book–ACC  it–CL.ACC  put–2S !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
14 The evidence she offers for the “emphasis on the event” interpretation of clauses involving 
preverbal stressed constituents includes: “(a) the absence of new information/contrastive readings 
for [Spec,CP] stressed constituents, (b) the availability of weak readings alone for quantifiers, and 
(c) the possibility of having a [Spec,CP] object existential quantifier and an object CP in the 
canonical object position” (Agouraki 2010:536). 
15 Having Rizzi’s (1997) articulated CP in mind, one may claim that there is a Topic position 
below Force/COp as well. Yet this position is not available in SMG/CG. Apparent COp–Topic 
orders, such as in example (1) are derived by movement of oti from COp to the higher C 
(Michelioudakis p.c.): 
1. Ksero         oti         ta   mila  (dh)en  ta                 efaen     i     Maria. 
Know–1S  COMP  the  apples–ACC NEG    them–ACC  ate–3S  the  Mary–NOM 
“(I) know that, as for the apples, Mary didn’t eat them” 
! 25 
 “The book, where did you put it?” 
 
Agouraki (2010) makes an interesting point about scope positions in the 
CG clause. She examines the position(s) and properties of preverbal stressed 
constituents and convincingly argues against the postulation of a Focus position in 
CG (2010:539–540). The basis of her argumentation is that clauses involving 
stressed constituents in CG do not have the characteristics shared by clauses 
involving syntactic foci, namely (a) across–the–board application, (b) contrastive / 
new information interpretation, (c) a Focus–Presupposition structure, and also that 
(d) it is not the case that only foci can bear stress in CG. In relation to (a), she 
argues that preverbal stressed constituents in CG form a small closed set and she 
gives an exhaustive list for constituents that can appear stressed in preverbal 
position. Her list comprises the following elements: universal quantifiers, also–
phrases, even–phrases, existential quantifiers, negative quantifiers, negative 
polarity items (NPIs), only–phrases and anaphoric/deictic proforms. As for (b), 
she claims that the contexts of use for preverbal stressed constituents in CG do not 
support a contrastive / new information interpretation. The existence of a Focus–
Presupposition structure is rejected as well. She argues that the semantic content 
of some of the preverbal stressed items, including existential quantifiers, negative 
quantifiers and NPIs, is incompatible with presupposition marking. As for 
anaphoric/deictic proforms and only–phrases, she claims that their contexts of use 
show that they do not have a presupposition structure. She also mentions that 
since [Spec,CP] is not a necessarily stressed position, it cannot be a Focus 
position. Agouraki’s (2010) proposal is in line with Grohmann et al.’s (2006:89) 
argument that CG, unlike SMG, “has bona fide cleft structures in lieu of syntactic 
focus movement”. 
Following Agouraki (2010), and departing from Roussou (2000), I assume 
that the left periphery of the CG CP has two scope positions in its surface syntax: 
the Topic position and the Force or COp position, but not a Focus position. The 
refined CG clausal structure including a Topic projection is outlined in (26). 
 
(26) [C pu [Topic [COp oti/an/na/as [Neg dhen/min [CM tha/na/as [I...]]]]]] 
! 26 
1.4. Wh–question formation 
 
This section discusses wh–question formation in CG as compared to SMG. In CG 
and SMG alike, wh–argument questions are headed by the pronoun pcos (“who”) 
inflected for nominative (27), accusative (28), or genitive (29) case. Examples 
(27), (28), and (29) constitute a subject–question, a direct–object question, and an 
indirect–object question respectively; the pronoun is appropriately inflected for 
number and gender. Pcos can be either used alone, in which case the translation 
offered in (a) applies, or as a quantifier that has scope over some NP, in which 
case the translation in (b) applies. Other pronouns used in both CG and SMG for 
the formation of wh–argument questions are posos (“how much”) (30) and ti 
(“what”) (31). Posos and ti pattern like pcos, namely they can either appear alone 
or as quantifiers that have scope over some NP. The wh–words in question share 
the same morphological properties in CG and SMG, modulo phonological 
differences. In CG alone, however, they optionally combine with the dialectal 
element embu (<en+pu) 16, which has the meaning of “(it) is that” in affirmative 
clauses and “is (it) that” in interrogatives. 
 
(27) Pcos             (mathitis)            (embu)17           irte? 
Who–NOM  (student–NOM)  is–3S  COMP  came–3S  
a. “Who came?” 
b. “Which student came?” 
(28) Pcon            (mathiti)            (embu)             idhes ? 
Who–ACC  (student–ACC)  is–3S  COMP  saw–2S 
a. “Whom did you see?” 
b. “Which student did you see?” !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
16 As for the derivation of embu–questions, there are two main approaches. The first one, 
suggested by Agouraki (2010), Grohmann et al. (2006) and Tsiplakou et al. (2005), assumes that 
embu–structures are cleft structures. The second one, adopted by Grohmann and Papadopoulou 
(2011) assumes that embu–structures involve a “fossilized” complementizer, which fills the 
interrogative C. I will return to the syntax of embu–clauses in the next section.!
17 It is assumed that constructions involving the dialectal element embu are clefts and that the form 
embu derives from the contraction of the copula en (“is”) with the factive complementizer pu 
(“that”). Therefore, the transliteration in the examples both in this section and the following one 
are in accordance with this assumption.!
! 27 
(29) Pcu           (mathiti)            (embu)             to                  edhokes? 
Who–GEN  (student–GEN)  is–3S  COMP  it–CL.ACC  gave–2S 
a.  “Whom did you give it to?” 
b. “Which student did you give it to?” 
(30) Posa                      (enikia)         (embu)             epleroses  ? 
How many–ACC  (rents–ACC)  is–3S  COMP  paid–2S 
a. “How much did you pay?” 
b. “How much rent did you pay?” 
(31) Ti       (traghudhi)  (embu)             ipes? 
What  (song)          is–3S  COMP  said–2S 
a. “What did you say?” 
b. “Which song did you sing?” 
 
 Wh–questions in both CG and SMG can also be headed by quasi–
arguments, like pu / pothen (“where”) and pote (“when”), as well as by true 
adjuncts, like pos (“how”), and jati (“why”) (32). The distribution of wh–words 
used for question formation in CG is the same in matrix and embedded clauses. A 
comparison of the matrix clause in (27) with its corresponding embedded 
structure in (33) is indicative. 
 
(32) Pu  /  Pothen  /  Pote  /  Pos   / Jati  (embu)              irte ? 
Where           / When / How / Why  is–3S  COMP  came–3S 
 “Where/When/How/Why did s/he come?” 
(33) Arotise       pcos    (mathitis)            (embu)             irte. 
Asked–3S  who–NOM  (student–NOM)  is–3S  COMP  came–3S   
a.  “S/he asked who came.” 
b. “S/he asked which student came.” 
 
As shown in the above examples, both CG and SMG typically exhibit the 
ex situ strategy for wh–question formation, i.e. the wh–element undergoes 
movement from the position where it is merged into the left periphery. The 
landing site of the moved wh–expression in CG is assumed to be the specifier of 
! 28 
COp. In situ questions are also licit in both CG (Grohmann & Papadopoulou 
2011) and SMG (Vlachos 2010), under certain pragmatic and/or discourse–related 
conditions. Examples (27–31) can be felicitously constructed as in situ questions 
as well; the only difference being that the dialectal element embu cannot be 
involved in in situ questions. Take for example the ex situ question in (29), which 
can also appear as an in situ question (34). Its interpretation remains unaltered. 
Observe, though, the different pattern for clitic placement in (29) as compared to 
(34), an issue to which I return in chapter 3.  
 
(34) Edhokes   to                  pcu          (mathiti)?                       
Gave–2S  it–CL.ACC  who–GEN  (student–GEN) 
“You gave it to which student?” 
 
As shown in examples (27–32), a number of wh–words used for question 
formation in CG are identical to their SMG counterparts, modulo phonological 
differences. However, there are also some CG–specific wh–words, namely inda18 
(“what” / “what for”, “why”) and indalos (“how”). Inda in CG corresponds to 
both an adverb meaning “why” (35) and a pronoun invariant in gender, number 
and case. Argumental inda can be used either pronominally (36) or pre–nominally 
in complex wh–expressions, i.e. as a quantifier that has scope over some NP (37). 
When argumental inda is used pronominally it is obligatorily followed by mbu 
(36) and they together form indambu19 (Grohmann et al. 2006, Grohmann & 
Papadopoulou 2011, Tsiplakou et al. 2005). Inda cannot appear in in situ 
questions, unless it is used pre–nominally (38), in which case its co–occurrence 
with (e)mbu is excluded. !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
18 Grohmann and Papadopoulou (2011) cite Simeonidis (2006:217), who indicates that historically 
inda derives from the interrogative pronoun tinda “what” that was used in Assizes (a corpus of 
texts that constituted the legislation for the island of Cyprus in the Middle Ages, when the island 
was under the dominance of Franks (1191–1489)).!
19 The status of indambu depends on the theory one adopts for its derivation. There is no 
consensus in the literature. Grohmann and Papadopoulou (2011) consider indambu a 
grammaticalized form, while Pavlou (2009) assumes that inda and (e)mbu combine syntactically 
to form a cleft. For Grohmann and Papadopoulou indambu constitutes a fossilized form merged in 
or moved to the left periphery of the cleft as a unit. Given that an account that satisfactorily 
accommodates indambu is still pending, I do not provide transliterations for mbu in the given 
examples so that their interpretation is not bound by any ad hoc assumptions concerning the status 
of indambu. 
! 29 
(35) Inda                   (mbu)  irte ?  
Why / what for  MBU   came–3S 
“What did s/he come for?” 
(36) Inda   *(mbu)  ipe? 
What  MBU    said–3S 
“What did s/he say?” 
(37) Inda    doro               (embu)              efere? 
What  present–ACC  is–3S  COMP  brought–3S 
“What present did s/he bring?” 
(38) Eferen           inda   doro                (*embu)? 
Brought–3S  what  present–ACC  is–3S  COMP                
“S/he brought what present?” 
 
Another CG wh–word used in question formation is indalo(i)s, a derivative of 
inda, a manner adverb that has the same meaning as pos (“how”) (39). Indalo(i)s 
can only appear in ex situ questions and optionally combines with (e)mbu. The 
distribution of inda and indalo(i)s is identical in matrix and embedded clauses.  
 
(39) Indalo(i)s  (embu)             to                   efere? 
How          is–3S  COMP  it–CL–ACC  brought–3S 
“How did s/he bring it?” 
 
In a nutshell, in cases in which CG and SMG share the wh–word, wh–
formation is also similar; this, of course, excludes the formation of interrogative 
clefts in CG. In this respect, it is not necessary to differentiate the CG CP from the 
SMG CP. However, wh–question formation in CG may also involve CG wh–
words, and/or the dialectal element (e)mbu. The syntax of questions headed by 
inda differs from that of other wh–structures. Even so, a satisfactory account of 
the properties of inda and the syntax of inda–clauses is still pending. Moreover, 
the availability of (e)mbu in CG is an important point of divergence from the 
! 30 
standard variety20. The co–occurrence of embu with wh–words points to important 
differences in the derivation of the relevant constructions. In particular, it is 
assumed that interrogatives involving (e)mbu employ the clefting strategy 
(Agouraki 2010, Grohmann et al. 2006, Tsiplakou et al. 2005). The next section 
discusses the syntax of clefts and revisits the structure of the CG CP.   
 
1.5. Clefts  
 
This section presents the clefting strategy in CG, which is absent from SMG, in 
order to illustrate whether further modifications in the CG clausal structure are 
required in order to accommodate these facts. The discussion revolves around the 
syntax of structures involving the dialectal element embu that have been treated as 
clefts by Agouraki (2010), Grohmann et al. (2006) and Tsiplakou et al. (2005). As 
mentioned in the previous section, CG has two strategies available for wh–
question formation: the first involves wh–fronting and is shared by both CG and 
SMG, and the alternative option, available in CG alone, involves the use of embu.   
Consider the examples (27–33) from the previous section. (33), cited 
below as (40), and (41) illustrate the distribution of (e)mbu, which is identical in 
matrix and embedded clauses. Its use in CG wh–questions is optional, unless the 
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
20 There are three important parameters concerning the presence of (e)mbu in CG wh–questions, 
as indicated in the literature (Grohmann et al. 2006, Grohmann & Papadopoulou 2011, Tsiplakou 
et al. 2005). First, embu is excluded from in situ questions, even in cases where the wh–expression 
involved allows in situ question formation (see (38) for example). Second, it may optionally 
appear in ex situ questions, unless the wh–word is bare and argumental inda, in which case the 
presence of mbu is obligatory (36). When mbu adjoins with inda, several phonological variants of 
mbu arise, such as nambu, tambu, ambu, and innambu (Pavlou 2009). These reduced forms of inda 
(’na, ’ta, ’a), together with the fact that inda is not a “stand–alone” form (1), could be used as 
evidence that inda displays clitic–like properties (Tsiplakou et al. 2005). Third, the distribution of 
(e)mbu is restricted, as it can only appear either immediately after the wh–word or immediately 
after the complex wh–expression, which consists of the wh–word and the quantified NP. 
Grohmann and Papadopoulou (2011) summarise this as in (2). 
1. [unintelligible] 
 Inda   *(mbon(i))? 
              mbon(i)<mbu+en 
 What  (it) that is 
 “What is it?” 
2. a. [CP WH ((e)mbu) ... t WH ... ] 
 b. *[CP ((e)mbu) XP ((e)mbu) ... WH ((e)mbu) ... ((e)mbu) ] 
! 31 
wh–word involved is bare and argumental inda, in which case mbu obligatorily 
follows inda21.  
 
(40) (Arotisen)  pcos     (embu)             irte? 
Asked–3S  who–NOM  is–3S  COMP  came–3S 
“S/he asked who (is it that) came” 
(41) (Arotisen)  indambu       ipe? 
Asked–3S  what  MBU  said–3S 
“S/he asked what (is it that) s/he said” 
 
Wh–questions involving any wh–word other than inda, and declarative 
clauses involving embu, such as (42–43), are unequivocally perceived as clefts in 
the literature (Agouraki 2010, Grohmann et al. 2006, Tsiplakou et al. 2005). This 
approach assumes that embu derives from the copula en (is) that (syntactically) 
combines with the factive complementizer pu and, arguably, in PF (Grohmann et 
al. 2006), contracts to yield embu. The clefted constituent may be an argument 
(44) or an adjunct (43), while argument–less embu–structures are also licit (42).  
 
(42) En       pu          etilefonise. 
Is–3S  COMP  called–3S 
 “It is because s/he called” 
(43) En       ehtes         pu         etilefonisame. 
Is–3S  yesterday  COMP  called–1PL 
 “It is yesterday that we called” 
(44) En  /  Itan   emis         pu         etilefonisame. 
Is/Was–3S  us–NOM  COMP  called–1PL 
 “It is / was us that called” 
 
The accounts of clefting in CG follow two different lines. The first 
assumes that the clefted constituent is base–generated in the cleft clause !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
21 This shows that inda, when used as an argument, does not allow regular wh–fronting like other 
wh–words but always employs the clefting strategy.!
! 32 
(Agouraki 2010). The second assumes that the clefted constituent is extracted 
from the pu–clause, where it is originally merged, undergoes movement and 
ultimately adjoins to the Specifier of the cleft clause (Grohmann et al. 2006, 
Tsiplakou et al. 2005).  
Agouraki’s (2010) proposal can be summarised as follows: the clefted 
constituent is not merged in its thematic position and movement does not occur. It 
is instead base–generated in the cleft clause (2010:553). The direct merging of the 
clefted constituent in its surface position is seen as an instance of late saturation 22. 
For Agouraki, if the C position of the cleft clause is filled by the complementizer 
pu, the Spec CP position cannot be also filled. She suggests instead that the 
clefted constituent adjoins to the cleft clause. The higher CP comprised of the 
clefted constituent adjoined to the lower CP is c–selected by a thematically null 
copula. The clause structure she proposes is shown in (45), following the tree–
diagram numbered (28) in Agouraki (2010:553).  
 
(45) [CPm en [TPm T [ASPPm ASP [vPm v [VPm V [CPc Clefted Constituent [CPc [C! pu 
[TPc]]]]]]]]]23 
 
Grohmann et al. (2006) capitalises on the fact that clefts are a focusing 
strategy and adopts a split–CP analysis involving a Focus projection. As shown in 
(46) (based on the tree structures numbered (25) and (26) in Grohmann et al. 
2006:90–91), the cleft clause is realised in the specifier of the FocP and the matrix 
clause appears as the complement of C0. In declarative clefts, the CP–domain of 
the cleft clause remains empty, while in wh–clefts it is filled with the wh–phrase 
and an interrogative C0. In both declarative and wh–clefts, the matrix clause is 
introduced by the complementizer pu. The structure of the cleft clause is given in 
(47) and involves a Small Clause; in this way the relationship of predication that 
holds between the focused element and the matrix clause is captured. Moreover, 
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
22 As for the semantic composition of CG clefts, Agouraki assumes that the cleft clause denotes an 
incomplete proposition, which is saturated by the clefted constituent (2010:551–552). The latter is 
interpreted as new information because it is filled in last in the structure. 
23!“Subscript m stands for matrix clause and subscript c stands for cleft clause” (Agouraki 
2010:553).!
! 33 
the predicate of the Small Clause behaves like a clause–selecting nominal D, such 
as the Greek determiner to (“the”) that selects subordinate clauses. 
 
(46) [FocP cleft clause [Foc! Foc0 [CP [C! C0 pu IP…]]]] 
(47) [CP [C! C0 [IP [I! I0 en [SC DP o Xambis DP O]]]]] 
 
Within such an analysis, declarative clefts like (44) are captured in a 
straightforward manner: the copula appears in the cleft clause, with the overt 
subject appearing inside the small clause, while the matrix clause is headed by pu. 
As for interrogative clefts (40), Grohmann et al. (2006) attribute their derivation 
to the application of sideward movement24. 
Tsiplakou et al.’s (2005) analysis capitalises on the fact that the verbal 
element en does not inflect for number, person, and tense and proposes that en 
involved in focus– and wh–clefts in CG is not copular, but existential. For them, a 
structure such as pcos embu irten is interpreted as “whoi [is it the case [that ti 
came]]”. In this way, the embedded clause headed by pu is not to be treated as a 
relative clause, but rather as the complement of en. Within their analysis the 
clefted wh– or focused expression is extracted out of the complement of en. They 
further claim that the extracted wh–word checks its disjunctive feature in the Spec 
of the matrix CP and its existential feature in the Spec of the embedded CP, the 
pu–clause. The movement of the wh–expression from the embedded clause to the 
matrix clause leaves a wh–feature on the C of the embedded clause, hence the 
realisation of the complementizer pu instead of oti or pos (“that”). 
Examples like (44) can be taken as evidence showing that extraction of the 
clefted constituent out of the pu–clause takes place (Tsiplakou et al. 2005). 
Observe the agreement between the verb of the pu–clause (etilefonisame–1PL) !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
24 Grohmann et al. (2006) propose that the wh–phrase in CG interrogative clefts moves sidewards 
and the derivation proceeds as follows: the wh–word, i.e. pcos in (40) for instance, is merged in its 
thematic position and it is assigned a "–role by the verb of the matrix clause. However, it bears an 
additional "–role and a wh–feature. Its first "–role is checked inside the matrix clause, but neither 
its second "–role nor its wh–feature can be checked (since the matrix clause is headed by pu, a 
non–interrogative complementizer), therefore the wh–word is copied and placed into the 
derivational workspace. Once re–merged with the predicate of the small clause, both its second "–
role and, upon subsequent movement to the SpecCP of the cleft clause, its wh–features are 
checked. As for the copula en, it appears as the predicate of the small clause and at PF it is 
contracted with pu to form embu.!
! 34 
and the clefted constituent (emis–1PL), on the one hand, and the lack of 
agreement between itan (was–3S) and the extracted constituent emis (us–1PL) on 
the other. This constitutes corroborative evidence for Tsiplakou et al.’s (2005) 
analysis. However, interrogatives headed by inda, arguably another type of cleft 
construction25, challenge all the analyses for clefting in CG including Tsiplakou et 
al.’s (2005), since none can satisfactorily account for the derivation of indambu. 
Turning now to the structure of the CG clause, as outlined in (26), the 
accounts by Agouraki (2010) and Tsiplakou et al. (2005) do not require the 
postulation of additional functional projections, while Grohmann et al.’s (2006) 
analysis requires a Focus projection. It is as yet unclear why the clefted 
constituent cannot adjoin to the COp head in Grohmann et al.’s (2006) analysis as 
well, together with other pre–verbal stressed constituents, following proposals in 
Agouraki (2010); this would render a Focus head unnecessary. In the absence of 
corroborative evidence for its presence in the CG CP, I assume that the clause 
structure in (26) suffices26 for accommodating CG clefts as well. The CG CP (26) 
differs from the SMG CP (1) in a crucial respect: only the latter involves a Focus 
head. Thus, while focused constructions in SMG involve movement of the 
focused element to FocP, in CG the clefting strategy is used instead. 
 
1.6. Summary 
 
The discussion in this chapter revolved around the properties and the distribution 
of modal and negative particles, the (non–)realisation of the Topic and Focus 
projections, wh–question formation, and the syntax of clefts in CG. This led to the 
formulation of the CG clause structure outlined in (26). Having sketched the basic 
aspects of the syntax of CG, I will now proceed to a discussion of the properties 
and the syntax of constructions involving clitic pronouns, in chapters 2 and 3 
respectively. !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
25 If inda followed either of the clefting strategies described above, it would co–occur with embu. 
However, it does not combine with embu either as an argument or as an adjunct. 
26 However, the specifics of the clefting strategy in CG point to important differences presumably 
concerning the feature specification of the relevant functional projections. These differences are 
only evident upon a careful investigation of the syntax of CG clefts. This issue is beyond the scope 
of the current thesis, and is left for future research. 
  
35 
 
CHAPTER 2: CYPRIOT GREEK CLITIC 
PRONOUNS 
 
2.1. Introduction 
 
This chapter describes the system of clitic pronouns in adult CG, which 
constitutes the target grammar for CG–speaking children. The discussion is 
organised as follows: the first section presents the main properties of CG clitics, 
including their morphological paradigm, their distribution and their behaviour 
with respect to coordination and modification, their semantic and phonological 
properties, and their categorial status. The second section deals with issues related 
to clitic placement. It constitutes an overview of the proclisis and enclisis contexts 
in CG for single clitics and clitic clusters in both matrix and embedded 
environments; the syntactic contexts that allow free ordering are also discussed. 
CG is compared with languages exhibiting second position (2P) restrictions, like 
Serbo–Croatian, and languages with similar clitic ordering, like European 
Portuguese and Galician.  
 
2.2. Basic properties of CG clitics 
 
This section presents the morphological paradigm of CG clitics as well as their 
phonological, semantic and syntactic properties. The discussion of the basic 
properties of clitics in CG uses the criteria for distinguishing clitics and strong 
pronouns proposed by Mavrogiorgos (2009) and Tsimpli and Stavrakaki (1999) to 
describe the counterpart clitic system in SMG. Based on the tripartite typology of 
Cardinaletti and Starke (1999) and Kayne’s (1975) criteria, Mavrogiorgos (2009) 
applies (a) morphological (morphological deficiency) (see also Tsimpli & 
Stavrakaki 1999), (b) syntactic (distribution, coordination, modification, X0 vs. 
XP properties), (c) semantic (animacy/human, referential properties, impersonal, 
expletive and non–referential uses) (see also Tsimpli & Stavrakaki 1999) and (d) 
(morpho)–phonological criteria (sandhi rules, prosodic restructuring, inherent 
  
36 
 
stress) to SMG clitics in order to demonstrate the respects in which they differ 
from strong pronouns. The application of Mavrogiorgos’ and Tsimpli and 
Stavrakaki’s criteria to the CG clitic system will highlight the points of 
convergence and divergence between CG and SMG. Pronominal clitics in CG and 
SMG share morpho–phonological and semantic properties, while their main 
differences lie on the syntactic level. 
 
2.2.1. The morphological paradigm 
 
CG, like SMG, has two types of personal pronouns: strong pronouns and clitics. 
The morphological paradigm of clitic and strong pronouns is given in tables 2.1 
and 2.2 respectively. Pronominal clitics in Greek, both in CG and SMG, function 
as (i) direct and indirect objects to verbs, as in (3), (ii) complements to nouns (to 
vivlio mu (the book my–CL.POSS)), adjectives (kaliteros mu (better (than) me–
CL.GEN)), determiners (enan tus (one (of) them–CL.GEN)) and quantifiers 
(kathenas mas (each (one of) us–CL.GEN)), and (iii) complements to adverbs 
(konda mu (close (to) me–CL.GEN)) (Holton et al. 1997:303–307). This thesis is 
concerned with the acquisition of object clitic pronouns in CG, and therefore the 
term clitic in the remainder of this thesis will be used to refer to clitic pronouns 
that function as (in)direct objects to verbs. 
 
Table 2.1: The morphological paradigm of CG clitic pronouns. 
 
 
 
 
 
 
 
Number / 
Case 
1st person 2nd person 3rd person 
Masculine Feminine Neuter 
Singular 
Genitive 
Accusative 
 
mu 
me 
 
su 
se 
 
tu 
to(n) 
 
tis 
ti(n) 
 
tu 
to 
Plural 
Genitive 
Accusative 
 
mas 
mas 
 
sas 
sas 
 
tus 
tus 
 
tus 
tes 
 
tus 
ta 
  
37 
 
 
Table 2.2: The morphological paradigm of CG strong pronouns. 
 
Both strong and clitic pronouns substitute for full NPs (1). Two 
representative examples with clauses involving strong and clitic pronouns are 
offered in (2) and (3) respectively. Taking into account that, on the one hand, CG 
(in fact, Greek in general) only has pronominal clitics, as it lacks the reflexive and 
auxiliary clitics of the Spanish/Italian and Serbo–Croatian type respectively (Terzi 
1999a:86, footnote 2), and on the other, that it only has object clitics, the term 
clitic(s) will henceforth exclusively refer to pronominal object clitics1. 
 
(1) Eferen           to    vivlio. 
Brought–3S  the  book–ACC 
  “S/he brought the book” 
(2) Tuton          eferen. 
This–ACC  brought–3S 
“S/he brought this” 
(3) Eferen           tu                     to. 
Brought–3S  him–CL.GEN  it–CL.ACC 
“S/he brought it to him” 
 
The discussion of the status and properties of clitics in CG will reveal their 
structural deficiency, in Cardinaletti and Starke’s (1999) sense, as compared to 
                                                
1 Both CG and SMG have pro subject pronouns as well, which will not, however, be discussed in 
this thesis. 
Number / 
Case 
1st person 2nd person 3rd person 
Masculine Feminine Neuter 
Singular 
Nominative 
Genitive 
Accusative 
 
egho 
emena(n) 
emena(n) 
 
esi/u 
esena(n) 
esena(n) 
 
tutos 
tutu 
tuto(n) 
 
tuti 
tutis 
tuti(n) 
 
tuto 
tutu 
tuto 
Plural 
Nominative 
Genitive 
Accusative 
 
emis 
emas 
emas 
 
esis 
esas 
esas 
 
tuti 
tutus 
tutus 
 
tutes 
tutus 
tutes 
 
tuta 
tutus 
tuta 
  
38 
 
strong pronouns. This deficiency is reflected in their morphological, syntactic, 
phonological and semantic properties. 
Let us start with their morphological composition. A comparison of tables 
2.1 and 2.2 shows that the two series of personal pronouns in CG carry almost 
identical nominal features: they are both marked for nominal inflectional features, 
such as case (nominative (for strong pronouns only), accusative and genitive, 
which has taken over the uses of the old dative), number (singular and plural) and 
person (1st, 2nd and 3rd). The 3rd person forms of both strong and clitic pronouns 
are also inflected for gender (masculine, feminine and neuter). The strong 
pronoun forms in 3rd singular and plural are in essence demonstrative forms, like 
their counterpart SMG forms (Mavrogiorgos 2009:14), and are declined as 
adjectival forms ending in –os, –i, –o. 
Mavrogiorgos indicates that although it is not a trivial task to match 
features such as number, gender or case with their morphological exponents due 
to the fusional character of the language, there is sufficient evidence that the 
segments /m/, /s/ and /t/ realise a person feature (2009:20–22) in SMG. His claim 
applies to CG as well: notice the presence of /m/ in all 1st person forms in tables 
2.1 and 2.2, with the exception of egho, as well as the presence of /s/ and /t/ in all 
2nd and 3rd person forms. As Mavrogiorgos points out, /m/, /s/ and /t/ mark the 
1st, 2nd, and 3rd person, respectively, across nominal and verbal paradigms in 
SMG; this claim holds for CG as well2. 
Turning now to case, the genitive (singular and plural) object clitic 
pronoun forms (table 2.1) are homophonous with the possessive pronouns. 
However, Mavrogiorgos (2009:34, footnote 21) offers evidence from Northern 
Greek dialects showing that their underlying structure differs. It is well known 
that in Northern Greek dialects the accusative clitic is used in verbal constructions 
                                                
2 Mavrogiorgos offers the following two arguments to justify his claim (2009:20–22), both of 
which apply to CG as well, and hence are exemplified in CG in the following: first, the segments 
/m/ and /s/ realise the 1st and 2nd person, respectively, in the verbal paradigm of the copula (1st 
person: ime, imaste, 2nd person: ise, isaste), and the medio–passive / non–active forms (1st 
person: pliniskome, pliniskumaste, 2nd person: pliniskese, pliniskeste). The segment /t/ realises the 
3rd person, as is evident from the corresponding medio–passive / non–active present forms in CG: 
pliniskete, pliniskunte. Second, the above segments are found in the corresponding persons in the 
possessive pronoun paradigm as well (/m/ in the 1st person: mu, mas, /s/ in the 2nd person: su, sas 
and /t/ in the 3rd person: tu, tus).  
 
  
39 
 
instead of the dative clitic, i.e. “Me (ACC) ekane” ((S/he) did to me) (Northern 
Greek dialects) vs. “Mu (GEN) ekane” (SMG) and “Ekame mu (GEN)” (CG). 
Nevertheless, with regard to the possessive pronouns, Northern Greek dialects 
pattern like SMG and CG in using the genitive form of the clitic, i.e. o pateras mu 
(the–NOM father–NOM my–POSS), to vivlio tis (the–NOM book–NOM her–
POSS). 
A direct comparison of the corresponding cells in tables 2.1 and 2.2 shows 
that clitic and strong pronouns in CG are morphologically distinct but related. In 
fact, the clitic forms constitute the reduced morphological variants of the 
corresponding strong forms3. For Tsimpli and Stavrakaki (1999:37–39) the third 
person clitic pronoun in SMG is a reduced form of the (strong) pronoun afton, 
aftin, afto and, in their view, this reduction in the number of morphemes of clitic 
pronouns illustrates their morphological deficiency as compared to strong 
pronouns. An analogous claim can be put forward for CG clitics as well. In both 
tables, there are some cells that share the same form. For example, in the clitic 
paradigm, the same form realises both the genitive and the accusative case of the 
1st (mas) and 2nd person plural (sas). Moreover, the form tus is used for the 
genitive case of all three genders in 3rd person plural, as well as for the accusative 
case of the masculine. The aforementioned cases constitute syncretic forms, 
namely forms that spell out more than one cell in the morphological paradigm.  
One of the most important and obvious differences between strong and 
clitic pronouns in CG and SMG alike, as indicated by Mavrogiorgos (2009:13–
15), is their morpho–phonological composition. Firstly, in terms of syllable 
length, while clitics are monosyllabic elements, strong pronouns are at least bi–
syllabic. Secondly, strong pronouns carry inherent lexical stress, while clitics are 
unstressed. Thirdly, only strong pronouns can stand alone in the clause, whereas 
clitics need some host to cliticise to (see example (14)). 
A second difference between strong and clitic pronouns, and in fact their 
only difference in terms of feature composition, is that CG clitics, unlike strong 
                                                
3 This asymmetric morphological relation between strong and clitic forms in the pronoun 
paradigm is not unusual for European languages. A representative example is the strong–weak–
clitic triplet found in Italian (3rd person plural forms: a loro, loro and gli) which reveals a gradual 
morphological deficiency from the strong to the weak and finally to the clitic. 
  
40 
 
pronouns, are not inflected for nominative case. This is a point of divergence of 
CG from SMG as well. Notably, 3rd person nominative clitic forms are used in 
SMG, but only in conjunction with either the interrogative pu(n) (< pu (where)  +  
’n (< ine (is)), “where is?”) or the deictic na (“here/there”) (Marinis 2000:261, 
Mavrogiorgos 2009:15–16). These points are illustrated in examples (4) and (6). 
Examples (5) and (7) highlight another important point: the interrogative pu(n) 
and the deictic na in SMG can also conjoin with a clitic in the accusative form4. In 
CG, on the other hand, no nominative forms of the clitic are used. Like their SMG 
counterparts, the accusative forms of CG clitics can be used together with pu(n) 
and e, the CG dialectal variant of na, as shown in examples (5) and (7). In this 
type of structures the clitic may be doubled by a full NP (4–7). 
 
(4) Pun ( < pu  +  ‘n (< ine))  tos                  (o     Stavros)?  (SMG) 
            where  is                he–CL.NOM  the  Stavros–NOM                      
(5)  Pun ( < pu  +  ‘n (< ine))  ton                (to   Stavro)?  (SMG & CG) 
            where  is                he–CL.ACC  the  Stavros–ACC                      
“Where is he (Stavros)?” 
(6) Na              tos                  (o    Stavros)!  (SMG)  
 Here/there  he–CL.NOM  the  Stavros–NOM                     
(7) E  (CG)  /  Na  (SMG)  to(n)             (to    Stavro)!  
 Here/there                     he–CL.ACC  the  Stavros–ACC                 
“Here/there he is (Stavros)”  
 
Leaving aside the nominative forms that exist only in SMG, the 
morphological paradigms for clitics in the two varieties extensively overlap5. This 
convergence is illustrated by a direct comparison between table 2.1 above for CG 
clitics and the morphological paradigm of SMG clitics given by Mavrogiorgos 
(2009:15, table 1), Revithiadou and Spyropoulos (2008), Tsimpli and Stavrakaki 
(1999:36, table 2) among others.  
                                                
4 Apart from 3rd person clitics, in SMG, 1st and 2nd person clitics can also conjoin with na as in 
Na me! (Here I am) and Na sas! (Here you (PL) are).  
5 Another difference concerns the genitive plural form of the feminine 3rd person clitic: while in 
SMG both tis and tes are available, in CG only the form tes is licit. 
  
41 
 
The genitive and accusative 3rd person singular forms, as well as the 3rd 
person accusative plural form of the clitic paradigm, are homophonous with the 
corresponding forms of the definite article (table 2.3). However, this is not a 
unique property of CG. Uriagereka (1995) observes that articles and 3rd person 
accusative clitics in Romance look alike morphologically and behave alike in 
syntactic and semantic respects. The morphological paradigms of determiners and 
(3rd person) clitic pronouns extensively overlap in SMG (Mavrogiorgos 2009:19–
20, Tsimpli & Stavrakaki 1999:36), Galician (Uriagereka 1995:81) and Italian 
(Guasti 1993/4:18) inter alia. 
 
 
 
 
 
 
 
 
 
 
 
Table 2.3: The morphological paradigm of definite articles in CG. 
 
 3rd person strong pronouns in CG and SMG alike pattern like real 
demonstratives. Mavrogiorgos (2009:26, footnote 14) and Tsimpli and Stavrakaki 
(1999:37) describe their properties as follows. First, they can be used either as 
pronouns (Tutos irte “This came”) or as adjectives (Tutos o psilos irte “This, the 
tall one, came”). Second, they may have a deictic or an anaphoric use. In their 
anaphoric use, as shown in (8), the referent is provided by the linguistic context. 
In their deictic use, their referent is identified by various indices in the extra–
linguistic context, such as locative adverbs dhame (‘here’) and dzame (‘there’) 
used together with strong pronouns; the former is used together with tutos (9), 
while the latter is used together with dzinos (10).  
 
(8) O     Stavros              en       tutos            o     psilos.  
The  Stavros–NOM  is–3S  him–NOM  the  tall–NOM   
“Stavros, he is the tall one” 
Number / Case Masculine Feminine Neuter 
Singular 
Nominative 
Genitive 
Accusative 
 
o 
tu 
to(n) 
 
i 
tis 
ti(n) 
 
to 
tu 
to 
Plural 
Nominative 
Genitive 
Accusative 
 
i 
ton 
tus 
 
i 
ton 
tes 
 
ta 
ton 
ta 
  
42 
 
(9) Tutos         dhame  irten          protos.  
He–NOM  here      came–3S  first–NOM  
“He over here came first” 
(10) Dzinos       dzame  irten         protos.  
 He–NOM  there    came–3S  first–NOM 
“He over there came first” 
 
As for the morpho–syntactic status of clitic pronouns in CG, it is well 
known that, cross–linguistically, clitics are both word–like and affix–like 
elements6. Clitics in CG pattern with their counterpart forms in Romance and 
SMG: they have at the same time properties typically related to word–level 
elements, such as the ability to bind antecedents, as well as properties typically 
found in affixes, such as the application of phonological rules in clitic–verb 
combinations. Crucially, CG clitics and SMG clitics alike can co–occur with full 
DPs in Clitic Doubling (CD) constructions. This has been used as an argument in 
favour of the claim that Greek clitics are not real arguments but agreement 
markers. Of course, being an agreement marker does not entail an affix–like 
status, since not all agreement markers are bound morphemes.  
Philippaki–Warburton and Spyropoulos (1999) argue against the affixal 
status of clitics in SMG on the basis of phonological and morphological evidence 
and claim that clitics behave as the arguments proper of the clause. Revithiadou 
and Spyropoulos (2008) provide further evidence for the argumental status of 
clitics on the basis of CD constructions. In particular, they argue that the clitic–
doubled DP objects are out–of–cycle elements and do not exhibit argument 
properties7, while the object role is undertaken by the clitic. 
                                                
6 The idea that clitics are affixes (either word–level or phrasal–level) and, hence, morphological 
parts of their hosts constitutes the basic tenet of lexicalist accounts. Syntactic accounts, on the 
other hand, do not consider clitics parts of their hosts, but rather independent words or morphemes 
(Mavrogiorgos 2009:61–62; see section 2.2.3 for an argumentation in favour of syntactic 
accounts). 
7 Revithiadou and Spyropoulos (2008) claim that the following facts justify their claim that clitic–
doubled DP–objects have been deprived of their argument status: first, they cannot be focalised. 
Second, no extraction can be manifested from within a clitic–doubled DP–object. Third, 
phonological evidence shows that the clitic–doubled DP–object is mapped onto a separate 
prosodic constituent from that which the verb is mapped onto. 
  
43 
 
Given the close resemblance between CG and SMG clitics at the 
morphological level, their similar behaviour in Clitic Doubling constructions, and 
the syntactic and semantic properties they share (as will be made obvious in 
subsequent sections), I take Mavrogiorgos’ (2009) and Revithiadou and 
Spyropoulos’ (2008) claim that SMG clitics are independent words / morphemes 
to hold for CG clitics as well. I now turn to their distribution. 
 
2.2.2. Distribution 
 
CG clitics (both single and clitic clusters) are obligatorily verb adjacent and no 
other element can appear between the clitic and its host (11–12). Hence, CG lacks 
interpolation, found in languages such as Serbo–Croatian (Terzi 1999b), where 
the clitic and the verb may be separated by some other element. Two patterns of 
clitic placement are available in CG: proclisis and enclisis, exemplified in (12) 
and (11) respectively. In this chapter, the notions of proclisis and enclisis have 
only descriptive value and refer to clitic placement with respect to the verbal host: 
the former refers to the pattern with the object clitic immediately preceding the 
finite verb and the latter refers to the pattern with the object clitic immediately 
following the finite verb.  
 
(11) Eferen           (*amesos        /  *i    Maria)           to 
Brought–3S   immediately      the  Mary–NOM  it–CL.ACC      
(amesos / i Maria).  
 “S/he (Mary) (immediately) brought it” 
(12) Prepi  na  (*amesos / *i Maria)  to                  feri           (amesos / i Maria)    
Must  M                                      it–CL.ACC  bring–3S  
“S/he (Mary) must bring it (immediately)” 
 
If we consider the requirement of CG clitics for verb adjacency, it 
straightforwardly follows that they cannot appear in isolation. As shown in 
examples (13–14), only strong pronouns or full NPs, but crucially not clitics, can 
  
44 
 
be involved in verb–less clauses used as answers in narrow–focus questions, even 
if their antecedents are established in the discourse. 
 
(13) Ti       eferes? 
What  brought–2S 
“What did you bring?” 
(14) *To              /  Tuto           /  To   aftokinito. 
it–CL.ACC  /  this–ACC  /  the  car–ACC  
“It (*CL / strong pronoun) / The car” 
 
As already mentioned in chapter 1, clitic pronouns in CG can appear in 
Clitic Doubling (CD) and Clitic Left (CLLD) and Right Dislocation (CLRD) in 
which they double overt DP–objects. Just as in the counterpart structures in SMG 
(Revithiadou & Spyropoulos 2008), the clitic–doubled DP–object can appear in 
clause initial, medial or final position (15). 
 
(15) (To  vivlio)          eferen           to      (to vivlio)  i      Maria          (to vivlio) 
The  book–ACC  brought–3S  it–CL.ACC         the  Mary–NOM   
“As for the book, Mary brought it” 
 
Unlike strong pronouns and full NPs, clitic pronouns cannot focalise. In 
terms of structure building, this implies that they never adjoin to a Focus 
projection. As has been argued in the previous chapter, CG lacks syntactic focus 
movement and the CG CP does not involve a Focus head (Agouraki 2010). In CG, 
in lieu of focus constructions, it–clefts are employed, while structures involving 
stressed constituents in the left periphery of the clause are also licit. Examples 
(16) and (17) constitute constructions involving stressed constituents and it–clefts 
respectively, and reveal the discrepancy between strong pronouns and full NPs, 
which are licit in both construction types, and clitic pronouns, which are banned 
from both. Moreover, as shown in (18) and (19), clitics are not compatible with 
focalised / stressed strong pronouns or full NPs involved in CD (19), CLLD (18) 
  
45 
 
or CLRD (19) constructions, as they cannot double focalised / stressed strong 
pronouns or full NPs (see Theophanopoulou–Kontou 1986/1987, Tsimpli 1995). 
 
(16) *TO             /  TUTO        /  TO  VIVLIO       thelo. 
it–CL.ACC  /  this–ACC  /  the  book–ACC  want–1S  
“It / This / The book I want” 
(17) En      *TO  /  TUTO  /  TO  VIVLIO  pu          thelo. 
Is–3S                                                     COMP  want–1S 
“It is it / this / the book that I want” 
(18) *TO  VIVLIO      /  To  vivlio  thelo         to. 
The     book–ACC                     want–1S  it–CL.ACC 
“The book, I want it” 
(19) Thelo  to  ([PAUSE])8  *TO  VIVLIO  /  to  vivlio. 
“I want it (,) the book” 
 
However, a clarification needs to be made in relation to clitics’ 
incompatibility with emphatic stress. Despite the fact that clitics are typically 
unstressed elements and cannot focalise or be used emphatically, under certain 
conditions they can be semantically focused and hence can bear emphatic stress. 
This can only happen when they are involved in constructions under the 
repair/correction interpretation, such as (20) for instance. Notably, in order for this 
type of construction to be felicitously uttered, both of its constituent parts must be 
present, i.e. the part including the old/known information (“na MU pis”) as well as 
the corrected part (“na TIS pis”). 
 
(20) Na  TIS                 pis,        oi       na  MU                pis. 
M   her–CL.GEN  tell–2S  NEG  M  me–CL.GEN  
“You should tell HER, not me”  
 
                                                
8 The presence or the absence of a pause immediately after the left dislocated NP renders the 
structure a CLRD or a CD construction (see chapter 1 for a discussion of Anagnostopoulou’s 
(2006:546!547) diagnostic for differentiating between the two). 
 
  
46 
 
Mavrogiorgos calls this type of focus “meta–grammatical” (2009:56, 
footnote 40), and he indicates that its difference from contrastive focus is that only 
the latter seems to have semantic properties. Notably, this type of focus may apply 
to sub–parts of words as well, as in “TheLIS, oi theLUN” ((You) want, not (they) 
want), which shows that there is no requirement for the focused constituent to be 
inherently stressed.  
 
2.2.3. Coordination and Modification 
 
Clitic pronouns in CG cannot be coordinated and cannot be modified, unlike 
strong pronouns and full NPs which pattern alike in this respect. Examples (21) 
and (22) illustrate the behaviour of clitics versus strong pronouns and full NPs 
with regard to coordination and modification respectively. This divergence 
between strong pronouns and full NPs, on the one hand, and clitic pronouns on the 
other has led Cardinaletti and Starke (1999) and Mavrogiorgos (2009) to include 
coordination and modification among the syntactic criteria for the classification of 
pronominal elements.  
 
(21) Ksero         tuton          /  to   mitsin         /  *to                    dze  
Know–1S  him–ACC  /  the  boy–ACC  /   him–CL.ACC  and  
tutin          /  tin   koruan       /  *tin. 
her–ACC  /  the  girl–ACC  /   her–CL.ACC  
“I know him (strong pronoun) / the boy / *him (CL) and her (strong 
pronoun) / the girl / *her (CL)” 
(22) Mono  tuton          /  to    mitsin        /  *to                    ksero.         
Only    him–ACC  /  the  boy–ACC  /  him–CL.ACC  know–1S 
“I only know him (strong pronoun) / the boy / him (CL)” 
 
2.2.4. Semantic properties 
 
CG clitics exhibit similar semantic properties to SMG clitics (Mavrogiorgos 2009, 
Tsimpli & Stavrakaki 1999): they can be either referential or non–referential, 
  
47 
 
while they can be also used deictically. As adequately described by 
Mavrogiorgos, in their referential use clitic pronouns are anaphoric, “in the sense 
that they pick up their referent via a prominent antecedent, which is provided by 
the previous immediate linguistic context” (2009:33), as in example (23). Tsimpli 
and Stavrakaki note that clitics, due to their lack of a referential index, refer by 
virtue of their ability to point (via indexing) to an antecedent, which, in turn, 
refers to an individual in the discourse, in Cardinaletti and Starke’s (1999) terms 
(1999:39). Clitics may be also used deictically, as in example (24), in which case 
their referent is identified in either of the following ways: the utterance may be 
accompanied by pointing or the antecedent may be sufficiently salient in the 
discourse (Tsimpli & Stavrakaki 1999:38).  
 
(23) O     Stavros              en       o     andras                mu.  
The  Stavros–NOM  is–3S  the  husband–NOM  my–POSS 
Eksanaghnorises  ton ? 
Re–meet–2S         him–CL.ACC 
“Stavros is my husband. Have you met him again?” 
(24) Kseris        tin? 
Know–2S  her–CL.ACC 
“Do you know her?” 
 
In relation to the anaphoric and deictic uses of the clitic pronouns, it is 
worth mentioning Heim and Kratzer’s (1998) contention that they are not really 
distinct, given that in both cases the clitic is assigned its referent via co–reference 
with a prominent antecedent (cited in Mavrogiorgos 2009:33, footnote 20). This 
ties in with the inherent property of clitic pronouns that they refer to entities 
already established in the discourse (Tsimpli & Stavrakaki 1999:38). It is in this 
sense that Mavrogiorgos considers them topicalizers. In his own wording, a clitic 
constitutes a definite head “optionally merged in the left periphery of the (direct 
or indirect) object DP of the clause rendering it a topic” (2009:9). 
CG clitics can also be non–referential. Mavrogiorgos (2009) mentions 
three types of non–referential uses in SMG (see also Tsimpli & Stavrakaki 
  
48 
 
1999:38), all of which are operative in CG as well. CG clitics may be used as 
variables bound either by non–referential antecedents, such as quantifiers as in 
(25), or by referential antecedents, such as proper nouns9. This is illustrated in 
example (26), where the clitic in the elided conjunct may receive, apart from a 
strict reading, a sloppy identity reading. In the former case, it will be bound by the 
subject of the first clause (a), while in the latter case, it will be bound by the index 
of the subject of the second clause (b). In addition, 3rd person clitics can be used 
in idioms. In these uses, the clitic alone has no antecedent, but it combines with 
the verb to form an idiom (27–29). Some expressions may be ambiguous between 
an idiomatic and a literal interpretation (28–29), while for others the only 
available interpretation is the idiomatic one (27). 
 
(25) Kathe    athropos     kamni    dzino         pu          ton                  /  tu                      
Every  man–NOM  do–3S   this–ACC  COMP  him–CL.ACC  /  DAT  
simferi. 
is–3S in his interest 
“Every man does what is in his interest” 
(26) O     Stavros              ipen        oti         ton                   peripezi.    
The  Stavros–NOM  said–3S  COMP  him–CL.ACC  fool–3S    
Dze   o    Marios             to     idhio. 
And  the  Marios–NOM  the  same 
a. “Stavrosi said that s/he fools himi. And Mariosii also [=said that s/he 
fools himi]” (strict) 
b. “Stavrosi said that s/he fools himi. And Mariosii also [=said that s/he 
fools himii]” (sloppy) 
(27) Idhen      tin                   arhighos. 
Saw–1S  her–CL.ACC  chief–NOM 
“S/he behaves as if s/he is the chief” (idiomatic)  
(28) Ipan           mas               tin. 
Told–3PL  us–CL.GEN  her–CL.ACC   
                                                
9 See Theophanopoulou–Kontou (1986/1987) and Tsimpli (1999) for a discussion on the 
distribution of clitic pronouns and null pro in restrictive and non-restrictive relatives and 
interrogatives. 
  
49 
 
a. “They told it to us” (literal)  
b. “We were told off (by them)” (idiomatic) 
(29) Epira        ta. 
Took–1S  them–CL.ACC   
a. “I took them” (literal)  
b. “I was furious” (idiomatic) 
 
Human reference and coordination are the two criteria that have been 
proposed by Cardinaletti and Starke (1999) for pronoun classification. According 
to them, both strong and clitic pronouns fall into the following two pronoun 
classes, regardless of whether they are morphologically distinct: class 1 pronouns 
can be coordinated but can only have human referents, while class 2 pronouns 
cannot be coordinated but can have both human and non–human referents. Class 1 
pronouns may only be strong pronouns, while both weak and clitic pronouns fall 
into class 2. Clitic pronouns in CG cannot be coordinated, while their referents 
can be either human or non–human entities. Therefore, CG clitics fall into class 2 
of Cardinaletti and Starke’s classification, together with SMG clitics 
(Mavrogiorgos 2009:26). 
The criteria for pronoun classification proposed by Cardinaletti and Starke 
(1999) are not applicable to all clitic pronouns in CG. 1st and 2nd person 
pronouns have the inherent property of realising human entities alone, as they 
refer to the speaker and the addressee respectively, thus the human reference 
criterion is not applicable in their case. However, if the aforementioned criteria for 
inclusion in class 2 are applied to 3rd person CG clitics, both are fulfilled.  
In sum, the semantics of clitic pronouns in CG can be characterised as 
follows: they can be referential or non–referential, can refer both to human and 
non–human entities, they can be bound variables or free pronouns and can be used 
idiomatically, and their antecedents must be prominent in the discourse. 
 
 
 
 
  
50 
 
2.2.5. Phonological properties 
 
Focusing on the phonological properties of CG clitics, I will try to illustrate the 
properties that determine their realisation as parts of a Prosodic Word (henceforth 
PrW). Following Revithiadou (2006, 2008), I will first briefly outline some basic 
aspects of CG phonology10 and will then focus on the prosodic properties of CG 
clitics. 
The description I give of the prosodic system in CG, largely based on 
Revithiadou (2006:83–84), focuses on various aspects of the phonological 
component at the lexical and post–lexical level. At the lexical level, CG, like 
SMG, has a three–syllable stress window, with default stress on the 
antepenultimate syllable, as in [eka'tharisen] (“(s/he) cleaned”), unless a lexically 
determined stress applies on the (pen)ultimate syllable, as in [katha'ros] (“clean”). 
Moreover, this rule states that no PrW may be stressed beyond the 3rd syllable 
from the right. An important difference between CG and SMG regarding the 
three–syllable stress window is that in SMG it is never violated. In contrast, in CG 
it may be violated, even within a single lexical word, as for example in the proper 
name Hadzikiriakos, that may be pronounced as follows [ha'dzikiriakos]. When 
CG clitics incorporate in their host’s PrW (Revithiadou 2006, 2008), the three–
syllable stress window rule is applicable, hence verb–clitic combinations also 
adhere to it11. In effect, they occasionally trigger a secondary/rhythmic stress in 
order to avoid violation. 
At the post–lexical level, a number of sandhi rules are applicable in CG. 
Such rules only apply within the Prosodic Word (PrW) domain and involve s–
voicing assimilation and e–deletion. The following sets of examples (Revithiadou 
                                                
10 With regard to the phonetics of CG, Arvaniti (2010:23–24) points out three characteristics 
which distinguish it from SMG. The first of these is the phonemic contrast between geminate and 
single consonants in CG, which is absent from SMG. For example, observe the contrast between 
the phoneme [n] in enna (modality marker used for future formation in CG; see chapter 1), 
pronounced as a geminate consonant, and in the numeral ena (“one”), pronounced as a single 
consonant. The second is the wide use of the prenasalised voiced stops [mb], [nd] and [!g] in CG 
that usually replace the SMG voiced stops [b], [d], and [g], respectively. The third is the post–
alveolar fricatives ["] and [#], which are part of the phonetic inventory of CG alone. 
11 Cf. Arvaniti (2010) who claims that while in SMG a word stressed on the antepenultimate 
syllable acquires an enclitic stress when followed by a clitic, CG lacks the enclitic stress of SMG; 
she gives the following example from CG: [tis ta'ftotitas tu] “his identity” (2010:25).  
  
51 
 
2006:84) illustrate this point. The first set shows the application of s–voicing, 
namely the assimilation of a voiceless /s/ to the [+voice] value of the following 
fricative or nasal. The following PrWs involve the voicing of [s] so that it is 
pronounced as [z] in front of [m] and [v]: zmili (“needle”), izmini (“Ismini”, 
proper name), prozvallo (“offend”), gherazmenos (“aged”). The next set of 
examples shows the application of e–deletion, namely the deletion of the first 
unstressed /e/ in a sequence of occurrences of the same phoneme, provided that 
the resulting cluster is phonotactically well–formed: puliste (<pulis–ete “sell”), 
ferte (<fer–ete “bring”), but ghrapsete/*ghrapste (<ghraps–ete “sell”). Both the 
above rules are applicable in clitic constructions as well.  
Turning now to CG clitics, I start from their well–known properties, 
namely their unstressed and monosyllabic status, and discuss some issues in 
relation to the former. Even though clitics are typically unstressed, under certain 
conditions they may appear stressed. One such case is when they are involved in 
corrective constructions; see discussion in section 2.2.2 and example (20).  
Another case in which clitics bear stress is when hiatus resolution applies. 
Hiatus resolution is the coalescence of the final vowel of the clitic and the initial 
vowel of the verb. The clitic–verb cluster in the following example /en tu to 
edhoka/ (NEG him–CL.GEN it–CL.ACC gave) may be pronounced as follows: 
['en.tu.to.'e.!o.ka]. However, if hiatus resolution is applied, vowel deletion (of 
phoneme [e] heading the initial syllable of the verb) takes place followed by 
prosodic restructuring, and the clitic–verb cluster is pronounced as follows: 
['en.tu.'to.!o.ka]. In the latter case, the clitic ends up carrying the lexical stress of 
the PrW within which it is realised together with its verbal host. 
Crucially, the clitic does not bear inherent/lexical stress either in corrective 
constructions or in utterances in which hiatus resolution applies. In corrective 
constructions it bears meta–grammatical focus, which, as was previously 
mentioned, is not contingent on the availability of inherent stress. In clauses 
involving hiatus resolution, stress shift applies from the verb to the clitic upon the 
application of a phonological process.  
CG clitics, on a par with their SMG counterparts, are monosyllabic and 
morpho–phonologically deficient elements (in terms of syllable structure), hence 
  
52 
 
they do not form a foot. In effect, they lack inherent/lexical stress. On the basis of 
the Metrical Stress Theory (Liberman & Prince 1977), stress is assigned to a 
syllable within a foot. Moreover, clitic pronouns in CG cannot form PrWs, which, 
according to the Prosodic Hierarchy (Selkirk 1995), are the minimal prosodic 
constituents onto which a morpho–syntactic constituent may be mapped. PrWs 
must at least contain a foot, hence the inability of CG clitics to form PrWs. Since 
CG clitics (along with their SMG counterparts) cannot form PrWs on their own, 
they obligatorily restructure into a (verbal) host. 
On the basis of the discussion so far, I assume that CG clitics are 
inherently unstressed and prosodically deficient elements, which need to cliticise, 
and hence prosodically restructure into their host. Evidence for the fact that CG 
clitics form a PrW together with their host is offered by the application of sandhi 
rules in clitic–host combinations. Recall that sandhi rules are only applicable 
within PrWs. S–voicing for example can apply in clitic–verb clusters, either in 
proclisis (30) or in enclisis (31) contexts. 
 
(30) Enna  mas               mundari  (  !  mazmu'ndari).  
M       us–CL.ACC  attack–3S   
“S/he will attack us” 
(31) Edhokes   mu  (  !  'edhokezmu).  
Gave–2S  me–CL.GEN 
“You gave me” 
 
E–deletion in CG typically applies to the ending of the 2nd person singular 
of the imperative verb (stress–less vowel [e]), when followed by a clitic pronoun, 
as in (32). This process may also be combined with s–voicing, as in (33). E–
deletion is an optional process in SMG (Mavrogiorgos 2009), whereas in CG it is 
subject to the phonotactic well–formedness of the resulting cluster; compare 
['afisto] (“leave it”), ['thkjavasto] (“read it”) and ['ferto] (“bring it”) with ['kapse 
to] (“burn it”), ['kopse to] (“cut it”) and ['filakse to] (“hide it”). Nevertheless, e–
deletion seems to have wider application on verb–clitic clusters in SMG than in 
CG: compare ['kopsto], ['kapsto] and ['filaksto] SMG/*CG. 
  
53 
 
(32) Fer (<fere)  mu                  lefta. 
Bring–2S    me–CL.GEN  money–ACC 
“Bring me money” 
(33) A(f)ise      mu ( !  'a(f)izmu)  lefta. 
Leave–2S  me–CL.GEN          money–ACC 
“Leave money for me” 
 
As Revithiadou (2006:84) observes, in CG (and SMG alike) a 
secondary/rhythmic stress is developed to repair violations of the three–syllable 
stress window caused by the addition of a clitic pronoun. Specifically, apart from 
the stress on the syllable prior to the antepenultimate syllable, a secondary stress 
is applied to the penultimate syllable, either the last syllable of a polysyllabic verb 
or the 1st syllable of the enclitic. The following examples taken from Revithiadou 
(2006:84) illustrate this point: /ekames to/ (“(you) did it”) parsed as ['eka'mes to], 
/ipe mu to/ (“(s/he) told me”) parsed as ['ipem 'mu to]. The former constitutes a 
single clitic and the latter a double clitic construction. The secondary/rhythmic 
stress is also triggered in noun–possessive pronoun combinations, as in /dhaskalos 
mu/ (“teacher my”) realised as ['dhaska'lozmu]. Notably, as indicated by 
Revithiadou, it is a matter of dialectal variation which stress peak is more 
prominent (2006:84, note 6). 
With regard to the prosodic organization of proclitics and enclitics in CG, 
a detailed description within Revithiadou’s (2006, 2008) approach is presented in 
chapter 3. The basic assumption is that CG clitics are prosodically organised in 
the following three ways: 
1. As internal enclitics: post–verbal clitics incorporate into the PrW of the verbal 
host; 
2. As affixal proclitics: pre–verbal clitics recursively adjoin with the unstressed 
function word into the PrW of the verb; 
3. As prosodic words: pre–verbal clitics incorporate into the PrW of the preceding 
stressed function word. 
It has been suggested, on the basis of acquisition data, that enclitics are 
more salient than proclitics (Mastropavlou et al. 2014).  Both cross–linguistic and 
  
54 
 
cross–dialectal evidence are justifying this suggestion. In particular it has been 
shown that post–verbal clitics emerge earlier than pre–verbal ones in early 
Romanian (Avram & Coene 2007) and in early SMG (Stephany 1997, Tzakosta 
2003, 2004a, 2000b). It has also been shown that the production of clitic pronouns 
in impaired populations in a language/variety that manifests predominantly the 
enclisis pattern is higher than clitic production in impaired populations in a 
language/variety that manifests predominantly proclisis (Mastropavlou et al. 
2014), which arguably again points to the salience of enclitics over proclitics. On 
the basis of these studies, the “salience” of the enclisis pattern over the proclisis 
pattern denotes the earliest emergence and the higher degree of accessibility of the 
former (as manifested in language–impaired populations) with no further 
theoretical implications at the discourse level. I will return to the issue of saliency 
in Chapter 7, in which I will elaborate in the light of new evidence from the L1A 
of clitics in CG.  
 
2.2.6. Phrase structure status 
 
An issue that has received a lot of attention in the generative literature on 
pronominal clitics is their syntactic status, namely whether they constitute 
maximal projections (XP) or heads (X0). This is related to the two kinds of 
approaches discussed in the literature on Romance cliticisation: the DP 
Hypothesis and the Affix Hypothesis. According to the former, clitics are 
generated as heads of DPs and they then move to a functional projection 
(Uriagereka 1995). Within the latter approach, clitics are considered as affix–like 
elements merged under a functional head (Duarte & Matos 2000). Focusing on 
CG, all the syntactic accounts so far proposed (Agouraki 2001, Terzi 1999a, 
1999b) assume that clitics in CG are heads. As for SMG, Mavrogiorgos (2009) 
has convincingly argued that SMG clitics have a number of head properties. 
I will apply the four diagnostics used by Mavrogiorgos (2009:57–60) for 
the “headness” of SMG clitics to clitics in CG. These diagnostics are based on 
arguments put forward by Cardinaletti and Starke (1999) to test the head 
properties of clitics cross–linguistically. The first is their ability to participate in 
  
55 
 
CD and CLLD (15) constructions. The second is that they should move together 
with their verbal host (34) and the third that they do not move across other heads; 
these properties trivially follow from their requirement for verb adjacency. Fourth, 
they should get deleted along with the auxiliary verb (35). Crucially, it is not 
possible for clitics to get deleted individually, no matter whether their verbal host 
is an auxiliary or a main verb (36). This is only possible when the verb of the 
second conjunct can be used intransitively as well, and if the clitic is deleted, a 
different interpretation arises (37).  
 
(34) To   vivlio           (akoma)  dhen   to               (*akoma)  eferen  (akoma).  
The  book–ACC  yet          NEG  it–CL.ACC                 brought–3S 
“The book, s/he hasn’t brought it (yet)” 
(35) Ichen      ta                        kanonismena  
Had–3S  them–CL.ACC  arranged–PP.ACC   
dze  simfonimena        pu     dzeron. 
and  agreed–PP.ACC  from  long time–ACC  
“S/he had them arranged and agreed a long time ago” 
(36) Eghorasa     to                  dze  efera          *(to)12.  
Bought–1S  it–CL.ACC  and  brought–1S 
 “I bought it and brought it” 
(37) Ethkjavasa  tin                   dze  ipoghrapsa13.  
Read–1S     her–CL.ACC  and  signed–1S 
 “I read it and signed” 
 
I will return to the phrase structure of CG clitics and their head properties in 
Chapter 3, in which I discuss the formal accounts for cliticisation in CG that have 
so far been proposed.  
 
                                                
12 In order for a grammatical (or the intended) interpretation to arise, clitics cannot be deleted 
individually and they should surface in the 2nd conjunct as well. In languages such as European 
Portuguese, in which clitics may be deleted in conjunction, clitics have been argued to have 
phrasal status (XP) (Mavrogiorgos 2009:69). In CG, however, only the verb–clitic cluster can be 
deleted. This confirms that CG clitics are not phrasal elements. 
13 The referent of the clitic tin could be i dhiatrivi (the thesis), while the verb in the second 
conjunct is interpreted as having a distinct referent, i.e. some other document. 
  
56 
 
2.3. Clitic placement in CG 
 
This section aims to describe clitic placement in CG. Enclisis and proclisis 
contexts are presented, as well as contexts allowing free clitic ordering. This 
overview summarises the descriptive sections of a number of papers (Agouraki 
2001, Chatzikyriakidis 2012, Revithiadou 2006; 2008, Terzi 1999a; 1999b), and 
also highlights particularly interesting data. 
CG is a language that adheres to the Tobler–Mussafia law, whose basic 
tenet is that clitics are banned from clause initial position14. Tobler (1875/1912) 
and Mussafia (1886, 1898) first recognised this pattern of clitic placement in 
medieval stages of Romance languages (Revithiadou 2006:80, footnote 2). 
Following Mavrogiorgos (2012), I assume that clitic languages fall within three 
categories with regard to the factors “regulating” the proclisis–enclisis alternation: 
(a) Finiteness–sensitive languages, in which the enclisis–proclisis 
alternation depends on the finiteness of the verbal host15. 
 (b) Tobler–Mussafia languages, in which clitics are banned from clause 
initial position and clitics appear pre– and post–verbally depending on the 
syntactic context. 
 (c) Languages exhibiting second position restrictions, or the so–called 
Wackernagel pattern, in which clitics typically occupy the 2nd (or Wackernagel) 
position in the clause. 
CG is a Tobler–Mussafia language, hence clitics cannot appear in the first 
position of the clause, but may appear in various other positions, as illustrated in 
(38). CG clitics must be verb–adjacent, and therefore they obligatorily precede or 
                                                
14 Note that in some Tobler–Mussafia languages (mainly old Romance), clitics may occupy clause 
initial position; see examples (19a) and (19b) in Mavrogiorgos (2012), taken from Fontana (1996) 
cited below as (1) and (2). 
1. S’est il donques corrouciez a nos? [Old French] 
              “Himself is he then vexed with us” 
2. Oy dia en Grecia lo    traen        por… [Old Spanish] 
Today  in Greece it.cl bring.3pl for… 
“Today in Greece they bring it as…” 
15 Mavrogiorgos (2009) suggested that proclisis in SMG correlates with non–restricted/full person 
agreement on T, while enclisis correlates with restricted person agreement on T. 
  
57 
 
follow the finite verb16, regardless of their linear order with respect to the rest of 
the constituents. Their placement is contingent on the syntactic context. The 
remainder of this section presents the relevant contexts for the manifestation of 
the enclisis and the proclisis pattern. 
 
(38) (I      (pio  poli)  fitites                 (mu))          (shedhon)  
The  most           students–NOM  my–POSS  almost 
eteljosan          tin  (tin  erghasia). 
finished–3PL  her–CL.ACC the  essay–ACC 
“Most of my students almost finished it (the essay)” 
 
2.3.1. Enclisis and proclisis contexts 
 
Declarative clauses in CG involve a post–verbal clitic, unlike Romance languages 
and SMG. This divergence between CG and SMG with respect to clitic placement 
is reminiscent of the divergence between Bulgarian and Macedonian: compare 
examples (39–40) (Franks & Bo!kovi" 2001:175) with (41–42). 
 
(39) (Bg: */ Mac: OK)  Mi                   go                 dade  Petko  vcera. 
       Me–CL.DAT  it–CL.ACC  gave  Petko  yesterday 
(40) (Bg: OK / Mac: *)  Dade mi go Petko vcera. 
(41) (CG: * / SMG OK) Mu                  to                 edhose  o Petros   (e)htes. 
        Me–CL.GEN  it–CL.ACC  gave     the Peter  yesterday 
(42) (CG: OK / SMG: *)  Edhoke mu to o Petros ehtes. 
 
CG clauses do not typically involve an overt subject, since CG is a pro–
drop language. Yet, even if an overt subject is realised, the clitic follows the finite 
verb (43), provided that the subject is not emphatically or contrastively focused 
(44). The pattern of clitic placement in yes–no questions resembles clitic 
placement in declaratives, with the clitic appearing post–verbally (43), unless an 
                                                
16 CG has no infinitives, while gerundive constructions, which are assumed to involve a defective 
person feature in SMG (Mavrogiorgos 2009), are occasionally used in CG (see next section), but 
not productively. 
  
58 
 
emphatically or contrastively focused subject is involved, in which case the clitic 
precedes the finite verb (44). The enclisis pattern is also manifested in imperative 
clauses (45); the imperative clause is one of the two types of structures (the other 
being gerundive constructions) in which both CG and SMG manifest enclisis.  
 
(43) Esi               eferes           to (?).                    
You–NOM  brought–2S  it–CL.ACC   
“You brought it / Did you bring it?” 
(44) ESI to eferes (?). 
“It was / Was it you who brought it?” 
(45) Fer(e)       to.  
bring–2S  
“Bring it!” 
 
Gerundive constructions are occasionally used in affirmative clauses in 
CG (46) and SMG alike. However, gerunds are not productively used in CG 
(Terzi 1999a:115, footnote 27): they do not have across the board application with 
different verbs and they sound unnatural when negated, unlike SMG gerunds. It 
is, thus, not surprising that none of the accounts of CG cliticisation takes into 
consideration clauses involving gerunds. 
 
(46) (*Mi)  Fkalondas  ton                    pu      ti    mesi,  
 NEG  taking        him–CL.ACC  from  the  middle–ACC    
enna  (e)shis     to    pedhion       elefthero. 
M      have–2S  the  field–ACC  free–ACC 
“(Without) Taking him out of the way, you’ll have the field free” 
 
At this point, it is worth discussing the behaviour of CG enclitics in 
indicative clauses when combined with certain suffixes in order to illustrate their 
status in relation to their verbal host. Specifically, I will offer some examples to 
show the incompatibility of CG clitics with certain suffixes. The first suffix with 
which CG enclitics can never co–occur is the 1st person plural suffix –te. The 
  
59 
 
suffix –te can be used with complement–less verbs, like eferamente (bring–
1PL.+PAST+PERF.nte), as well as with overt objects (47), or proclitics (49). 
Crucially, though, –te is incompatible with enclitics (48). According to Agouraki 
this suffix seems to behave as an “optional marker of V–in–C checking Emphasis 
on C” (2010:538, footnote 5)17. The second suffix with which enclitics are 
incompatible is –usi/–asi, which is a different ending for the 3rd person plural and 
is used interchangeably with –un in [–PAST] tense, e.g. fernun / fernusi (“bring”), 
and –an in [+PAST] tense, e.g. efer(n)an / efer(n)asi. The suffix –usi/–asi 
behaves similarly with –te, in that it can co–occur with overt objects (47) or 
proclitics (49) but not with enclitics (48). This behaviour of CG enclitics is 
indicative of their morpho–phonological status and the way they prosodify to their 
verbal host; I return to this in the next chapter. 
 
(47) Eferamente           /  Eferasin         to    krevati. 
brought–1PL.nte  /  brought–3PL  the  bed–ACC 
“We/They brought the bed” 
(48) *Eferamente / *Eferasin to (CL.ACC).  
“We/They brought it” 
(49) Na to ferumente / ferusi.  
“We/They should bring it” 
 
Turning now to CG proclitics, there are a number of environments in 
which CG clitics can only surface pre–verbally, immediately preceding the finite 
verb, patterning with the order manifested by SMG. CG clitics are preverbal when 
preceded by: 
1. Modal particles 
2. Wh–elements 
3. Negative particles 
4. The factive complementizer pu and complementizers an/otan/ama/afu 
5. Stressed constituents in the left periphery of the CG CP 
                                                
17 Agouraki (2010) postulates that in CG the specification of sentential force on C has to be 
checked in overt syntax. 
  
60 
 
These are discussed in the following subsections. 
 
2.3.1.1. Modal particles 
 
CG makes use of a number of modal particles, namely na, as, tha and enna18, 
which surface at the CM head, a C head specified for modality (Roussou 2000); 
na and as may move to COp as well. When clitic constructions are headed by na, 
as, tha or enna, the clitic obligatorily appears pre–verbally, as in (50). 
 
(50) Na  /  As  /  Enna  /  Etha  (<(dh)e+tha)  to                 fero.  
M                                           NEG  M    it–CL.ACC  bring–1S 
“I should / Let me / I will / won’t bring it”  
 
2.3.1.2. Wh–elements 
 
Matrix or embedded interrogative clauses19 headed by wh–elements require a pre–
verbal clitic in CG. This includes wh–argument questions (51) and quasi–
argument questions (52). Wh–elements in CG may optionally conjoin with the 
dialectal element embu, with no alternation with respect to clitic placement, while 
bare and argumental inda (“what”) obligatorily conjoins with mbu (53). Wh–
exclamatives also require pre–verbal clitic placement (54). 
 
(51) (Erotisen)   Pkjos            (embu)  to                  efere? 
Asked–3S  Who–NOM                it–CL.ACC  brought–3S 
“(S/he asked) Who brought it?” 
(52) (Erotisen)  Pothen  /  Pu  /  Pote  /  Indalos  (embu)  /  Pos  /  Jati  to  efere? 
                  From where      when    how                       how    why 
 “(S/he asked) From where/When/How/Why did s/he bring it?” 
(53)  (Erotisen)  Inda  *(mbu)  su                    eferen? 
                                                
18 See chapter 1 (1.2) for a thorough discussion on the properties and distribution of as, na, tha 
and enna. 
19 See chapter 1 (1.4) for a discussion on wh–question formation in CG. 
 
  
61 
 
                  What               you–CL.GEN                     
“(S/he asked) What did s/he bring to you?” 
(54)  Pos  /  Indalos  ta                        ekataferen     omos  o    atimos! 
 How                 them–CL.ACC  managed–3S  yet     the  crook–NOM    
“How he managed, yet, the crook!“ 
 
2.3.1.3. Negative particles 
 
In CG, two negation markers are used: the negator (dh)e(n), used with verbs in the 
indicative mood and immediately preceding the future particle tha20, and the 
negator men, used with clauses headed by modal particles. Both (dh)e(n) and men 
trigger proclisis (55–56). 
 
(55) (Dh)en  (tha)  to                 eferen. 
NEG      M     it–CL.ACC  brought–3S  
“S/he didn’t bring / wouldn’t have brought it” 
(56) Enna  /  Na  men    to                 feri. 
M                 NEG  it–CL.ACC  bring–3S  
“S/he will / should not bring it” 
 
The negative particles (dh)e(n) and men may combine with the 
coordinating21 conjunction dze (dialectal equivalent of the SMG conjunction ke) 
to form the clusters endze and mendze respectively. When endze and mendze head 
clitic constructions, enclisis is triggered, as illustrated in (57–58). 
 
(57) Edze  (<en + dze)  eferen  to  /  enna  to  feri. 
            NEG  and   
“S/he didn’t / won’t bring it” 
                                                
20 It usually combines with it to form the fossalized form etha, unless it is preceded by the 
complementizer an (see previous chapter). 
21 Dze can also function as a subordinating conjunction (Agouraki 2001). This is illustrated in 
example (13) in Agouraki (2001:8) repeated below as (1): 
1. Akui        tin                   dze  lali. 
   Hear–3S  her–CL.ACC  and  say–3S 
   “He heard her say” 
  
62 
 
(58) Medze  (<men + dze)  feri to. 
                NEG    and 
“S/he should not bring it” 
 
This pattern can be explained on the basis of the inherent properties of dze 
in CG, which can often conjoin with different particles. Apart from the negative 
particles, a number of other particles including temporal ones like ama, andan and 
oson (“when”) and the question markers ampa and (m)emba can appear conjoined 
with dze. As with the negative particles, while ama, andan, oson, ampa and 
(m)emba generally trigger proclisis, when conjoined with dze, they trigger 
enclisis. I assume, following Agouraki (2001:8–9), that dze typically functions as 
a coordinating conjunction that conjoins two CPs. When combined with a particle, 
the particle appears in the first conjunct, which constitutes an elliptical CP (with a 
missing predicate, possibly en (“is”)). Dze appears under the conjunction head and 
the verb surfaces in the second conjunct. Hence, if a clitic is realised in the second 
conjunct, it will not appear clause initially but will follow the finite verb. 
 
2.3.1.4. The factive complementizer pu and complementizers an/otan/ama/afu  
 
Pu is characterised as a factive complementizer as it follows factive predicates, 
contrasting with complementizers oti and pos which follow non–factive ones 
(Ralli 2006: 128–129)22. Following Roussou (2000), I assume that pu and an 
appear in the higher C head of the clause. I take her claim to hold for 
complementizers otan, ama and afu as well. As shown in examples (59–60), when 
either an/otan/ama/afu or pu head the clause, the clitic is realised pre–verbally. 
 
(59) An  /  Otan  /  Ama  /  Afu    ton                   dhi. 
If   /   When                After  him–CL.ACC  see–3S 
“If / When / After s/he sees him” 
                                                
22 Pu–clauses in Greek are argumental, yet they resist nominalisation (Roussou 1991, 1994). This 
may be due to the fact that pu is lexically specified as nominal. This can be indicated by the 
following facts: first, pu can occur in subject position and second, verbs that do not assign case 
cannot take pu–clauses as their complements (Tsakali 2006:179–180, footnote 2). 
  
63 
 
(60) Eharika               pu         se                     idha. 
Was–1S  happy  COMP  you–CL.ACC  saw–1S 
 “I was happy that I saw you” !
2.3.1.5. Stressed Constituents in the left periphery of the CG CP 
 
Full NPs, strong pronouns and adverbs that receive emphatic/contrastive focus or 
narrow/new information focus appear in the [Spec,CP] position of the CG clause 
(Agouraki 2010). When these constituents head a clause involving a clitic, 
proclisis is triggered. Notably, under the heading “stressed constituents”, a 
number of different phrasal constituents are subsumed. Thus, irrespective of the 
categorial status of the constituent in question, if it is stressed, proclisis is 
triggered. This constitutes evidence for the important role of prosody in clitic 
placement in CG. This is relevant for the discussion of syntactic and prosodic 
accounts of cliticisation in CG outlined in the next chapter. 
 
(61) A: Tuto  EPSES     to                 eferen           i      Maria.  [Narrow Focus] 
     This  last night  it–CL.ACC  brought–3S  the  Mary–NOM 
X: EGHO  to  efera,            oi       i  Maria!  [Contrastive Focus] 
      I                brought–1S  NEG  
“A: This (one) it was only last night that Mary brought it. 
X: I brought it, not Mary!” 
 
2.3.2. Free ordering of pronominal clitics in CG 
 
In some contexts, CG clitic pronouns can optionally appear either pre– or post–
verbally. These contexts include clauses headed by the complementizers oti and 
pos (Terzi 1999b:237, note 7), as illustrated in (62–63), the clausal conjunctions 
jati and epidhi (“because”) and the complementizer afu with the meaning of 
“because” (65–66) (Chatzikyriakidis 2010, 2012). However, if an unstressed overt 
subject is realised in the embedded clause, immediately following the COMP 
(67), the preferred position for the clitic is the post–verbal one (Terzi 1999a:109–
  
64 
 
110), unless the overt subject is stressed (see section 2.3.1.5 for a discussion of 
stressed constituents in the left periphery of the CG CP).  
 
(62) Lali       oti  /  pos  ton                   aghapa. 
Say–3S  that          him–CL.ACC  love–3S  
(63) Lali oti / pos aghapa ton. 
(64) Lali oti / pos dzini aghapa ton. 
                      She–NOM 
“S/he says that she loves him” 
(65) Jati / afu ton aghapa. 
Because 
(66) Jati / afu aghapa ton. 
(67) Jati / afu dzini aghapa ton. 
                     She–NOM 
“Because she loves him” 
 
Mavrogiorgos argues that the free ordering of object clitics in CG oti– and 
pos–constructions (62–63) can be attributed to the realisation of high and low 
complementizers; he also assumes that oti (and subsequently pos) occupies the F–
position only in constructions like (62)23 (2012:35–36). An alternative account is 
possible, however, if we adopt a prosodic or interface account of cliticisation in 
CG. Following ideas in Bo!kovi" (2000), we could argue that the optionality in 
clitic placement in the contexts under discussion is due to the fact that “It is 
generally possible to assign more than one prosodic structure to a single syntactic 
structure, depending on how it is pronounced” (Bo!kovi" 2000:108, footnote 35). 
For Bo!kovi" the domain of cliticisation is the Intonational Phrase (I–phrase). 
Therefore, a pause immediately after the complementizer in (63) would indicate 
the boundary of a different I–phrase and post–verbal clitic placement would 
follow. 
 
                                                
23 See chapter 3 for a detailed discussion of Mavrogiorgos’s proposal for cliticisation in Tobler–
Mussafia languages. 
  
65 
 
2.3.3. Clitic clusters in CG 
 
Single clitics and clitic clusters in CG adhere to the same positioning restrictions. 
Clitic clusters are placed pre– or post–verbally depending on the syntactic context, 
as described in 2.3.1, and they have free ordering in the environments discussed in 
2.3.2. The order in which clitics appear in CG clusters is strictly CLDAT–CLACC 
(contra Terzi 1999a), unlike in SMG, where the dative and the accusative clitic 
exhibit flexible ordering in imperative and gerundive constructions (both CLDAT–
CLACC and CLACC–CLDAT are possible). Certain person restrictions on the possible 
combinations between indirect (IO) and direct object (DO) clitics that are 
operative in SMG (Holton et al. 1997: 192–194, (Philippaki–)Warburton 1977) 
are applicable in CG as well. In particular, only a 3rd person clitic pronoun can 
appear as a DO within the cluster, whereas there is no person restriction for the IO 
(1st, 2nd and 3rd person clitics can all appear as IOs)24 (Revithiadou & 
Spyropoulos 2008). These facts are captured within the well–known Person–Case 
Constraint (Bonet 1994: 48), cited in (68)25. 
 
(68) If DAT–PERS, then ACC–3rd. !
2.3.4. CG clitics and second position restrictions 
 
Given that CG is a pro–drop language, the finite verb usually surfaces clause 
initially and clitics often appear in the second position (2P) of the clause. 
However, CG clitics are not subject to 2P restrictions and CG is not a 
Wackernagel language, unlike Serbo–Croatian and Ancient Greek26. This is an 
                                                
24 Ralli described the positional restrictions imposed on SMG clitic clusters as follows:  
1. 1st and 2nd person pronouns do not co–occur, 
2. 1st and 2nd person pronouns precede 3rd person ones, 
3. A genitive pronoun precedes an accusative one (2006:146). 
25 For the acquisition of the Person–Case constraint in Greek, see Tsakali and Wexler (2003). 
26 According to Horrocks (1997:59) clitic pronouns in Ancient Greek typically collocated with 
sentence connectives in 2P. In the Hellenistic period the verb would appear clause initially and the 
clitics would follow in 2P, unless clause initial position was occupied by an 
emphatically/contrastively focalised element or by some clausal operator expressing negation, 
interrogation or modality, in which case the order F(ocus)/O(perator)–CL–V would surface. This 
  
66 
 
issue originally brought to attention by Terzi (1999b), who offers ample evidence 
against an analysis of CG as a language with 2P restrictions. A comparison 
between an embedded clause in Serbo–Croatian and its CG equivalent trivially 
illustrates the divergence between the two languages with respect to clitic 
placement. In the Serbo–Croatian example in (69), taken from Terzi (1999b), the 
clitic (nam) in the embedded clause immediately follows the complementizer (da) 
in 2P. In the equivalent CG structure (70), however, the subject of the embedded 
clause is realised immediately after the complementizer (oti), the negator 
immediately follows it, while the clitic (mas) appears in the third position of the 
clause.  
 
(69) Ivan  kaze  da    nam       Olga  nista       ne      daje.  
Ivan  says   that  us–CL  Olga  nothing  NEG  gives 
                                                                      (Terzi 1999a:229, example 10) 
(70) O     Ivan  lali    oti    i     Olga   en      mas      dia      tipote. 
The  Ivan  says  that  the  Olga  NEG  us–CL  gives  nothing 
“Ivan says that Olga is not giving us anything”  
 
In order to accommodate the Serbo–Croatian facts, Rivero and Terzi (1995 
in Terzi 1999b:229) assume that the element that surfaces clause–initially moves 
to the Spec CP or to C, depending on its X–bar status, and C takes the 
Wackernagel Phrase (WP) as its complement. The clitic is realised in [Spec,WP] 
((8) in Terzi 1999b:229, repeated below as (71)). 
 
(71) [CP [C’ [WP CL [W’ [(NegP) [IP [I’ [VP [V’….]]]]]]]]] 
 
Bo!kovi" (2000/2001) offers a different account whose basic tenet is that 
clitic placement in Serbo–Croatian is regulated by some PF filters. In particular, 
he suggests that the following PF filters are applicable: a. #_ and b. suffix ((73) in 
Bo!kovi" 2000:107). The former indicates that Serbo–Croatian clitics appear in 
the initial Intonational Phrase (I–phrase) of their clause, and the latter that they are 
                                                                                                                                 
dual distribution of clitics continued into medieval Greek and is preserved in contemporary CG as 
well. 
  
67 
 
specified as suffixes in the lexicon. The relevance of Bo!kovi"’s analysis for 
cliticisation in CG is illustrated and discussed in the last chapter. 
 
2.3.5. CG, European Portuguese and Galician 
 
CG, European Portuguese (henceforth EP) and Galician adhere to the Tobler–
Mussafia law, whereby clitic pronouns are banned from clause initial position. 
Clitics appear post–verbally, unless a proclisis trigger heads the clitic clause. The 
proclisis–triggering contexts in CG, EP (Duarte & Matos 2000, Lobo & Costa 
2012) and Galician (Uriagereka 1995) extensively overlap. 
EP exhibits three patterns of clitic placement: proclisis (pre–verbal clitic), 
enclisis (post–verbal clitic) (72) and mesoclisis (clitic within the verb). Proclisis is 
contingent on specific triggers, mesoclisis is manifested in future and conditional 
tenses in enclitic contexts (73), and enclisis occurs in all other contexts. Proclisis 
triggers in EP include negation markers (74), negative (75) and quantified subjects 
(76), some preverbal adverbs (já, também, sempre, só, ainda) (77), finite 
subordinate clauses (78), a filled CP (wh–questions (79), wh–exclamatives, é que 
–clefts) and focused initial constituents (80). The following examples taken from 
Lobo and Costa (2012) illustrate these points.  
 
(72) O     João  lavou–se. 
The  João  washed_CL.3S.Refl  
“João washed himself” 
(73) O     João  lavar–se–á. 
The  João  wash_CL.3S.Refl_will 
“João will wash himself” 
(74) O     João  não     se                 lavou. 
The  João  NEG  CL.3S.Refl  washed 
“João didn’t wash himself” 
(75) Ninguém  se                 lavou. 
Nobody    CL.3S.Refl  washed 
“Nobody washed himself” 
  
68 
 
(76) Todos os meninos  se                    lavaram. 
All the boys            CL.3PL.Refl  washed 
“All the boys washed themselves” 
(77) O     João  já           se                 lavou  
The  João  already  CL.3S.Refl  washed 
“João already washed himself” 
(78) O     João  disse  que  se                 lavava   todos os dias. 
The  João  said   that  CL.3S.Refl  washed  all the days 
“João said that he washed himself every day” 
(79) Quem  se                 lavou?   
Who    CL.3S.Refl  washed 
“Who washed himself?” 
(80) MUITA ÁGUA  se                 perdeu! 
Much water        CL.3S.Refl  lost 
“So much water was lost!” 
 
The overlapping contexts for pre–verbal clitics in EP and CG include 
clauses headed by negative markers / subjects, wh–elements or focused initial 
constituents. Moreover, a filled CP and some preverbal adverbs trigger proclisis in 
EP, while in CG this applies to a confined set of complementizers/adverbs. 
Finally, quantified subjects trigger proclisis in EP (76) but not in CG (Terzi 
1999b:237, note 6) (81). With regard to negative quantifiers, CG and EP pattern 
alike, as in CG these are obligatorily followed by the negation marker (82) which 
triggers proclisis anyway. 
 
(81) O     kathenas                vlepi      to. 
The  everybody–NOM  see–3S  it–CL.ACC 
“Everybody sees it” 
(82) Kanenas            *(en)  to                  idhen. 
Nobody–NOM  NEG  it–CL.ACC  saw–3S 
“Nobody saw it”  
 
  
69 
 
As for Galician, pre–verbal clitics are triggered in wh–questions (84), 
negatives (85), clauses headed by quantifiers (86) or emphasised XPs, as well as 
in subordinate constructions (87); the examples are taken from Uriagereka (1995). 
All these contexts, except clauses headed by quantifiers, require pre–verbal clitic 
placement in CG as well. 
 
(83) Ouvimo–lo. 
Hear–1PL.it–CL 
“We hear it” 
(84) Quén  o         ten  ouvido? 
Who   it–CL  has  heard 
“Who has heard it?”  
(85) Non   o         ten   ouvido. 
NEG  it–CL  has  heard 
“S/he hasn’t heard it” 
(86) Todo o mundo  o         veu 
Everyone          it–CL  saw 
“Everyone saw it” 
(87) Quero      que  o         oiades. 
Want–1S  that  it–CL  hear–2S 
“I want you to hear it” 
 
Table 2.4 summarises the contexts for pre–verbal clitic placement in CG, EP and 
Galician and highlights the points of convergence and divergence between the 
three languages. 
 
 
Table 2.4: Proclisis contexts in CG, EP and Galician. 
Syntactic environments CG EP Galician 
Negation markers YES 
Wh–questions YES 
Focused/Stressed/Emphasised XPs YES 
Complementizers [YES] YES YES 
Quantifiers NO YES YES 
(Some) Pre–verbal adverbs NO YES YES 
  
70 
 
2.4. Summary 
 
This chapter discussed the basic properties and the syntax of clitic pronouns in 
CG. It outlined their morphological, semantic and phonological properties, their 
distribution, their behaviour with respect to coordination and modification and 
their categorial status. Clitic placement in CG was also discussed. An overview of 
the proclisis and enclisis contexts, as well as of contexts allowing free ordering of 
clitic pronouns, was followed by comparisons with 2P languages and languages 
exhibiting similar positioning restrictions to those of CG. A number of proposals 
put forward to account for clitic placement in CG will be discussed in the next 
chapter. 
  
71 
 
CHAPTER 3: ACCOUNTS OF CLITIC 
PLACEMENT IN CYPRIOT GREEK  
 
3.1. Introduction 
 
This chapter presents syntactic proposals, a Prosodic Inversion account and 
interface approaches 1  for clitic placement in CG and addresses a central 
theoretical dilemma: is the placement of clitic pronouns in CG the result purely of 
syntactic operations or is it constrained by certain phonological well!formedness 
constraints? The discussion is organised as follows: the first section presents 
syntactic accounts put forward by Agouraki (2001) and Terzi (1999a, 1999b), 
which suggest that enclisis in CG derives from proclisis with the manifestation of 
syntactic movement of the finite verb to a higher projection. The second section 
discusses a Prosodic Inversion (PI) account by Condoravdi and Kiparsky (2001) 
that assumes an enclitic template for CG clitics that triggers movement at PF. The 
third section presents two interface accounts, within which both syntax and PF are 
involved in cliticisation in CG. The first is an analysis of cliticisation in 
Tobler!Mussafia languages suggested by Mavrogiorgos (2012) and the second is 
a proposal by Revithiadou (2008) assuming a PF!controlled spell!out of copies 
of clitic pronouns. A number of interim conclusions are outlined in the last 
section. 
An important point should be made regarding the use of the terms proclisis 
and enclisis. These terms are typically used to indicate the directionality of the 
phonological/prosodic dependence of clitic pronouns upon their host. Specifically, 
they indicate whether clitics procliticise or encliticise to their host, independently 
of their relative order with respect to the verb (pre!/post!verbal). In this chapter, I 
will use the terms proclisis and enclisis descriptively to refer to clitics’ pre!, and 
post!verbal placement, correspondingly, unless otherwise indicated.  
                                                
1 This chapter discusses accounts that try to accommodate the phenomenon in synchronic terms. 
For a discussion of the diachronic development of clitic pronouns in Greek, which is also relevant 
for the distribution of CG clitic pronouns, see Pappas (2001, 2004). 
 
  
72 
 
3.2. Syntactic accounts 
 
To date, three syntactic analyses of CG clitic placement have been offered, 
namely those of Agouraki (2001), and Terzi (1999a) and (1999b). All assume that 
enclisis derives from proclisis upon the manifestation of verb movement past the 
clitic. However, they attribute verb movement to different operations and propose 
different landing sites for the finite verb. 
 
3.2.1. The V!to!M analyses for finite enclisis 
 
Terzi (1999a, 1999b) adopts Kayne’s (1994) antisymmetry proposals. She 
assumes that clitics are heads (X°s) 2,3 in both SMG and CG and that they adjoin 
to the same functional head. 
In her first proposal (1999a), she departs from the idea of a designated 
adjunction site for clitic pronouns, namely a functional head where clitics adjoin 
in all syntactic environments. Instead, she advocates “a bipartition with respect to 
the type of functional heads which serve as adjunction sites for clitics” (Terzi 
1999a:87), in both CG and SMG. She proposes that clitics adjoin to T in 
tense!defective contexts (when tense features are weak) and to F, an empty 
placeholder devoid of verb!related features, in all remaining contexts. The 
configuration for clitic constructions in non tense!defective contexts in CG (and 
SMG alike) is the following (Terzi 1999a:93): 
 
(1) [C…[F to [F 0 [T diavasai [Agr ei [V ei]]]]]] 
           it            read!1S 
 
                                                
2 Terzi adopts a line of reasoning that builds on the similarity between 3rd person clitics and 
determiners in Greek in order to argue for the head!ness of clitics (1999a:86, footnote 2). She 
takes into account clitic doubling facts such as Ton idha ton Yianni “him!CL saw!1S the 
John!ACC”, and assumes that the doubled phrase is the specifier of a determiner head, following 
Uriagereka (1995). Thus, she considers Greek clitics to be X°s. Notably, this similarity between 
3rd person clitics and determiners exists in both CG and SMG. So, provided this is the right 
criterion for deciding upon the status of clitics, any conclusion that can be drawn should apply to 
both varieties of Greek. 
3 See also section 2.2.6 of chapter 2. 
  
73 
 
Terzi’s F head is “a functional head partially reminiscent of the head of Clitic 
Voice of Sportiche (1996) (but significantly different from Uriagereka’s (1995) 
F)” (Terzi 1999a:934). Crucially, Terzi’s (1999a) F does not check features of the 
verb overtly, thus the verb does not move through F. 
In CG (as opposed to SMG), the finite verb undergoes movement past F to 
the M(ood) head, which gives rise to the order V!clitic (Terzi 1999a). M is the 
head of the Mood Phrase and encodes modality. V!to!M is manifested in all clitic 
constructions except clauses headed by functional heads with operator!like 
properties, such as M, Neg and the Focus head. According to Terzi (1999a), 
V!to!M movement in CG is related to the Infl make!up of CG, and specifically 
to the feature composition of M. 
With regard to imperatives, Terzi (1999a) assumes, following Beukema 
and Coopmans (1989), that the imperative Tense involves a [!Tense] Infl. 
(1999a:94). A T with weak verbal features, either specified as [!T] (as argued for 
the imperative verb), or involving a defective Tense (as is the case for infinitives), 
becomes a legitimate adjunction site for clitics within her analysis. The 
argumentation is developed as follows: since the imperative verb does not raise to 
T in the overt syntax for feature checking requirements, even if the clitic adjoins 
to T, a configuration of multiple adjunction is not created5. Terzi’s proposal is 
economical with respect to the adjunction site for clitics, in the sense that, in 
contexts where some functional head that constitutes an eligible host (T) for clitics 
is present, the F projection is precluded (1999a:94, footnote 12). 
The imperative verb will ultimately undergo movement to C. This 
movement is triggered by illocutionary features in C (1999a:89). As a by!product 
of the verb movement to C, the imperative verb will initially raise to T. Once it 
has its Tense features checked6 (Terzi 1999a:95!96), it ultimately raises to C. The 
                                                
4 Cf Terzi (1999a:93, footnote 10) for a comparison between Sportiche’s Clitic Voice and 
Uriagereka’s F. 
5 In finite contexts, T is not a legitimate adjunction site for clitics, since in languages like Greek 
the finite verb raises as high as T and adjunction of clitics to T would create a configuration of 
multiple adjunction, which is ruled out by Kayne’s Linear Correspondence Axiom (Terzi 
1999a:92). 
6 For this to happen, Terzi assumes that the Linear Correspondence Axiom does not apply after 
spell!out. In other words, she considers the LCA not applicable in LF, and hence multiple 
  
74 
 
movement of the imperative verb to C obligatorily proceeds via T because 
skipping the latter would violate the Shortest Move Requirement (Ferguson and 
Groat 1995). According to the Shortest Move Requirement, “a category moving to 
check feature(s) of a given type may not skip moving into an immediate relation 
with the closest c!commanding head which checks features of that type” (Terzi 
1999a:96). Therefore, for V!to!C movement to be manifested, the verb must 
left!incorporate into the clitic, which has already adjoined to T, and check its 
tense features. Then the complex consisting of the imperative verb and the clitic 
moves to C (1999a:97!98). 
Terzi’s first proposal is based on the idea that the Inflectional system in 
CG and SMG differs (1999a), and two pieces of evidence are offered to justify 
this claim (1999a:110, note 24). The first is the absence of the future particle tha 
and the second is the unavailability of compound tenses in CG. I will return to this 
argument after discussing her second proposal.  
In her second proposal, Terzi (1999b) leaves the imperative verb aside and 
discusses different finite contexts in CG. She suggests that CG clitics adjoin to the 
featureless functional head F which takes IP as its complement (2) (Terzi 
1999b:231). Her claim regarding the post!verbal position of clitics in CG remains 
unaltered. As in her original proposal (1999a), she assumes that the finite verb 
undergoes movement with M as its landing site. She rejects the possibility of 
V!to!C movement on the basis of constructions where lexical complementizers 
(like oti for example) that occupy the C position in the CG clause can co!occur 
with post!verbal clitics7. As to what triggers verb movement, Terzi (1999b) 
attributes V!to!M to the licensing requirements of CG clitics, which have to be 
satisfied. She proposes that CG clitics have strong features that must be licensed 
in the internal domain of a functional head with operator!like properties before 
                                                                                                                                 
adjunction is not ruled out (1999a:95!96). This assumption is contra Kayne (1994), who assumes 
that the LCA applies at all levels of representation, but is in accordance with Chomsky (1995).  
7 Terzi does not ignore the possibility of an analysis involving recursive CPs in CG, within which 
both a lexical complementizer and the verb can co!occur (1999b:238, note 14). She rejects it, 
though, as there is no independent evidence verifying either the presence of recursive CPs or the 
manifestation of V!to!C movement in CG. 
  
75 
 
spell!out 8 . Such functional heads are the Neg head, the M(ood) head, 
wh!elements and the head of the Focus phrase. In her own wording, this type of 
licensing involves a head!complement rather than a Spec!head relation (Terzi 
1999b:233). In the absence of such a functional head, the verb undergoes 
movement to M and clitics surface post!verbally. Verb movement in Terzi 
(1999b) is seen as an operation of last resort. 
 
(2) [CP [C’ COMP [MP [M’ [FP [F CL [F’ [IP [VP [V’….]]]]]]]]]] 
 
Terzi further compares clitic!left dislocation (CLLD) and clauses headed 
by focused XPs; in the former the clitic surfaces post!verbally and in the latter 
pre!verbally. In CLLD the dislocated XP is assumed to be either CP!adjoined 
(Cinque 1990) or IP!adjoined (Anagnostopoulou 1994). Terzi wants to show that 
only functional heads (i.e. the Focus head), and not just any kind of head, can 
license clitics. She compares structures with an empty C (an enclisis context) and 
clauses in which the C head has wh!features (a proclisis context), in both matrix 
and embedded environments, to show that it is a prerequisite for the functional 
head to have operator!like properties in order to license clitics9 (1999b:234!235).  
Terzi’s (1999a, 1999b) proposals raise a number of important issues. In 
the first place, with regard to the proposed movement of the verb past the clitic, it 
is not clear how the verb can skip the clitic on its way to M, without violating the 
Head Movement Constraint (HMC)10. Recall Terzi’s (1999a, 1999b) claims that 
clitics in both CG and SMG are heads (X°s). Recall also the HMC: “Head 
movement of X to Y cannot skip an “intervening” head Z”11. In this case, verb 
movement to M cannot skip the intervening12 clitic. 
                                                
8 Despite the key role of the notion of licensing in her analysis, some aspects of it are not clearly 
defined, i.e. the internal domain of the respective element, the set of proposed licensers, and the 
properties that operator!like elements have. 
9 As for clauses headed by the factive complementizer pu, she argues that the Specifier of pu!CPs 
is occupied by an empty operator (Anagnostopoulou 1994). 
10 But note that Terzi adopts the Shortest Move requirement which weakens/relativises HMC. 
11 Cf. Roberts’ claim that the Head Movement Constraint does not exist (2010:193). 
12 It is, of course, questionable what counts as a possible intervener (cf. Iatridou 1994) for Terzi 
(1999a, 1999b). Noteworthy, Terzi notes that with respect to this issue she follows proposals in 
Benincà and Cinque (1993). 
  
76 
 
Secondly, an important aspect of the movement analysis concerns the 
reason why it occurs. Terzi’s (1999b) assumption that the finite verb moves to 
satisfy requirements of the clitic pronoun is problematic. Following Chomsky 
(1991, 1993), the movement of any given element is driven by morphological 
requirements of the element that moves; this is something Terzi acknowledges 
(1999b:233) as well. On the other hand, if verb movement is triggered by strong 
verbal features, it is not easy to see how these features are satisfied when 
movement is not manifested, i.e. when M or Neg head the clause (Terzi 
(1999b:233). 
The third point concerns Terzi’s claim about the properties of CG and 
SMG Inflection. She argues that CG Infl is different from SMG Infl in several 
aspects (1999a:110, note 24; 1999b:238, note 18) and that what differentiates the 
two varieties is the feature composition of M. She provides two pieces of evidence 
to justify her claim: first, that CG lacks the future particle tha of SMG, and 
second, that compound tenses are not available in CG (1999a:110, note 24; 
1999b:237!238, note 5, 11 & 18).  However, CG does not lack a future particle. 
The dialectal equivalent of tha is enna13; this is shown on the basis of their 
identical distribution (see section 1.2 in chapter 1). Moreover, negated future 
clauses in CG are headed by etha, an amalgam form which combines the negator 
(dh)e(n) and the future particle tha. With regard to the compound tenses, I agree 
with Terzi that not all the compound tenses that are available in SMG are 
available and productive in CG as well. Greek Cypriots use simple past forms for 
events that SMG!speakers express in the present perfect. However, although 
present perfect forms are not productive in CG 14, as already pointed out by 
Arvaniti (2010) and Terzi (1999a, 1999b), past perfect forms are. Past perfect in 
CG is formed with the past form of the auxiliary eho (“have”) and the participle15. 
                                                
13 Terzi assumes that enna is a derived form that consists of the modal na and the copula. In her 
analysis future structures in CG utilise a sentential complement (headed by na) of an impersonal 
verb: E (is) na (M) pao (go) (1999b:238, note 11). 
14 Cf. Arvaniti (2010:28) for some examples in which the Present Perfect is used by Greek 
Cypriots, but with a concrete time reference, something that is ungrammatical for most speakers of 
SMG.  
15 Moreover, in CG and other Greek varieties, perfect tenses may be built with the finite forms of 
the auxiliary eho (“have”) or ime (“be”) and the past participle of the verb (Ralli 2006:133!135).  
1. Eho    /  Iha          ta              thkjavasmena. 
  
77 
 
Furthermore, the arguments for the bipartition with respect to the 
adjunction site for CG clitics in Terzi (1999a) are not strong. She argues that in 
CG the relative order of the dative and the accusative clitic in double clitic 
constructions depends on the finiteness of the verb involved: in finite contexts the 
dative clitic always precedes the accusative, while the imperative verb (bearing 
[!tense] Infl) allows free ordering (CLDAT!CLACC or CLACC!CLDAT 16). However, 
the argument is disconfirmed by the fact that in CG neither the finite verb nor the 
imperative verb allow free ordering of the dative and the accusative clitic in 
double clitic constructions. The relative order is fixed and is always 
CLDAT!CLACC. 
I agree with Terzi (1999a, 1999b) about a crucial aspect: the position of 
the finite verb in CG clitic constructions. Terzi claims that the finite verb does not 
move higher than M (1999a, 1999b). Following proposals in Alexiadou (1997) for 
SMG, verb placement has an overt correlate in the relative ordering of adverbs. 
Hence, Terzi’s (1999a, 1999b) claims with regard to verb placement can be 
(dis!)confirmed on the basis of adverb placement in the relevant structures. The 
relevant position of the verb–clitic combination in CG in relation to the position 
of adverbs (3), and, in particular, of aspectual adverbs such as schedon (4), in root 
and na!clauses shows that the finite verb in CG clitic constructions does not 
move higher than M. Corroborative evidence is offered by the identical order of 
the verb–clitic combination in CG and SMG (3). Recall the standard assumption 
that in SMG clitic constructions the finite verb adjoins to a functional head lower 
than M (T in Mavrogiorgos 2009 and Terzi 1999a, I in Philippaki–Warburton 
1998). 
                                                                                                                                 
Have  /  Had!1S  them!CL  read!PP.3PL.ACC 
       “I have/had done my homework” 
16 Within Terzi’s (1999a) analysis, double clitics may either adjoin each to a different functional 
head (CLACC to the lower and CLDAT to the higher), or both to the same, with the CLDAT adjoined 
to the accusative and their complex to F (in finite contexts) or T (in imperatives) (see (32) and (33) 
in Terzi 1999a:102). In CG finite contexts, the V!to!M movement is not manifested via F (where 
the clitics adjoin), because F is devoid of verbal features. So verb movement causes no alternation 
in the relative order of the dative and the accusative clitic. In contrast, the imperative verb moves 
to C through T. If the former is manifested, the imperative verb incorporates into the lower clitic, 
namely CLACC, they skip over CLDAT and the following order surfaces: CLACC!CLDAT. If the latter 
option is manifested, the imperative verb incorporates into the lower clitic, namely the dative, they 
move as a complex to C and the V!CLDAT!CLACC order surfaces.  
     
  
78 
 
(3) Panda    prosehis          ton          (CG)  /  na  ton  prosehis  (CG & SMG). 
Always  take care!2S  him!CL.ACC       M 
“You always take care of him” 
(4) Shedhon  edhere    me                   /  na  me  deri. 
Almost    beat!3S  me!CL.ACC 
“He almost beat me” 
 
3.2.2. The filled!C requirement 
 
Agouraki (2001) offers a different analysis, which is built on three tenets. First, 
CG clitics head clitic phrases located between CP and IP17. Second, clitic 
placement in CG does not involve syntactic cliticisation, but it only depends on 
verb placement; in fact, it is an epiphenomenon of verb placement. Third, CG has 
a filled C requirement for clause!typing purposes. Let us start from the last point. 
Agouraki analyses CG as a verb second (V2) language. However, she dissociates 
the two component parts of V2, namely V!to!C and filled [Spec,CP], arguing 
that they constitute two distinct requirements18, and suggests that CG only has the 
former requirement. For Agouraki (2010:542) the filled!C requirement is 
basically a requirement for the specification of sentential force on C to be checked 
in the overt syntax19. 
Thus, in CG clitic constructions, if the C position is not filled, V!to!C is 
manifested to satisfy the filled!C requirement, whereupon the verb appears in C 
and the clitic surfaces post!verbally. If, on the other hand, the C position is filled, 
V!to!C raising is blocked, hence the verb adjoins to I and the clitic appears 
pre!verbally. For Agouraki (2001, 2010) the default pattern for clitic placement is 
enclisis, and proclisis obtains when some element fills the C position. 
                                                
17 In Agouraki (2010) she elaborates on this idea and argues that clitics are AGR!O(bject) heads, 
in line with Sportiche (1996), and AGR!O in Greek is above T and lower than NEG and C 
(2010:531). 
18 In support of her proposal, Agouraki mentions the case of Old Irish, a language for which it has 
also been proposed that it only has a filled C but not a filled [Spec,CP] requirement (Carnie, 
Harley & Pyatt 2000). 
19 Agouraki (2001) assumes that the C position in CG has the morphological feature [Declarative], 
[Interrogative] or [Emphasis], which is checked either by the verb or by a complementizer with the 
relevant feature. In Agouraki (2010:542) [Emphasis] as a specification on C appears to be in 
complementary distribution with [Assertion], [Question] and [Situation]. 
  
79 
 
Agouraki (2001) further assumes that the C position in CG can host either 
overt or null complementizers, both of which block V!to!C raising and yield 
proclisis. As shown in chapter 2, apart from the factive complementizer pu and 
the complementizer an, a number of other elements trigger proclisis in CG, 
including negative markers, wh!elements, focalised XPs and modality markers. In 
order to account for these facts, Agouraki makes a number of assumptions for the 
relevant heads. For Neg, she argues that it is either generated in or raised to C in 
complementizer!less clauses. As for modality markers, she assumes that they 
occupy the C position. For wh!questions and sentences with XP!foci, she 
suggests that the C position is occupied by the null complementizers WH and F 
respectively, and the [Spec,CP] position by wh!phrases and focalised XPs 
respectively. Hence, V!to!C is blocked and proclisis obtains. She further assumes 
that the [Spec, CP] position can be filled with pre!verbal existential quantifiers 
and pre!verbal stressed operators (with the exception of universal quantifiers, 
also!phrases and even!phrases (2010:531)). The structure in (5), based on the 
clause structure in (7) in Agouraki (2001:5), illustrates her assumptions with 
regard to the architecture of the CG CP. 
 
(5) [CP WH/F/Top [C’ C ([WH/F]) [NegP [Neg’ Neg [ClP [Cl’ Cl [IP [I’ I VP]]]]]]]]] 
 
Agouraki (2001) discusses the syntax of verbal modifiers in Hungarian 
aiming to show that verbal placement can determine the placement of some other 
constituent. Verbal modifiers in Hungarian typically appear pre!verbally, as in 
(6). In negative clauses (7), syntactic foci and wh!clauses, however, they appear 
post!verbally. These are syntactic contexts that trigger proclisis in CG. Even 
though the aforementioned heads trigger the reverse pattern of placement for CG 
clitics compared to Hungarian verbal modifiers with respect to the verb, namely 
proclisis in CG and enclisis20 in Hungarian, Agouraki focuses on the fact that in 
both cases verbal placement determines the order between the verb and some 
other constituent.  
                                                
20 In Hungarian, V!to!C is manifested when some constituent (i.e. focused XP, wh!phrase) fills 
the [Spec,CP] position (Agouraki 2001:11). 
  
80 
 
(6) Péter  be     ment  a     hazba. 
Peter  into  went   the  house 
“Peter went into the house”  
(7) Péter  nem   ment  be     a     hazba. 
Peter  NEG  went  into  the  house 
“Peter didn’t go into the house”  (Agouraki 2001) 
 
Agouraki (2010) further develops her account of the filled!C requirement 
of CG. The essence of this proposal is that the C head in CG has a clause!typing 
feature that must be checked in narrow syntax. This feature can be checked by an 
appropriate complementizer, the Negation head or the verb. Otherwise, an 
appropriate operator in [Spec,CP] can also satisfy the requirement; recall that 
Agouraki (2001, 2010) takes filled C and filled [Spec,CP] to be in complementary 
distribution in CG. Notably, Agouraki (2010) argues that this proposal is in line 
with Rizzi’s (1997) proposal concerning what marks clause!typing 
cross!linguistically21. 
Some points need to be raised with respect to Agouraki’s (2001) proposal. 
First, there is a lack of independent evidence that CG has one of the two 
component parts of V2 languages, namely the filled!C requirement. Second, the 
assumption that V!to!C is manifested unless the C position is filled is challenged 
by CG clauses where complementizers like oti and pos co!occur with post!verbal 
clitics (see section 2.3.2 in chapter 2). Third, one of the basic tenets of Agouraki’s 
(2001) analysis, namely the assumption of V!to!C movement in CG finite 
clauses, is challenged on the basis of the similarity between CG and SMG with 
respect to adverb placement (see previous section). 
  
3.3. Prosodic inversion account 
 
Condoravdi and Kiparsky (2001) offer a prosodic account for clitic placement in 
Greek dialects aiming to capture the syntactic variation and change with respect to 
                                                
21 In Rizzi’s wording “Force is expressed sometimes by overt morphological encoding on the head 
(special C morphology for declaratives, questions, relatives, etc.), sometimes by simply providing 
the structure to host an operator of the required kind, sometimes by both means” (Rizzi 1997:283). 
  
81 
 
cliticisation. They take a diachronic perspective and adopt a 
comparative!historical approach. They suggest that finite enclisis in varieties like 
CG22 is due to prosodic inversion that alters the relative order between the clitic 
and the verbal host. 
Condoravdi and Kiparsky (2001) argue that clitics in modern Greek 
dialects are of three distinct types. Type A clitics are Xmax elements and 
syntactically adjoin to a maximal projection. Clitics of this type appear in pre! 
and post!verbal position, but they always encliticise to their host. Type B clitics 
are X0s, hence they syntactically adjoin to a lexical head, and they are either 
proclitics (Kozani) or enclitics (Pontic). Type C clitics (as in SMG) are lexical 
affixes and they attach to prosodic words in the lexicon. It is assumed that CG 
clitics are of type A on the basis of their similarity in distribution with type A 
clitics in other Greek varieties. Specifically, clitics of type A comparable to CG 
clitics immediately follow the finite verb, unless they appear in negative clauses, 
in subjunctive or future tense clauses, in wh!phrases, after relative pronouns 
(Janse 1998), or after subordinating complementizers and preverbal phrases in 
focus; in all the aforementioned syntactic contexts clitics immediately precede 
their verbal host. 
Condoravdi and Kiparsky (2001:5) assume the clause structure in (8) for 
all Greek dialects with Xmax clitics. 
 
(8) [CP Spec [C’ C ["P Spec ["’ "0 [TnsP Cl [TnsP Spec [Tns’ Tns0  VP]]]]]]] 
 
They further assume that finite verbs move from V to the head of TnsP and that 
there is a higher inflectional projection "P, composed of NegP, MoodP, and 
FocP23 (cf. Laka 1990, Piñón 1993), dominating TnsP. "P is headed by negative 
and modal particles, while focused XPs and emphatic negatives surface in its Spec 
position. According to Condoravdi and Kiparsky (2001), clitics adjoin to TnsP, 
                                                
22 The proposal does not discuss CG data in particular, but rather data from a set of dialects that 
includes CG. The analysis presented applies to varieties spoken in the following areas: inland Asia 
Minor (Cappadocia, Bithynia), the Cyclades, some Dodekanese islands (Karpathos, Kos, 
Astypalaia), two localities on Lesbos (Ajassos, Plomari), and the Taur!Roumeic dialects of 
Ukraine (Marioupoli/Azov). Late Medieval Greek is a Greek variety of this type as well. 
23 Condoravdi and Kiparsky indicate that their analysis would be consistent with an expansion of 
"P into its component parts, namely the relevant heads (2001:34, note 10). 
  
82 
 
the functional projection to whose head the verb moves to. Their placement is 
attributed to a prosodic requirement according to which a prosodic word should 
appear on their left within the same CP. This prosodic word should be a 
non!adjoined constituent, i.e. a lexical (overt) complementizer (in C0), a 
wh!element (in SpecCP) in matrix or embedded interrogatives, a negative or 
modal particle (in "0), or a focused constituent (in Spec"P). If no suitable 
prosodic host is available, the clitic pronouns encliticise onto the verb on their 
right by prosodic inversion (Halpern 1995). 
From an Optimality theoretic perspective, Condoravdi and Kiparsky view 
prosodic inversion as an optimisation strategy which aims at satisfying both the 
prosodic and the syntactic requirements of clitic pronouns. To satisfy the former 
clitics have to encliticise to some constituent on their left; to satisfy the latter the 
input order has to be preserved in the output, with clitics remaining within the 
same CP. 
As Condoravdi and Kiparsky indicate (2001:7), within their account the 
default case for clitic positioning in type A dialects is the pre!verbal one. The 
post!verbal placement involves an extra step. Moreover, they assume the same 
clause structure for type A and type C dialects, such as CG and SMG respectively. 
They attribute the differences in clitic placement to the different prosodic 
requirements of clitics in the two types of dialect (2001:16). Specifically, they 
claim that type C clitics are affixes, which subcategorize for a phonological word 
on their right, while type A clitics subcategorize for a phonological word on their 
left. Hence, type C clitics do not attach syntactically to TnsP but lexically to the 
left of a finite verb24. The clitic–verb cluster then moves as a whole to TnsP. In 
contrast, type A clitics move to TnsP in the overt syntax. Within Condoravdi and 
Kiparsky’s account (2001), CG clitics are inherently enclitic and attach 
phonologically to any preceding constituent that meets their prosodic 
requirements. 
However, Condoravdi and Kiparsky’s account (2001) can be challenged 
on the basis of the following facts. First, pre–verbal NPs in Topic position trigger 
                                                
24 Their assumption for the phrase structure status of type C clitics, hence for SMG clitics as well, 
is that they are word!level affixes; that is, affixes that attach to words, not to stems (like subject 
agreement morphemes) (2001:16). 
  
83 
 
enclisis, whereas complementizers like oti and pos can co–occur with post–verbal 
clitics. Second, operations such as movement and linearisation of constituents are 
standardly perceived as syntactic. Within such an analysis, we need to assume a 
very powerful PF that is able to move syntactic constituents and “regulate” the 
linear order of the clause. This is not theoretically desirable and is something that 
diverges from typical PF operations.  
  
3.4. Interface accounts 
 
Mavrogiorgos (2012) and Revithiadou (2008) have proposed analyses that 
capitalise on the role of the syntax–PF interface for clitic placement in CG. These 
are discussed in the following subsections. 
 
3.4.1. The syntax–PF interface analysis for Tobler–Mussafia languages 
 
Mavrogiorgos (2012) discusses the proclisis–enclisis alternation in finiteness– 
sensitive and Tobler–Mussafia languages. This section summarises his proposals 
focusing on the aspects of the analysis that are relevant for CG. 
Following Pancheva (2005), Mavrogiorgos assumes that Tobler–Mussafia 
languages impose the following PF requirement on clitics: a ban from the initial 
position of a prosodic domain, e.g. of the utterance. He suggests that finite enclisis 
in Tobler–Mussafia languages in general, and in CG in particular, derives via verb 
movement across the cliticization site, which is yet triggered by PF–requirements. 
To elaborate, according to Mavrogiorgos (2012), finite enclisis involves a 
F(unctional) head, an abstract syntactic category, which immediately c–
commands the cliticisation site (T). This C–head, which is linked to enclisis in 
both finiteness–sensitive and Tobler–Mussafia languages, is a phase head; a 
syntactic Agree relationship is established between the F–head and the lower T–
head (2012:31–32). The F head carries (CP–related25) syntactic features, which 
                                                
25 For Mavrogiogos what differentiates the F–head involved in finite enclisis and the F–head 
involved in non–finite enclisis is the fact that the latter carries features that are linked to subject, 
tense and/or mood agreement, while the former carries CP–related features (2012: 29–30). 
  
84 
 
are taken by Mavrogiorgos to be affixal26, and a PF requirement. 
If an appropriate XP or X is merged at the edge of the F head and has 
established a syntactic relationship with it, the PF–requirement of these features is 
satisfied; notably, the relevant XP or X must be related in terms of syntactic 
features with the feature(s) on the head. The F head may surface with different 
features, including the following: neg, foc, wh, inter, topic. In CG negative 
clauses, for example, as well as in clauses headed by wh–elements or modal 
particles, inter alia, the features of the F head are satisfied, so the verb does not 
move higher than the cliticization site and proclisis obtains. When no appropriate 
XP or X is available to satisfy the PF–requirement of the relevant feature(s), the 
verb cannot be spelled out in the lower T position, and moves to F at PF to satisfy 
the PF–requirement; this is possible as both the verb in T and the F head share a 
V–feature. Verb movement is manifested in another two instances. First, when an 
XP or X is present but cannot establish a syntactic relationship with the F–head. 
Second, when an XP is present and able to establish a syntactic relationship with 
the F–head but unable to satisfy the PF requirement of the relevant feature(s) (i.e. 
when it is not merged at the edge of the F–projection) (Mavrogiorgos 2012:33). 
As a result, reordering of the clitic–verb cluster is manifested at PF and enclisis 
obtains. 
Crucially, the main assumption within Mavrogiorgos’ account is the PF 
nature of verb movement in finite enclisis: the verb moves to F, if the 
PF/morphological requirement imposed by the F–head has not been satisfied at 
syntax. 
 
3.4.2. The PF!controlled spell!out of copies 
 
Revithiadou (2006) offers an interface account for cliticisation in CG in the spirit 
of Franks (1998, 2000), Bo!kovi" (2000, 2001) and Franks and Bo!kovi" (2001). 
She proposes that the syntax deals with movement operations and provides pairs 
of equally well!formed syntactic structures (with the clitic placed both pre! and 
                                                
26 Mavrogiogos takes these inflectional features to be affixal, hence, the triggering of V–
movement to the relevant C–head at PF (2012:30, footnote 11). 
  
85 
 
post!verbally) and phonology has a filtering role on the syntactic output. A basic 
tenet of her analysis is that the prosodic system not only determines how clitics 
are to be incorporated in the prosodic structure (as Prosodic Words or 
Phonological Phrases), but also whether they will precede or follow the verb.  
Revithiadou claims that syntax alone cannot provide a satisfactory account 
for the CG facts and bases her argumentation on examples like (9!10), which 
show that a complementizer does not always provide a context for a pre!verbal 
clitic (9) and that the presence of a heavy wh!phrase like pjos pu ulus tus 
athropus in (10) cannot guarantee the clitic!verb order (Revithiadou 2006:93). 
She takes these facts to indicate that constraints on phonological phrasing are 
involved in clitic placement. 
 
(9) Lali        oti  /  pos  ton                    aghapa  /  aghapa  ton. 
            Say!3S  that           him!CL.ACC  love!3S  
 “S/he says that s/he loves him” 
(10) Pjos   pu  ulus                (tus athropus)               ethkjavasen  to? 
Who  of   all!ACC.PL  (the people!ACC.PL)  read!3S        it!CL.ACC  
 “Which one out of all (the people) read it?” 
 
Revithiadou extends Bo"kovi#’s (2001) analysis of second position effects 
in Serbo!Croatian and Bulgarian to account for clitic placement in CG. Bo"kovi# 
(2001) proposes that syntax places the clitics both pre! and post!verbally and 
pre!specified phonological matrices select the optimal case. This analysis is built 
upon ideas in Franks (1998) and Franks and Bo!kovi" (2001), where it is 
suggested that pronunciation of chains27 created by movement of some element is 
decided in PF, and hence the deletion of copies is phonologically motivated. 
Within this account, the default case is for a chain to be pronounced in the head 
position28, 29, unless this leads to a PF violation. In the latter case, a lower member 
                                                
27 Chomsky (1993, 1995) argues that an element that undergoes movement leaves behind a copy 
so that syntax does not create new elements (Inclusiveness Condition). Syntactic operations 
conform to the IC and only re!arrangement of elements already inserted from the lexicon is 
possible. 
28 This is imposed by the HEAD CHAIN, a constraint that determines that the highest copy of the 
chain is realised. Revithiadou claims that this constraint is based on PRONOUNCE HIGHEST (see 
footnote 29). The “head of a chain” is the highest copy in a series of copies left behind by the 
  
86 
 
of the chain is pronounced, which best satisfies the prosodic requirements of the 
language. 
Within Revithiadou’s (2006) account, clitics in CG are generated in VP 
and they then left!adjoin to the inflectional head to which the verb raises. The 
decision of which copy of the clitic will be pronounced relies on the prosodic 
system, which performs the overall prosodic organisation of the syntactic strings. 
The assignment of prosodic structure to all syntactic strings ! to clitic 
constructions as well ! adheres to the language!specific hierarchy of prosodic 
constraints (p!constraints), comprised of prosodic domination and alignment 
constraints. The copy of the clitic that best satisfies these constraints is ultimately 
spelled out. 
In CG, when no other material is located before the verb, the lower copy 
of the clitic, namely the one immediately following the finite verb, is obligatorily 
realised. Conversely, when some element that can satisfy the clitic’s prosodic 
requirements is present, the head of the chain is pronounced, namely the clitic that 
immediately precedes the finite verb. In imperatives, however, where the verb 
raises as high as C (Terzi 1999a), the phonology has no option but to parse the 
lower clitic. 
Turning now to the specifics of the prosodisation of pronominal clitics in 
CG, I offer a summary of Revithiadou’s (2006) approach, which is formulated 
within Optimality Theory (Prince & Smolensky 1993) and is built on Selkirk’s 
(1995) assumptions on the prosodisation of function words. Revithiadou’s (2006) 
central claim is that the ranking of p!phrasing constraints determines not only 
how CG clitic pronouns are prosodically organised, but also how they are ordered 
with respect to the verb. She uses Selkirk’s (1995) prosodic typology of clitics, 
cited below (11) (Revithiadou 2006:84), to show how CG clitic!host 
combinations are prosodically organised.! 
 
(11) a. [word cl]PrW internal clitic 
 b. [[word]PrW  cl]PrW affixal clitic 
                                                                                                                                 
movement of some element (Revithiadou 2006:96, note 18). 
29 Franks assumes the constraint PRONOUNCE HIGHEST that states that “lower identical copies 
are silent” (2000:28). 
  
87 
 
 c. [[word]PrW  cl]PPh free clitic 
 d. [word]PrW  [cl]PrW prosodic word 
 
According to the four major patterns outlined in (11), clitic pronouns can be 
realised as: 
1. Internal clitics (11a), when dominated by the PrW of the lexical word, i.e. of 
the verb. 
2. Affixal clitics (11b), when they adjoin to the PrW of their host, forming a 
nested PrW structure. 
3. Free clitics (11c), in which case they skip the intermediate PrW!level and are 
parsed by the higher phonological phrase (PPh). 
4. Independent prosodic words, as in (11d). 
The patterns in (11) arise from the interaction of Selkirk’s constraints on 
prosodic domination (Selkirk 1995:443), outlined in (12) [where Cn = some 
prosodic category], and a set of alignment constraints outlined in (13) 
(Revithiadou 2006:85). 
 
(12) a. LAYEREDNESS: No Ci dominates a Cj, j >  i. 
 b. HEADEDNESS: Any Ci must dominate a Ci-1. 
c. EXHAUSTIVITY: No Ci immediately dominates Cj, j <  i - 1. 
d. NON RECURSIVITY: No Ci dominates Cj, j = i. 
(13) a. WORD CONSTRAINT  (WCON): Align (LexWord, L/R; PrW, L/R) 
b. PRW CONSTRAINT  (PCON): Align (PrW, L/R; LexWord, L/R) 
 
The alignment constraints control the mapping of morpho!syntactic 
constituents to prosodic structure and require that the edges of a prosodic 
constituent are aligned with the edges of a morpho!syntactic constituent 
(Revithiadou 2006:85). For Revithiadou (2006:85, 2008:1399 and references 
therein) the alignment constraints refer only to the edges of lexical but not 
functional categories. Hence, with respect to clitic constructions, the alignment 
constraints are biased towards an output in which the verb sustains alignment 
between its morphological and phonological edges, and therefore an output in 
  
88 
 
which the clitic prosodises with the function word and not with the lexical word, 
i.e. the verb, is favoured.  
Revithiadou adopts the constraints on prosodic domination in (12) and the 
alignment constraints in (13) to develop the constraint ranking for clitic 
prosodisation in CG, outlined in (14). Apart from the constraints in (12) and (13), 
(14) also includes FAITH, a constraint that spells out the following rule: “Preserve 
an inherent accent of the input in the output” (Revithiadou 1999; 2006:87). 
Moreover, two constraints on prosodic domination are precluded from (14), 
namely LAYEREDNESS and HEADEDNESS. This is because both constraints are, 
according to Revithiadou, universally inviolable, and can therefore not be 
dominated, in Optimality theoretic terms. In contrast, EXHAUSTIVITY and NON 
RECURSIVITY are violable and favour an alignment that causes the fewest 
violations. In effect, the ranking of EXHAUSTIVITY and NON RECURSIVITY has 
an important role to play in clitic prosodisation and placement. All in all, the 
hierarchy in (14) determines the prosodisation of clitic pronouns in CG, namely 
whether they prosodically incorporate into their host or adjoin to it, as well as 
their ordering with respect to the verb. Notably, when alignment constraints are 
outranked by the constraints on prosodic domination, clitic incorporation into the 
PrW of the host is induced. 
 
(14) FAITH (acc), EXH, WCON (L), NON REC >>  PCON, WCON (R) 
 
On the basis of Selkirk’s (1995) clitic typology, Revithiadou (2006, 2008) 
argues that CG enclitics are identified as internal clitics, in that they incorporate 
into the same prosodic word (PrW) as their verbal host. She justifies her claim on 
the basis of the following facts. Firstly, sandhi rules apply in verb!clitic clusters; 
recall that the domain of application of such rules is the PrW. Examples such as 
/'ipes mu/ (“you told me”) parsed as: ['ipezmu]PrW, where s!voicing applies, and 
/'fere to/ (“bring it”) parsed as: ['ferto]PrW, where e!deletion applies, offer the 
necessary pieces of evidence. Secondly, a secondary/rhythmic stress develops to 
repair violations of the three!syllable stress window rule caused by clitic 
insertion. For example, a double clitic construction like /'ipen mu to/ (“s/he said it 
  
89 
 
to me”) will be stressed as follows: /'ipem 'mu to/ (Revithiadou 2006:84, 2008: 
section 3.2.2)30; see also the discussion in section 2.2.5 of chapter 2. 
Within Revithiadou’s (2006, 2008) analysis, clitic pronouns in CG need to 
have some other constituent on their left in order to appear pre!verbally. Function 
words such as modal markers, negation particles and complementizers, provided 
they lack stress, incorporate together with the clitic into the verbal host. Crucially, 
the function word–clitic sequence in CG remains un!footed and adjoins 
recursively to the PrW of the verb as follows: [fnc cl [V]PrW]PrW. In this case, 
clitics are realised as affixal elements.  Evidence that the function word, the clitic 
and the verbal host are parsed into the same PrW is offered by the application of 
sandhi rules, such as s!voicing and intervocalic voiced fricative deletion 
(Revithiadou 2006; 2008) in this type of construction 31 . Examples (15), 
(Revithiadou 2006:86) and (16) (Revithiadou 2008:1403) respectively offer 
evidence for the application of these operations. 
 
(15) /pos      mas                ghi'refki/ 
 COMP  us!CL.ACC  looks for!3S 
 “That s/he looks for us” 
Parsed as: [poz maz ghi'refki] 
(16) /na  mu                  'dhokis/ 
 M   me!CL.GEN  give!2S 
 “To give me” 
Parsed as: [na mu 'okis] 
                                                
30 The use of this secondary or rhythmic stress in verb!clitic combinations is also mentioned by 
Mavrogiorgos as evidence in favour of the argument that enclitics in SMG attach to their host’s 
PrW (2009:54!55). 
31 Revithiadou (2006:86, 2008: section 3.2.2) assumes that in CG, apart from s!voicing and 
intervocalic voiced fricative deletion, another rule is applicable, namely the fusion of the /u, o/ + 
/e/ sequence into [o]. I use example (1) (Revithiadou 2006:86) to illustrate this point. However, in 
my grammar, the fusion of /u/ + /e/ sequence into [o] in such an example is not felicitous, unlike 
the fusion of the /o/ + /e/ sequence into [o], which is; see example (2).   
1.        /pos       tu                      e'thkjavasen/ 
 COMP  him!CL.GEN  read!3S 
 “That s/he read to him”  
Parsed as: [pos to 'thkjavasen] 
2. /pos tu to e'thkjavasen/ 
 “That s/he read it to him”  
Parsed as: [pos tu (t)o 'thkjavasen] 
 
  
90 
 
Revithiadou points out the difference between an enclitic, which 
incorporates into the same PrW as its verbal host, and a proclitic with an 
unstressed function word, which joins into a recursive PrW structure with the 
verb. She takes the following observation as evidence for her claim: 
re!syllabification is only allowed with enclitics, as in /'ipes mu/ ['ipe.zmu] (“you 
told me”), but is blocked with proclitics, as illustrated in /pos mas ghi'refki/ 
[pozmaz.ghi'refki] (“that s/he looks for us”) (2006:86). 
If the function word that heads the clitic construction is stressed, a 
different picture arises: the clitic attaches to the function word to its left to form 
with it an independent prosodic constituent and the verb is realised into a separate 
PrW. Examples (17!18) taken from Revithiadou (2006:87) illustrate this point. 
 
(17) /'pcos  tu                     e'thkjavasen/ 
 who   him!CL.GEN  read!3S 
 “Who read to him?” 
Parsed as: ['pjostu]PrW  [e'thkjavasen]PrW  
(18) /'enna  sas                    'dhoki / 
 M        you!CL.GEN  give!3S  
 “S/he will give you” 
Parsed as: ['enna sas]PrW ['dhoki]PrW 
 
A piece of evidence that supports the proposed prosodic pattern comes 
from vowel epenthesis (Revithiadou 2008:1402). When epenthesis occurs in CG 
the epenthetic vowel i is inserted between the clitic and the verb in order to repair 
illicit consonant clusters, as in (19) (Revithiadou 2008:1402). Crucially, vowel 
epenthesis does not apply within the PrW domain. This suggests that the 
functional word–clitic string is prosodically hosted by the verb but only at a 
recursive level; if the clitic were prosodically incorporated into the verb, 
assimilation rather than vowel epenthesis would apply.  
 
(19) /'enna  ton                  – i – 'ftaso/ 
 M        him!CL.ACC         reach–1S  
  
91 
 
 “I will reach him” 
  
Revithiadou (2006) claims that further evidence in favour of the argument 
that the clitic and the verb do not belong to the same PrW in constructions with 
stressed functional elements is offered by examples like: /'pcos tu e'thkjavasen/ in 
which vowel!fusion (of an underlying /u, o/+/e/ sequence into [o]) between the 
final vowel of the clitic and the initial vowel of the immediately following verb is 
blocked, and /'enna sas 'dhoki/ where s!voicing of the final /s/ of the clitic is 
blocked, even though it is immediately followed by a fricative. Recall that 
s!voicing and vowel fusion are PrW rules. 
The alignment for the clitic!verb cluster when preceded by a stressed 
function word is determined by the ranking of the constraints on prosodic 
domination outlined in (14). Specifically, since FAITH is ranked above PCON, the 
prosodic boundaries of the clitic pronoun cannot fuse with the prosodic 
boundaries of the verbal host. Moreover, the incorporation of the pronominal 
clitic into the stressed function word elicits fewer violations of PCON, since the 
latter prohibits the matching of non!lexical morphological boundaries to prosodic 
ones.  
Coming back to example (9), let us see how this can be accommodated 
within Revithiadou’s (2006) account. The alternation in clitic placement is the 
result of variation in p!phrasing and, in particular, of the enforcement of the 
binarity requirement. This constraint requires that pairs of PrWs are grouped 
together in p!phrases. The p!phrasing when the clitic is post!verbal is performed 
as follows: [[la'li]PrW ['oti]PrW ]PPh, with the first p!phrase complying with binarity. 
The p!phrasing when the clitic is pre!verbal is performed as follows: [['oti 
ton]PrW [agha'pa]PrW]PPh, with the second p!phrase complying with binarity. 
Following the same line of reasoning we can accommodate (10) as well: while a 
wh!element and the clitic form a single p!phrase, the heavy wh!phrase in (10) 
forms a p!phrase individually, forcing the verb and the clitic to form another 
p!phrase, within which the clitic occupies the position on the right of the verb. 
The same line of reasoning applies to constructions involving adjuncts. The 
blocking of cliticisation across an adjunction boundary can be attributed to the 
  
92 
 
fact that peripheral adjuncts, such as clitic!doubled DPs, form independent 
p!phrases (Revithiadou 2006:93). 
Finally, under the proposed framework, the dialectal variation between CG 
and SMG is captured as well32. SMG adheres to the HEAD CHAIN constraint, 
hence the highest copy of the chain is always realised and proclisis obtains. The 
lower copy is only realised when the verb raises to a functional projection higher 
than I, i.e. in imperatives and gerunds. In CG the surface position of clitics is 
regulated by the prosodic system, on the basis of the language!specific constraint 
ranking outlined in (14). 
In sum, within Revithiadou’s (2006) analysis, CG clitics appear 
pre!verbally when a functional constituent is realised in the same CP on the left 
of the clitic and post!verbally when no such element is present. With respect to 
the prosodisation of clitic constructions in CG, it is suggested that this is regulated 
by the constraint ranking in (14), specifically concerning the hierarchy of 
constraints on prosodic domination (12) and the alignment constraints (13). Clitic 
pronouns in CG are prosodically organised in the following three ways 
(Revithiadou 2006:99; 2008:1404): 
1. As internal enclitics: [V cl]PrW 
2. As affixal proclitics (with unstressed function words): [fnc cl [V]PrW]PrW 
3. As prosodic words (with stressed function words): [fnc cl]PrW [V]PrW 
Post!verbal clitics are always incorporated into the PrW of their verbal 
host. On the basis of the constraint ranking in (14), repeated here for ease of 
reference: [FAITH (acc), EXH, WCON (L), NON REC >> PCON, WCON (R), 
WCON L], the optimal structure for the clitic is to incorporate into the PrW of its 
verbal host: [V cl]PrW and not to recursively adjoin to it: [cl [V]PrW ]PrW (or its 
mirror image). Pre!verbal clitics are either parsed together with the preceding 
(stressed) function word or they join into a recursive PrW structure with the 
(unstressed) function word and the verb. When a stressed function word is 
involved, the template [fnc cl]PrW [V]PrW is preferred over [fnc]PrW [cl [V]PrW]PrW, 
                                                
32 Revithiadou (2008) takes this claim one step further and suggests that the phonological 
component determines the evolution of cliticisation in Greek and, more specifically, the transition 
from a system with (mostly) second position clitics (2P), like CG, to a non!2P system, like SMG.  
 
  
93 
 
so that the verb (lexical word) will keep the transparency of its edges 
(Revithiadou 2008:1408). With unstressed function words, the template [fnc cl 
[V]PrW]PrW is preferred over [fnc [V]PrW cl]PrW because in the former case the verb 
preserves at least its right alignment with the PrW (Revithiadou 2008:1408!09). 
All in all, Revithiadou’s (2006) proposal underlines the truly interface!dependent 
nature of cliticisation in CG and offers an account of exactly how the syntax 
interacts with phonology. 
 
3.5. A dynamic account  
 
Chatzikyriakidis (2010, 2012) has recently put forward a dynamic account for 
clitic positioning in Cypriot Greek. This account is developed within the 
framework of Dynamic Syntax (DS), a processing oriented framework. Its basic 
tenet is that natural language syntax can be seen as “the progressive accumulation 
of transparent semantic representations with the upper goal being the construction 
of a logical propositional formula” (Chatzikyriakidis 2012:648). For DS, syntactic 
differences across languages derive from different lexical specifications. As 
regards clitic placement in CG, and in general, it is treated as a lexical 
phenomenon. In particular, it is assumed that clitic pronouns are inserted as 
lexical entries and three generalized parsing strategies (general actions of structure 
building, i.e. the unfixed node strategy) are employed, depending on the structural 
context, as triggers for parsing the clitic pronouns. Positioning restrictions are, 
thus, viewed as a number of triggering points that, if satisfied, regulate clitic 
positioning. This approach won’t be further discussed as the framework within 
which is developed is at odds with the assumptions regarding the computational 
system adopted in this thesis. The interested reader is directed to Chatzikyriakidis 
(2010, 2012) for a thorough presentation of both the framework of Dynamic 
Syntax and the proposed analysis for clitic positioning in CG.   
 
 
 
 
  
94 
 
3.6. Interim conclusions for clitic placement in CG 
 
On the basis of the above discussion and taking into account the various aspects of 
clitic placement in CG, a number of conclusions can be drawn: 
1. The requirement of CG clitics for verb adjacency illustrates their structural 
and/or prosodic dependency on their verbal host. 
2. Their ban from clause initial position illustrates the effect of prosodic 
requirements on the well!formedness of clitic constructions in CG. 
3. CG clitics are not inherently enclitics. 
4. Syntax plays an important role in clitic placement in CG. 
5. The enclisis!proclisis alternation in CG is not contingent on the finiteness of 
the verbal host. 
6. The enclisis!proclisis alternation in CG cannot be attributed either to pure 
syntactic or to pure prosodic operations. 
Let us briefly discuss these concluding remarks one by one. 
 
3.6.1. Verb adjacency and structural and/or prosodic dependency 
 
The requirement for verb adjacency imposed on CG clitics illustrates the 
structural and/or prosodic dependency of clitic pronouns on their verbal host. This 
property is not adequately accounted for in the majority of the accounts proposed 
so far. Terzi (1999a, 1999b) and Mavrogiorgos (2012) assume that the clitic and 
the verb adjoin to distinct functional projections. Condoravdi and Kiparsky (2001) 
suggest that clitics in CG are prosodically required to encliticise to a constituent 
on their left, with no special reference to the (always adjacent) finite verb. 
Agouraki’s proposal (2001) is even weaker in this respect, as it treats clitic 
placement as a mere epiphenomenon of verb placement. However, within 
Revithiadou’s (2006) account this is adequately accommodated. 
 
3.6.2. The ban from clause initial position 
 
CG clitics can never appear clause initially. This Tobler–Mussafia effect indicates 
  
95 
 
that their placement depends on prosodic constraints as well. However, the 
position that clitics occupy in the clause does not seem to depend on any inherent 
preference for enclisis, as argued by Condoravdi and Kiparsky (2001). The 
approach I adopt with respect to the ban of clitic pronouns from clause initial 
position in CG is in line with Mavrogiorgos’ (2012) proposal: this seems to be 
related to an absolute PF requirement stating that clitics cannot appear in the first 
position of a prosodic domain represented by the utterance. 
 
3.6.3. CG clitics are not inherently enclitics 
 
Revithiadou (2006, section 3.1.) discusses the prosodic behaviour of CG clitics 
and shows that they do not always encliticise to a constituent at their left, but 
occasionally procliticise to a constituent on their right. She bases her 
argumentation on constructions such as: /pos mas ghi'refki/ parsed as follows: 
[poz maz ghi'refki] (“That s/he looks for us”) (2006:86). In such a construction 
s!voicing, whose domain of application is the PrW, is applied on the clitic–verb 
cluster, showing that the clitic ultimately procliticises to the verb; it recursively 
adjoins to the PrW of the verb together with the function word. These facts 
question the assignment of an enclitic template to CG clitics (as per Condoravdi 
and Kiparsky 2001).  
 
3.6.4. The role of syntax 
 
A general theoretical conclusion regarding the proclisis!enclisis alternation in CG 
is that it complies with various structural rules and restrictions imposed on clitic 
pronouns. As already discussed, CG clitics must be verb!adjacent. Moreover, 
clitic placement in CG is regulated by the presence/absence of a functional head 
c!commanding the cliticisation site (Agouraki 2001, Mavrogiorgos 2012, Terzi 
1999a, 1999b). The selective blocking effect of different types of XPs/Xs on 
enclitic orders establishes the crucial role syntax plays in clitic placement in CG. 
Finally, as observed by Revithiadou (2006:92), clitic constructions obey locality 
restrictions. 
  
96 
 
3.6.5. Enclisis!proclisis alternation and finiteness 
 
In Mavrogiorgos’ (2012) classification, clitic languages belong to two different 
groups depending on whether finiteness determines clitic placement. In 
finiteness–sensitive languages, such as SMG, Romanian, Italian, Catalan and 
Spanish, proclisis is manifested with finite verbs and enclisis with non!finite 
verbs (or verbal hosts which are defective in terms of tense, mood and/or subject 
agreement). In Tobler–Mussafia languages, on the other hand, both finite and 
non!finite enclisis is attested, indicating that clitic placement is not contingent on 
the finiteness of the verbal host. CG, European Portuguese, Old Romance and 
Medieval Greek belong to the latter type. 
However, with regard to the derivation of non!finite enclisis, CG 
resembles SMG and other finiteness–sensitive languages in that it involves 
syntactic movement of the verb past the cliticisation site. The assumption of the 
manifestation of V!to!C by the imperative verb is in line with the majority of the 
approaches discussed above (Agouraki 2001, Terzi 1999a, Revithiadou 2006). I 
adopt Mavrogiorgos’ (2009) proposal that non!finite enclisis correlates with the 
presence of an unvalued person feature in T (suggested for SMG imperatives), 
and argue that this also accounts for CG imperatives on the basis of the similarity 
between CG and SMG imperative clauses (20). According to Mavrogiorgos, the 
blocking effect of modal and negative particles on the movement of the 
imperative verb in CG can be explained in syntactic terms: it is either an 
intervention effect or an instance of complementary distribution between the 
(modal or negative) particle and verb!related features (2012:19!20). 
 
(20) (*Na  /  *Min)  Fer           to                  piso   amesos.  (CG / SMG) 
 M         NEG   bring!"#  it!CL.ACC  back  immediately 
“(*To/Not) Bring it back immediately” 
 
In sum, finite and non!finite enclisis in CG cut across different lines, in 
spite of the similarities in the spell!out structure. The V!to!C movement of the 
imperative verb in non!finite enclisis is attributed to syntax, while finite enclisis 
  
97 
 
is the result of operations at the syntax!phonology interface. 
 
3.6.6. Enclisis!proclisis alternation: neither pure syntax nor pure PF 
 
Clitic placement in CG cannot be attributed either to purely syntactic or to purely 
prosodic operations. A purely prosodic account, like the one put forward by 
Condoravdi and Kiparsky (2001), is challenged on the basis of the following facts. 
First, CG clitic pronouns can be both proclitic and enclitic to their host (cf. section 
3.6.3). Second, it is not the case that any preceding element can satisfy their 
PF!requirements: some of the elements that surface in the CP domain trigger 
proclisis (e.g. the factive complementizer pu), while others trigger enclisis (e.g. 
topics). Moreover, complementizers like oti and pos trigger both patterns. In 
addition, as Revithiadou (2006:92) observes, it is theoretically undesirable to have 
an excessively powerful phonology that can move syntactic entities. 
However, prosody cannot be disregarded either and a purely syntactic 
account is also ruled out on the basis of the following facts. First, 
PF!requirements that are independent of the clitic PF properties play a crucial 
role in clitic placement in CG, as clitic constructions are subject to certain 
phonological well!formedness constraints, e.g. the ban from clause initial position 
(section 3.6.2). Second, all the syntactic analyses so far proposed lack any real 
motivation for verb movement33, which is claimed to constitute the triggering 
force for post!verbal placement in finite enclisis (cf. section 3.2.1, 3.2.2). 
Notably, as Terzi (1999b:232!233) observes, verb movement in CG 
cannot be attributed to strong features of the finite verb, in line with proposals for 
EP (another Tobler!Mussafia language). For EP there is independent evidence 
                                                
33  Mavrogiorgos (2012:23!24) has shown that for SMG the verb movement across the 
cliticisation site is not a sufficient condition for enclisis. Compare examples (1) and (2). It is 
obvious that no re!ordering of the verb and the clitic is possible irrespective of the position that 
the clitic!V / V!clitic cluster occupies. 
(1) Ta    kerdhi                 tu              schedhon  ta                        triplasiase       
The  winnings!ACC  his!POSS  almost     them!CL.ACC  tripled!3S  
o     Jianis.  
the  John!NOM 
(2) Ta kerdhi tu ta triplasiase schedhon o Jianis. 
 “John almost tripled/already tripled his winnings” 
 
  
98 
 
justifying this line of reasoning. First, EP is a language with inflected infinitives. 
Second, EP, unlike CG, allows VP!deletion in examples like (21) (Terzi 
1999b:232). Thus, it is not unjustified to assume that the strong " or F features of 
the EP verb trigger V!to!" or V!to!F movement, which results in enclisis. 
However, given the absence of both inflected infinitives and verb deletion, this 
assumption cannot be extended to CG. 
 
(21) Deste!lhe   o     livro 
Gave  him  the  book   
(Sim),  dei. 
Yes,     gave. 
 
Finally, as Revithiadou (2006:82!83, 89) observes, both the prosodic and 
syntactic approaches fall short in explanatory power with respect to constructions 
headed by wh!elements. While wh!words typically trigger pre!verbal clitic 
placement, in some (heavy) wh!phrases the clitic pronoun encliticises on the 
verbal host (cf. section 3.4.2). 
 
3.7. Summary 
 
This chapter has discussed a number of accounts of clitic placement in CG ! 
syntactic, prosodic and interface–based – offered over the last twenty years. The 
review has included the V!to!M proposals for finite enclisis put forward by Terzi 
(1999a, 1999b), Agouraki’s (2001) proposal for the filled!C requirement of CG, 
the syntax!PF interface analysis for Tobler–Mussafia languages by Mavrogiorgos 
(2012), Revithiadou’s (2006) proposal for a PF!controlled spell!out of copies, 
and the Prosodic Inversion account by Condoravdi and Kiparsky (2001), 
summarised in the last section. The discussion in this chapter provides the 
background for the hypotheses and predictions to be tested in the empirical 
section. 
  
99 
CHAPTER 4: THE L1 ACQUISITION OF CLITICS 
 
4.1. Introduction 
 
This chapter discusses first language acquisition (L1A) of clitic pronouns cross–
linguistically. Recall that, following Mavrogiorgos (2012), I adopt a tri–partition 
with respect to the types of clitic languages, distinguishing finiteness–sensitive 
languages (like Catalan, Italian, Romanian, Spanish and Standard Modern Greek 
amongst others), Tobler–Mussafia languages (like Cypriot Greek, European 
Portuguese and Galician), and second position languages (e.g. Serbo–Croatian). 
Clitic production in early language has two aspects: the emergence of 
clitic pronouns and their placement. The focus of the majority of studies so far 
conducted on clitic L1A has been clitic realisation and omission and the attested 
dichotomy among early clitic languages: some exhibit clitic omission and some 
others exhibit adult–like clitic production1. The discussion of early CG and early 
EP data reveals another interesting, albeit under–studied, aspect of early clitic 
production: the phenomenon of clitic misplacement. This phenomenon is absent 
in the majority of clitic languages, with the interesting exceptions of CG and EP. 
The presence/absence of clitic misplacement in early grammars offers an 
alternative way of classifying clitic languages based on the placement rather than 
on the emergence of clitic pronouns in child language. 
The organisation of the current chapter follows the aforementioned tri–
partition: the first section discusses acquisition studies in finiteness–sensitive 
languages, including Catalan, Italian, Romanian, Spanish and SMG. The second 
section focuses on Serbo–Croatian, a representative example of a language with 
second position restrictions with regard to clitic placement. The third section 
presents acquisition data from Dutch, a language exhibiting clitic scrambling, and 
the fourth section discusses clitic L1A in Tobler–Mussafia languages, including 
CG, EP and Galician. 
 
                                                
1 See Tsakali and Wexler (2004), Tsakali (2006) and Wexler et al. (2004) for a proposal put 
forward to account for this divergence between clitic languages.  
  
100 
4.2. Finiteness–sensitive languages 
 
Clitic placement in most Romance languages, including Italian, Romanian, 
Spanish and Catalan (with the interesting exception of EP), as well as in SMG 
depends on the finiteness of the verbal host2, hence reference will be made to 
finiteness–sensitive languages, a term suggested by Mavrogiorgos (2012). A 
number of influential studies on clitic L1A in Catalan, Spanish (Wexler et al. 
2004), Italian (Guasti 1993/94, Schaeffer 2000), Romanian (Avram 1999, Avram 
& Coene 2007, Babyonyshev & Marin 2006) and SMG (Marinis 2000, Stephany 
1997, Tsakali 2006) have revealed patterns of clitic omission for some of these 
languages, to a greater or lesser degree, whereas instances of clitic misplacement 
were sparsely attested. 
 
4.2.1. Italian 
 
The L1 acquisition of clitic pronouns in Italian was investigated by Guasti 
(1993/94) and Schaeffer (2000), amongst others, on the basis of spontaneous and 
experimental data respectively. Clitics in Italian, parallel to other finiteness– 
sensitive languages, surface pre–verbally in finite clauses and post–verbally in 
infinitival clauses. When the infinitive is governed by modal, causative or 
aspectual verbs the clitic may either immediately follow the infinitive or 
immediately precede the finite verb; the latter pattern is known as clitic climbing 
(Guasti 1993/94:13). 
Guasti takes clitic placement as a reliable cue to establish children’s 
knowledge of the (non–)finite nature of the relevant verbal host, and studied 
natural production data from 3 monolingual Italian children: Martina (1;8–2;7), 
Diana (1;10–2;6) and Guglielmo (2;2–2;7) (CHILDES database). Her analysis is 
based on a total of 534, 660, and 217 utterances for Martina, Diana, and 
                                                
2 An interesting proposal put forward by Mavrogiorgos (2009) builds on the correlation of non–
finite enclisis with the presence of an unvalued person feature in T to explain the enclisis–proclisis 
alternation in SMG, a proposal that may be applicable to other languages that fall within this 
group. 
  
101 
Guglielmo, respectively, and the individual results are summarised in table 4.13, 4. 
The table presents raw numbers of (pre–verbal) clitics produced or omitted in 
finite contexts. As for infinitival contexts, no child produced any post–verbal 
clitics following an infinitive before the age of 2; Martina produced 1 clitic 
between ages 2;1 and 2;7, Diana produced 2 clitics from age 2;1 to 2;6 and 
Guglielmo produced 5 clitics between 2;2 and 2;7 years of age (see table 8 in 
Guasti 1993/94:18). 
 
Table 4.1: Production of pre–verbal clitic pronouns in early Italian (based on 
tables 5–7 in Guasti 1993/94:15–16). 
 
The results obtained reveal that Italian children have low clitic production 
up to the age of 2;3. From approximately age 2;5 onwards (2;3 for Martina; 2;5  
for Diana; 2;7 for Guglielmo) their clitic production is good. Guasti attributes the 
rather delayed acquisition of cliticisation and the optional use of clitics at the 
onset of L1A to difficulties in forming A–chains5 (in line with Antelmi 1992) and 
                                                
3 Guasti disregards from her calculations clitics for which a corresponding complement does not 
exist, including impersonal si, inherent si or the clitic ci, which is not used in Standard Italian 
(1993/94:17). 
4 I only report the results for the production of clitic pronouns, leaving aside the other two types of 
syntactic elements produced by children and reported by Guasti, namely cliticisable complements 
and pre–syntactic devices. In Guasti’s wording the latter constitute “undifferentiated phonetic 
segments that are likely to be the precursors of certain syntactic entities” (1993/94:17). 
5 Guasti assumes that clitics are originally inserted as heads of a maximal projection in the 
complement position. Then, they undergo A–movement to the Specifier of AgrOP and from there 
the head containing the clitic undergoes head–movement to the designated functional head I 
(1993/94:18–19). 
Age Martina Diana Guglielmo 
Omission Pre–verbal Omission Pre–verbal Omission Pre–verbal 
1;8 3 8   
1;9 – 2 
1;10 5 4 1 4 
1;11 14 2 2 1 
2;0  3 3 
2;1 3 5 4 13 
2;2   2 7 
2;3 8 17 1 3 
2;4 6 18 1 14 
2;5 1 11 2 47 – 10 
2;6  7 107  
2;7 2 25  4 42 
  
102 
to the optionality of functional categories in child Italian grammar6. 
A legitimate objection to Guasti’s argumentation is that the absence of A–
chains from child Italian grammar should result in clitic realisation in base–
generation positions, as already pointed out by a reviewer (1993/94:19, footnote 
9). Guasti’s response to this criticism is that Italian children are aware of the 
phonological properties of clitics, which force them to move, and hence clitics 
never appear where originally inserted. Moreover, it remains unclear why clitic 
production in infinitival contexts in early Italian is delayed as compared to finite 
contexts, given that clitic placement in infinitival contexts is assumed to involve a 
one–step A–movement similar to that in finite contexts. Guasti refers to the 
delayed production of clitics with infinitives (1993/94:20, footnote 12) and 
assumes that this is due to the fact that relevant infinitival contexts, namely those 
in which the infinitive is not governed by some matrix verb and the use of a post–
verbal clitic is obligatory, are rather infrequent in child language. This results in a 
restricted number of contexts imposing the use of a post–verbal clitic, hence the 
low production rate.  
Returning to clitic placement, Guasti reports that systematic errors of clitic 
placement were not attested in the database examined: when Italian children use 
clitics, their placement is adult–like (1–2) (Guasti 1993/94:14). 
 
(1) Lo       naccondi  su. 
It–CL  hide–2S   up 
“(You) hide it up” (Martina, 1;8) 
(2) Mi  vieni          prendere? 
Me  come–2S  pick–Inf  up 
“Do (you) come (to) pick me up?” (Guglielmo, 2;3) 
 
Schaeffer’s (2000) study confirms Guasti’s (1993/94) findings. In fact, 
Schaeffer’s results are more robust in showing that clitics are omitted in early 
                                                
6 The claim advanced in Guasti (1993/94) is that Italian children optionally project functional 
categories, such as clitics, at the onset of L1A. She attributes this optionality to their incomplete 
mastery of the referential system associated with the relevant category. 
 
  
103 
Italian. Schaeffer studied clitic production in 35 Italian children, aged 2;1 to 5;11. 
Her methodology was a combination of a truth value judgement task with an 
elicited production task. She elicited constructions involving single and double 
clitics in the following contexts: present tense constructions, restructuring verb 
constructions and passato prossimo constructions (with and without agreement)7. 
Table 4.2 summarises Schaeffer’s results.  
 
 
Table 4.2: Overall proportion of overt and omitted direct object clitics and 
production of full NPs (in obligatory clitic contexts) (Table 7 in Schaeffer 
2000:76) 
 
The most striking results are the significantly low proportion of obligatory 
clitics produced by 2–year–olds (22%) and the huge developmental leap from 2 to 
3 years of age in which the proportion of obligatory clitics produced nearly trebles 
to 62%. An example of an elicited clause with an omitted clitic is offered in (3) 
(Schaeffer 2000:78). Schaeffer attributes this outcome to the optional marking of 
referentiality in early Italian grammar8. 
 
(3) Raja: Il     coniglio  lava       il     pupazzo!   
         The  rabbit      washes  the  puppet 
         “The rabbit is washing the puppet” 
Child: No,  pettina! 
 No  combs. 
“No, (she) is combing” (A 2;5) 
                                                
7 In adult Italian in root clauses clitics precede the finite verb, in passato prossimo they precede 
the auxiliary and in clitic climbing constructions they either precede the finite modal or follow the 
infinitive (Schaeffer 2000). 
8 In particular, Schaeffer claims that “if the non–overt pro object is not marked for referentiality, it 
does not move to SpecRefP and therefore cannot license the [referential] feature of the clitic. If the 
clitic’s [referential] feature is not licensed, it cannot be spelled out” (2000:76). 
Age (in years) Overt clitics Omitted clitics Full NPs 
2 .22 (22/99) .64 (63/99) .14 (14/99) 
3 .62 (179/290) .15 (43/290) .23 (68/290) 
4 .89 (237/265) 0.0 (0/265) .11 (28/265) 
5 .91 (227/250) 0.0 (0/250) .09 (23/250) 
adults 1.0 (439/439) 0.0 (0/439) 0.0 (0/439) 
  
104 
Clitic omission in Italian 2–year–olds is attested in both Schaeffer’s (2000) 
and Guasti’s (1993/94) studies. On the basis of Schaeffer’s results, even 3–year–
olds are not adult–like with respect to clitic production, while 4– and 5–year–olds 
perform at ceiling. With respect to clitic placement, Guasti explicitly mentions 
that no systematic errors are attested in early Italian. 
 
4.2.2. Romanian 
 
Romanian resembles Italian and other finiteness–sensitive languages with respect 
to clitic placement: direct object clitics are pre–verbal, as in (4), except when 
combined with gerund and positive imperative forms, where they are post–verbal. 
However, the accusative 3rd person singular feminine clitic o (“her”) exhibits an 
exceptional pattern of placement: in periphrastic constructions that utilise the 
auxiliary avea (“have”) it surfaces post–verbally (5)9 (Babyonyshev & Marin 
2006:21). 
 
(4) Elefantul         l–            /  i–                  /  le–                a     stropit  
Elephant–the  him–CL  /  them–CL.M  /  them–CL.F  has  sprinkled 
(pe  b!iat  /  b!ie"i  /  fete) 
on   boy   /  boys    /  girls 
“The elephant sprinkled him (the boy/boys/girls)” 
(5) Elefantul         a     stropit–o                 pe  fat!. 
  Elephant–the  has  sprinkled–her–CL  on  girl 
“The elephant sprinkled the girl”         
                                                
9 Marin (2004) attributes the exceptional behaviour of o to morpho–phonological requirements on 
clitic combinations. Avram (2000) attributes it to a coalition of factors and, building on evidence 
in Avram (1986), claims that one of them is phonological. Avram (1986) points out that the sandhi 
rule is optional when the masculine clitic precedes the lexical verb avea, whereas in the auxiliary 
uses of avea the use of the sandhi variant is obligatory. This divergence is exemplified in (1–2) 
(Avram & Coene 2007:13–14, footnote 8). On the basis of this observation, Avram (2000) 
suggests that the feminine clitic o cannot appear pre–verbally because it lacks a sandhi variant.  
1. îl     am     /  l –am 
CL  have 
“I have it” 
2. l      –am  vazut  /  *îl am vazut 
CL  have  seen 
“I have seen him”  
  
105 
Avram (1999) was the first to study the L1A of clitic objects in Romanian 
(Babyonyshev & Marin 2006, Tsakali 2006:147–148). She administered an 
elicitation experiment modeled after Schaeffer (2000) to 16 monolingual 
Romanian children aged 2 to 5, comprising 3 age groups. She reports an omission 
rate of 42% (15/25) for 2–year–olds (N=3), 25% (32/133) for 3–year–olds (N=8) 
and 10% (7/56) for 4–year–olds (N=5) and concludes that clitic omission in child 
Romanian is initially high and decreases with age. With regard to clitic placement, 
Avram (1999) does not report target–deviant occurrences, even with the clitic o 
(Petinou & Terzi 2002:17, footnote 13). 
Avram and Coene (2007) studied the emergence of accusative direct 
object clitics (ADOCs) in Romanian on the basis of longitudinal data from 2 
monolingual Romanian children, Bianca (1;05–2;10) and Antonio (1;09–3;05). 
They calculated clitic omission and clitic production rates against the number of 
identified obligatory clitic contexts and report early emergence of clitics in both 
corpora (at age 2;0 for Bianca and at age 1;9 for Antonio), clitic omissions at the 
onset of L1A and substitution errors. Children in Avram and Coene’s study have 
90% target–like clitic production by age 3 (Bianca at age 2;10 and Antonio at age 
2;11). This outcome challenges Avram’s (1999) claim that 3–year–olds omit 
clitics 25% of the time.  
Contrastingly, Avram and Coene’s (2007) results are supported by 
Babyonyshev and Marin’s (2006) study. Babyonyshev and Marin administered an 
elicitation task for direct object clitics based on Schaeffer (2000) to 25 
monolingual Romanian children, aged 2;0 to 3;10. The independent variables 
tested were tense [+/–PAST], gender [FEM/MASC] and type of direct object 
([+DEF] DP/Proper Name); a sample experimental item is offered in (6) 
(Babyonyshev & Marin 2006:29). Gender is of special interest for us, given that, 
as mentioned above, the position of the feminine clitic differs depending on the 
tense of the clause.  
 
(6) Experimenter: “Look what else I have here, a bad dinosaur and a snake 
and look, the snake is swallowed, he is in the dinosaur's mouth” 
 
  
106 
Ce      i–     a      f!cut  dinozaurul     la  "arpe? 
What  him  has  done  dinosaur–the  to  snake? 
“What did the dinosaur do to the snake?” 
Child (2;4):  L–         a–   nghi#it. 
                  Him–CL  has  swallowed 
     “S/he swallowed/ate him” 
 
They report significantly higher production rates for direct object clitics in 
3–year–olds (N=13) (93%, 361/387), while clitic production in the group of 2–
year–olds (N=12) only reached 38% (94/193), with an object omission rate of 
60%10 (96/193), a figure much higher than the 42% reported in Avram (1999). In 
a follow–up study Babyonyshev and Marin (2006) administered an elicitation task 
for direct and indirect object clitics to 18 monolingual Romanian children aged 
2;5 to 3;10. They report comparable production rates for 2– and 3–year–olds for 
both direct clitics (86% (76/88) for 2–year–olds and 86% (73/86) for 3–year–olds) 
and indirect clitics (82% (114/133) for 2–year–olds and 74% (101/122) for 3–
year–olds).  
In sum, the results reported for clitic L1A in early Romanian differ: 
Babyonyshev and Marin (2006) and Avram and Coene (2007) report adult–like 
clitic production by age 3, while Avram (1999) reports that even 3–year–olds omit 
clitics 25% of the time. Taking into account that Avram has tested only 16 
children, while Babyonyshev and Marin’s (2006) database comprises data from 
43 children, the latter seems to be a safer basis for conclusions. Finally, the 
substitution errors reported in Avram and Coene (2007), as well as their indication 
that post–verbal clitics emerge prior to pre–verbal ones, are of crucial importance 
for the purposes of the current investigation: I return to them in Chapter 7.  
 
 
 
 
                                                
10 Babyonyshev and Marin (2006) attribute the instances of clitic omissions to production 
limitations, namely to the inability of very young children to produce clitic constructions of the 
required length. 
  
107 
4.2.3. Spanish / Catalan 
 
 
Clitics in Spanish and Catalan manifest the pattern attested in other finiteness– 
sensitive languages: they precede the finite verb and follow the non–finite verb. 
Wexler, Gavarró and Torrrens (2004) have studied clitic production in early 
Catalan and Spanish aiming to highlight the correlation between clitic omission 
and participle agreement. Spanish and Catalan are two closely related languages 
that differ in that only the latter exhibits (optional) participle agreement with the 
preceding object. Wexler et al.’s (2004) proposal is that in languages with 
participial agreement like Catalan, object clitics need to enter a double checking 
relation with two functional projections: the Clitic Phrase (checking of a 
D(efiniteness) feature) and AgrOP (checking of an agreement feature). By 
adopting Wexler’s (1998) UNIQUE CHECKING CONSTRAINT (UCC) according to 
which “the D–feature of a DP can only check against one functional category”, 
they assume11 that checking of more than one feature in clitic constructions is 
impossible in child language, and one of the functional categories is not projected, 
hence leading to optional clitic omission.  
Wexler et al. (2004) administered an elicited production task modelled 
after Schaeffer (2000) to 31 Catalan and 28 Spanish children aged 1;10 to 5;1. An 
example set of the experimental material is offered in (7). 
 
(7) Experimenter 1:  Aquest matí el cuiner ha començat a preparar el dinar. Ha  
  agafat el trencanous i les nous i mira què ha fet. 
“This morning the cook started preparing lunch. 
He took the nutcracker and the walnuts and look at what he 
did” 
Experimenter 2: Ja sé què ha fet: s’ha menjat les nous. 
  “I know what he did: he ate the walnuts” 
Experimenter 1: No! Digues–l’hi tu: Què ha fet el cuiner amb les nous? 
                                                
11 In fact, Wexler, Gavarró and Torrrens (2004) assume that the UCC acts in conjunction with the 
Minimise Violations constraint, which requires that the derivation violates as few grammatical 
properties as possible. 
 
  
108 
“No! You tell her: What did the cook do with the walnuts?” 
Expected response: Les ha trencades. 
       “He broke them” 
 
The results obtained showed the following three patterns: 
1. Ceiling percentages of correct clitic placement in both language groups, in 
all syntactic contexts: children placed clitics before finite verbs and after 
infinitives. The authors take this to indicate children’s sensitivity towards the 
finite/non–finite distinction, and the raising of the finite verb to T. 
2. Significant differences between Catalan and Spanish children with regard 
to the rates of clitic omission in all age groups. Clitic production with [–PAST] 
verbs reached 22.6% in Catalan 2–year–olds, 68.2% in 3–year–olds and 95.7% in 
4–year–olds; clitic/object omission reached 74.2%, 25%, and 4.2%, respectively, 
for the 3 age groups. Clitic production was at ceiling in Spanish children; a single 
instance of clitic omission was attested in the group of 3–year–olds. Clitic 
production with the verb in the present perfect occurred at a rate of 12.9% in 
Catalan 2–year–olds, 71.4% in 3–year–olds, and 85.1% in 4–year–olds12; the rates 
of clitic/object omission were 83.9%, 19%, and 6.4% respectively for the 3 age 
groups. The clitic production rate for Spanish 4–year–olds was 100%, 97.5% for 
3–year–olds and 81.2% for 2–year–olds. 
3. Spanish children from all age groups produced no non–target clitic forms.  
A few errors were attested in Catalan children: 3–year–olds produced 3 incorrect 
clitic forms in [–PAST] clauses, while in the present perfect clauses, errors in 
clitic form reached 100% for 2–year–olds, 76% for 3–year–olds and 11.4% for 4–
year–olds. All the errors attested in early Catalan involved substitution of the 
feminine, plural form les with the masculine, singular (unmarked) form l’. 
In sum, the outcome of Wexler, Gavarró and Torrrens’ (2004) study is that 
the distribution of object clitics with respect to the verb in Catalan and Spanish is 
                                                
12 Catalan children did not produce agreement between the participle and the direct object clitic, 
but they produced the default masculine singular form for the participle (Wexler et al. 2004). In 
particular, 2–year–olds produced clauses with agreement in 23.8% of the relevant contexts and 
without in 76.1% of them; the rates for 3–year–olds were 10.7% and 89.3%, respectively, and for 
4–year–olds 28.9% and 71.1%, respectively. Notably, no statistically significant difference was 
found between the clitic omission rate in children who produced participle agreement and those 
who did not. 
  
109 
adult–like from the earliest stages of L1A. As for clitic production, Spanish 
children are target–like by age 3, while Catalan 3–year–olds still omit clitic 
pronouns 19% of the time with a verb in the present perfect and 25% of the time 
with [–PAST] verbs. 
 
4.2.4.  Standard Modern Greek 
 
Clitic production in early SMG has been studied by Marinis (2000), Stephany 
(1997) and Tsakali (2006) among others. SMG patterns like other finiteness–
sensitive languages and exhibits pre–verbal clitic placement with finite verbs and 
post–verbal clitic placement with gerunds and imperatives13. 
Stephany (1997) built a corpus of spontaneous data, the Stephany Corpus, 
consisting of the recordings of 4 monolingual Greek children: Spiros, Janna, Mairi 
and Maria, available from the CHILDES database, on the basis of which she 
studied the acquisition of clitic pronouns. With respect to clitic production, she 
points out (1997:239) that at age 1;10 Spyros omits 3rd person neuter accusative 
singular to in 91% of obligatory contexts (N=34), as in (8), or uses a phonetic 
placeholder, while in imperatives he correctly uses to enclitically, as in (9). Mairi 
and Janna at age 2;4 and Maria at age 2;10 use proclitic object pronouns in more 
than 90% of obligatory contexts (Stephany 1997:239). 
 
(8) Aniki        Ula. 
Open–3S  Ulla–NOM 
[Intended utterance: Na  to                  aniksi       i      Ula] 
              M   it–CL.ACC  open–3S  the  Ulla–NOM 
“Ulla shall open it” (Spiros, 1;10) 
(9) Pa(r)       to! 
take–2S  it–CL.ACC 
“Take it!” 
 
                                                
13 I follow Mavrogiorgos (2009) in assuming that the imperative verb in SMG has an unvalued 
person feature in T. 
  
110 
With respect to clitic placement, Stephany notes that by age 1;10, Mairi, Janna 
and Spyros use accusative and genitive clitics enclitically, and proclitics occur 
only in the data from Mairi and Spyros. She takes this to indicate that enclitics are 
used productively before proclitics (1997:238). She also reports a few errors with 
clitic placement. In particular, 3 tokens of misplaced clitics occurred in Mairi’s 
data at age 1;10 and in Mairi’s and Maria’s data at age 2;4, all of which involved 
an enclitic used with a non–imperative verb, as in (10) (Stephany 1997:272). 
 
(10) Epese     me. 
Fell–3S  me–CL.ACC  (instead of: “mu–CL.GEN epese”) 
“I dropped it” (Mairi 2;4) 
 
 Marinis (2000) studied the emergence and placement of object clitics in 
SMG in single clitic, Clitic Doubling and Clitic Left Dislocation structures; here I 
report figures for the overall clitic production, irrespective of the construction 
type. His study is based on two longitudinal corpora, the Christofidou corpus, 
consisting of (69) weekly recordings of spontaneous data from a monolingual 
Greek child, Christos, from age 1;7 to 2;8, and the Stephany Corpus mentioned 
above, covering ages 1;9 to 2;9. The raw numbers of pre– and post–verbal clitics 
in the Stephany corpus and the Christofidou corpus are reported in tables 4.3 and 
4.4 respectively. 
 
Table 4.3: Clitic production in early SMG: the Stephany Corpus (based on table 4 
in Marinis 2000:269 & table 14 in Tsakali 2006:152). 
Child Age MLU Pre– / Post– 
Verbal 
Total  
 
Total Rate of Clitic 
Omission 
(Marinis 2000) (Tsakali 2006) 
Spiros 1;9 1.6 3 3 6 10 .195 
Janna 1;11 
2;5 
2;9 
1.4 
2.4 
2.8 
1 
46 
37 
5 
4 
0 
6 
50 
37 
10 
50 
– 
.155 
.011 
– 
Mairi 1;9 
2;3 
2;9 
2.0 
2.2 
2.5 
102 
122 
151 
41 
62 
11 
143 
184 
162 
143 
184 
– 
.056 
.048 
– 
Maria 2;3 
2;9 
2.3 
2.9 
18 
67 
13 
20 
31 
87 
31 
87 
.059 
.006 
  
111 
 
Table 4.4: Clitic production in early SMG: the Christofidou Corpus (based on 
table 3 in Marinis 2000:267) and the Doukas Corpus (based on table 16 in Tsakali 
2006:153/154). 
 
The most relevant results from Marinis’ study (2000) are summarised 
below. 
1. A stage in which no clitics are present is found in Christos’ data alone: no 
clitics are produced from age 1;7 to 1;10, while only one clitic is produced 
from age 1;11 to 2;0. 
2. No correlation is attested between the early non–finite verb forms in SMG 
(realised with the suffix –i; see Varlokosta et al. 1998) and object/clitic 
omission14.  
3. Pre– and post–verbal clitics emerge simultaneously: pre–verbal clitics are 
used in the indicative (11) and subjunctive (2000:267) and post–verbal 
clitics follow the imperative verb (12) (2000:268). No gerunds are 
produced in the corpora examined. 
4. No instances of clitic misplacement are attested. 
 
                                                
14 Marinis (2000) subsumes clitic omission under object omission and claims that object drop is 
attested in early SMG. However, these do not unequivocally constitute instances of clitic omission, 
with the exception of three cases (2000:270, footnote 33). 
[Christos] 
Age / MLU 
Pre– / Post– 
verbal 
Total 
 
[Maria] 
Age 
Total Rate of Clitic 
Omission  
Christofidou Corpus (Marinis 2000) Doukas Corpus (Tsakali 2006) 
1;7 
1;8 
1;9 
1;10 
1;11 
2;0 
2;1 
2;2 
2;3 
2;4 
2;5 
2;6 
2;7 
2;8 
1.2 
1.1 
1.1 
1.3 
1.4 
2.0 
2.1 
2.2 
2.2 
2.0 
2.4 
2.6 
2.6 
2.9 
0 
0 
0 
0 
0 
0 
23 
13 
22 
26 
49 
79 
134 
181 
0 
0 
0 
0 
1 
0 
3 
3 
5 
6 
4 
2 
6 
14 
0 
0 
0 
0 
1 
0 
26 
16 
27 
32 
53 
81 
140 
195 
– 
– 
– 
– 
– 
2;0.24 
– 
2;2.8 
2;3.18 
– 
2;5.4 
2;5.24 
2;7.1 
2;8.27 
– 
– 
– 
– 
– 
6 
– 
39 
46 
– 
46 
36 
37 
35 
– 
– 
– 
– 
– 
.041 
 – 
.025 
.037 
– 
.020 
0.0 
0.0 
0.0 
527 41 571 245 
  
112 
(11) To                chalacie  (target verb: chalase).  
It–CL.ACC  destroyed–3S 
“S/he destroyed it” (Christos, 2;1.14) 
(12) Pa (target verb: par(e))  to. 
Take–2S                         it–CL.ACC 
“Take it” (Spiros, 1;9.11 & Janna, 1;11.6) 
 
Marinis (2000) argues, on the basis of the observations that (i) pre– and 
post–verbal clitics emerge simultaneously and (ii) clitic misplacement is not 
attested, that Greek children’s clause structure is adult–like (at least with respect 
to the projections involved in clitic placement) and that it projects, at least, up to 
M(ood) (following Philippaki–Warburton 1998) or up to C (following Terzi 
1999a). The unavailability of clitics in Christos’ data does not falsify this claim, 
since within Marinis’ argumentation it may reflect an incomplete lexicon or object 
omission15 in general. 
Tsakali (2006) investigated the rates of clitic production and omission in 
early SMG on the basis of naturalistic and experimental data. She examined the 
files of the 4 children from the Stephany Corpus and the Doukas Corpus, 
consisting of data from one monolingual Greek child, Maria, recorded monthly 
from age 2;0.24 to 2;8.27. The results of her analysis of the Stephany corpus and 
Doukas Corpus are reported in tables 4.3 and 4.4 respectively. Tsakali’s outcome 
(2006) confirms Marinis’ (2000) claim as to the low level of clitic/object 
omission16 in early SMG.  
Tsakali also performed an elicitation task with 25 monolingual Greek 
children aged 2;4 to 3;6 to test clitic production in obligatory contexts. She used a 
picture–based task modelled after Schaeffer (2000) in which children were 
prompted to produce constructions involving a single direct object clitic. A 
sample experimental item is offered in (13). The experiment aimed at eliciting 5 
                                                
15 The fact that object omission in early SMG does not correlate with the use of early non–finite 
verbal forms shows that the former is not the result of a non–adult–like phrasal marker (Marinis 
2000:277). 
16 The only point of divergence between Marinis’ (2000) and Tsakali’s (2006) analysis of the 
Stephany corpus are the number of clitics reported for Spiros at age 1;9 and Janna at age 1;11. 
Marinis (2000) reports that each child produced 6 clitics, while Tsakali (2006) reports that each 
produced 10. 
  
113 
clitic constructions per child, hence 125 overall. Children produced 124 clitics, 
with only one child (aged 2;6) omitting a clitic on one occasion.  
 
(13) Experimenter Question: Ti kani edho to agoraki sto koritsaki?  
   What is the boy doing here to the little girl?  
Expected Answer: To (CL) filai.  
(He) is kissing her. 
  
On the basis of the three studies (Marinis 2000, Stephany 1997, Tsakali 
2006) carried out on clitic L1A in SMG, it can be claimed that SMG–speaking 
children have good clitic production from age 2;2 onwards, with low rates of clitic 
omission, while their clitic placement is adult–like, with the exception of the 3 
misplacement errors reported in Stephany (1997). 
  
4.3. Second position clitic languages 
 
Languages like Serbo–Croatian, a South–Slavic pro–drop language, adhere to 
second position (2P) restrictions with respect to clitic placement: clitic pronouns 
undergo obligatory cliticisation and appear raised in 2P. The reason why clitics 
never appear unraised in utterance–final position is that they cannot bear focus 
(Ilic & Ud Deen 2004). 
Ilic and Ud Deen (2004) investigated clitic production and clitic placement 
in early Serbo–Croatian on the basis of naturalistic and experimental data from 3–
year–old monolingual speakers of Serbo–Croatian. Naturalistic data were 
collected from Marija, Marko, and Ivan, while Marija, Ivan and Lana took part in 
an elicited production task; a sample test item is offered in (14). The results 
obtained from the analysis of naturalistic data are summarised in table 4.5 and the 
experimental results in table 4.6. 
 
(14) Researcher: Zmiya Ka hochye da poyede Mogliya. Shta ce zmiya Ka da 
uradi Mogliyu?  
  
114 
“The snake Kaa wants to eat Mowgli. What is the snake going to do with 
Mowgli?”  
Child: Da ga (CL) poyede. 
“To eat him up”  
 
 
Tables 4.5 & 4.6: Clitic placement in child Serbo–Croatian (based on tables 5 and 
6 respectively in Ilic & Ud Deen 2004) 
 
 
 
 
 
 
Table 4.5: Naturalistic data 
 
 
 
 
 
 
 
Table 4.6: Experimental Data 
 
The results reported show high rates of clitic production in Serbo–Croatian 
children. Moreover, they appropriately raise clitic pronouns to the clause 2P from 
as early as 3 years of age. Only sparse instances of unraised clitics occurred in 
naturalistic data (2/127; the 2 erroneous productions occurred in Marko’s data; see 
table 4.5), while the results of the experimental task show that whenever a clitic is 
produced, it appears in the raised position.  
In sum, the study carried out by Ilic and Ud Deen (2004) shows that 
Serbo–Croatian children have high rates of clitic production and adult–like clitic 
placement by age 317. The few instances of omitted or unraised clitics may 
indicate that they omit or leave clitics unraised at earlier stages of L1A; as of yet, 
no suitable data are available to test this hypothesis.  
 
                                                
17 Ilic and Ud Deen (2004) take this result to indicate that Serbo–Croatian children have learned 
the correlation between specificity and raising in lexical objects and object personal pronouns. 
 
 
Clitic 
pronouns 
Raised *Unraised Total 
Marko 32 2 34 
Marija 76 0 76 
Ivan 17 0 17 
Clitic 
pronouns 
Raised *Omitted Target 
Marija 33 2 35 
Lana 35 0 35 
Ivan 33 2 35 
  
115 
4.4. Clitic scrambling languages 
 
In Dutch all direct objects which co–refer with an antecedent in the preceding 
discourse obligatorily scramble over a number of elements including negative 
particles and adverbs. Schaeffer (2000) investigated the distribution of direct 
objects, including clitic pronouns, in 49 Dutch children, whose age ranged 
between 2;4 and 6;10. I report the results for pronominal elements alone, since 
this is the relevant aspect of Schaeffer’s study for our investigation. Schaeffer 
employed a task combining truth value judgement and elicited production and 
tested 3 different types of constructions, each involving one of the following 
elements (over which an object can be scrambled): low adverbs (manner), high 
adverbs (temporal, locative) and negation. The results obtained are summarised in 
table 4.7, showing the proportions of pronouns produced per age group. 
 
 
Table 4.7: Proportion of full objects, demonstrative pronouns, strong personal 
pronouns and clitics in Dutch clitic scenarios (Table 8 in Schaeffer 2000:80)  
 
As shown in table 4.7, Dutch children never omit objects18, even at age 2. 
However, Schaeffer reports low rates of scrambled personal (33%) and 
demonstrative pronouns (17%) in Dutch 2–year–olds. Considerable development 
towards adult–like performance is observed at age 3, when the production rates 
for scrambled personal and demonstrative pronouns reach 95% and 78% 
respectively. Note that Dutch adults always scramble personal and demonstrative 
pronouns. 
What is relevant for the current thesis is that all personal pronominal 
objects that appear unscrambled in Schaeffer’s study are full pronouns, not clitic 
pronouns. This indicates that clitic placement in early Dutch is adult–like (15) 
                                                
18 Schaeffer claims that the type of the verbs used in the elicitation experiment contributed to the 
low rates of object omission: all the verbs involved in the Dutch clitic scenarios were particle 
verbs, hence telic, which require an overt object (2000:81). 
Age 
(years) 
Full objects Demonstrative 
pronouns 
Strong personal 
pronouns 
Clitics 
2 .52  .26 .06 .16 
3 .15 .26 .01 .58 
4 .07 .07 .0 .86 
  
116 
(Schaeffer 2000:83), irrespective of the low rates of clitic production19 for the 2– 
and 3–year–olds. 
 
(15) Bert  gaat  ’m   niet  kleuren,      de   kikker. 
Bert  goes  CL  not   color–INF  the  frog 
           “Bert is not going to color him, the frog” (L 2;8) 
 
4.5. Tobler–Mussafia languages 
 
Cypriot Greek, European Portuguese and Galician are languages that adhere to the 
Tobler–Mussafia law; hence, clitic pronouns are banned from clause initial 
position. All these languages exhibit finite and non–finite enclisis and finite 
proclisis (see section 2.3.1 for a discussion on the relevant syntactic contexts for 
CG and 2.3.5 for the corresponding contexts in EP and Galician). This section 
focuses on studies investigating clitic production and/or comprehension in early 
CG, early EP and early Galician.  
 
4.5.1 European Portuguese 
 
European Portuguese is a Tobler–Mussafia language which exhibits both pre– and 
post–verbal clitics and resembles CG in many respects with regard to contexts that 
trigger the one or the other; see section 2.3.5 in chapter 2. The current section 
discusses a number of experimental studies conducted on clitic L1A in EP. The 
focus of the majority of these studies is clitic omission in early EP. Various 
aspects of the phenomenon are studied, including its relation to null objects, and 
its contingency on clitic type (reflexive, 1st & 2nd person clitics) and syntactic 
context (i.e. strong islands), while a proposal is put forward to accommodate the 
data. In recent work on clitic placement in early EP Lobo and Costa report 
instances of clitic misplacement. An overview of their work is also provided. 
                                                
19 Schaeffer (2000) attributes the low rates of produced clitics in Dutch children to the optional 
marking of referentiality in child language. Scrambling is not allowed with non–referential objects 
in Dutch, since discourse–relatedness is not established, and the movement (scrambling) of the 
element in discussion to SpecRefP/SpecDiscP is not motivated. When the pragmatic system 
develops, this optionality is no longer attested. 
  
117 
Costa and Lobo (2007a) investigated whether the correlation that holds for 
many clitic languages between clitic omission and participial agreement (Tsakali 
& Wexler 2004, Tsakali 2006) holds for EP as well, thus testing the applicability 
of Wexler’s UCC in early EP. They modelled a production task after Schaeffer 
(2000), with a puppet commenting on an acted–out story, and elicited accusative 
3rd person clitics in enclitic and proclitic environments. Strong islands were also 
included in the test in order to control for the difference between target–like null 
objects and target–deviant clitic omission: given the availability of null objects in 
adult EP, strong islands, in which null objects are ruled out, are a suitable context 
for disambiguating between object drop and clitic omission. An example item 
from the enclisis condition is offered in (16) (Costa & Lobo 2007a). 
 
(16) Experimenter: Olha! Está aqui o Urso Pooh. Ele hoje encontrou o Tigre e 
achou que o Tigre estava muito despenteado... Ah! Ele tem uma escova! 
Olha para o que o Pooh fez ao Tigre. 
“Look! Here’s Pooh. Today, he met Tigger and he thought his hair was not 
nice…Ah! He has a comb! Look at what Pooh did to Tigger” 
Puppet: Eu sei! Ele lavou o Tigre! 
  “I know! He washed Tigger!” 
Experimenter: Não...não lavou nada. Diz!lhe lá o que o Pooh fez ao Tigre! 
   “No…he did not. Tell him what Pooh did to Tigger.” 
Expected response: penteou–(o) 
            “combed–(him)” 
 
The test was administered to 21 monolingual speakers of EP from 2 age 
groups: 2– to 3–year–olds and 4–year–olds. High rates of null forms were attested 
in all contexts (simple clauses and strong islands) in both age groups: clitic 
production reached 10% for 2– to 3–year–olds and 13.9% for 4–year–olds. Clitic 
production within strong islands was at 2.3% for the former age group and at 0% 
for the latter. Costa and Lobo take the massive proportion of null complements in 
strong island contexts to indicate that these are instances of omitted clitics rather 
than null objects. They base their claim on the significantly higher DP production 
  
118 
in strong island contexts (as compared to simple declaratives) observed in both 
age groups, which shows some sensitivity to the special characteristics of these 
domains. 
Clitic omission in early EP challenges the basic tenet of the UCC. 
However, Costa and Lobo (2007a) do not interpret these results as counter–
evidence for the UCC. Instead, they assume that clitic omission in EP is of a 
different nature than clitic omission in other languages. For them, UCC is 
maturational in nature (cf. the correlation with the Root Infinitive stage). So, in 
languages in which omission is a result of the UCC, a developmental effect is 
attested: by age 3 children have adult–like production. In EP, on the other hand, 
4–year–olds still omit clitics. Therefore, the UCC is argued not to be responsible 
for the EP data. For Costa and Lobo (2007a), clitics and null objects in EP 
compete to convey the same message, while the choice between the two is post–
syntactic and discourse–conditioned. 
Moreover, Costa and Lobo (2007a) attribute the rates of clitic omission in 
early EP to complexity factors. On the one hand, the availability of null objects in 
adult EP obliges the children to learn the contexts in which null objects are 
legitimate and the ones in which they are not. On the other, the realisation of both 
enclisis and proclisis in finite contexts (Duarte & Matos 2000) results in extra 
burden for children. Costa and Lobo (2007a) assume, following Reinhart (1999), 
that this complex clitic system forces young children to decide between multiple 
convergent derivations, which results in problems in production. 
What is not clear from Costa and Lobo’s (2007a) study is whether clitic 
placement is a relevant variable for determining rates of omission in early EP. The 
results reveal different rates of clitic production in enclisis and proclisis contexts: 
7.3% and 12.8% respectively for 2– to 3–year–olds and 4.8% and 22.7% 
respectively for 4–year–olds. However, a definitive answer would require the 
results to be statistically validated. Finally, a further aspect of Costa and Lobo’s 
results which they did not further investigate, but which is relevant for the 
purposes of our investigation, is that some of the few clitics attested in proclitic 
environments were realised post–verbally. This outcome confirms Duarte and 
  
119 
Matos’ (2000) observation that there is a tendency for EP–speaking children to 
overuse enclisis. 
Costa and Lobo (2009) looked at comprehension data in order to offer 
corroborative evidence for their claim that the absence of clitic pronouns in the 
early production of EP is the result of the overgeneralisation of target–like null 
object construction (2007a), rather than the result of target–deviant clitic 
omission. They adapted to Portuguese a truth value judgment task developed by 
Grüter (2006) and administered it to 20 EP–speaking children aged 3;2 to 5;0. 
Grüter’s task aimed at assessing the comprehension of structures with a null 
argument in French and English children. The experimental material involved 
intransitive structures, structures with superfluous DPs, superfluous clitics, object 
clitics and null objects. All these constructions were tested in two contexts: in 
simple clauses and in strong islands. Costa and Lobo hypothesised that the EP 
children’s ability to assign a transitive interpretation to a complement–less verb 
would constitute an indication that they have acquired the null object construction.  
The results obtained showed that: (1) EP children master transitivity, since 
they reject superfluous arguments and their interpretation of intransitive structures 
is adult–like, (2) they interpret clitics in an adult–like way, (3) they accept null 
objects in simple clauses like adults do, but (4) they accept null objects within 
strong islands as well, unlike adults. Overall, children performed very similarly to 
adults (above 80% correct responses in all conditions) with the exception of the 
condition testing null objects in strong islands, in which children assigned 
transitive interpretations to complement–less verbs. 
For Costa and Lobo (2009) these results are compatible with the 
assumption that clitic omission in early EP is due to the overgeneralisation of the 
null object construction: EP children are aware of the availability of this type of 
construction, yet, unlike adults, they overgeneralise it to contexts with strong 
islands as well. The reason for doing so is that they have not yet developed the 
more restrictive adult–like grammar, within which the distribution of null objects 
is constrained. 
The symmetry between production and comprehension data in early EP, as 
revealed by the similar results reported in Costa and Lobo (2007a) and (2009), is 
  
120 
at odds with the asymmetry observed in early French, as reported in Grüter 
(2006). While French children omit accusative clitics, they reject the transitive 
interpretation for null object constructions (rejection rate: 85%). This divergence 
in the performance between EP and French children reveals that clitic omission 
does not seem to be a uniform phenomenon cross–linguistically (Costa & Lobo 
2009). 
The hypothesis developed by Costa and Lobo (2007a, 2009) that EP 
children overgeneralise the null object construction at the onset of L1A is fully 
justified on the basis of the evidence from two other studies reported in Costa et 
al. (2008). The first one conducted by Costa and Lobo (2007b) elicited reflexive 
(1st, 2nd and 3rd person singular) and non–reflexive (3rd person singular) 
accusative clitics. The second one carried out by Silva (2008) elicited non–
reflexive 1st, 2nd and 3rd person singular dative clitics in two contexts: simple 
clauses (enclitic and proclitic contexts) and strong islands. Reflexive clitics, as 
well as 1st and 2nd person clitics, do not freely alternate with null objects, thus 
patterning with clitics in strong islands. The methodology of both studies was 
based on Schaeffer (2000). 24 EP–speaking children aged 3 to 4 participated in 
Costa and Lobo’s (2007b) study and 11 in Silva’s (2008) study. 
The results obtained from both studies showed that EP children omit 
clitics in all construction types, even in contexts in which this is illegitimate in 
adult language, i.e. reflexive clitics and 1st and 2nd person clitics. This outcome 
confirms Costa and Lobo’s (2007a, 2009) hypothesis that the null object 
construction is overgeneralised in early EP. Yet, some important asymmetries 
were detected between the different clitic subtypes. Costa and Lobo (2007b) 
report a significantly higher production of reflexive clitics (47.4%, 104/219), as 
compared to non–reflexive ones (13%, 13/100). Higher rates of omission were 
attested for 3rd person non–reflexive clitics as compared to 1st and 2nd person 
non–reflexives (Silva 2008). In enclitic contexts, clitic production was low for 1st 
and 2nd person clitics (31.8%, 7/22), while no 3rd person (0/22) clitics were 
produced. In proclitic contexts a few 1st (9.1%, 2/22) and 2nd person (31.8%, 
7/22) clitics were produced, while no 3rd person (0/22) clitics were produced. The 
rates of production for non–reflexive dative clitics (8.8%, 35/396) (Silva 2008), 
  
121 
and for non–reflexive accusative clitics (13%, 13/100) (Costa & Lobo 2007b), are 
comparable. Finally, no difference in dative clitic production was attested in 
enclitic (10.6%, 14/132) versus proclitic (6.8%, 9/132) contexts (Silva 2008). 
This outcome is indicative for the two hypotheses that are invoked to 
explain the pattern of clitic omission in EP: Wexler’s (1998) UCC and Reinhart’s 
(1999) post–syntactic complexity. Given that all clitic types tested are similar in 
terms of feature specification, UCC predicts similar omission rates in all 
conditions. Contrastingly, on the basis of Reinhart’s theory of post–syntactic 
complexity, production rates are expected to differ depending on the clitic 
subtype. Recall that reflexive clitics and 1st and 2nd person clitics do not freely 
alternate with null objects. Thus, no multiple convergent derivations arise and 
post–syntactic choices are not imposed on children. As a result, the system’s 
complexity is reduced, hence higher production rates are expected for 1st and 2nd 
person clitics as well as for reflexive clitics. The results obtained clearly favour 
the latter hypothesis, since considerably lower omission rates were attested for 
reflexive clitics as compared to non–reflexive ones (Costa & Lobo 2007b) as well 
as for 1st and 2nd person clitics as compared to 3rd person clitics (Silva 2008). 
Crucially, though, it seems that children have not yet reached the state of adult 
knowledge concerning the illegitimacy of null objects in these contexts. 
Turning now to clitic placement in early EP, in a recent study Lobo and 
Costa (2012) report instances of clitic misplacement in EP children at 5 years of 
age. Recall that EP exhibits 3 patterns of clitic placement: proclisis, enclisis and 
mesoclisis20. Lobo and Costa (2012) elicited se clitics, a type of clitic less 
frequently omitted in early EP (cf. Costa & Lobo 2007b), in enclitic and proclitic 
contexts in 20 EP children aged 5 to 6. The following proclitic contexts were 
elicited: clauses with negative markers, negative subjects, quantified subjects, the 
adverb já (“already”), as well as embedded complement and adverbial clauses. 
Simple and coordinate clauses that constitute enclitic contexts were also elicited. 
The control group, which consisted of adult speakers of EP, produced 217/240 
clitics in enclitic contexts all of which were placed post–verbally. In proclitic 
                                                
20 Cf. section 2.3.5 in Chapter 1 for a discussion on the syntactic contexts in which each pattern of 
clitic placement is manifested. 
  
122 
contexts the control group produced 465/480 pre–verbal clitics (88.8%). Children 
produced 167/240 post–verbal clitics in enclitic contexts and 362/480 in proclitic 
ones. Interestingly, pre–verbal clitic placement in proclisis contexts in children 
(25.8%) was much lower than in the adult group (88.8%), but also much lower 
than post–verbal placement (74.2%). Notably, some cases of double clitic 
production, with the clitic placed both pre– and post–verbally, were also attested, 
but only very rarely (0.4%, 3/720). 
Closer examination of the results reveals that pre–verbal clitic placement 
in proclitic contexts varies depending on the syntactic environment. On the basis 
of the results obtained the following ranking with respect to the percentages of 
target–like responses (in parenthesis) emerges: 
1. negatives (60.8%) 
2. negated subjects (48%) 
3. embedded complements (47.1%) 
4. adverbial clauses (34.3%) 
5. embedded adverbials (8.7%) 
6. quantified subjects (6.2%). 
Lobo and Costa (2012) conclude that only proclitic contexts are truly problematic 
for young children. They hypothesise that variation in the input may be a factor 
that can explain the delay in the acquisition of clitic placement in EP. However, 
they do not assume that child production is merely a reflex of a variable input. 
Instead, they take the one–way tendency to indicate that this output is constrained 
by grammatical factors as well. 
 
4.5.2. Galician 
 
Galician resembles CG and EP in terms of clitic placement. It is a Tobler–
Mussafia language exhibiting post–verbal clitic placement in root clauses: 
bicouno (CL) (“(s/he) kissed him”), bicalo (CL) (“to kiss him”) (Grohmann et al. 
2012) and pre–verbal clitic placement in wh–questions, negatives, and clauses 
headed by quantifiers or focused XPs, as well as in subordinate constructions 
(Uriagereka 1995). 
  
123 
Children residing in Galicia, in the northwest of Spain, are exposed to both 
Galician, a Tobler–Mussafia language, and Spanish, a finiteness–sensitive 
language. Grohmann et al. (2012) tested clitic placement in 12 Galician–speaking 
children aged 5 to 6 residing in the area of Pontevedra. They used the COST 
Action A33 tool adapted to Galician. However, they elicited simple declaratives 
rather than because–clauses, since most subordinate constructions, such as those 
used in the original A33 tool, trigger pre–verbal clitic placement in Galician. An 
example of the experimental material is offered in (17) (Grohmann et al. 2012). 
The elicited clause could involve either a finite or a non–finite verb.  
 
(17) Neste debuxo temos un neno e un balón. 
“In this picture, we have a boy and a ball” 
O neno fixo algo. 
“The boy did something”   
Qué fixo o neno co balón? 
“What did the boy do with the ball?” 
 
Overall, 144 constructions were elicited, 131 of which involved a clitic 
pronoun (91%). Only 55 of the 131 elicited clitic constructions (42%) involved a 
finite verb, while the remaining 76 (58%) constituted infinitival constructions. 
Children’s clitic placement was target–like 97% of the time, irrespective of the 
context, with only 4 misplacement errors out of the 131 responses. 
This result is a clear indication that Galician–speaking children acquire 
adult grammar with respect to clitic placement by age 5. An earlier age of 
acquisition is not precluded, and this is an issue worth exploring. Likewise, it is 
worth exploring whether clitic misplacement is manifested in the course of 
Galician L1A. For this purpose, younger children should be tested and the 
experimental material should involve proclisis–triggering contexts as well.  
 
 
 
 
  
124 
4.5.3. Cypriot Greek 
 
Petinou and Terzi (2002) were the first to study the acquisition of clitic pronouns 
in CG. They were also the first to observe and report the phenomenon of clitic 
misplacement. A second study carried out by Grohmann and colleagues (2011, 
2012) studied clitic production and placement in CG–speaking children aged 3 to 
5 and offered useful insights regarding the reflection of the sociolinguistic 
situation in Cyprus in language use. 
Petinou and Terzi (2002) investigated the L1A of CG clitics on the basis 
of a corpus consisting of data from 5 typically developing (TD) children21 and 5 
children diagnosed with specific language impairment (SLI). The TD group was 
followed longitudinally and recorded bimonthly over a period of 4 months (at 32, 
34, and 36 months), while the SLI group, whose age ranged between 48 and 60 
months, was tested once; the two groups were matched for mean length of 
utterance in words (MLUW), gender, and socioeconomic status. Petinou and Terzi 
calculated misplaced clitics out of children’s overall clitic production in na–
clauses and negatives; recall that both contexts require a pre–verbal clitic in adult 
grammar (cf. Chapter 2). Examples of clauses involving a misplaced clitic are 
offered in (18–19) (2002:8).  
 
(18) (N)a  kolisume    ta              !             tetradhio mu. 
M      stick–1PL  them–CL  (on–the)  notebook  my 
“(To) stick them on my notebook” (LK) 
Adult Production: “Na ta kolisume sto tetradhio mu” 
(19) Oi,  en      aresi           mu. 
No  NEG  please–3S  me–CL 
“No, I don’t like (it)” (LK) 
 Adult Production: “Oi, en mu aresi” 
 
                                                
21 Petinou and Terzi (2002) call this group children with normal language development (NLD). I 
use the term typically developing children for the same subset of the overall population. 
  
125 
 The phenomenon of clitic misplacement was attested in both groups with 
clitics placed post–verbally. The relevant results are summarised in tables 4.8 and 
4.9. Table 4.8 shows the proportion of clitic misplacement (PCM) and MLUW in 
the children with SLI and table 4.9 gives the same information for the TD 
children. 
 
 
 
 
 
 
 
Table 4.8: Proportion of clitic misplacement in children with SLI (based on table 
3 in Petinou & Terzi 2002:13). 
 
Child 32 mo. 34 mo. 36 mo. 
PCM MLUw PCM MLUw PCM MLUw 
OX 0.10 (3/30) 3.0 0.02 (1/52) 3.6 0.0 (0/61) 3.3 
NA 0.66 (16/24) 2.8 0.28 (5/18) 3.1 0.12 (4/33) 3.3 
AM 0.07 (2/26) 2.8 0.0 (0/17) 3.2 0.0 (0/34) 4.0 
AI 0.21 (3/14) 2.4 0.14 (3/21) 3.0 0.05 (2/37) 3.4 
AX 0.62 (20/32) 2.9 0.44 (13/29) 3.1 0.0 (0/38) 4.0 
 
Table 4.9: Proportion of clitic misplacement in TD children (based on table 2 in 
Petinou & Terzi 2002:13). 
 
Paired comparisons revealed significant differences regarding PCM in the 
TD group between 32 and 36 months, t(8)=1.92, p<.05. No significant differences 
were found between 32 and 34 months, t(8)=0.74, p>.05, or between 34 and 36 
months, t(8)=0.98, p>.05. Finally, PCM was significantly larger (and invariant) in 
the group with SLI (M=1.0, SD=.00) when compared to the TD group (M=0.31, 
SD=.29), t(8)=6.93, p<.05. 
In a follow–up test, the PCM and MLUW of 3 additional TD children were 
measured, but no instances of clitic misplacement were attested; the relevant data 
appear in Table 4.10. 
 
                                                
22 The first number in parentheses refers to the number of misplaced clitics, and the second to the 
overall clitic production (misplaced and correctly placed) per child (Petinou & Terzi 2002:13). 
Child Age (in mo.) MLUw PCM22 
LK 60 3.8 1.0 (33/33) 
FI 48 2.7 1.0 (47/47) 
GK 60 4.0 1.0 (20/20) 
SK 48 2.4 1.0 (31/31) 
EP 54 2.9 1.0 (16/16) 
  
126 
 
 
 
 
Table 4.10: Proportion of clitic misplacement in 3 younger TD children (based on 
table 4 in Petinou & Terzi 2002:15). 
 
Petinou and Terzi (2002) interpret their results following proposals in 
Terzi (1999a). Specifically, they assume that finite enclisis in CG derives from 
proclisis with the manifestation of verb movement past the cliticisation site and 
that verb movement from T(ense) to M(ood) is an alternative mechanism 
employed to satisfy the strong verbal features of M: these features are satisfied by 
the inflectional particles heading MP, i.e. negators, modal particles. If no such 
particle heads the clitic clause, V–to–M obtains. Clitic misplacement is interpreted 
along these lines as the result of the overgeneralisation of verb movement to M. 
This behaviour is attributed to children’s misanalysis of the X–bar status of the 
inflectional particles heading MP: while these particles in adult CG are heads (of 
MP) and thus able to satisfy the feature checking requirements of M, in child 
grammar they are perceived as phrasal specifiers (located in the SpecMP) or 
adjuncts, hence their inability to check the verbal features of M. As a result, V–
to–M is manifested even in their presence and the following order derives: 
particle–verb–clitic.  
Petinou and Terzi (2002) offer two pieces of evidence in support of this 
analysis. First, they take the na omissions in children with SLI23 as evidence for 
problems with the M(ood) head. Second, they take some uses of the inappropriate 
negative marker by a single child with SLI to indicate the phrasal status of 
negative particles in early CG24. However, their claim is weakened by the fact that 
both pieces of evidence are found in SLI data alone. Petinou and Terzi mention 
that none of the TD children either omitted the modal particle na or made the 
                                                
23 The raw numbers and the relevant percentages of na omissions in total numbers of na–clauses 
involving clitic pronouns for the group with SLI are as follows: GK: 12% (4/33); LK: 15% (7/47); 
FI: 27% (8/30); EP: 6% (2/31) and SK: 28% (9/32) (Petinou &Terzi 2002:17). 
24 This child used constituent negation (ohi), which is phrasal, instead of sentential negation (dhen 
and min), which is considered an X (Petinou & Terzi 2002, and references therein); but, crucially, 
not vice versa. Petinou and Terzi (2002) take the nature and the direction of the substitution as 
evidence that the child has attributed phrasal status to all the negative particles and they extend 
their claim to the modal particle na. 
Child Age (in mo.) MLU PCM 
AI 28 2.8 0.0 (0/4) 
IP 28 3.0 0.0 (0/7) 
OK 28 3.0 0.0 (0/11) 
  
127 
wrong choice of negative particle; moreover, only 1 out of the 5 children with SLI 
used the inappropriate negative marker (2002:24). 
A number of conclusions can be drawn on the basis of Petinou and Terzi’s 
(2002) study. First, CG–speaking children seem to learn the basic grammatical 
properties of clitics quickly; no inflectional errors have been reported, not even 
for the SLI population. This constitutes corroborative evidence for Petinou and 
Terzi’s claim that children have no problems with clitics per se. In the second 
place, a negative correlation seems to hold between MLUW and PCM: PCM 
decreases as MLUW increases. A third point is that the first set of data (TD 
children) points to a developmental path, with the age of 32 months (or younger) 
as the starting point, with (fewer or more instances of) clitic misplacement, and 
the age of 36 months as the end–point, with adult–like clitic placement. However, 
this outcome is not confirmed by the follow–up study, with 3 children as young as 
28 months exhibiting target–like clitic placement. This result raises issues 
regarding the generalisability of the phenomenon, especially if the small number 
of participants is taken into account. Moreover, the results reported in table 4.9 
vary with regard to children’s performance at the age of 32 months: PCM ranges 
between 7% and 66%; in fact only 2 children (N.A. & A.X.) have relatively high 
PCM. At the subsequent two stages (34 and 36 months) PCM decreases rapidly 
for all the participants relative to their performance at the age of 32 months. The 
latter point raises issues regarding the robustness of the phenomenon observed. In 
the fourth place, there is a striking difference between TD and SLI children: only 
the latter group misplaces clitics across the board. Petinou and Terzi take this 
observation one step further and suggest that ceiling percentages of incorrect clitic 
placement across syntactic contexts constitutes a clinical marker for SLI in CG–
speaking children. Finally, the proposal put forward by Petinou and Terzi relates 
the overgeneralisation of enclisis with the X–bar status of the inflectional 
particles. On the basis of such a proposal, clitic misplacement is expected to 
correlate with the omission or wrong use of the relevant particles, i.e. negative and 
modal particles, as well as any other particle heading MP. However, the evidence 
provided does not adequately justify this claim. 
Let us now turn to the studies conducted by Grohmann (2011) and 
  
128 
colleagues (Grohmann et al. 2012) within COST Action A3325 on CG clitics. 
Grohmann (2011) administered A33’s clitics–in–islands test (adapted for CG) to 
24 TD children aged 5 to 6 years, as well as to a group of 10 TD children aged 3 
to 4 years. Grohmann et al. (2012) administered the same test to 117 TD children 
aged 2 to 7 years. 
The test developed within COST Action A33 and used by Grohmann and 
colleagues (2011, 2012) is an elicited production task for 3rd person singular 
accusative object clitics within syntactic islands. An example of the experimental 
material used is offered in (20): after the introductory sentence followed by a 
question, children were confronted with an embedded jati (“because”)–clause, 
which they were prompted to complete; the bracketed part indicates the target 
structure.  
 
(20) I mama xtenizi ti gorua t!e i korua en omorfi. 
“Mommy is combing the girl and the girl is beautiful” 
Jati i korua en omorfi?   
“Why is the girl beautiful?” 
I korua en omorfi "ati i mam:a tis [htenizi tin–CL] 
The girl is beautiful because mommy [combs her]. 
 
All the 8 adults in the control group of Grohmann’s study (2011) produced 
exclusively post–verbal clitics. Grohmann takes this outcome to indicate that the 
target–like grammar for CG–speaking children requires the enclitic pattern. On 
the other hand, Leivada et al. (2010), quoted in Grohmann et al. (2012), report a 
mixed pattern of clitic placement in the adult control group with a preference for 
enclitic placement (76.6%). The results reported in Grohmann (2011) and 
Grohmann et al. (2012) are summarised in tables 4.11–4.13. 
 
 
 
                                                
25 COST Action A33 is a project aiming to investigate the linguistic performance of typically 
developing children at the age of 5 across Europe(an languages) with respect to five areas of 
grammar including clitic production. 
  
129 
 
Table 4.11: Clitic production in TD children aged 3 to 6 years for clitics–in–
islands test (Grohmann 2011: tables 1 & 2) 
 
 
 
 
 
 
 
Table 4.12: Clitic placement in TD children aged 3 to 6 years for clitics–in–
islands test (Grohmann 2011: tables 1 & 2) 
 
 
Table 4.13: Clitic production and clitic placement for all age groups for clitics–in–
islands test (Grohmann et al. 2012: table 3). 
 
Grohmann (2011) and colleagues (2012) take the enclisis pattern 
manifested by children aged 2 to 4 (tables 4.12–4.13), which resembles the adult 
pattern, to indicate that CG–speaking children acquire adult–like clitic placement 
at around 3 years of age. They attribute the mixed pattern exhibited by 5– and 6–
year–olds26 to the fact that the CG grammar of these children gets “muddled” by 
the interference from SMG at school; recall that at age 5;6–6;0 children in Cyprus 
enter the school system, where SMG is the language of instruction. In sum, they 
interpret their results within the proposed (domain–specific) Socio–Syntax of 
Development Hypothesis that primarily concerns the domain of schooling. 
Nevertheless, the study reported in Grohmann (2011) and Grohmann et al. 
(2012) raises some issues. In the first place, the objective of COST Action A33 
was to test clitic production rather than clitic placement. The choice of clitic 
                                                
26 As mentioned by Grohmann (2011), the 24 children aged 5 to 6 who performed the clitics–in–
islands test can be grouped as follows: (i) 10 children who use predominantly proclisis, (ii) 10 
children who use predominantly enclisis and (iii) 4 children who use the two patterns. 
Age group  N Clitics Omission NP No answer Other 
3;0–4;0 
(N=10) 
120 110 (.92) 2 (.02) 3 (.02) 0 (0.0) 5 (.04) 
5;0–6;0 
(N=24) 
288 276 (.96) 2 (.01) 2 (.01) 1 (0.0) 7 (.02) 
Clitic placement  3;0–4;0 (N=10) 5;0–6;0 (N=24) 
Pre–verbal 0 (0.0) 137 (.496) 
Post–verbal 110 (1.0)  139 (.503) 
Overall 110 (1.0) 276 (1.0)  
Clitic  2;0–2;11 3;0–3;11 4;0–4;11 5;0–5;11 6;0–6;11 Adults 
Overall .986 .867 .885 .943 .873 1.0 
Post–verbal .90 .89 .88 .68 .47 1.0 
  
130 
constructions involving an embedded jati–clause in the clitics–in–islands test is 
perfectly suited for eliciting clitics from languages with high percentages of clitic 
drop, like European Portuguese. Jati–clauses, however, are not ideal to test clitic 
placement in CG. Pappas (2011) carried out a magnitude estimation analysis of 
acceptability for clitic constructions in CG and showed that, in subject–less causal 
clauses headed by jati, both enclisis and proclisis are equally acceptable. 
However, if a pre–verbal subject is realised, as in the experimental material in the 
clitics–in–islands test used by Grohmann and colleagues (2011, 2012), focus is of 
crucial importance for clitic placement. Pappas (2011) tests the effect of focus for 
pre–verbal subjects and reports the following. With contrastive focus, there is no 
preference between the proclitic/enclitic pattern; with information focus, on the 
other hand, proclisis is preferred; and when there is no emphasis on the DP, 
enclisis is strongly preferred27. Grohmann (2011) reports that the 8 adults of the 
control group produced only post–verbal clitics in jati–clauses, whereas Leivada 
et al. (2010, quoted in Grohmann et al. 2012) report a mixed pattern of clitic 
placement in the adult control group. The latter outcome confirms Pappas’ (2011) 
results: the CG target–grammar allows for both enclisis and proclisis in jati–
clauses; this is also independently reported by Chatzikyriakidis (2010, 2012). 
Moreover, what is of crucial importance for an acquisition study is the 
grammaticality of a given structure. Since jati–clauses exhibit dialectal variation 
with regard to clitic placement (pre–verbal in SMG and (mainly) post–verbal in 
CG) both proclisis and enclisis are grammatical in some variety. Hence, the 
results reported by Grohmann and colleagues (2011, 2012) are relevant for issues 
pertaining to code–switching or to sociolinguistic parameters in linguistic 
environments such as that of Greek–speaking Cyprus, which Grohmann (2011) 
calls “bi–x”. In relation to the former point, he mentions that extensive code–
switching takes place (in situations that require formal and polite styles) by Greek 
Cypriots, who use (some form of) SMG or what Arvaniti (2010) calls Standard 
Cypriot. From an acquisitionist’s perspective the most important outcome of the 
aforementioned study is that CG–speaking children as early as 2 years of age have 
                                                
27 Grohmann et al. (2012) acknowledge the possibility of a bias against post–verbal clitic 
placement in CG in subject initial declaratives. 
  
131 
adult–like clitic production, i.e. they do not omit clitics (Grohmann et al. 2012). 
 
4.6. Interim conclusions  
 
A number of conclusions can be drawn on the basis of the various studies of clitic 
L1A in a number of European languages. 
With respect to clitic production, a divergence is observed in the 
performance of children acquiring different languages. In some languages, clitics 
are omitted, while in others they are produced from very early on. Children 
acquiring Cypriot Greek (Grohmann 2011, Grohmann et al. 2012), Spanish 
(Wexler, Gavarró & Torrrens 2004), Standard Modern Greek (Marinis 2000, 
Stephany 1997, Tsakali 2006), Serbo–Croatian (Ilic & Ud Deen 2004), and, 
according to Babyonyshev and Marin (2006) and Avram and Coene (2007), 
Romanian have adult–like clitic production by age 3. Clitics have been found to 
be problematic and omitted by children older than 3 in Catalan (Wexler, Gavarró 
& Torrrens 2004), European Portuguese (Costa & Lobo 2007a et seq., Costa et al. 
2008), Italian (Guasti 1993/4, Schaeffer 2000) and, according to Avram (1999), 
Romanian. Low clitic production is also reported for early Dutch (Schaeffer 
2000). 
Early clitic omission has been attributed to different factors. Costa and 
Lobo’s (2007a) analysis of EP refers to complexity, following ideas in Reinhart 
(1999), while Guasti (1993/94) mentions difficulties in forming A–chains. 
According to Tsakali and Wexler (2003) and Wexler, Gavarró and Torrrens 
(2004), the UCC in early grammars results in clitic omission in languages with 
past participle agreement alone. Schaeffer attributes the low rates of produced 
clitics at the onset of L1A of Dutch to the optional marking of referentiality in 
child–Dutch. 
Clitic placement in most early languages, including Catalan, Italian, 
Romanian, Spanish, SMG and Serbo–Croatian has been reported to be adult–like, 
with the interesting exceptions of CG (Petinou & Terzi 2002) and EP (Lobo & 
Costa 2012), in which clitic misplacement has been attested. This is an intriguing 
and understudied phenomenon, which deserves closer examination in order for its 
  
132 
nature to be fully understood. This is precisely the focus of the empirical section 
of the current thesis. The methodology implemented and the results obtained are 
outlined in the following two chapters respectively.    
 133 
CHAPTER 5: METHODOLOGY 
 
5.1. Introduction 
 
The discussion in this chapter revolves around various methodological aspects of 
the two studies conducted to investigate clitic L1 acquisition in Cypriot Greek: a 
spontaneous speech study, and an elicited–production experiment. The research 
questions posed, the hypotheses sketched and the predictions made are outlined in 
the first section. The second and third sections present the profile of the 
participants, the method and the materials used, as well as the analysis conducted 
on both the naturalistic and the experimental data. 
 
5.2. Hypotheses of the present study 
 
The goal of the present study is twofold. On the one hand, it aims to explore clitic 
L1 acquisition in CG and to answer a number of research questions in relation to 
the developmental stages that CG–speaking children pass through until they reach 
the target grammar. On the other hand, it aims to investigate the theoretical 
implications of the developmental patterns attested and to accommodate the child 
data within a formal account of cliticisation in CG. 
The results obtained by Petinou and Terzi (2002) for early CG, as well as 
those by Lobo and Costa (2012) for early EP, are taken as the starting point. In 
Petinou and Terzi’s (2002) study, only 2 (out of 5) typically developing (TD) 
children had relatively high percentages of misplaced clitics, even at the earliest 
stage reported (2002:13, table 2), which raises several questions about the 
generalisability of the phenomenon attested across participants. The current study 
investigates the acquisition of object clitics in a systematic and thorough fashion 
with a large number of children. 
The first set of research questions is addressed from a developmental 
perspective: 
1. Is the phenomenon of clitic misplacement attested in early CG? 
2. What is the age range of the children exhibiting clitic 
 134 
misplacement?  
3. Is it generalised across participants within the relevant age range? 
4. If not, what proportion of the overall population exhibits the 
phenomenon? 
Concerning the theoretical implications of the above questions, the following 
hypothesis is developed: if the phenomenon of clitic misplacement is generalised 
across all (or a large proportion of) the participants within a defined age range, it 
marks a distinct developmental stage in the course of L1 acquisition of typically 
developing CG–speaking children1  (Hypothesis 1). 
In the second place, what is of crucial importance is to establish whether 
the target–deviant pattern attested is part of child grammar. Hence, I next ask 
whether the children’s production is characterised by real optionality (see Parodi 
& Tsimpli (2005) for a definition of real and apparent optionality in the context 
of L2 acquisition) or not: 
5. Is clitic misplacement attested in both enclisis and proclisis 
contexts? 
6. If it is only attested in either proclisis or enclisis contexts alone, 
does it occur in all the syntactic contexts exhibiting the relevant pattern? 
I am assuming that the manifestation of real optionality in children’s clitic 
placement will result in the arbitrary choice of either the pre– or the post–verbal 
position irrespective of the syntactic context. As a result, clitic misplacement will 
occur in both enclisis and proclisis contexts. On the other hand, the 
overgeneralisation of either the one or the other pattern will show that children are 
consistent, albeit non–adult–like, in their clitic (mis)placement. On the basis of 
the above considerations, the following hypothesis is developed: if clitic 
misplacement occurs in both enclisis and proclisis contexts in comparable 
percentages, it shows real optionality in children’s clitic placement. The 
overgeneralisation of either the enclisis or the proclisis pattern in all clitic contexts 
provides evidence that child grammar is not characterised by real optionality 
                                                
1 Petinou and Terzi (2002) claim that across–the–board clitic misplacement is attested in SLI 
populations alone and, presumably, constitutes a clinical marker for language impairment(s). It 
would be, therefore, interesting to check whether across–the–board clitic misplacement is attested 
in typically–developing (TD) populations on the basis of a much larger database. 
 135 
(Hypothesis 2). 
If a systematic pattern is attested, the next step is to identify which aspect 
of child grammar appears to be defective. Following Marinis’ (2000:259–260) 
views in assuming the Minimalist Program (Chomsky 1995, 1998) as well as the 
Syntax–Morphology Interface as defined in the framework of Distributed 
Morphology (Halle & Marantz 1993), the formal features of clitic pronouns can 
be checked prior to spell–out, while lexical items are inserted after spell–out. 
Within such a framework, clitic omission could potentially be attributed both to 
an impoverished lexicon and/or to an impoverished grammar (cf. Guasti 1993/94, 
Wexler et al. 2004 inter alia). Clitic misplacement, however, could only be 
attributed to an impoverished grammar. Moreover, it has been claimed that CG–
speaking children have adult–like clitic production from as early as 2 years of age 
(see table 3 in Grohmann et al. 2012). This can be taken as evidence for the early 
construction of the relevant lexical items in the child’s lexicon. Hence, children’s 
target–deviant performance in clitic placement would unequivocally constitute 
evidence for a defective computational system.  
Clitic placement, as shown in chapter 3, is directly related to the 
inflectional domain of the clause. What is therefore of crucial importance is to 
investigate whether clitic misplacement correlates with defective I. In many 
European languages the use of optional infinitives indicates that the IP is not fully 
specified. Varlokosta, Vainikka and Rohrbacher (1998) claim that the early non–
finite verbal forms attested in early Greek correspond to the optional infinitives 
attested in other early languages. I adopt Varlokosta et al.’s (1998) proposal and I 
hypothesise that if the proportion of early non–finite verbal forms in constructions 
involving misplaced clitics is significantly higher than that of finite verbs, target–
deviant clitic placement may correlate with a non–adult–like I(nflection) 
(Hypothesis 3). If, however, the proportion of finite verbs in these constructions is 
higher, no such correlation holds. The (non–)existence of a correlation between 
clitic misplacement and early non–finite forms would provide useful indications 
for the second set of research questions. 
From a theoretical perspective this investigation aims at explaining the 
child data in the context of a formal account of clitic placement in CG. A number 
 136 
of different proposals have been put forward over the past two decades to account 
for clitic placement in CG. Chapter 3 discussed purely syntactic accounts 
proposed by Agouraki (2001) and Terzi (1999a, 1999b), a purely prosodic 
account by Condoravdi and Kiparsky (2001) and interface accounts by 
Mavrogiorgos (2012) and Revithiadou (2006, 2008). Acquisition data constitute a 
good tool for testing the adequacy of these proposals. All the syntactic accounts 
assume that finite enclisis derives from proclisis. Thus, for enclisis to obtain, the 
manifestation of additional syntactic (movement) operations is required, namely 
V–to–C in Agouraki (2001) and V–to–M in Terzi (1999a, 1999b). On the basis of 
this account, a legitimate prediction is that proclisis, being the less demanding 
pattern derivationally, will emerge first in child grammar. The same prediction 
follows with respect to Condoravdi and Kiparsky’s (2001) account but on 
different grounds. In their account, proclisis appears as the default pattern, while 
enclisis results from the application of Prosodic Inversion. Hence, proclisis is less 
demanding derivationally in the sense that it does not require the application of 
Prosodic Inversion. 
In Mavrogiorgos’ (2012) and Revithiadou’s (2006, 2008) accounts, on the 
other hand, the derivation of proclisis and enclisis are two independent 
procedures. The former account suggests that, in proclisis, the syntactic features 
and the PF requirement of the F head are satisfied by a feature!congruent XP or X 
and clitics surface in a functional head immediately c!commanded by F, while in 
enclisis the verb moves to F to satisfy F’s features and the PF requirement. 
Revithiadou (2006) assumes that syntax provides two equally well!formed 
structures with a pre! and a post!verbal clitic, respectively, and PF filters the 
syntactic output on the basis of prosodic constraints. Thus, in all the above 
accounts except those of Revithiadou’s and Mavrogiorgos’, proclisis is considered 
the default pattern, as the less demanding derivationally, while for enclisis to be 
manifested additional operations are required (verb movement or Prosodic 
Inversion). Consequently I take the claim that proclisis is the default pattern as 
Hypothesis 4 in this study. 
The present study is based on corpora of spontaneous speech, both cross–
sectional and longitudinal, as well as on experimental data, and assumes that not 
 137 
only the larger number of participants but the different types of methodologies 
will also increase the reliability of the results obtained. The database built for the 
purposes of the current investigation consists of data from 58 children2 overall. 
 
5.3. Spontaneous data 
 
The first study is based on a corpus of naturalistic data. Given that no corpora of 
early CG are available in databases accessible to the research community (e.g. 
CHILDES), the first step for the current investigation was the construction of a 
corpus of spontaneous speech.  
 
5.3.1. Participants 
 
Eight typically developing (TD) Greek–Cypriot children, 3 females and 4 males, 
whose age ranged between 2;3 and 3;4 years took part in this study, and one of them 
was also followed longitudinally. It should be noted that the original design of the 
study involved the building of longitudinal corpora for all the eight TD Greek–
Cypriot children. However, the building of such a large corpus of longitudinal data 
ended up being much more time–consuming than originally estimated. So, I had to 
rethink what methodology would better serve the aims of the current investigation. 
Our firm view was that a larger pool of participants would increase the reliability of 
the results obtained. This was relevant for our study since one of the main research 
questions it aimed to answer was whether the phenomenon under investigation was 
manifested by a large proportion of the population in discussion. I also thought that 
the inclusion of both experimental and spontaneous data was of equal importance. 
So, I decided to collect cross–sectional data, both naturalistic and experimental. In 
addition, I decided to build a longitudinal corpus for a single child. The underlying 
rationale for this was the following: even though following longitudinally a single 
child comes with the usual limitations that case studies imply, it could still offer 
useful indications for the robustness and the duration of the phenomena under 
                                                
2 The children that participated in the spontaneous speech study are different from the children 
who performed the elicited production experiment. 
 138 
investigation. The selection of the child who was followed longitudinally was not 
arbitrary: he was the youngest participant with ceiling percentages of misplaced 
clitic pronouns. 
All the participants were monolingual speakers of CG3, with a monolingual 
CG–speaking background and no history of cognitive deficits or language 
impairments. They were all residing in Limassol, Cyprus. They all belonged to 
middle–class households but no detailed socio–economic information was collected. 
The profiles of the participants of the spontaneous speech study and the longitudinal 
study are sketched in tables 5.1 and 5.2 respectively. 
 
5.3.2. Method 
 
Samples of spontaneous speech were collected for the corpus. In order to achieve the 
optimal level of interest and comfort for the participants in the study, the recordings 
took place at each child’s home. The experimenter met the child’s parents well in 
advance, so that they could introduce her to the children as a friend of theirs. A 
number of prompts, including picture books, wooden puzzles and stickers, were used 
by the experimenter for the elicitation of naturalistic data, with a focus on the 
elicitation of constructions involving object clitics. The child was audio–recorded 
while s/he interacted with the experimenter. The recordings were performed with a 
digital voice recorder, introduced to the children as the experimenter’s mobile phone. 
In this way, the required proximity of the voice recorder to the interlocutors was 
maintained, which resulted in good quality recordings, while the distraction for the 
participant was kept at low levels. Each recording session lasted approximately an 
hour and each child was recorded once. The youngest participant of the study [S1] 
was followed longitudinally and recorded every 1–1.5 months over a period of six 
months. 
 Data transcription was performed in accordance to the CHAT conventions 
of the Child Language Data Exchange System (CHILDES), the largest database for 
                                                
3 All children born in Greek Cypriot families and brought up in Cyprus are exposed to two 
varieties of Modern Greek, namely CG and SMG (see Arvaniti 2010 and a host of references 
therein). Therefore, the term monolingual in the Cyprus linguistic context refers to children that 
were not exposed to any other language(s). 
 139 
child language, as outlined in MacWhinney (2012a (electronic edition), 2000 (last 
printed edition)). To be precise, the data were transcribed following the standards set 
by mid–CHAT, as described in MacWhinney (2012a:20–21). These transcripts 
constituted the corpus of utterances for each child. A sample of the transcribed data 
is offered in (1).  
 
(1) [Neokleous corpus: S5, 2;10] 
@Begin 
@Languages: cgr 
@Participants: CHI S5 Child, EXP Theoni Investigator, MOT Mother, FAT 
 Father 
@ID: cgr|theoni|CHI|2;10.0||Target||Child|| 
@Date: 20–JAN–2010 
*CHI: &ina epa(r)kara !u to . 
*EXP: eparkares mu to ? 
*EXP: pundo "iko mu ? 
*CHI: eto "ame (.) en tuto . 
*EXP: u (.) estamatise (.) jati: ? 
*CHI: jati elip!e i pe"ina tu <pale> [?] . 
*EXP: e ti ena kamis kalo ? 
*CHI: pai !i#a [/] !i#a . 
*EXP: tora pu elipsen i pezina tu ti ena kamis ? 
*CHI: o: pai !i#a [/] !i#a . 
*EXP: pai si#a [/] si#a . 
*CHI: ena vano k' ali pe"i(na) jati elip!e . 
*CHI: to(r)a ena [/] ena pai "inata . 
*EXP: emena ti ena mu kamis tora pu elipse dz' emena ? 
*CHI: eva(l)a !u te !ena . 
[…] 
*EXP: ti na tis ka(m)umen tora tis mamas tus ? 
*CHI: to [/] to(r)a ena et!eta!o . 
*EXP: ti: ? 
*CHI: to(r)a ena eteta!o ti . 
*EXP: ena tin eksetasis ? 
*EXP: ise jatros tipote ? 
*CHI: ne . 
*EXP: ne:, <ti tin kamnis> [/] ti tin kamnis ? 
*CHI: ena va(l)o ti xame te na ete:ta!o ti . 
*EXP: (n)a ti valis xame dze (n)a tin eksetasis ? 
*EXP: hm (.) ti eshi, xtipa i kar"ja tis ? 
*CHI: ne . 
*EXP: e simeni e(n) zondani (.) a ? 
*EXP: tak [/] tak [/] tak xtipa i kar"ula tis (.) a ? 
*CHI: to kefalin ti! ? 
*EXP: to kefalin tis en kala ? 
 140 
*EXP: ma ti ena tis kamis ? 
*CHI: ena va(l)o . 
*CHI: poni <t' eva(l)a ti(s)> [/] t' evala ti(s) !lipi . 
*EXP: ti tis evales ? 
*CHI: eva(l)a ti(s) !lipi . 
*FAT: !rapse tis farmaka [CHILD’S NAME] . 
*CHI: a: [/] afu e(!)lata ti(s) . 
*FAT: !rapse (t)is farmaka na pkjasi . 
*CHI: <(e)na felo k' alo> [/] ena felo k' ala xxx jati ena . 
*CHI: e [/] e [/] e(n) "iko"i . 
*CHI: ja [/] jati ena po(n)i"i pola . 
*EXP: ti ena kamis tis arku#as ? 
*CHI: &ep ena te [/] te ena felo alo jati e [/] en ek"eta"a to akoma . 
*CHI: jati ena po(n)i"i <pola> [>] . 
 
It was essential to strictly adhere to the conventions of the CHAT 
transcription format in order for the CLAN (Computerized Language Analysis) 
commands to run successfully on these files. CLAN is a programme purpose–
designed for the analysis of data transcribed in the CHAT format (MacWhinney 
2012a). All the calculations reported in the current chapter were performed by 
CLAN. Tables 1 and 2 report the number of turns, utterances and words produced 
per recording session by the participants in the spontaneous speech study and the 
longitudinal study respectively. The three rightmost columns report the ratio of 
words per turn, utterances per turn and words per utterance. The Mean Length of 
Utterance in Words (MLUw) for the purposes of the current study is identified 
with the number of words per utterance. The MLUw rate is considered a very 
good way to measure children’s linguistic development. The unit of linguistic 
analysis used for the calculation of the Mean Length of Utterance is, thus, the 
word rather than the morpheme. This issue is discussed in the following section. 
   
 141 
 
Table 5.1: Responses elicited in the spontaneous speech study. 
 
 
 
Table 5.2: Responses elicited in the longitudinal study. 
 
5.3.3. MLT analysis 
 
The computation of the ratio of words to turns, utterances to turns and words to 
utterances was performed by the CLAN program, as part of the Mean Length of 
Turn (MLT) analysis, described in detail in the CHILDES manual (MacWhinney 
2012b:45–47, 111–113). The following definition is adopted for the term turn: a 
sequence of utterances spoken by a single speaker. Following standard practice in 
the acquisition literature, I am using the ratio of words to utterances to assess the 
child’s language development. The computation of MLUw is essential for the 
purposes of our investigation, as the chronological age is not always a reliable 
indicator for the developmental stage of young children in terms of language 
abilities. 
The reason for calculating the mean length of utterance in words (MLUw) 
rather than the mean length of utterance in morphemes (MLU) is twofold. First, a 
Child/Age Number Ratio 
Utterances Turns Words Words/ 
Turns 
Utterances/ 
Turns 
Words/ 
Utterances 
S1 2;4 464 464 1104 2.379 1.000 2.379 
S2 2;9 462 400 1285 3.213 1.155 2.781 
S3 2;9 440 392 1704 4.347 1.122 3.873 
S4 2;10 732 629 2190 3.482 1.164 2.992 
S5 2;10 469 461 1028 2.230 1.017 2.192 
S6 2;11 503 437 1399 3.201 1.151 2.781 
S7 3;4 440 396 1302 3.288 1.111 2.959 
S8 3;4 828 644 3258 5.059 1.286 3.935 
Child/Age Number Ratio 
Utterances Turns Words Words/ 
Turns 
Utterances/ 
Turns 
Words/ 
Utterances 
S1 
 
2;4 464 464 1104 2.379 1.000 2.379 
2;5 422 409 886 2.166 1.032 2.100 
2;8 437 421 1159 2.753 1.038 2.652 
2;10 520 482 1338 2.776 1.079 2.573 
 142 
morpho–syntactic (MOR) lexicon4 for CG has not yet been constructed. MLU 
rates correspond to the ratio of morphemes over utterances, and thus MLU 
depends on the correct morphemic analysis of words. The calculation of MLU in 
files transcribed in the CHAT format is performed by the MOR program. It is 
essential that the MOR and POST programs are used to construct a morphemic 
analysis of the language of the transcribed data on which the analysis is to be 
performed; this requires a good MOR grammar and a POST disambiguation 
database  (MacWhinney 2012b:143–182). Since neither a MOR lexicon nor a 
POST database is available for CG, the automatic tagging of the corpora in 
question in terms of morphemes could not be performed. Secondly, the 
morphemic analysis takes the morpheme, which usually corresponds to a lexical 
or functional word, as the unit of analysis. However, defining morphemes across 
languages has been proven to be a controversial topic in the acquisition literature, 
especially in highly inflected languages such as Greek. Contrastingly, taking the 
word as the unit of linguistic analysis has the positive result that the respective 
computations are not theory–bound. Moreover, the computation of MLUw is 
expected to correlate with the MLU rates. MacWhinney (2012b:116) indicates 
that for English (Malakoff, Mayes, Schottenfeld & Howell 1999), Spanish 
(Aguado 1988) and Irish (Hickey 1991) a very strong correlation was found 
between MLU rates and MLUw rates (a correlation of 97% for English, of 99% 
for Spanish, and of 99% for Irish).  
MLT calculations, as performed by the CLAN program, included all 
utterances and words produced by the child in a single file. Turns that involved 
unintelligible vocalisations transcribed as xxx were also counted as utterances 
and/or turns, but the unintelligible strings were excluded from word counts. MLT 
calculations also excluded material followed by [/], [//] and [///], as well as the xxx 
symbols from the word counts. Moreover, pairs of utterances that use the +, and 
+. continuation codes were counted as single utterances. 
 
 
                                                
4 MOR grammars exist for the following languages: Cantonese, Chinese, Danish, Dutch, English, 
French, German, Hebrew, Japanese, Italian and Spanish and are available in: 
http://childes.psy.cmu.edu/morgrams/. 
 143 
5.3.4. Data analysis 
 
Data analysis included only (a) fully intelligible, (b) multi–word, and (c) 
spontaneously used utterances (imitations and immediate self–repetitions were 
discarded), and investigated clitic pronoun production and placement in early CG. 
The first stage of data analysis focused on clitic production. In order to 
evaluate and analyse children’s performance in the employment of clitics, all the 
contexts requiring them in adult language had to be measured and analysed. The 
discussion on the L1 acquisition of European Portuguese (Costa & Lobo 2007a et 
seq., Silva 2008) in chapter 4 has shown that it is not trivial to disambiguate 
between object drop and clitic omission in child data. Consider example (1), 
which constitutes an experimental question–answer pair taken from Costa and 
Lobo’s (2007a) study.  
 
(1) Experimenter: Diz–lhe lá o que o Pooh fez ao Tigre! 
   “Tell him what Pooh did to Tiger” 
Expected response: Penteou–(o) 
             “combed–(him)” 
 
The expected response is obviously one which involves a clitic pronoun. 
However, if the elicited response is penteou, what is unequivocally missing is the 
direct object, which could have either been a clitic pronoun or a full DP. Given 
that the referent (Tigre) has already been introduced into the discourse, the target 
response should involve a clitic pronoun and not a full DP whose usage would 
render the response pragmatically inappropriate. However, Costa and Lobo 
(2007a) report significantly higher DP production in strong island contexts where 
null objects are ruled out in adult language. This shows that children may use a 
full DP even in contexts where adults don't. In this case, we will get a 
grammatical, albeit pragmatically over–informative, utterance. In a similar 
fashion, as for utterances lacking an overt object, it is difficult to decide whether 
they constitute instances of object drop or clitic omission. Taking all the above 
into account, I consider all these cases as instances of object omission, without 
 144 
disambiguating between clitic omission and object drop. The computations in the 
corpora examined were, hence, performed as follows: the absolute numbers and 
the relevant percentages of object realisation versus object omission were 
calculated and only cases that unequivocally involved a missing clitic pronoun, 
i.e. structures that required clitic doubling or CLLD, were identified as instances 
of clitic omission. 
All the clitic constructions produced were then classified as proclisis– or 
enclisis–triggering contexts on the basis of adult language. Hence, contexts where 
an adult would produce a pre–verbal clitic were taken as proclisis contexts, while 
the ones in which an adult would produce a post–verbal clitic were taken as 
enclisis contexts. The absolute numbers and the respective percentages of clitic 
constructions produced per condition were calculated. The absolute numbers of 
clitics produced in different syntactic contexts falling within these two conditions 
were also counted. In particular, enclisis contexts involved root clauses and 
imperatives, while proclisis contexts involved negatives, clauses headed by the 
modality markers na and enna, wh–elements and other proclisis–triggering 
elements. The absolute numbers were then computed of Single Clitic (SC), Clitic 
Doubling (CD) and Clitic Left Dislocation (CLLD) structures produced. 
The second stage of data analysis focused on clitic placement in enclisis 
and proclisis contexts. In the first place, the absolute numbers and the respective 
percentages of clauses in which clitic placement was target–like and target–
deviant per condition were calculated. The next stage investigated clitic placement 
by syntactic context: on the one hand, root clauses and imperatives, and on the 
other, negatives, clauses headed by the modality markers na and enna, wh–
elements and any other proclisis–triggers. The absolute numbers of clitics in 
adult–like and non–adult–like position were counted. 
 
5.3.5. Clitic misplacement and finiteness 
 
The final stage of data analysis investigated whether a correlation holds between 
clitic (mis)placement and finiteness in early CG. Clitic pronouns are taken to be 
inflectional elements and it has been assumed that their emergence is a result of a 
 145 
fully–fledged IP. In the literature on L1A a correlation has been found to hold 
between clitic omission and the optional infinitive stage (Guasti 1993/94, Hamann 
et al. 1996, Haegeman 1996): it has been shown that, during this stage, clitic 
objects are omitted. Clitic production in early CG, according to Grohmann et al. 
(2012), is at ceiling from as early as 2 years of age. However, clitic misplacement 
has been attested (Petinou & Terzi 2002), a phenomenon absent from most 
European clitic languages. 
An intriguing question is whether this target–deviant performance is the 
result of an impoverished I(nflection) in the early clause. It is, thus, of crucial 
importance to investigate the role of finiteness in the manifestation of clitic 
misplacement. The question is whether a correlation holds between (non–)adult–
like clitic placement and a (non–)fully–fledged I in early CG. Such a correlation 
would strongly suggest that clitic misplacement is the result of an impoverished 
clause structure. 
CG, like SMG, is a language with no infinitival forms. Varlokosta et al. 
(1998) proposed that the stage at which SMG–speaking children over–use 
perfective verbs with the suffix –i (Tsimpli 1992 inter alia)5 corresponds to the 
root infinite stage in languages with infinitives6. The suffix –i in SMG and CG 
alike constitutes the 3rd person singular [–PAST] suffix, as well as the active 
participle. For Varlokosta et al. (1998) perfective –i forms in child Greek 
constitute early non–finite forms. They convincingly argue in favour of the non–
finiteness of the –i forms on the basis of their distribution (i.e. overgeneralisation 
to non–3rd singular contexts), as well as their incompatibility with overt subjects 
(only rare occurrences attested) and finiteness7,8. The fact that the overuse of the–i 
                                                
5 This stage is characterised by the use of the –i form for over half of the occurrences of a verb, 
while a large proportion of these occurrences involve non–3rd singular contexts. Moreover, the 
use of non–i forms is relatively low and there is no evidence for the acquisition of tense or modals 
(Varlokosta et al.1998:197–198). 
6 See Hyams (2002) for a sound argumentation against Varlokosta, Vainikka and Rohrbacher’s 
(1998) prefunctional grammar hypothesis for early Greek, namely the hypothesis that Greek–
speaking children pass through a stage during which their grammar lacks functional categories. 
7 Verbal forms in –i do not emerge along with modals and verbs with productive tense and 
agreement morphology, while their use correlates with the absence of an overt subject; in the 
acquisition literature it is assumed that overt subjects are licensed by a functional projection 
(Varlokosta et al. 1998:198–199). 
8 Hyams (2002) argues that such a hypothesis is not empirically supported by acquisition data in 
the light of cross–linguistic findings of early morphosyntactic convergence on the target grammar. 
 146 
form at the onset of L1A in child Greek is more prominent with the perfective 
stem than with the imperfective stem has been taken by Varlokosta et al. (1998) to 
indicate that the –i form attested in child Greek corresponds to the active 
participle9 (which in adult Greek requires the perfective stem) rather than the 3rd 
singular suffix. Their proposal is based on the fact that in Greek both verbal stems, 
the perfective and the imperfective, may occur with the 3rd person –i suffix. If it 
was assumed, however, that the –i form attested in child Greek constitutes the 
finite 3rd singular form, the fact that the perfective stem is favoured over the 
imperfective stem in child speech would remain unaccounted for. The view that 
the early verbal forms in –i are identified with the participle is also supported by 
the non–finite properties of the participle in Greek, i.e. no agreement with the 
subject. 
Varlokosta et al. (1998) exploit Rizzi’s (1993/94) truncation theory and 
the notion of markedness to account for the emergence of early non–finite forms 
in child Greek. According to their markedness hierarchy, infinitives are the least 
marked verbal forms followed by participles followed by finite verbs; 
markedness, in their analysis, relates to syntactic complexity. They hypothesise 
that young children initially prefer the verbal form that allows them to use (or 
project) as little of the functional hierarchy as possible. A (child) structure 
truncated at the VP–level yields an infinitive, a structure truncated at the ASPP 
level yields a participle, and the projection to AGRS yields a finite verbal form. 
According to Varlokosta et al., children make use of the smallest tree that is 
compatible with a well–formed item of the verbal paradigm. Since Greek has no 
                                                                                                                                 
She, further, claims that Greek children show an early convergence on the adult target as well. She 
puts forward the following two arguments. First, the rates of correct agreement with 1st and 2nd 
person verbs and 3rd person imperfective verbs for all the children in the corpora examined by 
Varlokosta et al. (1998) are higher than predicted under the hypothesis that children lack 
functional projections. Second, the use of modal particles in these corpora is greater than predicted 
by the prefunctional grammar model. However, the high proportion of 3rd person perfective verbs 
occurring in non–3rd person contexts, on the one hand, and the considerable change in frequency 
of modals between children in Stage I and children in Stage II, on the other, seek for an 
explanation.  
9 Hyams (2002) rejects Varlokosta et al.’s proposal that the perfective –i form attested in child 
Greek is a participle on the basis of its interpretation. For Hyams, if the –i form was a participle it 
should have had a perfective, i.e. completive, meaning contrary to fact, since verbal forms in –i 
receive a modal interpretation. In particular, she argues that since children do not typically assign 
wrong aspectual or modal meanings to inflectional forms, showing an early convergence on the 
target form–meaning correlates, the mismatch between the modal meaning and participle form is a 
strong argument against analyzing the –i form as a participle. 
 147 
infinitival forms, Greek–speaking children cannot have a functional structure 
truncated at the VP level. The least marked form that is available in Greek is the 
participle, hence the truncation at the ASPP level10.    
Following proposals in Varlokosta et al. (1998) for early SMG, I take the 
perfective verbal forms ending in –i to constitute early non–finite forms in early 
CG as well. The analysis conducted here aimed to investigate whether clitic 
misplacement in CG correlates with the use of these early non–finite forms. For 
the purposes of this analysis, the data from children exhibiting ceiling percentages 
of misplaced clitics was used, and all constructions involving a misplaced clitic 
pronoun found in their data were analysed. In line with Varlokosta et al.’s (1998) 
proposal, the verbal forms used in constructions with misplaced clitics were 
classified as finite or early non–finite. All the verbal forms in –i were classified as 
follows: the forms with the perfective stem as early non–finite and the forms with 
the imperfective stem as finite. All verbal forms overtly marked for any other 
feature, including [person] and [number], were also classified as finite. 
It should be noted, though, that in (Cypriot) Greek, as well as Standard 
Modern Greek, some verbs lack a perfective stem, i.e. verbs like ka(m)ni (“do”) 
for which the stem ka(m)n– has both perfective and imperfective uses. These 
verbs were excluded from the analysis; however, the raw number of such verbs 
that appear in structures involving misplaced clitics is reported. The absolute 
numbers and the respective proportions of finite and early non–finite verbs were 
computed with the aim of showing whether the use of early non–finite verbal 
forms correlates with target–deviant clitic placement. 
 
 
 
 
                                                
10 Hyams (2002) rejects Varlokosta et al.’s characterization of perfective forms in –i as participles 
and proposes an alternative that primarily aims to capture the modal interpretation of these 
structures. She assumes that under appropriate structural conditions certain temporal/aspectual 
features may license the MoodP and proposes that the perfective forms in –i in Greek is an 
instance of aspectually licensed mood; in particular, the perfective feature in the verb licenses the 
active MoodP under Attract. Such an analysis involves the interaction of the following projections: 
Tense, Aspect, and Mood, hence the rejection of Varkolosta et al.’s proposal that child structure is 
truncated in ASPP. 
 148 
5.4. Elicited production experiment 
 
Elicited production experiments allow the experimenter to construct a situation 
that requires the production of the target structure. On the basis of the results 
obtained from the spontaneous speech study, the aim of the experimental 
investigation was to elicit object clitics in specific types of constructions. A 
picture–based production experiment was designed and implemented for the 
purposes of the current study. 
 
5.4.1. Participants 
 
Fifty–one Greek Cypriot children were recruited from 5 nursery schools in 
Limassol, Cyprus, after approval from the directors and upon written parental 
consent. In a subsequent stage, one of the participants (S28) was removed from 
the data analysis, since he was unable to perform the task. Thus, ultimately, data 
from 50 children were analysed. The participants were divided into 3 age groups, 
namely Age group A=2;6–3;0, Age group B=3;1–3;6 and Age group C=3;7–4;0. 
Table 3 shows the age range (in years and in months), the mean age (in months) 
and the standard deviation (STDEV), as well as the number of children (N) 
included in each age group. The overall number of male and female participants is 
also reported in the table.  
 
GROUP AGE RANGE N MEAN AGE STDEV SEX 
A 2;6 – 3;0 
(30 – 36 mo.) 
18 33.5 2.20 F: 12 / M: 6 
B 3;1 – 3;6 
(37 – 42 mo.) 
22 39.7 1.91 F: 8 / M: 14 
C 3;7 –  4;0 
(43 – 48 mo.) 
10 
 
44.2 1.32 F: 6 / M: 4 
Overall 2;6 – 4;0 
(30 – 48 mo.) 
50 38.4 4.48 F: 26 / M: 24 
 
Table 5.3: Participants: elicited production experiment 
 
 
 
 149 
5.4.2.  Method 
 
Each experimental session involved the implementation of a picture–based task, 
preceded by a warm–up session. The teacher introduced the experimenter to all 
the children in the classroom. The teacher then informed them the experimenter 
had some nice puzzles and books with which they could play for a while. Many 
children volunteered to perform the task and the ones who had parental 
permission were chosen. The testing took place in a quiet room in the respective 
nursery schools. Each participant was tested individually in a single session that 
lasted for approximately 20 minutes. The experimental sessions were audio–
recorded, as were the spontaneous production sessions. 
 
The puzzle task 
A semi–structured elicitation task, the puzzle task, introduced by Eisenbeiss 
(2009), was used as a warm–up session. Two puzzles were used each consisting 
of a wooden board with slots for pieces, and the puzzle pieces that fitted in the 
slots. One of them had a number of jungle animals depicted on its pieces, and the 
other had the characters of the well–known cartoon “Dora the Explorer”. The 
children were initially encouraged to familiarise themselves with the puzzles and 
they were then asked to guide the naïve experimenter to place the pieces in the 
corresponding slots. Throughout their interaction with the experimenter, there 
were many prompts for the use of clitic constructions, including both enclisis (2) 
and proclisis contexts (3).  
 
(2) EXP: Ti      na  kamo    ton  elefanda?  
          What  M  do–1S  the   elephant–ACC 
         “What shall I do with the elephant?” 
CHI: Var        ton                   dhame. [Elicited Clause] 
          Put–2S  him–CL.ACC  here 
“Put it here”  
(3) EXP: Hori      ti    Dora            dhame? 
           Fit–3S  the  Dora–ACC  here 
 150 
 “Does Dora fit here?” 
CHI: Oi,   en      ti                      hori. [Elicited Clause] 
          No,  NEG  her–CL.ACC  fit–3S 
         “No, she doesn’t fit” 
 
The picture–based task 
The picture–based task is an elicited production task for 3rd person singular 
object clitics. Seven pictures chosen from the book “First Hundred Words in 
English” (edited by Amery & Cartwright 2009) were matched with 8 questions 
related to the situation depicted on the corresponding picture. The pictures 
presented two–role transitive activities, e.g. the dog doing something to the girl, 
the boy doing something with his socks, etc. In particular, situations were selected 
in which only a specific sentence, the target construction, was felicitous. 
The task was designed for the elicitation of two types of constructions: (1) 
root clauses and (2) clauses headed by the modal particle na and the future 
particle enna11, within two conditions: enclisis and proclisis contexts. Root 
clauses constitute enclisis contexts, while clauses headed by modal particles 
constitute proclisis contexts. Four clauses per condition were elicited. The verbs 
that were chosen for the elicited constructions had to fulfil two requirements: they 
had to be part of the child’s vocabulary and they had to be transitive. The 
following 4 verbs were selected: thkjevazo (“read”), foro (“wear”), pkjano 
(“take”) and valo (“put”). The experimental questions for the elicitation of root 
clauses involved the verb kamno (“do”) and the experimental questions for the 
elicitation of modality clauses involved the modal particles enna and na and the 
verb kamno. No particular attention was paid to the tense of the verb involved in 
the prompts. Mastropavlou, Petinou and Tsimpli (2014) have shown that typically 
developing (TD) CG–speaking children, as well as children diagnosed with SLI, 
were not affected by the verb tense in the production of object clitics. 
Only 3rd person singular object clitics were elicited; see the forms in bold 
in table 5.4. I controlled for both genitive and accusative case and all the three 
                                                
11 Structures headed by particles na and enna are subsumed under the same category, as it is 
assumed that both particles are realised under CM; see the discussion in chapter 1 of this thesis. 
 151 
genders to be represented in the task. A complete list of the experimental 
materials used can be found in Appendix I. 
 
Table 5.4: The morphological paradigm of CG clitics 
 
The testing procedure was simple; the children were shown the pictures one 
at a time. The experimenter would point at a picture (see sample picture below) 
and ask a question related to the situation depicted on it (4), in order to elicit a 
clitic construction (5). Notably, the antecedent of the elicited clitic pronoun was 
introduced by the experimenter, i.e. in the example material in (4–5) the DP to 
kadhro is the antecedent of the clitic to. This is essential given that only a D–
linked definite object appears in a clitic form.  
In the case of a non–answer, the question would be repeated once. If the 
child still gave no response, no further help would be provided and the 
experimenter would proceed to the next item. Over the duration of the experiment, 
the child was prompted to answer 2–3 filler questions that aimed to elicit 
constructions involving intransitive verbs. A pilot study was conducted with adult 
speakers of CG and two young children prior to the administration of the test to 
the participants of the experiment. 
 
 
 
CONTENT REMOVED FOR COPYRIGHT REASONS 
 
 
 
 
 
 
Number Case 1st 
person 
2nd 
person 
3rd person 
Masculine Feminine Neuter 
Singular Genitive 
Accusative 
mu 
me 
su 
se 
tu 
to(n) 
tis 
ti(n) 
tu 
to 
Plural Genitive 
Accusative 
mas 
mas 
sas 
sas 
tus 
tus 
tus 
tes 
tus 
ta 
 152 
(4) EXP: Kita dhame tutin tin ikona. I mama krata mia sponda. 
          “Look at this picture. Mum is holding a nail.” 
          Ti theli na kami to koritsaki to kadhro? 
          “What does the girl want to do with the frame?” 
(5) CHI: Na  to                 vali  /  kremasi    (sto       diho) . 
          M   it–CL.ACC  put      hang–3S  on_the  wall–ACC 
“To put / hang it on the wall”  [Elicited Clause: na–clause]        
          
Children’s responses to the experimental questions were entered in answer 
sheets (see Appendix I). No responses and other answers (i.e. production of an 
interrogative or a root clause in a context that required a modal clause) were 
calculated but disregarded from further analysis. All the modal clauses produced 
after an experimental question aiming to elicit a modal clause and all the root 
clauses produced after a corresponding experimental question were categorized as 
target and non–target responses. Only structures involving a clitic pronoun were 
categorised as target responses. Non–target responses included errors of (clitic) 
omission, e.g. [na kremasi] instead of [na to kremasi], and the use of a full  DP 
object instead of a clitic, e.g. [na kremasi to kadhro] instead of [na to kremasi]. 
Inflectional errors related to the features of the clitic, i.e. person, number, case and 
gender, were not taken into consideration for the inclusion of the elicited clause in 
the relevant category. To give an example, if the child produced a clause headed 
by a modal particle such as (7) as a response to the corresponding triggering 
question (6), the elicited clause would be categorised as enna–clause, irrespective 
of the inflection of the clitic pronoun; example (6–7) is a question–answer pair 
taken from one of the experimental sessions (participant P37). In the experimental 
question (6), a neuter definite object is involved, i.e. tu moru (the baby–GEN). 
However, the child arbitrarily assigns female gender to the baby in his response 
and uses a female clitic to refer to it (7). Irrespectively of the gender mismatch 
between the anteceded and the object clitic, the child utterance was categorised as 
enna–clause since it was produced in the appropriate triggering context. 
 
 
 153 
(6) EXP: Ti       enna  kami    o     papas          tu    moru ?  
What  M      do–3S  the  dad–NOM  the  baby–GEN 
“What is dad going to do to the baby?” [Experimenter] 
(7) CHI: Enna  ti               voithisi    na  vali       ti    fanelan           tis. 
M       her–ACC  help–3S  M  put–3S  the  t–shirt–ACC  her–POSS 
“He will help her to put on her t–shirt” [Elicited Clause] 
 
Elicited clauses were then coded as correct and incorrect as follows: a 
clitic construction was coded as correct if the clitic was placed pre–verbally in a 
proclisis environment (na/enna–clause) or post–verbally in an enclisis 
environment (root clause), and incorrect if the clitic was placed post–verbally in a 
proclisis environment or pre–verbally in an enclisis environment. Finally, the 
absolute numbers and the respective percentages were calculated. 
 
5.4.3. Double realisation of the clitic pronoun 
 
An unexpected pattern attested in the spontaneous data of some of the children, 
the double realisation of the clitic pronoun in two positions: immediately 
preceding and immediately following the verbal host, required further scrutiny. 
Thus, it was deemed necessary to study whether this particular pattern occurred in 
the data of more children. In order to explore this issue, the entire recorded 
experimental session from a subset of the children that participated in the 
experimental investigation was transcribed and examined. 
The first step was choosing one of the 5 nursery schools from which 
participants were recruited. The selection was based on two criteria. First, all the 
age groups had to be equally represented in the chosen group of participants, and, 
second, the representation of the two genders in the chosen group had to be 
balanced. At the chosen school, 11 children had been tested, 5 male and 6 female, 
whose age ranged from 2;8 to 3;9. All 3 age groups were represented: 3 children 
from age group A, 5 children from age group B and 3 children from age group C. 
Table 5 provides information about their age and the number of turns, words and 
utterances they produced per experimental session. The ratio of words to turns, 
 154 
utterances to turns and words to utterances (MLUw)12 are reported in the table. 
This subset represents one–fifth of the overall group of participants in the 
experiment. 
 
 
Table 5.5: Subset of participants: elicited production experiment 
 
 The audio–recorded material for each of these children, comprising the 
implementation of both the puzzle task (warm–up session) and the elicited–
production task, was transcribed according to the conventions of the mid–CHAT 
transcription format (MacWhinney 2012a); see section 5.3.2 of the current 
chapter. These transcripts constituted the corpus of utterances for each child and 
were examined in order to identify instances of double realisation of the clitic 
pronoun in two positions: pre– and post–verbally. 
Having now introduced the methodology, the next chapter will present the 
results obtained. 
                                                
12 A detailed discussion for MLT calculations is given in section 5.3.3 of this chapter. 
Child 
Age 
Number Ratio 
Utterances Turns Words Words/ 
Turn 
Utterances/ 
Turn 
Words/ 
Utterances 
P6 2;8 347 341 757 2.220 1.018 2.182 
P13 2;11 317 310 588 1.897 1.023 1.855 
P18 3;0 200 195 407 2.087 1.026 2.035 
P22 3;2 362 352 833 2.366 1.028 2.301 
P26 3;2 151 146 486 3.329 1.034 3.219 
P30 3;3 231 231 515 2.229 1.000 2.229 
P35 3;5 126 113 462 4.088 1.115 3.667 
P37 3;6 119 114 362 3.175 1.044 3.042 
P44 3;7 148 142 496 3.493 1.042 3.351 
P48 3;9 131 125 377 3.016 1.048 2.878 
P50 3;9 74 73 197 2.699 1.014 2.662 
 155 
CHAPTER 6: RESULTS  
 
6.1. Introduction 
 
This chapter presents the results and analysis of the two studies conducted: the 
spontaneous speech study, involving both cross–sectional and longitudinal data, 
and the elicited production experiment. The results are presented in the form of 
tables and graphs, while the statistical analysis is reported following standard 
practice. The chapter is organised as follows: the first section presents clitic 
production and clitic placement in naturalistic data. The second section 
investigates the phenomenon of clitic misplacement on the basis of experimental 
data. The third section presents an unexpected phenomenon attested in the corpora 
examined, namely the double realisation of the clitic pronoun in both pre– and 
post–verbal position. The last section summarises the findings.  
 
6.2. Spontaneous data 
 
Two corpora of spontaneous speech, the first consisting of cross–sectional and the 
second consisting of longitudinal data were examined; see tables 5.1–5.2 in 
chapter 5 for more information about the corpora in question. The focus of the 
investigation was clitic placement and, in particular, the phenomenon of clitic 
misplacement attested in early CG (Petinou & Terzi 2002). Initial inspection of 
spontaneous production offered useful indications as to the nature, the generality 
and the robustness of the phenomenon. Further analysis of spontaneous data 
provided information about the use of simple and complex clitic constructions, 
with a focus on the production and placement of single clitics, as well as clitic 
clusters, in three types of constructions: Single Clitic (SC), Clitic Doubling (CD) 
and Clitic Left Dislocation (CLLD). The presence or absence of a correlation 
between the target–deviant clitic placement attested in a large group of children 
and the production of early non–finite forms was also investigated.   
 
 
 156 
6.2.1. Clitic production and clitic placement 
 
Tables 6.1–6.5 report the outcome of the first stage of data analysis on the first 
corpus consisting of cross–sectional data from 8 monolingual CG–speaking 
children. In these tables, as in all the tables in this chapter, the participants are 
identified by a serial number that appears in the first column. These serial 
numbers were assigned to the participants of both studies on the basis of their 
chronological age: a serial number starting with an “S” for the participants of the 
spontaneous speech study and with a “P” for the participants of the experimental 
study. The chronological age and the MLUw calculated by the CLAN tool for 
each participant (as described in section 5.3.3) are given in the second and third 
columns respectively. 
The first stage of data analysis focused on clitic production and clitic 
placement. To evaluate and analyse children’s performance in the employment of 
clitics, all the contexts requiring them in adult language were identified and 
analysed.! The discussion in chapter 5 has shown that it is not trivial to 
disambiguate between object drop and clitic omission in child data, hence, the 
analysis in the current chapter does not disambiguate between the two. There are, 
however, cases where the missing complement is unequivocally a clitic pronoun, 
such as Clitic Doubling (CL) or Clitic Left Dislocated (CLLD) structures. These 
cases were identified and are reported in table 6.1. It should also be noted that 
some of the structures with a missing object produced by the children are 
grammatical in adult language, like the one in (1). These structures were 
disregarded from the analysis performed. 
 
(1) *EXP: eshis kane(n)an arku!akin pkjo oreo pu to !iko mu pu enen oreo?  
 “Do you have a nicer teddy bear than mine which is not nice?” 
*CHI: pa(o) na  fero.  
 go–1S  M  bring–1S (S4, 2;10) 
 “I’m going to bring (one)” 
 
 157 
Repetitions were also disregarded from the calculations provided that the 
exact form of the verb was repeated. To exemplify, consider examples (2) and (3). 
Example (2) involves two consecutive child utterances: while the first utterance 
involves a clitic, the second one lacks one. The use of the clitic pronoun in the 
first instance is an indication that the child is aware that a clitic pronoun should be 
provided; such cases may arguably constitute performance errors. Hence, for the 
purposes of the current analysis, the second child utterance in (2) was disregarded 
from data analysis. Example (3) involves two similar child utterances: the first 
one lacks a clitic, while the second involves one. However, the form of the verb in 
the two utterances differs: the first utterance involves the subjunctive form of the 
verb (na valo) while the second involves the past tense of the verb (evala). For the 
purposes of the current analysis the utterance lacking a clitic pronoun in (3) was 
calculated as an instance of object/clitic omission. 
 
(2) *CHI: kame to 
 do–2S  it–CL.ACC 
*EXP: ti? 
 what 
*CHI: kame. 
 do–2S (S5, 2;10) 
(3) *CHI: (n)a  valo? 
 M     put–1S 
*CHI: evala  su. 
 put–1S  you–CL.DAT (S2, 2;9) 
 
Table 6.1 reports the raw numbers of overt and omitted objects/clitics and 
the respective percentages in parentheses. The percentages of clitic production 
were calculated by dividing the raw number of clitics produced by the raw 
number of clitic contexts identified in the respective corpora. And, the 
percentages of clitic/object omission were calculated by dividing the raw number 
of complement–less structures by the raw number of clitic contexts identified in 
the respective corpora. The cases that unequivocally involve a missing clitic 
 158 
pronoun, such as clauses involving Clitic Doubling or Clitic Left Dislocation, are 
reported in square brackets. 
 
Table 6.1: Clitic production and clitic/object omission (Spontaneous data). 
 
Examples of children’s utterances involving an omitted clitic/object are 
given in (4–10). In examples (4–7), the experimenter asks a question aiming to 
elicit a clitic structure, given that the antecedent is already introduced. Yet, 
children’s responses involve a verbal form alone with a missing clitic. In example 
(8), the child fails to use the standard expression “kataferno ta” (“I can make it”) 
which obligatorily involves a clitic pronoun. The strong pronouns tuto and kino in 
examples (9) and (10) are clitic doubled in adult CG when used as direct or 
indirect objects. These children, however, failed to produce the adult–like 
structure and omitted the clitic. These cases unequivocally constitute instances of 
clitic omission. 
 
(4) *EXP: ti enna tin kamis tin kamilopar!ali pu kratas? 
 “What are you going to do with the giraffe you are holding?”  
*CHI: epetat"a. 
 threw–1S (S4, 2;10) 
Target structure: epetaksa   tin. 
   threw–1S  her–CL.ACC  
Child Age MLUw Clitic Production Clitic/Object 
Omission 
S1 2;4 2.38 69/74 (.93) 5/74 (.07) [1/5] 
S2 2;9 2.78 69/82 (.84) 13/82 (.16) [6/13] 
S3 2;9 3.87 110/116 (.95) 6/116 (.05) [1/6] 
S4 2;10 2.99 134/156 (.86) 22/156 (.14) [2/22] 
S5 2;10 2.19 68/75  (.91) 7/75 (.09) [1/7] 
S6 2;11 2.78 120/131 (.92) 11/131 (.08) [2/11] 
S7 3;4 2.96 63/68 (.93) 5/68 (.07) [3/5] 
S8 3;4 3.93 279/296  (.94) 17/296 (.06) [6/17] 
 159 
(5) *EXP: ama su !oko mja fanela (ti ena tin kamis)? 
 “If I give you a t–shirt (what are you going to do with it)?”  
*CHI: (n)a  foriso. 
 M     wear–1S (S6, 2;11) 
Target structure: (n)a  tin         foriso. 
   M     her–CL.ACC  wear–1S     
(6) *EXP: ti na ka(m)ume ta nihja tu? 
 “What shall we do with his nails?”  
*CHI: tipote (.)  edze   "a  kotsume. 
 nothing     NEG  M  cut–1PL (S2, 2;9) 
Target structure: edze  ena  ta         kopsume. 
   NEG  M   them–CL.ACC  cut–1PL   
(7) *EXP: ne (.) ala ti ena tu kami o kini#os? 
 “Yes, but what will the hunter do to him?”  
*CHI: ena  kami. 
 M    do–3S (S5, 2;10) 
Target structure: ena  tu/ton (+verb) 
   M    him–CL.DAT/ACC 
(8) *EXP: (n)a se voi"iso? 
 “Shall I help you?” 
*CHI: kataferno. 
 make–1S (S7, 3;4) 
Target structure: kataferno  ta. 
   make–1S  them–CL.ACC 
   “I can make it” 
(9) *CHI: na  valo      tuto. 
 M  put–1S  this–ACC 
 “I shall put this” (S6, 2;11) 
Target structure: (n)a  to       valo      tuto. 
   M     it–CL.ACC  put–1S  this–ACC 
(10) *CHI: ena  pkjahjo   kino. 
 M    take–1S  him–ACC 
 160 
 “I will take him” (S3, 2;9) 
Target structure: ena  ton         pkjaso     kino. 
   M    him–CL.ACC  take–1S  him–ACC 
 
S21 and S4 have the lowest rate of clitic production (84–86%). As for S2, 
almost half (46%) of the structures involving a missing complement identified in 
her corpus constitute cases of clitic omission (6/13). In the vast majority of the 
structures with an omitted clitic (5/6), the child does not clitic double the strong 
pronoun dzinos, as in (11). In a similar fashion, the accusative form of the first 
person singular strong pronoun emena in (12) is normally clitic doubled in adult 
language unlike in S2’s utterance. S4 has the second lowest rate of clitic 
production (86%) but, unlike S2, only 9% of the structures with a missing 
complement identified in his database constitute cases of clitic omission (2/22).  
 
(11) *CHI: !elo         (n)a  valo      dzino          dzame   (s)to     poi 
 want–1S  M     put–1S  that–ACC  over there  to–the  leg 
 mu. 
 my–POSS 
 “I want to put that over there on my leg” (S2, 2;9) 
Target structure: !elo        (n)a  to                 valo       dzino…  
   want–1S  M    it–CL.ACC  put–1S  that–ACC 
(12) *CHI: jati    e(n)    "exete  n’  akani      emena? 
 why  NEG  agree–3S  M  bite–3S  me–ACC 
                                                
1 It should be noted that S2 produced 3 complement–less clauses involving the verb fkalo (“take 
out”), which is a transitive verb in adult language. These structures, however, were not taken as 
instances of object/clitic omission because a closer inspection revealed that the verb fkalo (“come 
out”) was used in her corpus in contexts which required the use of the verb fkeno (“take out”); 
corroborative evidence for this conclusion is provided by the fact that no form of the verb fkeno 
(“come out”) was used by S2. An illustrative example of this weird syntax of fkalo is offered in 
(1).       
(1) *EXP: #iris to (.) en pirazi. 
  “Turn it. It doesn’t matter”  
 *CHI: e(n)   (f)kali? 
  NEG  take out–3S (S2, 2;9) 
 Intended structure: e(n)    fkeni? 
    NEG  come out–3S 
    “Doesn’t it come out?” 
 
 161 
 “Why doesn’t he agree to bite me?” (S2, 2;9) 
Target structure: jati    e(n)    !exete  na  me        akani 
   why  NEG  agree–3S  M  me–CL.ACC  bite–3S 
   emena? 
   me–ACC 
 
Table 6.2 reports the absolute number of constructions involving single 
clitics (SC) and clitic clusters (CC) in enclisis and proclisis contexts. The 
proportion of the overall clitic production that arises in each type of context is 
given in parentheses. It can easily be seen from table 6.2 that a comparable 
number of enclisis and proclisis contexts are found in naturalistic child data. This 
indicates that the contexts requiring pre– and post–verbal clitics are approximately 
equally represented in child production, a crucial factor for the purposes of the 
current investigation. 
 
 
Table 6.2: Production of single clitics and clitic clusters in enclisis and proclisis 
contexts (Spontaneous data). 
 
Table 6.3 reports clitic production in imperatives and root clauses, both 
requiring post–verbal clitic placement. The table presents the absolute number of 
Child Age MLUw Enclisis Contexts Proclisis Contexts Overall 
SC CC SC CC SC CC 
S1 
 
2;4 2.38 
 
26 9 33 1 59 10 
35 (.51) 34 (.49) 69 (1.0) 
S2 
 
2;9 2.78 
 
26 1 39 3 65 4 
27 (.39) 42 (.61) 69 (1.0) 
S3 
 
2;9 3.87 
 
49 2 49 10 98 12 
51 (.46) 59 (.54) 110 (1.0) 
S4 2;10 2.99 
 
79 2 53 0 132 2 
81 (.60) 53 (.40) 134 (1.0) 
S5 2;10 2.19 
 
31 3 30 4 64 4 
34 (.50) 34 (.50) 68 (1.0) 
S6 
 
2;11 2.78 51 6 62 1 113 7 
57 (.47) 63 (.53) 120 (1.0) 
S7 3;4 2.96 
 
27 4 29 3 56 7 
31 (.49) 32 (.51) 63 (1.0) 
S8 3;4 3.93 110 17 130 22 240 39 
127 (.46) 152 (.54) 279 (1.0) 
 162 
clitics produced per construction type, while the number in parentheses indicates 
the number of clitic clusters. Examples of children’s production of root clauses 
and imperatives are given in (13) and (14) respectively. 
 
(13) Oi,     vali        mu                  to                 i      mama           mu .  
NEG  put–3S  me–CL.DAT  it–CL.ACC  the  mum–NOM  my–POSS 
“No, my mum puts it for me”  (S3, 2;9) 
(14) Pe          mu                   to                 tuto.   
Tell–2S  me–CL.DAT  it–CL.ACC  this–ACC  
 “Tell me this one” (S8, 3;4) 
 
 
Table 6.3: Clitic production in enclisis contexts: imperatives and root clauses 
(Spontaneous data). 
 
The absolute numbers of clitic structures produced in proclisis contexts are 
reported in table 6.4. In particular, the number of clauses headed by the modality 
markers na and enna (M), negative particles (NEG) and wh–elements (WH) is 
reported in separate columns, while all the other contexts that require pre–verbal 
clitic placement, including clauses headed by focalized XPs, the factive 
complementizer pu and other particles, are subsumed under the category Other. 
The number of clitic clusters is given in parenthesis. 
 
 
 
 
Child Age MLUw Clitic Production 
Imperatives Root Overall 
S1 2;4 2.38 15 (6) 20 (3) 35 (9) 
S2 2;9 2.78 2 (0) 25 (1) 27 (1) 
S3 2;9 3.87 1 (0) 50 (2) 51 (2) 
S4 2;10 2.99 33 (1) 48 (1) 81 (2) 
S5 2;10 2.19 11 (3) 23 (0) 34 (3) 
S6 2;11 2.78 10 (3) 47 (3) 57 (6) 
S7 3;4 2.96 3 (2) 28 (2) 31 (4) 
S8 3;4 3.93 60 (13) 67 (4) 127 (17) 
 163 
 
Table 6.4: Clitic production in proclisis contexts: clauses headed by modality 
markers and other proclisis–triggering elements2, negatives and wh–questions 
(Spontaneous data). 
 
 Examples of children’s production of negative clauses and clauses headed 
by na and enna are given in (15–16) and (17–18) respectively. The reason why 
clauses headed by the modal particle na and the future particle enna are subsumed 
under the same category for the purposes of data analysis is threefold. First, both 
construction types constitute proclisis contexts. Second, it is assumed, and has 
been adequately justified in chapter 1, that na and enna occupy the same 
functional head in the CG clause, namely CM (Roussou 2000). Third, in many 
cases it is difficult to identify which of the two forms is used in child data, given 
that na is often pronounced as [a], while enna is often substituted by its reduced 
form [na] or [a] (as in (17)). Moreover, in some cases the child would repeat an 
utterance with the particle enna using the reduced form [na] the second time. 
 
(15) E(n)   me                  afini     o     papa(s)       tora.   
NEG  me–CL.ACC  let–3S  the  dad–NOM  now 
 “Dad doesn’t let me now” (S2, 2;9) 
(16) Papa [/] papa [/] papa [/] papa  e(n)    m’                   afini    na  pao  
Dad                                           NEG  me–CL.ACC  let–3S  M  go–1S 
(s)ti      &ia (!ia)    ti     Lenja.   
to–the  aunt–ACC  the  Lenia–ACC 
                                                
2 The constructions subsumed under the category Other in the table above are headed by the 
following elements: for S1, [2 pu]; for S2, [focalized XP]; for S3, [1 ama, 2 pu]; for S7, [1 pu]; for 
S8, [4 ama(n), 1 an, 1 embu, 1 pu]. 
Child Age MLUw Clitic Production 
M NEG WH Other Overall 
S1 2;4 2.38 30 (1) 2 (0) 0 2 (0) 34 (1) 
S2 2;9 2.78 29 (2) 7 (1) 5 (0) 1 (0) 42 (3) 
S3 2;9 3.87 52 (10) 4 (0) 0 3 (0) 59 (10) 
S4 2;10 2.99 49 (0) 4 (0) 0 0 53 (0) 
S5 2;10 2.19 21 (3) 12 (1) 1 (0) 0 34 (4) 
S6 2;11 2.78 60 (1) 0 3 (0) 0 63 (1) 
S7 3;4 2.96 25 (3) 5 (0) 1 (0) 1 (0) 32 (3) 
S8 3;4 3.93 125 (19) 18 (3) 2 (0) 7 (0) 152 (22) 
 164 
“Dad, s/he doesn’t let me go to aunt Lenia” (S5, 2;10) 
(17) (Enn)a  to           valo      e!o         dzame [/] tame. 
M         it–ACC  put–1S  I–NOM  there 
“I will put it there” (S2, 2;9) 
(18) Enna  tu(s)                   kedisi     tuto(s)         (o)  po(r)tokalis. 
M       them–CL.ACC  win–3S  this–NOM  the  orange–NOM 
“He will win them this one, the orange” (S3, 2;9) 
 
Table 6.5 reports the absolute numbers of correctly placed and misplaced 
clitics in enclisis and proclisis contexts, while the respective proportions are given 
in parentheses.  
 
Child Age Enclisis Context Proclisis Context 
COR INCOR COR INCOR 
S1  2;4 35/35 (1.0) 0/35 (0.0) 0/34 (0.0) 34/34 (1.0) 
S2 2;9 27/27 (1.0) 0/27 (0.0) 38/42 (.90) 4/42 (.10) 
S3  2;9 51/51 (1.0) 0/51 (0.0) 58/59 (.98) 1/59 (.02) 
S4  2;10 81/81 (1.0) 0/81 (0.0) 1/53 (.02) 52/53 (.98) 
S5  2;10 34/34 (1.0) 0/34 (0.0) 32/34 (.94) 2/34 (.06) 
S6  2;11 57/57 (1.0) 0/57 (0.0) 1/63 (.02) 62/63 (.98) 
S7  3;4 31/31 (1.0) 0/31 (0.0) 30/32 (.94) 2/32 (.06) 
S8  3;4 127/127 (1.0) 0/127 (0.0) 146/152 (.96) 6/152 (.04) 
 
Table 6.5: Correct and incorrect clitic placement in enclisis and proclisis contexts 
(Spontaneous data). 
 
One conclusion easily drawn on the basis of the results reported in table 
6.5 is that, in enclisis contexts, all the children performed in an adult–like way and 
placed the clitic post–verbally; not a single instance of clitic misplacement is 
attested. In proclisis contexts, on the other hand, a subset of the participants (S2, 
S3, S5, S7, S8) performed in an adult–like way, with correct performance levels 
ranging between 90% and 100%. However, the remaining participants (S1, S4 and 
S6) exhibited ceiling percentages of incorrect clitic placement, as they were 
producing post–verbal clitics at rates of 98% and 100%. This bimodal distribution 
of children’s clitic production in proclisis contexts is illustrated in figure 6.1; the 
figures reported on the graph show the proportions of incorrect placement. 
 165 
 
Figure 6.1: Clitic placement in proclisis contexts (Spontaneous data). 
 
Tables 6.6–6.8 report clitic placement in different syntactic contexts. They 
report the absolute numbers and, in parentheses, the proportions of correctly 
placed and misplaced clitics. Table 6.6 presents the figures for imperative and root 
clauses, both requiring post–verbal clitic placement. Table 6.7 reports figures for 
clauses headed by the modality markers na and enna and the negative particles 
dhen and men, requiring pre–verbal clitic placement, and table 6.8 reports figures 
for clauses headed by wh–elements and other proclisis–triggers, also requiring 
pre–verbal clitic placement. It is worth remarking that children’s performance 
with respect to clitic placement is similar irrespective of whether the construction 
involves a single clitic or a clitic cluster; hence, the number of clitic clusters is not 
reported separately. 
 
Table 6.6: Correct and incorrect clitic placement in enclisis contexts: imperatives 
and root clauses (Spontaneous data). 
Child Age Imperatives Root Clauses 
COR INCOR COR INCOR 
S1 2;4 15/15 (1.0) 0/15 (0.0) 20/20 (1.0) 0/20 (0.0) 
S2 2;9 2/2 (1.0) 0/2 (0.0) 25/25 (1.0) 0/25 (0.0) 
S3  2;9 1/1 (1.0) 0/1 (0.0) 50/50 (1.0) 0/50 (0.0) 
S4 2.10 33/33 (1.0) 0/33 (0.0) 48/48 (1.0) 0/48 (0.0) 
S5  2;10 11/11 (1.0) 0/11 (0.0) 23/23 (1.0) 0/23 (0.0) 
S6  2;11 10/10 (1.0) 0/10 (0.0) 47/47 (1.0) 0/47 (0.0) 
S7  3;4 3/3 (1.0) 0/3 (0.0) 28/28 (1.0) 0/28 (0.0) 
S8  3;4 60/60 (1.0)  0/60 (0.0)  67/67 (1.0) 0/67 (0.0) 
 166 
Child Age Modality Markers Negatives 
COR INCOR COR INCOR 
S1 2;4 0/30 (0.0) 30/30 (1.0) 0/2 (0.0) 2/2 (1.0) 
S2 2;9 26/29 (.90) 3/29 (.10) 6/7 (.86) 1/7 (.14) 
S3  2;9 52/52 (1.0) 0/52 (0.0) 4/4 (1.0) 0/4 (0.0) 
S4 2.10 1/49 (.02) 48/49 (.98) 0/4 (0.0) 4/4 (1.0) 
S5  2;10 20/21 (.95) 1/21 (.05) 11/12 (.92) 1/12 (.08) 
S6  2;11 1/60 (.02) 59/60 (.98) 0 0 
S7  3;4 23/25 (.92) 2/25 (.08) 5/5 (1.0) 0/5 (0.0) 
S8  3;4 120/125 (.96) 5/125 (.04) 18/18 (1.0) 0/18 (0.0) 
 
Table 6.7: Correct and incorrect clitic placement in proclisis contexts: negatives 
and clauses headed by modality markers (Spontaneous data). 
 
Child Age Wh–questions Other proclisis–triggers 
COR INCOR COR INCOR 
S1 2;4 0 0 0/2 (0.0) 2/2 (1.0) 
S2 2;9 5/5 (1.0) 0/5 (0.0) 1/1 (1.0) 0/1 (0.0) 
S3 2;9 0 0 2/3 (.67) 1/3 (.33) 
S4 2;10 0 0 0 0 
S5  2;10 1/1 (1.0) 0 (0.0) 0 0 
S6  2;11 0/3 (0.0) 3/3 (1.0) 0 0 
S7 3;4 1/1 (1.0) 0 (0.0) 1/1 (1.0) 0/1 (0.0) 
S8 3;4 2/2 (1.0) 0 (0.0) 6/7 (.86) 1/7 (.14) 
 
Table 6.8: Correct and incorrect clitic placement in proclisis contexts: wh–
questions and other proclisis–triggering elements3 (Spontaneous Data). 
 
Examples of children’s production of misplaced clitics are given in (19–
25). Negative clauses involving a dative and an accusative clitic are given in (19) 
and (20) respectively, while enna– and na–clauses involving a dative and an 
accusative clitic are given in (21) and (22) respectively. Clauses headed by the 
complementizer pu are given in (23–25).   
 
(19) E(n)   mila       mu.  
NEG  talk–3S  me–CL.DAT 
 “S/he doesn’t talk to me” (S4, 2;10) 
(20) En      eplinamen      to.   
                                                
3 The constructions subsumed under the category Other in the table above are headed by the 
following elements: for S1 [pu (2 correct)]; for S2, [focalised XP (correct)]; for S3, [ama (correct), 
pu (1 correct & 1 incorrect)]; for S7, [pu (correct)]; and for S8, [an (1), aman (4), embu (1) correct, 
pu (incorrect)]. 
 167 
NEG  washed–1PL  it–CL.ACC 
“We didn’t wash it” (S4, 2;10) 
(21) <Ena  !it"o        "u> [/] na !it"o "u. 
 M       show–1S  you–CL.DAT 
“I will show you” (S4, 2;10) 
(22) E(n)   mboro    na  (f)kalo         to.  
NEG  can–1S  M   take out–1S  it–CL.ACC 
 “I can’t take it out” (S4, 2;10) 
(23) Ke: [/] ke: e: [/] eklee                  kini             i      #ineka 
And                    was crying–3S  that–NOM  the  lady–NOM  
pu:        e: [/] e(s)tavrohja(n)  to(n)                 i     kaki 
 COMP           crucified–3PL  him–CL.ACC  the  wicked–NOM 
 to    hri(s)tuli                          k’     eva(l)a(n)  to(n). 
 the  Christ–ACC.diminutive  and  put–3PL    him–CL.ACC 
“And that woman was crying that the wicked crucified the Christ and put 
him” (S3, 2;9) 
(24) (N)a  !i           ta                       i     mami                  pu 
 M      see–3S  them–CL.ACC  the  mummy–NOM  COMP 
 feramen          ta 
 brought–1PL  them–CL.ACC 
 “(For) mummy to see them that we brought them” (S1, 2;4) 
(25) Kafki         tora  pu          ap"a   to. 
 Burns–3S  now  COMP  lit–1S  it–CL.ACC 
 “It burns now that I lit it” (S8, 3;4)  
 
The youngest participant of the spontaneous speech study [S1] was 
followed longitudinally for a period of 6 months from age 2;4 to 2;10. S1 had 
good clitic production from age 2;4 but exhibited ceiling proportions of misplaced 
clitic pronouns in negatives, clauses headed by modality markers and other 
proclisis triggers. The aim of the longitudinal study was to investigate the 
robustness of the phenomenon of clitic misplacement in terms of magnitude and 
duration in the speech of S1. This participant was recorded 4 times and his data 
 168 
were transcribed and analysed in the same way as the cross–linguistic data. Table 
6.9 reports clitic production as well as clitic/object omission in S1’s speech in 
contexts that require an overt clitic/object in adult CG. 
 
Table 6.9: Clitic production and clitic/object omission (Longitudinal data). 
 
Examples of S1’s utterances involving an omitted clitic/object are given in 
(26–31). In examples (26) and (27), the experimenter asks a question aiming to 
elicit a clitic pronoun while the child’s response involves a structure with an 
omitted clitic/object. In the second case, even though the correct form of the clitic 
pronoun is used in the experimenter’s question, the child does not use it. In 
examples (28) and (29) the child freely utters two structures exhibiting 
object/clitic drop. The child utterances in (30) and (31) involve the strong 
pronouns tuto (tutos, –i, –o) and ulo (ulos or olos, –i, –o) respectively as direct 
objects. In adult CG, tuto and ulo in contexts such as the ones in (30) and (31) are 
normally clitic doubled, unlike in the child’s productions. 
  
(26) *EXP: ti na kanume ta molivakja, dze to svistiri, dze ti ksistra, hm? 
  pe mu !inata dz’ en akusa. 
“What shall we do with the pencils, and the rubber, and the 
scraper, hm? Tell me louder, I couldn’t hear you” 
  *CHI: (n)a  valo       mesa. 
  M     put–1S  in (S1, 2;5) 
Target structure: (n)a  ta             valo mesa. 
   M     them–CL.ACC  put–1S  in 
    “I shall put them in”  
(27)  *EXP: to psali!aki pu e(n) spasmeno ti prepi na to kamume tora? 
   “The scissors that are broken, what shall we do with them?” 
  *CHI: fkjaksume. 
Age MLUw Clitic Production Clitic/Object Omission 
2;4 2.38 69/74 (.93) 5/74 (.07) [1/5] 
2;5 2.10 99/109 (.91) 10/109 (.09) [1/10] 
2;8 2.65 76/83 (.92) 7/83 (.08)  [0/7] 
2;10 2.57 99/107 (.92) 8/107 (.08) [0/8] 
 169 
  fix–1PL (S1, 2;10) 
Target structure: (n)a  to       fkjaksume. 
   M     it–CL.ACC  fix–1PL 
    “We should fix it”  
(28) *CHI: oi  (n)a  svisis. 
  no  M   erase–2S (S1, 2;10) 
Target structure: oi  (n)a  to                  svisis. 
   no  M    it–CL.ACC  erase–2S 
    “You should not erase it”  
(29) *CHI: e!o    pu         vazis. 
  here  COMP  put–2S (S1, 2;8) 
Target structure: e!o   pu          to         vazis. 
   here  COMP  it–CL.ACC  put–2S  
    “Here you should put it” 
(30) *CHI: efkjaksa   mama  tuto. 
  fixed–1S  mum   this–ACC (S1, 2;4) 
Target structure: efkjaksa   to   mama  tuto. 
   fixed–1S  it–CL.ACC  mum  this–ACC 
    “I fixed that, mum” 
(31) *CHI: ulo (.)             "elo         ulo. 
  all–ACC.MASC  want–1S  all–ACC. MASC (S1, 2;5) 
Target structure: ulo (.)             "elo         to                 ulo. 
   all–ACC.MASC  want–1S  it–CL.ACC  all  
    “All, I want it all” 
 
Table 6.10 shows his clitic production in enclisis and proclisis contexts. 
 
Table 6.10: Clitic production (Longitudinal data). 
Age MLUw Enclisis Contexts Proclisis Contexts Overall  
2;4 2.38 35 (.51) 34 (.49) 69 (1.0) 
2;5 2.10 64 (.65) 35 (.35) 99 (1.0) 
2;8 2.65 39 (.51) 37 (.49) 76 (1.0) 
2;10 2.57 40 (.40) 59 (.60) 99 (1.0) 
 170 
On the basis of the results reported, S1 has good clitic production in both types of 
contexts. His data resemble the cross–sectional data in exhibiting comparable 
percentages of enclisis and proclisis contexts. 
S1, as already shown, has ceiling percentages of misplaced clitics in all 
proclisis contexts from his first recording at age 2;4. As revealed by the analysis 
conducted on his data, this non–adult–like pattern is manifested in his clitic 
production up to the age of 2;10. Table 6.11 reports his clitic production in 
different syntactic contexts. The leftmost columns report the number of clitics 
produced in imperatives and root clauses (enclisis contexts) and the rightmost 
columns report the number of clitics produced in negatives, wh–questions, pu–
clauses and clauses headed by the modality markers na and enna (proclisis 
contexts). No figures for correct/incorrect clitic placement are given in the table, 
since all the clitics S1 produced were placed post–verbally in both enclisis and 
proclisis contexts, with only one exception discussed below. This applies to both 
single and double clitic constructions; the figures in parentheses show the absolute 
numbers of clitic clusters produced.  
 
Table 6.11: Clitic production in enclisis and proclisis contexts (Longitudinal data). 
 
The figures reported in table 6.11 reveal that S1 has good production of root 
clauses and clauses headed by na and enna, while his production of negative 
clauses, pu–clauses and wh–questions is relatively low. S1 produced a few wh–
questions4, all of which involved a misplaced clitic. Examples involving an 
                                                
4 See section 1.4 in chapter 1 for a thorough discussion on wh–formation in CG. 
Age MLUw Enclisis Contexts Proclisis Contexts 
IMP Root M NEG WH PU–clauses 
2;4 2.38 
 
15 (6) 20 (3) 30 (1) 2 (0) 0 2 (0) 
35 (9) 34 (1) 
2;5 2.10 
 
25 (1) 39 (5) 33 (2) 2 (0) 0  0 
64 (5) 35 (3) 
2;8 2.65 6 (0) 33 (2) 21 (0) 9 (0) 5 (1) 2 (0) 
39 (2) 37 (1) 
2;10 2.57 8 (4) 32 (1) 53 (1) 3 (0) 2 (0) 1 (0) 
40 (5) 59 (0) 
 171 
argumental wh–question and a question headed by the quasi–argument pu are 
given in (33) and (32) respectively. 
 
(32) *EXP: a: (.) kita (.) alon a!rio zoon . 
 “Oh, look, another wild animal.” 
*EXP: ti en tuto ? 
 “What is this?” 
*CHI: pu        vazun       to ? 
 where  put–3PL  it–CL.ACC 
*EXP: vazun to: +/. 
 (They) put it… 
*CHI: e"o ! 
 here! (S1, 2;8) 
(33) Pkjos            etarakse      mu                  to ?   
who–NOM  moved–3S  me–CL.DAT  it–CL.ACC 
“Who moved it for me?” (S1, 2;8) 
 
There is a single instance of correct clitic placement in a proclisis context, 
in S1’s third file (age 2;8). This is a clause headed by the negator men5 and it is 
cited in (34). However, this is an extract of a well–known folk song; the adult 
version (intended utterance) is cited below. Thus, it apparently does not constitute 
productive use of the clitic pronoun involved, but is part of a memorised unit.   
 
(34) Palakalo  sas   ta   kimata  me(n)  mu                   ksipnate .   
Ask         you  the  waves  NEG    me–CL.DAT  wake–up–2PL 
Parakalo sas kimata me mu tin (CL)–e–ksipnate. [Intended utterance] 
“Waves, I’m asking you not to wake her up” (S1, 2;8) 
 
The results obtained from the longitudinal study indicate that clitic 
placement in enclisis contexts is target–like, and, in a reversely analogous pattern, 
                                                
5 See section 1.2 in chapter 1 for a thorough discussion on the interaction of the negators dhen and 
men with the modal particles na, tha, enna and as. 
 172 
clitic placement in proclisis contexts is target–deviant throughout the period 
studied. S1’s consistency with regard to incorrect clitic placement for as long as 6 
months is a strong indication of the systematic nature of the pattern attested, its 
magnitude and its potential duration. In sum, this outcome mirrors the robustness 
of the phenomenon in child grammar. 
 
6.2.2. Clitic ordering within clitic clusters 
 
Turning to clitic ordering within clitic clusters, the vast majority of the clauses 
produced adhere to the adult placement. Recall that the ordering of clitic pronouns 
within clusters in CG is strictly CLDAT–CLACC (see section 2.3.3, chapter 2). No 
child produced a clitic cluster with the order CLACC–CLDAT except for S1. S1 
produced 27 double clitic constructions overall: 25 with the target–like order 
(CLDAT–CLACC) and 2 with a target–deviant order (CLACC–CLDAT). At age 2;4, he 
produced 10 constructions involving clitic clusters, 6 in imperatives, 3 in root 
clauses and 1 in a na–clause. Of these, all except one imperative clause, given in 
(35), adhere to the adult ordering. At age 2;6, he produced 8 clitic cluster 
constructions, 1 in an imperative clause, 5 in root clauses and 2 in na–clauses, all 
of which had correct ordering with the exception of one root clause repeated twice 
(36). At age 2;8, he produced 3 double clitic constructions, 1 in a wh–question 
and 2 in root clauses, and at age 2;10, 6 double clitic constructions, 1 in a root 
clause, 4 in imperatives and 1 in a na–clause, all with correct ordering. 
 
(35) Anikse      ta                       mu . 
Open–3S  them–CL.ACC  me–CL.DAT 
“Open them for me” (S1, 2;4) 
(36) Espasen     to                 mu . 
Broke–3S  it–CL.ACC  me–CL.DAT 
“S/he broke it for me” (S1, 2;6) 
 
Two points can be raised with respect to (35). First, the verb involved in 
the clause in (35) appears on another 4 occasions in the same file (S1, 2;4) in the 
 173 
same type of construction (the respective occurrences are cited below as (37–40)), 
in all of which the order is the correct one (CLDAT–CLACC). Second, an identical 
clitic construction in terms of feature specification is found 14 lines after (35) and 
on the second occasion the clitics in the clitic cluster appear in the correct order 
(39).  
 
(37) Anikse mu to. 
(38) Anikse mu ti. 
(39) Anikse mu ta. 
(40) Anikse mu to. 
 
These data show that the utterance in (35) does not constitute strong evidence that 
clitic ordering within child CG clusters is incorrect. It may be taken, together with 
(36), as an indication that clitic clusters in early CG may not always be target–
like. However, no robust conclusions can be drawn on the basis of these sparse 
instances of incorrect ordering, since the vast majority of the clitic clusters 
produced adhere to the adult pattern. 
 
6.2.3. Production of Clitic Doubling and Clitic Left Dislocation  
 
The purpose of the analysis in this section is to investigate whether the 
participants in the study are able to produce more complex construction types 
involving clitic pronouns. Following Marinis (2000:272, footnote 34), the 
calculations reported below included clauses in which the clitic was doubled by 
either a full DP (D+NP), as in (41), or a strong pronoun, as in (42). The absolute 
numbers of the relevant structures (single clitic (SC), CD and CLLD) produced 
are summarised in table 6.12. Table 6.13 reports the absolute numbers of full DPs 
and strong pronouns produced in CL and CLLD constructions, while tables 6.14 
and 6.15 present the different syntactic structures that appear in CD and CLLD 
respectively.  
 
 
 174 
(41) (N)a  to                  rotiso    ton  papa          mu              an  ehi        pilo. 
M      it–CL.ACC  ask–1S  the  dad–ACC  my–POSS  if   has–3S  concrete 
“I should ask my dad whether he has concrete” (S8, 3;4) 
(42) Ehu(m)e     to                  ke   mi(s)        (s)piti            ma(s)        (e)kino . 
Have–1PL  it–CL.ACC  and  us–NOM  home–ACC  us–POSS  that–ACC 
“We have it, us too, at home that one” (S3, 2;9)  
 
 
Table 6.12: Production of Single Clitic (SC), Clitic Doubling (CD) and Clitic Left 
Dislocation (CLLD) (Spontaneous data). 
 
 
Table 6.13: Production of strong pronouns and DPs in Clitic Doubling (CD) and 
Clitic Left Dislocation (CLLD) (Spontaneous data). 
 
Child Age Clitic Production 
SC CD CLLD Overall 
S1 2;4 57 (.83) 6 (.09) 6 (.09) 69 (1.0) 
S2 2;9 63 (.91) 3 (.04) 3 (.04) 69 (1.0) 
S3 2;9 84 (.76) 26 (.24) 0 110 (1.0) 
S4 2;10 123 (.92) 11 (.08) 0 134 (1.0) 
S5 2;10 62 (.91) 6 (.09) 0 68 (1.0) 
S6 2;11 89 (.74) 23 (.19) 8 (.07) 120 (1.0) 
S7 3;4 52 (.82) 8 (.13) 3 (.05) 63 (1.0) 
S8 3;4 250 (.90) 27 (.10) 2 (.01) 279 (1.0) 
Child Age Clitic Doubling CLLD 
Pronoun DP Pronoun DP 
S1 2;4 6/6 0/6 4/6 2/6 
S2 2;9 3/3 0/3 2/3 1/3 
S3 2;9 9/26 17/26 0 0 
S4 2;10 6/11 5/11 0 0 
S5 2;10 4/6 2/6 0 0 
S6 2;11 22/23 1/23 5/8 3/8 
S7 3;4 4/8 4/8 3/3 0/3 
S8 3;4 11/27 16/27 0/2 2/2 
Child Age Clitic Doubling 
IMP Root M Neg Other 
S1 2;4 2/6 1/6 2/6 1/6 0/6 
S2 2;9 1/3 0/3 1/3 1/3 0/3 
S3 2;9 0/26 11/26 13/26 1/26 1/26 (pu–clause) 
S4 2;10 1/11 5/11 5/11 0/11 0/11 
S5 2;10 1/6 4/6 0/6 1/6 0/6 
 175 
 
Table 6.14: Production of Clitic Doubling (CD) per syntactic context 
(Spontaneous data). 
 
 
Table 6.15: Production of Clitic Left Dislocation (CLLD) per syntactic context 
(Spontaneous data). 
 
The same calculations were performed on the longitudinal data (6.16–
6.18). The purpose of this analysis is twofold. Firstly, it aims to investigate 
whether more complex constructions, such as CD and CLLD clauses, emerge later 
than less complex ones. Secondly, it aims to offer some characterisation of the 
emergence of CD as compared to CLLD structures, taking into account the debate 
in the literature regarding their underlying nature (cf. the Uniformity Hypothesis 
of Sportiche (1992) and arguments against it in Anagnostopoulou (1994)).  
 
Table 6.16: Production of Single Clitic (SC), Clitic Doubling (CD) and Clitic Left 
Dislocation (CLLD) (Longitudinal data). 
 
 
S6 2;11 1/23 5/23 16/23 0/23 1/23 (wh–question) 
S7 3;4 1/8 2/8 5/8 0/8 0 
S8 3;4 7/27 4/27 10/27 5/27 1/27 (aman–clause) 
Child Age Clitic Left Dislocation 
IMP Root M Neg 
S1 2;4 1/6 4/6 1/6 0/6 
S2 2;9 0/3 1/3 0/3 2/3 
S6 2;11 0/8 6/8 2/8 0/8 
S7 3;4 0/3 3/3 0/3 0/3 
S8 3;4 0/3 2/3 0/3 0/3 
Age Clitic Production 
SC CD CLLD Overall 
2;4 57 (.83) 6 (.09) 6 (.09) 69 (1.0) 
2;5 94 (.95) 4 (.04) 1 (.01) 99 (1.0) 
2;8 70 (.92) 6 (.08) 0 76 (1.0) 
2;10 92 (.93) 7 (.07) 0 99 (1.0) 
Age Clitic Doubling CLLD 
Pronoun DP Pronoun DP 
2;4 6/6 0/6 4/6 2/6 
2;5 3/4 1/4 1/1 0/1 
2;8 4/6 2/6 0 0 
2;10 5/7 2/7 0 0 
 176 
Table 6.17: Production of strong pronouns and DPs in Clitic Doubling (CD) and 
Clitic Left Dislocation (CLLD) (Longitudinal data). 
 
 
Table 6.18: Production of Clitic Doubling (CD) per syntactic context 
(Longitudinal data). 
 
 
Table 6.19: Production of Clitic Left Dislocation (CLLD) per syntactic context 
(Longitudinal data). 
 
The results show that CG–speaking children have a good command of CD 
from as early as 2;4. Moreover, while SC and CD structures are available and 
productively used by all the participants of the study, CLLDs are not attested in 
all the corpora examined. 
 
6.2.4. Clitic misplacement and finiteness 
 
As mentioned in the previous chapter, what is crucial for our study is to 
investigate whether the (mis)placement of clitic pronouns in early CG correlates 
with finiteness. For the purposes of the current analysis the corpora of the 3 
children who exhibited clitic misplacement across the board (S1, S4 and S6; see 
table 6.5 and figure 6.1) were examined and all the constructions involving a 
misplaced clitic pronoun were identified. Following proposals in Varlokosta et al. 
(1998) for early SMG, I take the perfective verbal forms with the suffix –i to 
constitute early non–finite forms in early CG. Data analysis was performed as 
follows: all the verbal forms in –i with the perfective stem were classified as non–
finite forms and all the other verbal forms (including imperfective forms with the 
Age Clitic Doubling 
IMP Root M Neg Pu–clause 
2;4 2/6 1/6 2/6 1/6 0/6 
2;5 1/4 1/4 1/4 1/4 0/4 
2;8 0/6 1/6 2/6 0/6 3/6 
2;10 1/7 4/7 2/7 0/7 0/7 
Age Clitic Left Dislocation 
IMP Root M Neg 
2;4 1/6 4/6 1/6 0/6 
2;5 0/1 1/1 0/1 0/1 
 177 
suffix –i as well as verbal forms overtly marked for the [person] or the [number] 
feature) were classified as finite6. As for the verbs that lack a perfective stem and 
appeared in structures involving misplaced clitics, they were calculated but 
excluded from further analysis.  
Overall only 4 verbs lacking a perfective stem were identified in the 
corpora examined: fkali (“take out”), kam(n)i (“do”), vali (“put”), v!ali (“take 
out”), while plini (“wash”) was also disregarded from data analysis; these verbs 
appeared 22 times. The verb kam(n)i (“do”) is the dialectal equivalent of the SMG 
verb kani (“do”), which also lacks a perfective stem. The dialectal equivalents of 
the SMG verb v!azi (“take out”) are fkali and v!ali and the dialectal equivalent of 
the SMG verb vazi (“put”) is vali. For the SMG verb v!azi, v!al– is the perfective 
stem and v!az– the imperfective stem, while as regards its dialectal equivalents, 
v!ali and fkali, the only difference between their perfective and imperfective 
stems, v!al– and fkal– respectively, is the following: their perfective stems 
involve a lateral [l] while their imperfective stems involve a geminate lateral [l]. 
The same holds for the CG verb vali, whose imperfective stem involves a 
geminate lateral. It is not, however, easy to discriminate between the two types of 
lateral sounds in child language, hence, all these verbal forms were taken as 
lacking a perfective stem. As regards the form plini (“wash”), it was used as the 
dialectal equivalent of the SMG verb pleni in the corpora examined. However, in 
adult CG the verb pliniski is in use as the dialectal equivalent of pleni and plin– is 
its perfective stem. However, the verb pliniski did not occur in child data. It was, 
thus, unclear whether plini was used as a shortened form of pliniski or as a 
perfective form and was, therefore, disregarded from data analysis. 
As regards S1’s earliest two corpora, the one at age 2;4 and the second at 
age 2;5, only one form from each corpus, which only appeared once, was 
excluded for further analysis, namely plini (“wash”) which appeared in the first 
corpus and v!ali (“take out”) which appeared in the second. Thus, while S1 
produced 34 verbal structures with a misplaced clitic at age 2;4, and 35 at age 2;5, 
only 33 and 34 forms respectively were further analysed. As regards his latest two 
                                                
6 A single verbal form from S4’s corpus whose inflectional ending could not be identified was 
excluded from the analysis. 
 178 
corpora, the one at age 2;8 and the other at age 2;10, only two forms, each of 
which appeared once in each corpus, were excluded from further analysis, namely 
vali (“put”) and fkali (“take out”). Thus, while S1 produced 37 verbal structures 
with a misplaced clitic at age 2;8 and 59 at age 2;10, only 35 and 57 forms 
respectively were further analysed. As regards S4’s corpus, the following 2 forms 
that appeared 6 times were excluded from the analysis: vali (“put”) and plini 
(“wash”). Thus, only 46 verbal structures out of the 52 that involve a misplaced 
clitic were further analysed. As for S6’s corpus, kami (“do”), plini (“wash”) and 
vali (“put”), which appeared 11 times, were excluded from the analysis; thus, only 
51 structures out of the 62 structures with a misplaced clitic were analysed.  
Table 6.20 reports the absolute numbers of early non–finite as well as 
finite forms used in constructions in which the clitic is misplaced; the proportions 
of each type of form are given in parentheses. 
 
 
Table 6.20: Use of finite and early non–finite forms in constructions involving 
misplaced clitics. 
 
As shown in table 6.20, the overall percentage of structures involving a 
misplaced clitic and a finite verbal form (71%) outnumber the percentage of 
structures involving a misplaced clitic and an early non–finite form (29%). In fact, 
the use of inflectional morphemes in these structures is productive. Table 6.21 
reports the absolute numbers of finite verbal forms used in structures involving a 
misplaced clitic broken down by person and number. 
 
 
 
Child Age MLUw Early non–finite 
forms 
Finite forms Overall 
S1 2;4 2.38 9 (.27) 24 (.73) 33 (1.0) 
2;5 2.10 16 (.47) 18 (.53) 34 (1.0) 
2;8 2.65 11 (.31) 24 (.69) 35 (1.0) 
2;10 2.57 21 (.37) 36 (.63) 57 (1.0) 
S4 2;10 2.99 4 (.09) 42 (.91) 46 (1.0) 
S6 2;11 2.78 14 (.27) 37 (.73) 51 (1.0) 
Overall 75 (.29) 181 (.71) 256 (1.0) 
 179 
 
 
Table 6.21: Finite forms in constructions involving misplaced clitics. 
 
As shown in table 6.21, children mostly use the 1st person singular suffix, 
while the least used inflectional suffix is the 2nd person plural suffix. The 
distribution of these inflectional affixes in out study resembles the distribution of 
non–3SG verb forms in Varlokosta et al.’s (1998) study. For the sake of 
comparison, I cite their results on the distribution of the non–3SG verb forms in 
table 6.22. Note, however, that Varlokosta et al. exclude the copula, modals and 
imperatives from their analysis, while I report all proclisis contexts with a 
misplaced clitic (including modals). 
 
Child Age Singular Plural 
1st 2nd 1st 2nd 3rd 
Spi, St. I 1;9 4 2 5 0 4 
Jan, St. I 1;11 7 0 11 0 3 
Jan, St. II 2;5 58 20 19 2 6 
Mai, St. II 1;9 29 8 19 0 1 
 
Table 6.22: The distribution of the non–3SG verb forms (excluding the copula, 
modals and imperatives) (Table 7 in Varlokosta et al. 1998:198). 
 
The results summarised in table 6.20 show that children who misplace 
clitics use both finite and non–finite forms and, in fact, the percentage of 
structures involving a misplaced clitic and a finite verbal form outnumbers the 
percentage of structures involving a misplaced clitic and an early non–finite form. 
Hence, no correlation is evident between the use of early non–finite forms and 
target–deviant clitic placement. 
 
Child Age Singular Plural Overall 
1st 2nd 3rd 1st 2nd 3rd  
S1 2;4 15 0 0 9 0 0 24 
2;5 11 1 1 4 0 1 18 
2;8 6 1 8 2 1 6 24 
2;10 24 4 3 4 0 1 36 
S4 2;10 29 0 3 10 0 0 42 
S6 2;11 30 1 1 5 0 0  37 
 180 
 
6.3. Elicited production experiment 
 
This section presents the results from the experimental investigation, focusing on 
clitic production and clitic placement. Recall that the children who performed the 
experimental task are different from those who provided the spontaneous data and 
were divided into 3 age groups: A: 2;6–3;0, B: 3;0–3;6 and C: 3;6–4;07. Proclisis 
contexts appear to be problematic for a subset of children from age group A, who 
misplace clitic pronouns. Between–group comparisons will show that 
chronological age is an important factor for clitic placement in early CG. 
 
6.3.1. Clitic production 
 
The experiment aimed to elicit structures involving clitic pronouns in two 
syntactic contexts: root clauses and clauses headed by the modal particles na and 
enna. As mentioned in the previous chapter, only structures that were produced in 
the relevant contexts were categorised as target responses. To be precise, after an 
experimental question aiming to elicit a root clause, only a root clause was 
categorised as a target response. Likewise, only a clause headed by a modal 
particle was categorised as a target response after an experimental question 
aiming to elicit a na–/enna–clause. 
The participants would occasionally remain silent (even after the 
experimental question was repeated once). Absence of response was recorded as 
No Answer. The participants would sometimes produce a different type of 
structure than the required one. To exemplify, the experimental question in (43) 
aimed at eliciting a root clause. The child in (44) produced a negative utterance as 
a reply to the experimental question aiming to elicit a root clause and the child in 
(45) produced a non–clitic structure; responses of this type were coded as Other 
Answers. 
 
                                                
7 See table 5.3 in the previous chapter for full demographic information on the participants in the 
experiment. 
 181 
(43) *EXP: Ti ekamen i korua tin alin klatsa? 
 “What did the girl do with the other sock?” 
Target Structure: Evalen/Eforisen  tin. 
   put on/wore–3S   it–CL.ACC   
   “She put it on/wore it” 
(44) *CHI: en       tin          evalen     kala 
 NEG  her–CL.ACC  put on–3S  well (S28, 3;3) 
“She didn’t put it on well”  
(45)  *CHI: e(n)    foli   alin     klatsa  
 NEG  wear–3S  another  sock–ACC 
 “She doesn’t wear another sock” (S11, 2;11) 
 
Table 6.23 reports the raw numbers of no–answers, other answers and 
clauses of the required type produced per experimental condition, as well as the 
relevant percentages in parentheses. It should be noted that table 6.23 as well as 
table 6.24 report group results; the individual results are offered in Appendix II 
and Appendix III respectively. 
 
Table 6.23: Production of target responses, other answers and no answers per age 
group per experimental condition (Elicited production experiment). 
 
 The results summarised in table 6.23 show that overall 282 target 
responses were elicited: 151 root clauses and 131 na–/enna–clauses. There was a 
Age 
Group 
N Root Clauses 
Root Clauses No Answers Other Answers 
A 18 53/72 (.74) 6/72 (.08) 13/72 (.18) 
B 22 68/88 (.78) 3/88 (.03) 17/88 (.19) 
C 10 30/40 (.75) 5/40 (.12) 5/40 (.12) 
Overall 50 151/200 (.76) 14/200 (.07) 35/200 (.17) 
Age 
Group 
N Na– / Enna–clauses 
Na– / Enna–
clauses 
No Answers Other Answers 
A 18 41/72 (.57) 21/72 (.29) 10/72 (.14) 
B 22 61/88 (.69) 23/88 (.26) 4/88 (.05) 
C 10 29/40 (.73) 8/40 (.20) 3/40 (.07) 
Overall 50 131/200 (.66) 52/200 (.26) 17/200 (.08) 
 182 
relatively high proportion of No Answers and Other Answers on both 
experimental conditions. In particular, after a question aiming to elicit a root 
clause, the participants would provide no answer or another answer 24% of the 
times on average, while after a clause aiming to elicit a na–/enna–clause, the 
percentage of no answers/other answers is even higher, representing the 34% of 
all responses.  
At a second stage of data analysis, all the modal clauses produced after an 
experimental question aiming to elicit a modal clause, as well as all the root 
clauses produced after a corresponding experimental question were categorized as 
target and non–target structures. Only structures involving a clitic pronoun were 
categorised as target sructures. Non–target structures included errors of (clitic) 
omission (see examples 46–47) and the use of a full noun phrase (NP) instead of a 
clitic. 
 
(46) Experimental Question: What does mum do with the baby? 
*CHI: (!)kevasi  palami!i. 
 read–3S    fairy–tale.ACC (S12, 2;11) 
Target structure: !kevazi   tu             parami!i. 
   read–3S  it–CL.DAT  fairy–tale.ACC  
  “S/he reads him a fairy–tale” 
(47) Experimental Question: What does the baby want to do with the cow? 
*CHI: Na  vali       mesa. 
 M   put–3S  in  (S12, 2;11) 
Target structure: Na  ti                     vali       mesa. 
   M   her–CL.ACC  put–3S  in 
   “(She wants) to put her in” 
 
Table 6.24 reports the raw numbers of clitic production, clitic omission 
and the production of full noun phrases (NP) per experimental condition, as well 
as the relevant percentages in parenthesis. 
 
 183 
 
Table 6.24: Clitic production and clitic omission per age group per syntactic 
context (Elicited production experiment). 
 
The results summarized in table 6.24 show that overall 261 target 
structures were elicited: 138 root clauses and 123 na–/enna–clauses involving a 
clitic pronoun.  The rates of clitic omission were relatively low, ranging between 
5% and 7%. The use of a full noun phrase (NP) instead of a clitic occurred in 
merely two cases, one in age group A and another in age group B, as a response to 
an experimental question aiming to elicit a root clause.  
 
6.3.2. Clitic placement 
 
The second part of data analysis focused on clitic placement. All the elicited target 
structures were coded as correct and incorrect as follows. As regards the first 
experimental condition, only root clauses involving a post–verbal clitic were 
coded as correct, while root clauses involving a pre–verbal clitic were coded as 
incorrect. As regards the second experimental condition (the modality markers 
condition), only na–/enna–clauses involving a pre–verbal clitic were coded as 
correct while na–/enna–clauses involving a post–verbal clitic were coded as 
incorrect. Table 6.25 summarises the results of children’s clitic placement in the 
two experimental conditions (first column) per age group (second column). The 
third and fourth columns report the raw numbers of structures with correct and 
incorrect clitic placement respectively, while the figures in parentheses show the 
relevant proportions. 
Age 
Group 
N Root Clauses 
Clitic Production Clitic Omission NP 
A 18 48/53 (.91) 4/53 (.07) 1/53 (.02) 
B 22 62/68 (.92) 5/68 (.07) 1/68 (.01) 
C 10 28/30 (.93) 2/30 (.07) 0/30 (0.0) 
Overall 50 138/151 (.91) 11/151 (.07) 2/151 (.01) 
Age 
Group 
N Na– / Enna–clauses 
Clitic Production Clitic Omission NP 
A 18 38/41 (.93) 3/41 (.07) 0/41 (0.0) 
B 22 58/61 (.95) 3/61 (.05) 0/61 (0.0) 
C 10 27/29 (.93) 2/29 (.07) 0/29 (0.0) 
Overall 50 123/131 (.94) 8/131 (.06) 0/131 (0.0) 
 184 
CONTEXT 
 
AGE GROUP PLACEMENT 
 COR INCOR 
Root Clauses A 48/48 (1.0) 0/48 (0.0) 
B 61/62 (.98) 1/62 (.02) 
C 28/28 (1.0) 0/28 (0.0) 
Overall 137/138 (.99) 1/138 (.01) 
Modality 
Markers 
A 25/38 (.66) 13/38 (.34) 
B 55/58 (.95) 3/58 (.05) 
C 27/27 (1.0) 0/27 (0.0) 
Overall 107/123 (.87) 16/123 (.13) 
 
Table 6.25: Clitic placement per experimental condition (Elicited production 
experiment). 
 
The children performed in an adult–like way with respect to their clitic 
placement in root clauses. To be precise, the overall number of clitics produced in 
root clauses was 138, of which 137 were placed post–verbally, as in the adult 
language, with only a single use of a pre–verbal clitic. This outcome reveals that 
children’s performance in enclisis contexts is target–like from as early as age 2;6. 
Turning now to na– and enna–clauses, a different picture emerges. While children 
from age groups B and C produce pre–verbal clitics, as required in these contexts, 
at ceiling percentages (95% for age group B and 100% for age group C), the rate 
of adult–like performance in age group A reaches only 66%, with 34% of clitics 
misplaced. Figure 2 plots the proportion of children who manifest correct (grey) 
and incorrect (black) clitic placement in na– and enna–clauses for each age group. 
This figure shows that the number of children who consistently produce 
misplaced clitics in proclisis contexts decreases with age, with no child older than 
3;6 producing any misplaced clitics. 
 
 185 
 
Figure 6.2: Proportion of children exhibiting adult–like and non–adult–like clitic 
placement in clauses headed by modality markers per age group (Elicited 
production experiment). 
 
Further analysis was performed on these results in order to validate them 
statistically, as described in the following subsection. 
 
6.3.3. Between–group analysis for clitic placement 
 
A factorial ANOVA was used for a between–group analysis of children’s 
performance in clauses headed by the modal particles na and enna. It was shown 
that children’s incorrect clitic placement differs significantly among age groups, 
F(2)=3.64, p=.034. Scheffé post–hoc comparisons of the three age groups 
indicated that the performance of age group A (M=.72) differs significantly from 
that of age group B (M =.14) as well as that of age group C (M=.00). With regard 
to the production of misplaced clitics, comparisons between age groups A and B 
(MD =.59, 95% CI [–.06, 1.23]), and between age groups A and C (MD =.72, 95% 
CI [–.08, 1.52]) show a difference approaching significance, with p=.08 in each 
case. However, the difference between age groups B and C (MD = 0.14, 95% CI 
[–.64, .91]) is not significant (p= .91). 
 
6.4. Double realisation of the clitic pronoun 
 
An unexpected pattern attested in the spontaneous data of some of the children, 
namely the double realisation of the clitic pronoun in two positions: immediately 
 186 
preceding and immediately following the verbal host, has crucial relevance to the 
interpretation of the results obtained. Thus, further scrutiny is required. As 
described in the previous chapter, the transcripts of another 11 children, a subset 
of the children that participated in the experimental investigation, were examined. 
An exhaustive list of these occurrences is cited below (48–53) and the 
phenomenon is extensively discussed in the next chapter. 
 
(48) *EXP: ti ena kami o (p)apas ton kafen tu ? 
 “What is dad going to do with his coffee?” 
*CHI: (n)a  to                 (p)ki         to    <ston ka> [/] ston      kanape . 
 M     it–CL.ACC  drink–3S  it–CL.ACC          on–the  sofa–ACC 
“He will drink it (lying) on the sofa” (S1, 2;8) 
(49) *CHI: <ti                  po          su                        ti    nirika> [//] ti po su ? 
   it–CL.ACC  say–1S  you–ACC.DAT  the  nirika–ACC     
  Na su tin po ti ririka [//] na su tin po?  [Intended Utterance] 
  “Shall I sing Ririka for you?” 
 *CHI: ka!ete i rika (s)ti veranta t' ehi te to "ataki ti(s) konta . 
 “Ririka is sitting on the balcony and has her cat nearby”    (S2, 2;9)  
(50) *EXP: ne (.) ti prepi na kami ta paputsha tu ? 
“Yes. What shall he do with his shoes?” [Uttered while pointing at 
a picture with a boy holding his shoes and a dog nearby] 
*CHI: (en)a:  ta                        fai        ta                        o    (s)kilos . 
 M        them–CL.ACC  eat–3S  them–CL.ACC  the  dog–NOM 
 “The dog will eat them” (P6, 2;8) 
(51) *EXP: ama su #oko to maherin . 
*EXP: ti ena kamis to milo ? 
“If I give you the knife, what will you do with the apple?” 
*CHI: (en)a  to                  kopso   to . 
 M       it–CL.ACC  cut–1S  it–CL.ACC  
“I will cut it” (P6, 2;8) 
(52) *EXP: esi trois tiri ? 
 “Do you eat cheese?” 
 187 
*CHI: o:i (.)  e(n)    m'                  are(s)ki   mu. 
 no      NEG  me–CL.DAT  like–3S  me–CL.DAT   
 “No, I don’t like it” (P22, 3;2) 
(53) *EXP: ta paputsha tu, <hm:> [>] ? 
  “His shoes? Hm?” 
*CHI: <ne> [<] (.) ki     ena  ta                       vali       ta     
  yes              and  M    them–CL.ACC  put–3S  them–CL.ACC   
tuta             ta . 
them–ACC  the  
  “Yes. And s/he will put them these the” (P48, 3;9) 
 
6.5. Summary 
 
Based on the results obtained from the analysis of naturalistic data, both cross–
sectional (table 6.5), and longitudinal (table 6.10), as well as experimental data 
(table 6.25), I report ceiling percentages of target–like clitic placement in root 
clauses, that constitute an enclisis context, from as early as age 2;6. However, this 
does not apply to proclisis contexts. Children’s clitic placement in these contexts 
shows a bimodal distribution (displayed in figure 6.1): most children perform in 
an adult–like way, whereas a subset of children younger than 3 years 
(approximately one third of the children aged 2;6–3;0) use exclusively post–
verbal clitics in both enclisis and proclisis contexts. The longitudinal study shows 
that these children may consistently misplace clitics for a prolonged period. The 
experimental investigation shows that this phenomenon diminishes with age (see 
figure 6.2) with no child older than 3;6 producing misplaced clitics. An interesting 
phenomenon sparsely attested in the data examined, namely the double realisation 
of the clitic pronoun, offers useful indications for the interpretation of the results 
obtained. Building on ideas in Franks (1998), Bo!kovi" (2000, 2001) and Franks 
and Bo!kovi" (2001) I account for the developmental patterns attested by appeal 
to the PF–controlled spell–out of copies (Revithiadou 2006), which is discussed in 
the next chapter. 
 
 
 
 
188 
CHAPTER 7: THE L1 ACQUISITION OF CLITIC 
PLACEMENT IN CYPRIOT GREEK  
 
7.1. Introduction 
 
This chapter discusses the theoretical implications of the results obtained for the 
developmental stages of clitic L1 acquisition, as well as for formal approaches for 
cliticisation in CG. The first section recapitulates the results of the two studies 
conducted. The second, third and fourth sections discuss the manifestation of 
clitic misplacement and the double realisation of the clitic pronoun in early 
grammars. The fifth section shows whether and to what extent formal accounts on 
clitic placement in CG can accommodate the developmental patterns attested, 
while the last section offers an account within the hypothesis of PF!controlled 
spell!out of copies. 
 
7.2. Recapitulation: the L1 acquisition of clitic placement in CG 
 
The main findings of the two corpus studies (cross–sectional and longitudinal) 
and the experimental investigation are summarised below: 
1. Most CG–speaking children have good clitic production from as 
early as age 2;4. This is in line with the findings of Grohmann et al. (2012), who 
report adult–like clitic production in children acquiring CG from age 2. I take the 
emergence of clitics in child speech to reflect the representation of the relevant 
lexical items in the child lexicon. 
2. The corpora examined show that CG–speaking children produce 
Single Clitics (SC) and clitic clusters from as early as age 2;4, while Clitic 
Doubling (CD) in which the clitic is doubled by either a full DP or a strong 
pronoun was also attested in children as young as 2;4. Notably, the examination of 
the corpora of the 8 children showed that, while SC and CD constructions are 
available and productively used by all the participants of the study, CLLD 
constructions were not attested in all the corpora examined. This observation may 
 
 
 
 
189 
have implications for the debate between approaches that take SC, CD and CLLD 
to involve the same underlying structure (see Sportiche’s (1996) Uniformity 
Hypothesis) and approaches that assume different structure for CD and CLLD 
(Anagnostopoulou 19941). The acquisition data in the current study seem to pose 
a problem for the Uniformity Hypothesis (cf. Marinis 2000). However, this issue 
requires further scrutiny, and is left for further research. What is relevant for the 
current thesis is that, irrespective of the complexity of the clitic construction 
involved, clitic placement remains intact. In particular, children with target–like 
clitic placement exhibit this in SC, CD and CLLD, while children with target–
deviant clitic placement misplace clitics in all the different types of structures. 
3. Both finite and non–finite enclisis in CG is acquired from the onset 
of L1A: CG–speaking children exhibit target–like clitic placement in root clauses 
and imperatives from as early as age 2;4. CG enclitics, both in finite and non–
finite contexts, emerge earlier than proclitics.  
4. The acquisition of proclisis is delayed in approximately 30% of 
CG–speaking children aged 2;6 to 3;0, while the remaining 70% of children place 
clitics pre–verbally in the relevant contexts by age 2;6. Children who have not 
acquired proclisis overgeneralise post–verbal clitic placement, i.e. enclisis, to 
proclisis contexts as well. The delayed acquisition of proclisis reveals the salience 
of the enclitic order, which appears to be the default pattern of clitic placement in 
CG. Recall that the “salience” of the enclisis pattern over the proclisis pattern is 
understood as the earliest emergence and the higher degree of accessibility of the 
former as manifested in TD and language–impaired populations (see Avram & 
Coene 2007 for Romanian, Mastropavlou et al. 2014, Stephany 1997, Tzakosta 
2003, 2004a, 2000b for SMG) with no further theoretical implications at the 
discourse level. 
5. The large number of children exhibiting this target–deviant pattern 
reveals that this phenomenon marks a distinct developmental stage in the course 
of L1 acquisition of CG by typically developing (TD) children (Hypothesis 1). 
                                                
1 Anagnostopoulou (1994) assumes that clitics in CD are nominal agreement morphemes and the 
clitic–doubled object DPs appear within VP, while in CLLD clitics are topic markers and the 
clitic–doubled object DPs are base–generated IP–adjuncts.  
 
 
 
 
190 
6. The second developmental stage attested in early CG is 
characterised by the manifestation of both pre– and post–verbal clitic placement 
in proclisis contexts, while sparse occurences of double realisation of the clitic 
pronoun in both pre– and post–verbal position are also attested.  
7. The third developmental stage involves target–like clitic placement 
in all syntactic contexts. TD CG–speaking children reach this stage around age 3, 
while no child in the corpora examined exhibited target–deviant clitic placement 
after the age of 3;6. 
8.  The robustness of the phenomenon of clitic misplacement, as 
documented by the ceiling proportions of misplaced clitics in the speech of some 
TD children, challenges Petinou and Terzi’s (2002) claim that across–the–board 
clitic misplacement is attested in SLI populations alone.  
9. Children may consistently exhibit non–adult–like clitic placement 
in all the proclisis–triggering syntactic contexts for a prolonged period. 
Longitudinal data showed that this may last for as long as six months, and does 
not preclude the possibility that it may be longer. This reveals the robustness of 
the phenomenon both in its consistency in the direction of the target–deviant 
placement (only enclisis–pro–proclisis was attested but not proclisis–pro–enclisis) 
and its duration. I take this finding to reflect the systematic nature of the 
phenomenon attested and the absence of real optionality from child grammar  
(Hypothesis 2). 
10. Clitic misplacement does not correlate with the use of early–non–
finite forms. The mean proportion of co–occurrences of misplaced clitics with a 
perfective verbal form ending in –i in the corpora examined amounts to 29%, with 
the proportion ranging between 9% and 47%, while the respective mean 
proportion for finite forms co–occurring with misplaced clitics reaches 71%, with 
the proportions ranging between 53% and 91%. If target–deviant clitic placement 
were an epiphenomenon of a non–fully–fledged or non–fully–specified I2, the 
                                                
2 In fact, children with ceiling percentages of misplaced clitics, like S1, seem to have a fully–
fledged clause structure. A representative example in support of this claim is cited below: an 
(arguably) bi–clausal cleft clause (cf. chapter 1 for the syntax of it–clefts in CG) is used by S1 at 
age 2;10, while his clitic placement is target–deviant.  
(1) *CHI: e(n)  dhame  pu         (en)a  psa(r)epsi . 
 is      here     COMP  M       fish–3S   
 
 
 
 
191 
proportion of early–non–finite verbal forms co–occurring with misplaced clitics 
would have been higher than the proportion of finite verbs (Hypothesis 3). 
 
7.3. Clitic misplacement in CG 
 
One of the main findings of the current study is that the acquisition of proclisis is 
delayed in approximately 30% of CG–speaking children aged 2;6 to 3;0. As 
regards the remaining 70% of the population in discussion, they use proclisis and 
enclisis in an adult–like way in the relevant contexts. The individual variability 
attested in CG–speaking children aged 2;6 to 3;0 as regards clitic placement seeks 
for an explanation. 
 One of the hypotheses this thesis has explored was the possibility that 
clitic misplacement in CG is an epiphenomenon of a defective inflectional domain 
(Hypothesis 3). As outlined in chapter 5, if this line of argumentation is correct, 
the proportion of early non–finite verbal forms used in constructions involving 
misplaced clitics should be significantly higher than that of finite verbs. The 
analysis performed on the corpora of all the children who consistently misplaced 
clitic pronouns revealed that children who misplace clitics use both finite and 
non–finite forms and, in fact, the percentage of structures involving a misplaced 
clitic and a finite verbal form (71%) outnumbers the percentage of structures 
involving a misplaced clitic and an early non–finite form (29%). This piece of 
evidence shows that no correlation seems to hold between a defective Inflection 
and target–deviant clitic placement. 
Another hypothesis is that non–adult–like clitic misplacement co–occurs 
with non–adult–like clitic production. Recall that the following 3 participants: S1, 
S4 and S6 had ceiling percentages of incorrect clitic placement in proclisis 
contexts (see table 6.5). Let us consider the results reported in table 6.1 (repeated 
below for ease of reference) for clitic production and clitic/object omission in the 
spontaneous data. S1 and S6 produced omitted clitics/objects 7% to 8% of the 
time, a rate of omission that was at comparable levels with the rest of the 
participants, with the exception of S4 who had the second highest rate of 
                                                                                                                                 
  “It is here that s/he will fish” (S1, 2;10) 
 
 
 
 
192 
clitic/object omission (14%). One may argue that S4’s relatively high rate of 
clitic/object omission relates to his ceiling proportions of clitic misplacement. 
This, however, does not hold for S1 and S6, who consistently misplaced clitics as 
well, while the participant who exhibited the highest rate of clitic/object omission 
(16%), S2, misplaced clitics only 10% of the time. Thus, clitic misplacement does 
not seem to be contingent on clitic/object omission either. 
 
Table 6.1: Clitic production and clitic/object omission (Spontaneous data). 
 
 The discussion so far has revealed that clitic misplacement in early CG is 
neither related to non–adult–like clitic production nor to a defective inflectional 
domain. Does clitic misplacement correlate with another target–deviant 
phenomenon in early CG? The examination of the corpora of spontaneous 
production has offered no indications. However, based on parental observations as 
regards earlier stages in their children’s development we speculate that the 70% of 
the population in discussion that did not manifest this target–deviant pattern of 
clitic placement between age 2;6 to 3;0 must have passed through this stage at an 
earlier time in development.  This remains to be confirmed on the basis of more 
data and remains an open question for future research. 
 An important conclusion we can, yet, draw on the basis of the results 
obtained as regards the individual variability attested in CG–speaking population 
aged 2 to 4 is that the transition from the target–deviant to the target–like clitic 
Child Age MLUw Clitic Production Clitic/Object 
Omission 
S1 2;4 2.38 69/74 (.93) 5/74 (.07) [1/5] 
S2 2;9 2.78 69/82 (.84) 13/82 (.16) [6/13] 
S3 2;9 3.87 110/116 (.95) 6/116 (.05) [1/6] 
S4 2;10 2.99 134/156 (.86) 22/156 (.14) [2/22] 
S5 2;10 2.19 68/75  (.91) 7/75 (.09) [1/7] 
S6 2;11 2.78 120/131 (.92) 11/131 (.08) [2/11] 
S7 3;4 2.96 63/68 (.93) 5/68 (.07) [3/5] 
S8 3;4 3.93 279/296  (.94) 17/296 (.06) [6/17] 
 
 
 
 
193 
placement in the course of L1A in CG seems to be a relatively rapid process. This 
is evident from the results obtained from both the spontaneous and the 
experimental study. The participants of the spontaneous study had either ceiling 
percentages of adult–like clitic placement or non–adult–like clitic placement in 
proclisis contexts as illustrated in table 6.5 (repeated below for ease of reference). 
In addition, all the 50 children who performed the experimental task were 
consistent as regards their clitic placement in clauses headed by na/enna (proclisis 
contexts): they would place the clitic either pre–verbally or post–verbally in all 
the modal clauses they produced. If our study had revealed bimodal distribution as 
regards clitic placement in proclisis contexts, this could have been taken as 
evidence of a rather prolonged transitional stage from the target–deviant to the 
target–like pattern. Most importantly, no participant of either the spontaneous or 
the experimental study has manifested a bimodal distribution as regards his clitic 
placement. On the basis of this outcome, we can safely conclude that this 
transitional stage is a relatively rapid process in the acquisition of CG.  
 
Child Age Enclisis Context Proclisis Context 
COR INCOR COR INCOR 
S1  2;4 35/35 (1.0) 0/35 (0.0) 0/34 (0.0) 34/34 (1.0) 
S2 2;9 27/27 (1.0) 0/27 (0.0) 38/42 (.90) 4/42 (.10) 
S3  2;9 51/51 (1.0) 0/51 (0.0) 58/59 (.98) 1/59 (.02) 
S4  2;10 81/81 (1.0) 0/81 (0.0) 1/53 (.02) 52/53 (.98) 
S5  2;10 34/34 (1.0) 0/34 (0.0) 32/34 (.94) 2/34 (.06) 
S6  2;11 57/57 (1.0) 0/57 (0.0) 1/63 (.02) 62/63 (.98) 
S7  3;4 31/31 (1.0) 0/31 (0.0) 30/32 (.94) 2/32 (.06) 
S8  3;4 127/127 (1.0) 0/127 (0.0) 146/152 (.96) 6/152 (.04) 
 
Table 6.5: Correct and incorrect clitic placement in enclisis and proclisis contexts 
(Spontaneous data). 
 
7.4. Clitic misplacement in early grammars 
 
CG–speaking children at the first developmental stage in the course of clitic L1 
acquisition overgeneralise the enclisis pattern to all syntactic contexts, including 
proclisis contexts as well. This target–deviant pattern is not exclusively attested in 
early CG. Lobo and Costa (2012) report clitic misplacement in proclisis contexts 
 
 
 
 
194 
in early European Portuguese (EP), while Stephany (1997) and Tzakosta (2003, 
2004a, 2004b) report sparse instances of clitic misplacement at the onset of the L1 
acquisition of SMG. 
Lobo and Costa (2012) offer ample evidence that EP–speaking children 
pattern like CG–speaking children and misplace clitic pronouns in proclisis 
contexts, while some instances of misplacement were attested in enclisis contexts 
as well. The examples cited below from early EP (taken from Lobo & Costa 
2012) illustrate the target–deviant productions of young children in negative 
clauses (1–2), wh–questions (3–4) and subordinate clauses (5–6). Similar 
productions by the CG–speaking participants of the current study in negative 
clauses, wh–questions and subordinate clauses appear in chapter 6 as examples 
(19–20), (32–33) and (23–25) respectively.  
 
Negative Clauses 
(1) O     mano     não    deixa–me   dormir.   
 The  brother  NEG  let–CL.1S  sleep 
 “My brother does not let me sleep” (J., 3;8) 
(2) Não  chama–se         nada.   
 not    call–CL.3Refl  nothing 
 “It isn’t called anything” (M., 20 months) (Duarte et al. 1995) 
Wh–questions 
(3) Porque  partiu–se,             mãe?  
 Why      broke–CL.3Refl  mum 
 “Why did it break, mum?” (J., 3; 4)  
(4) Porque  é   que   foste–me       interromper?   
 Why      is  that  went–CL.1S  interrupt 
 “Why did you interrupt me?” (R., 2;5) (Duarte et al. 1995) 
Subordinate Clauses 
(5) Foste  tu     que  daste–me.  
 Were  you  that  gave–CL.1S 
 “It was you that gave it to me?” (J., 4;8) 
(6) Foi    a     Mariana  que  deu–me         este?   
 
 
 
 
195 
Was  the  Mariana  that  gave–CL.1S  this  
 “It was Mariana that gave me this one”  (Sandra, 3;1) (Soares 2006:375) 
(7) Foi    alguém    que  meteu–me   nesta   fotografia?   
 Was  someone  that  put–CL.1S  in this  picture 
 “It was someone that took this picture for me”  (J.G, 3;3) (Duarte et al. 
1995) 
 
However, Lobo and Costa (2012) mention that proclisis is also attested in 
enclitic contexts in early EP. They offer three examples of such clauses, which I 
cite in (8–10). Notably, all these examples involve a pre–verbal constituent, 
specifically either a subject (9–10) or an object (8). Recall that European 
Portuguese requires a pre–verbal clitic with focused initial constituents (cf. 
section 2.3.5 in chapter 2). Moreover, EP is a pro–drop language. Overt subjects 
in pro–drop languages are either topicalised or focused constituents. Thus, it is not 
clear whether the clause initial constituent in (9–10) constitutes a topicalised or a 
focused DP in subject position; the former would require a post–verbal clitic, 
while the latter a pre–verbal clitic, as in (9–10). Nevertheless, example (8) is an 
uncontroversial case in which proclisis is manifested instead of enclisis.  
 
(8) Uma  carta  me        caiu,  do            pokémon.   
 A       letter  CL.1S  fell    from_the  pokemon 
 “A letter fell from my pokemon” (J., 4;8) 
(9) Se  queres      levar      isto,  eu  te        empresto.  
 If    want–2S  to_take  this,  I    CL.2S  lend 
 “If you want to take this, I will lend it to you” (J., 3;6) 
(10) Eu  te         empresto  um,  pai. 
 I     CL.2S  lend         one   daddy 
 “I will lend you one, daddy” (J. 3;7) 
 
 In CG, on the other hand, only 4 cases of proclisis in root clauses, which 
constitute enclisis contexts in CG, were attested. The exhaustive list of these 
occurrences is cited in (11–14); the experimenter’s utterance is included as well so 
 
 
 
 
196 
that the linguistic context is accurately represented. However, only two of these 
occurrences can be unequivocally claimed to be true instances of proclisis–pro–
enclisis: namely (11) and (12), both of which were produced by P44 at age 3;7.  
 
(11) *EXP: ne:, <ti ekaman tu:> [/] ti ekaman tu: [/] tu simba ? 
  “Yes, what have they done, what have they done to Simba?” 
 *CHI: en iks(ero) +//. 
  “I don’t know” 
 *CHI: +, ekaman tu etsi . 
  “They have done this to him” 
 *CHI: tu                     evalan      etsi         to   fteron              tu. 
  him–CL.DAT  put–3PL  like this  the  feather–ACC  his–POSS 
  “They put him his feather like this” 
 *EXP: ti:? 
  what? 
 *CHI: ekaman    tu                     etsi         to    ftero . 
  did–3PL  him–CL.DAT  like this  the  feather–ACC (S44, 3;7) 
(12) *EXP: pkjes en tutes? 
  “Who are they?” 
 *CHI: i       files                tis               (.)  kapu            tes  
  the  friends–NOM  her–POSS       somewhere  them–CL.ACC  
  idhame     dze  tes                      iksera  . 
  saw–3PL  and  them–CL.ACC  knew–1S 
          “Her friends. We have seen them somewhere and I knew them” 
(S44, 3;7) 
 
 The utterance in (11) involves a clitic that occupies the first position in the 
clause. This pattern violates the basic tenet of Tobler–Mussafia clitic languages 
according to which clitics are banned from clause initial position. However, two 
lines below, the child repeats the utterance in a slightly modified form but, 
interestingly, with post–verbal clitic placement. The child’s utterance in (12) 
consists of two instances of a clitic and a finite verb. The first one follows a 
 
 
 
 
197 
locative adverb, a proclisis context, and the second appears immediately after a 
coordination conjunction. Even though the latter is not a proclisis trigger, proclisis 
occurs. 
The two other child utterances involving a pre–verbal clitic in an enclisis 
context that occurred in the corpora examined are cited in (13) and (14). In (13) 
the clitic pronoun immediately follows an overt subject; hence, the same 
observations apply as for the EP examples in (9–10). In (14) the child uses a 
single clitic followed by a target–deviant verb form, realised as plini. It is not 
clear whether this constitutes a truncated form of the CG verb pliniski (“washes”) 
or a variant of the SMG equivalent pleni (“washes”). 
  
(13) *EXP: pe mu aghapi mu . 
  “Tell me, my love” 
 *CHI: <o mako xxx mba(njo)> [?] . 
  “Mako [unintelligible] ba(th)” 
 *EXP: o ma ? 
  “Ma(ko)?” 
 *CHI: o     mako             ton                    kamni mbanjo . 
  the  mako–NOM  him–CL.ACC  bath–3S  
  “Mako baths him” (S6, 2;8) 
(14) *EXP: ti kamni ta pkjata ? 
  “What does s/he do with the dishes?” 
 *CHI: ta                        plini .  
  them–CL.ACC  wash–3S 
  “S/he washes them” (S30, 3;3) 
 
The fact that the child in (14) opts for a verb form like plini that resembles 
the SMG equivalent pleni is not without explanation. This is related to the 
linguistic situation in Cyprus that has been identified as a state of diglossia 
(Newton 1972), in the sense of Ferguson (1959). SMG is the official language of 
the island (together with Turkish and English), which is used in education, in the 
media and administration, and sociolinguistically has the status of the high 
 
 
 
 
198 
variety. CG, on the other hand, is used in informal conversations and everyday 
communication and has the status of the low variety. Moreover, the variety used 
for written production in education, in the press and administration is SMG alone. 
CG–speaking children are exposed to SMG (or to the Cypriot Standard Greek in 
Arvaniti’s (2010) terminology3) especially through schooling (see Ioannidou 
2009, Ioannidou & Sophocleous 2010, Pavlou & Papapavlou 2004, Yiakoumetti 
et al. 2005 for teachers’ and students’ language use in primary and secondary state 
education in Cyprus). As a result, Greek–Cypriot children may occasionally 
switch to SMG. 
Let us, now, return to the 4 cases of pre–verbal clitic placement in root 
clauses attested in our database and presented in examples (11–14): in the light of 
the current linguistic situation in Cyprus, examples (11), (12) and (14) can be 
easily accounted for as instances where the SMG pattern of clitic placement is 
employed by the children. In (13) the presence of an overt subject makes the pre–
verbal placement of the clitic pronoun a legitimate option within the CG variety as 
well. All in all, the very few instances of pre–verbal clitic placement in enclisis 
contexts as well as the fact that all the children from both the spontaneous study 
and the experimental study had ceiling percentages of post–verbal clitic placement 
in enclisis contexts (see tables 6.5 and 6.25) invites us to conclude that young 
CG–speaking children manifest post–verbal placement in enclisis contexts. In any 
case, even if larger numbers of pre–verbal clitic placement were attested in 
enclisis contexts, this would have been explained as instances of code–switching 
between the high (SMG) and the low (CG) variety. 
In contrast, the large number of cases with post–verbal clitic placement in 
proclisis contexts offers ample evidence for the claim that (one third of) young 
CG–speaking children (aged 2;6–3;0) overgeneralise the enclisis pattern to 
proclisis–triggering contexts as well. Most importantly, these cases are of 
different nature: post–verbal clitic placement in negatives, in clauses headed by 
modal particles, wh–elements and other proclisis triggers is illicit in any variety of 
                                                
3 Arvaniti (2010) argues that SMG as used in Cyprus has been increasingly diverging from the 
Standard variety as spoken in Greece up to a level where the two are recognizably different, and 
uses the term Cypriot Standard Greek to describe the former. 
 
 
 
 
199 
Greek I am aware of. Thus, these instances clearly constitute cases of a target–
deviant grammar.  
Stephany (1997) and Tzakosta (2003, 2004a, 2004b) report misplacement 
errors in early SMG as well, and in accordance with the data from early CG, these 
are instances of the manifestation of enclisis–pro–proclisis alone, while proclisis–
pro–enclisis was not attested. In particular, 3 tokens were attested in the speech of 
2 out of 5 children whose corpora were examined by Stephany. In these tokens, an 
enclitic was used with a non–imperative verb (see example (62b) in Stephany 
1997:272, repeated as (10) in chapter 4), while Tzakosta (2003, 2004a) reports 
instances in which an enclitic is used instead of a proclitic (but never vice versa), 
and provides the examples in (15–17) in support of her claim. 
 
(15) /!a.to.'v"alo/ # ['"a.lo.to] “I will take it (CL) out”  
(16) /!a.to.'valo/ # ['va.lo.to] “I will put it (CL)” 
(17) /!a.to.'paro/ # ['pa.lo.to] “I will take it (CL)” 
 
7.5. Double realisation of the clitic pronoun in early grammars 
 
Additional to the cases of clitic misplacement, it is worth devoting attention to the 
phenomenon of double realisation of the clitic pronoun, as it will also help shed 
some light on the former phenomenon. As mentioned in the previous chapter, 
instances of double realisation of the clitic pronoun are attested in the corpora 
examined. The exhaustive list is offered in section 6.4. However, this target–
deviant pattern has not only been attested in early CG. Lobo and Costa (2012) 
report a few instances of double realisation of the clitic pronoun in both pre– and 
post–verbal position in early EP. Examples (18–19) are taken from Lobo and 
Costa (2012). Notably, these occurrences were attested in proclisis contexts alone, 
showing that these contexts are problematic not only for CG–speaking children 
but also for EP–speaking children. 
 
(18) Eu  disse  que  não     se                 põe–se                em  pé.   
 I     said   that   NEG  CL.3S.Refl  put–CL.3S.Refl  in   foot 
 
 
 
 
200 
 “I said that it doesn’t stand up” (J., 3; 4) 
(19) Não    te         engasgas–te     nada! 
 NEG  CL.2S  choke–CL.2S  nothing 
 “You don’t choke at all!” (R., 2;5) (Duarte et al. 1995) 
 
 This target–deviant pattern with the clitic pronoun realised in two 
positions in the clause will play an important role in the evaluation of the formal 
accounts of clitic placement in CG. I return to this type of structure in section 7.7 
and, building on ideas in Franks (1998) and Franks and Bo!kovi" (2001), I argue 
that the clitic is realised in its raised position (pre–verbal placement) while a copy 
left behind upon the manifestation of clitic movement is realised in a lower 
position (post–verbally), causing the dual realisation of the clitic pronoun.  
 
7.6. Developmental patterns and formal accounts of clitic placement in CG 
 
The acquisition data and the developmental patterns attested are expected to 
contribute to a better understanding of the nature of cliticisation in CG from a 
theoretical perspective and facilitate the evaluation of competing analyses of the 
proclisis–enclisis alternation in CG. In the light of new evidence for clitic 
misplacement and double realisation of the clitic pronoun in early CG, we can 
now evaluate the syntactic, prosodic and interface accounts of clitic placement in 
CG presented in chapter 3. 
Purely syntactic accounts like Agouraki’s (2001) and Terzi’s (1999a, 
1999b) share the assumption that enclisis derives from proclisis with the 
manifestation of verb movement. Agouraki (2001) postulates V–to–C, while Terzi 
(1999a, 1999b) postulates V–to–M. This assumption implies that derivationally 
enclisis involves an extra step, realised as movement of the verb to C or to M, as 
compared to proclisis. This predicts that the latter should be easier to acquire than 
the former. Data from early CG, however, have revealed that enclisis is acquired 
first. This preference for enclisis, realised as its overgeneralisation in all syntactic 
contexts, requires explanation. If we follow either Agouraki’s (2001) or Terzi’s 
(1999a, 1999b) line of reasoning we would have to argue that the derivationally 
 
 
 
 
201 
more complex pattern, namely enclisis, is in fact easier to acquire. Such a position 
is unequivocally counter–intuitive, and hence undesirable. 
An alternative line of thinking would be to attribute the overgeneralisation 
of the enclisis pattern to the overgeneralisation of verb movement and to assume, 
following proposals in Petinou and Terzi (2002), that children initially misanalyse 
the X–bar status of the inflectional particles heading MP, which regulate finite 
verb movement. The basic tenet of Petinou and Terzi’s (2002) account is that even 
though these inflectional particles are heads (of MP) in adult language and thus 
able to satisfy the feature checking requirements of M, in child grammar they are 
perceived as phrasal specifiers (located in the SpecMP) or adjuncts, and hence 
unable to check the verbal features of M. Such a claim lacks empirical 
justification (see discussion in chapter 4), since the evidence provided by Petinou 
and Terzi (2002) is found in SLI data alone. However, even if this analysis were 
valid, it would still fail to account for the instances of double realisation of the 
clitic pronoun in child Cypriot Greek. 
The purely prosodic account put forward by Condoravdi and Kiparsky 
(2001) states that enclisis derives by Prosodic Inversion (PI), if no suitable 
prosodic host is available on the left of the clitic pronoun. If, however, a non–
adjoined constituent appears on the left of the clitic pronoun within the same CP, 
no PF operations need to take place and proclisis follows. In order to account for 
the acquisition data within Condoravdi and Kiparsky’s (2001) account, one must 
postulate redundant applications of PI in all syntactic contexts and assume that 
children have not yet acquired the restrictions imposed on this operation. Even so, 
once again, the phenomenon of clitic double realisation would remain 
unaccounted for. To clarify, PI has an effect on the relative order of constituents, 
but there is no obvious way in which it can result in the realisation of the clitic 
pronoun in two positions. 
Turning now to Mavrogiorgos’ (2012) analysis, it must be noted that while 
superficially it may seem similar to Agouraki’s (2001) and Terzi’s account in that 
it assumes verb movement to F in finite enclisis, it differs in a crucial respect: 
Mavrogiorgos assumes that enclisis and proclisis derive independently. He, 
further, indicates (p.c.) that enclisis in CG involves the following operations: 
 
 
 
 
202 
agree or select, clitic movement and PF inversion, whereas proclisis involves 
agree and incorporation. Hence, agree is involved in the manifestation of both 
enclisis and proclisis. For enclisis, two additional operations are implemented, 
clitic movement and PF inversion, whereas for proclisis incorporation (of the 
clitic pronoun to its verbal host) applies. From an acquisition perspective the 
question is whether clitic movement and PF inversion are easier for children to 
acquire than incorporation. In the absence of suitable evidence, this remains an 
open question. However, even if one could prove that children acquire the former 
set of operations earlier than incorporation, Mavrogiorgos’ (2012) analysis faces 
the same problem as the aforementioned analyses: under no condition does the 
clitic pronoun appear in a double position in the clause. 
Having found no entirely satisfactory answer in all the accounts discussed 
so far, I will examine Revithiadou’s (2006) account in the next section. 
 
7.7. Developmental patterns and the PF!controlled spell!out of copies 
  
Within Revithiadou’s analysis (2006), clitic pronouns in CG move from the VP to 
the inflectional head to which the verb raises and copies are left behind (Chomsky 
2000). In the spirit of Franks (1998) and Franks and Bo!kovi! (2001), the 
pronunciation of such a chain is decided in the PF: the chain is produced in the 
head position, unless this leads to a PF violation. When a functional constituent is 
realised in the same Intonational phrase (henceforth I–phrase) (Selkirk 1995) as 
the clitic, the higher copy is realised, whereas when no such constituent is present, 
the lower clitic is realised. With respect to their prosodisation, Revithiadou (2006, 
2008) proposes that post! and pre!verbal clitics in CG differ with respect to the 
way they prosodise to their verbal host. Post!verbal clitics are always 
incorporated into the PrW of their verbal host and appear as internal enclitics: [V 
cl]PrW, while pre!verbal clitics are either parsed together with the preceding 
(stressed) function word as prosodic words: [fnc cl]PrW [V]PrW, or they join into a 
recursive PrW structure with the (unstressed) function word and the verb and they 
 
 
 
 
203 
appear as affixal proclitics: [fnc cl [V]PrW]PrW4 . 
Revithiadou’s (2006) account seems promising for accommodating the 
two main findings of the current study: the overproduction of the enclisis pattern 
at the onset of L1 acquisition and the occasional realisation of the clitic pronoun 
in two positions, i.e. both preceding and following the finite verb. In the first 
place, the proposal put forward by Franks (1998) and Franks and Bo!kovi" (2001) 
and adopted by Revithiadou (2006) for the production of copies in non!trivial 
chains can straightforwardly account for the double realisation of the clitic 
pronoun in child data. In the second place, the different prosodic organisation of 
the clitic!verb cluster depending on their relative order explains the salience of 
the enclisis over the proclisis pattern.  
Nevertheless, some scrutiny is required in order to clarify these aspects. 
Let us start from the latter point. Revithiadou (2006) claims that while enclitics 
are always incorporated into the PrW of their verbal host, proclitics are either 
incorporated into the PrW of the preceding (stressed) function word5 or they are 
parsed as affixal proclitics with the preceding (unstressed) function word. Thus, at 
PF, the verb!enclitic cluster differs from the proclitic!verb cluster in a crucial 
respect: only the former is parsed as a single prosodic word. Due to the 
divergence in the prosodisation of the enclitic versus the proclitic pronouns, the 
salience of the enclisis over the proclisis pattern in early data follows. Evidence 
justifying the salience of enclisis over proclisis across languages/varieties is 
offered by Avram and Coene (2007) for Romanian, Mastropavlou, Petinou and 
Tsimpli (2014), Stephany (1997) and Tzakosta (2003, 2004a, 2000b) for SMG, 
and Bermúdez–Otero and Luís (2009) for EP. 
 Avram and Coene (2007) report that post–verbal clitics emerge earlier in 
child Romanian than pre–verbal ones. They carried out a longitudinal study on the 
emergence of accusative direct object clitics (ADOCs) in Romanian (see chapter 
4) and report that the first attested clitic in both corpora examined (Bianca 1;05–
                                                
4 Modal particles like na and tha carry no stress and procliticise to the non–imperative verb form 
(Revithiadou & Spyropoulos 2008); this applies to enna, the dialectal equivalent of tha in CG 
non–negated clauses as well.  
5 In the current chapter, the term function word is used pre–theoretically to refer to modal and 
negative particles, complementizers and wh–elements (see an identical use of the term in 
Revithiadou & Spyropoulos 2008). 
 
 
 
 
204 
2;10 & Antonio 1;09–3;05) was the feminine singular clitic o (“her”). Recall that 
o is the only clitic that surfaces post–verbally in restricted finite contexts, namely 
in periphrastic constructions with the auxiliary avea (“have”)6. Tables 7.1 and 7.2 
show the use of pre– and post–verbal clitics in Antonio’s and Bianca’s corpora 
respectively. 
 
 
Table 7.1: Pre– and post–verbal ADOCs in Romanian (Antonio corpus) (Avram 
& Coene 2007:20, table 2) 
 
Table 7.2: Pre– and post–verbal ADOCs in Romanian (Bianca corpus) (Avram & 
Coene 2007:20, table 3) 
 
As shown in tables 7.1 and 7.2, post–verbal o is the only clitic used until 
age 2;3 by Antonio and until age 2;1.23 by Bianca. Avram and Coene report than 
Antonio and Bianca exclusively used post–verbal o in the early stages of L1A 
irrespective of the feature specification of the clitic’s antecedent. Examples (20–
21) illustrate the substitution errors reported in Avram and Coene (2007) with the 
feminine clitic (o) used in contexts in which a masculine clitic is required. 
 
(20) Adult: Ce–ai facut cu ligheanul (MASC)? 
           “What have you done to the bowl?” 
 
                                                
6 The exceptional behaviour of the singular feminine clitic has been attributed to morpho–
phonological requirements on clitic combinations (Marin 2004) as well as to a coalition of factors, 
some of which are phonological (Avram 2000); see the discussion in section 4.3.2 of chapter 4 
and, in particular, footnote 8 of that chapter. 
Age 1;9 2;0 2;1 2;2 2;3 2;4 2;5 2;6 2;6.30 2;7 
Post–verbal o 6 4 2 2 8 5 5 2 9 9 
Other post–verbal 0 0 0 0 0 0 1 0 0 0 
Pre–verbal 0 0 0 0 3 0 5 5 12 14 
Total 6 4 2 2 11 5 11 7 21 23 
Age 2;0 2;0.21 2;1.11 2;1.23 2;1.29 2;2 2;3 
Post–verbal o 2 4 2 0 0 0 0 
Other post–verbal 0 0 0 1 0 0 0 
Pre–verbal 0 0 0 4 1 2 1 
Total 2 4 2 5 1 2 1 
 
 
 
 
205 
Child: Spart–o. 
             Broken_CL.ACC.FEM 
          “I broke it” (A., 1;9) 
(21) Child: Pun       pestele    si     o               p(r)ind     asa. 
           Put–1S  fish_the  and  CL.ACC  catch–1S  like this 
“I put the fish and I catch it like this” (B., 2;5.18) 
 
Turning now to SMG, Stephany (1997) and Tzakosta (2003, 2004a, 
2004b) show that enclitics are acquired earlier than proclitics7. The results of 
these studies can be taken as indirect evidence for the salience of enclitics over 
proclitics within the same language, while Mastropavlou, Petinou and Tsimpli 
(2014) offer indirect evidence for the salience of enclitics over proclitics between 
languages/varieties.  
Stephany (1997) reports that by age 1;10, 3 out of the 5 children she 
studied (Mairi, Janna and Spyros) use accusative and genitive clitics only 
enclitically, while proclitic pronouns occur in the speech of 2 children only (Mairi 
and Spyros). She takes this to indicate that enclitics are used productively before 
proclitics (1997:238). 
Tzakosta’s (2003, 2004a, 2004b) results confirm Stephany’s (1997) 
observations. Tzakosta dealt with phonological aspects of the acquisition of clitics 
in SMG on the basis of longitudinal data from 6 children: Melitini 
(1;07.05!2;04.27), Bebis 1 (1;09.22!2;10.23), Bebis 2 (1;10!2;01.05), Felina 
(1;11.07!3;09.19), Dionisis (2;01!2;09) and Marilia (2;07.06!3;05.23)8. Tzakosta 
observes an asymmetry in the acquisition of proclisis as compared to enclisis: 
enclitics, both single clitics and clitic clusters, emerge first, while proclitics 
emerge in a subsequent stage. The developmental stage in which SMG!speaking 
children produce only enclitics covers the age range from 1;07.05 to 1;10. Table 
7.3 summarises the results for the production of enclitics and proclitics in child 
Greek reported in Tzakosta (2003, 2004a, 2004b).  
                                                
7 However, note that Marinis (2000) reports simultaneous emergence of both pre– and post–verbal 
clitics (see chapter 4 for an extensive discussion). 
8 These data come from two corpora (Tzakosta & Metaxaki) from the database of the University 
of Leiden Centre for Linguistics. 
 
 
 
 
206 
 
Table 7.3: Production and omission of single enclitics and single proclitics (based 
on tables 1 & 2 in Tzakosta 2003). 
 
Tzakosta’s (2003) claim for the salience of the enclisis pattern as 
compared to the proclisis pattern is supported by the results reported in table 7.3. 
The results show that the proportion of overt clitics, which she calls percentage of 
preservation, is lower in proclisis than in enclisis contexts, and, conversely, the 
proportion of omission is higher in proclisis than in enclisis contexts. In line with 
Revithiadou (2006), Tzakosta (2003) argues that enclitics are word internal. She 
stresses that enclitics are post!stress elements that can perfectly fit the minimal 
prosodic word 9  in early production. She offers the fused productions of 
verb!enclitic clusters, as in /'dhos(e)mu/" ['dho.mu] “give me.DAT”, as 
evidence for that claim. Furthermore, she observes that, during the stage in which 
children truncate their prosodic words, the stressed and rightmost syllables are 
faithfully kept, as they signal word boundaries. For Tzakosta (2003, 2004a, 
2004b), phonology outranks syntax with respect to the acquisition of clitic 
pronouns. Moreover, she attributes the observed enclisis!proclisis asymmetry to 
phonological, segmental, and prosodic, as well as perceptual factors. 
The salience of enclisis over proclisis within the same language/variety is 
adequately justified on the basis of the studies conducted by Stephany (1997) and 
Tzakosta (2003, 2004a, 2004b), while Mastropavlou, Petinou and Tsimpli (2014) 
show the salience of enclisis over proclisis to hold between languages/varieties as 
well. They tested clitic production in root clauses with [+/!PAST] verb forms in 
two groups of SLI children: CG! and SMG!speaking. As already shown, CG and 
                                                
9 Tzakosta (2003) assumes that Greek children employ the Minimal Word Template (see Tzakosta 
2003 and references therein) at the onset of L1A. 
Clitics Enclitics Proclitics 
Produced Omitted Produced Omitted 
Melitini 90/94 (.96) 4/94 (.04) 304/328 (.92) 24/328 (.07) 
Bebis 1 107/110 (.97) 3/110 (.03) 332/376 (.88) 44/376 (.12) 
Bebis 2 16/16 (1.0) 0/16 (0.0) 9/26 (.35) 17/26 (.65) 
Felina 67/71 (.94) 4/71 (.06) 369/387 (.95) 18/387 (.05) 
Dionisis 23/23 (1.0) 0/23 (0.0) 205/220 (.93) 15/220 (.07) 
Marilia 38/38 (1.0) 0/38 (0.0) 195/195 (1.0) 0/195 (0.0) 
 
 
 
 
207 
SMG display an interesting divergence with respect to clitic placement in root 
clauses, with the former exhibiting enclisis and the latter exhibiting proclisis. 
Mastropavlou et al. administered a speech elicitation task to CG! and 
SMG!speaking children from 3 groups: an SLI group aged 5 to 6, an TD 
age!matched group, and a TD language!matched group aged 3;4 to 4;4. They 
showed that enclitic production in CG!speaking children with SLI was 
significantly better than proclitic production in their SMG!speaking peers. 
Table 7.4 summarises the results obtained from the speech elicitation task 
performed by CG! and SMG!speaking children from these 3 groups: an SLI 
group, a TD age!matched and a TD language!matched group.  
  
Group SLI/SMG SLI/CG TD/SMG 
(age) 
TD/CG 
(age) 
TD/SMG 
(lang) 
TD/CG 
(lang) 
PRES .70 .89 .96 .97 .91 .94 
PAST .64 .81 .92 .80 .87 .92 
 
Table 7.4: Mean clitic production per group in the present and past condition 
(based on figure 1 in Mastropavlou et al. 2014). 
 
The performance between the CG and SMG control groups (both the 
age!matched and the language!matched) seems to differ. However, these 
differences are not statistically significant. The performance of the two SLI 
groups, on the other hand, differs significantly, with the CG group performing 
better than the SMG group (p=.016), but only in constructions involving present 
tense verbal forms. 
Mastropavlou et al. (2014) do not take this outcome to indicate that 
morpho!phonology plays a role in clitic acquisition in impaired populations 
alone. Instead, they suggest that similar effects may be attested in typically 
developing children before syntactic knowledge is fully acquired. Hence, their 
prediction is essentially that if younger TD children were tested in clitic 
production, CG!speaking children would have performed better than the 
age!matched SMG group. Hence, the crucial factor affecting the performance of 
the two groups for Mastropavlou et al. (2014) is the pattern of placement. One 
implication of such a proposal is that the advantage in salience for the enclisis 
pattern over the proclisis pattern boosts clitic production.  
 
 
 
 
208 
The different status of enclitics as compared to proclitics with regard to 
their relation to their verbal host seems to hold in another language that exhibits 
clitic misplacement in early stages of L1A, namely EP. Bermúdez–Otero and Luís 
(2009) have convincingly argued that EP enclitics belong to the same grammatical 
word (GWd)10 as their verbal host, while EP proclitics lie outside the GWd 
containing the verb. In other words, EP enclitics are incorporated into their verbal 
host at the word–level, while EP proclitics are incorporated at the phrase–level. 
Lobo and Costa (2012) offer some examples that can be taken as corroborative 
evidence for the above claim. In particular, they cite examples in which the 
enclitic (incorrectly) precedes the verbal agreement. This reveals that EP enclitics 
are incorporated into their host at the word–level, while EP proclitics, for which 
no such evidence is available, are presumed to be incorporated at the phrasal–
level. 
 
(22) Dá–me–s                      uma moeda no meu porquinho? 
 Give–CL.1S–AGR.2S  a coin in the my little pig 
 “Will you give me a coin for my little pig?” (J., 3; 4) 
(23) Ai, duas pessoas a agarrar–me–m!   
 Oh, two people to  grab–CL.1S–AGR.3PL 
 “Oh, two people grabbing me!” (J., 3; 5) 
 
The salience of enclisis in child grammar both within as well as between 
languages/varieties has thus been established on the basis of various studies 
(Avram & Coene 2007, Lobo & Costa 2012, Mastropavlou et al. 2014, Stephany 
1997, Tzakosta 2003, 2004a, 2004b). 
Turning now to the other aspect of Revithiadou’s (2006) analysis, namely 
the availability of copies of the clitic pronoun both pre! and post!verbally, I offer 
evidence from the acquisition of verbal morphology in English. It is well!known 
that double–auxiliary constructions of the type shown in (24) are widely attested 
in child English. 
 
                                                
10 Grammatical word (GWd) defines a word–level phonological domain. 
 
 
 
 
209 
(24) What did the smurf didn’t buy? 
(Actual example collected by Hiramatsu 1997; 2000a; 2000b in Bo!kovi" 
2001:117) 
 
Hiramatsu (1997, 2000a; 2000b) examines this type of structure and 
claims that such a non–adult–like production can be attributed to two phenomena: 
the misanalysis of n’t as constituent rather than sentential negation, and the double 
realisation of the copy of the auxiliary. Let us clarify this point, following 
Hiramatsu. The clause structure of (24) appears in (25), which shows the 
instances of movement and the copies left behind: the subject moves to the 
specifier of IP, the wh–phrase to the specifier of CP, and I to C. 
 
(25) [CP What [C# I [IP the smurf [I# I [VP n’t [VP the smurf buy what]]]]]] 
(Bo!kovi" 2001:117) 
 
Upon the manifestation of the wh– movement and the subject movement, 
the heads of the chains are pronounced. As for the I–to–C movement, the head is 
pronounced and do–support takes place in order for the Stranded Affix Filter11 not 
to be violated. Hence, the raised I is spelled out as did. However, the tail of the 
chain I–to–C cannot be deleted, because, if it was, the negative clitic n’t would 
not have a proper lexical host, and the construction would crash in PF. In order to 
avoid a PF violation, the tail is pronounced as well and did serves as the host for 
n’t. 
 
(26) [CP What [C# did [IP the smurf [I# did [VP n’t [VP the smurf buy what]]]]]] 
(Bo!kovi" 2001:118) 
 
The double auxiliary construction in (24) is reminiscent of the double 
clitic constructions attested in my database, cited in examples (48–53) in chapter 
6. Let us try to capture exactly how the derivation of such a construction proceeds. 
                                                
11 The Stranded Affix Filter proposed by Lasnik (1981) states that a morphologically realised affix 
must be a syntactic dependent of a morphologically realised category at the surface structure. 
 
 
 
 
210 
Following Mavrogiorgos’ (2009) movement proposal for the derivation of 
proclisis in SMG, it is assumed that clitic pronouns are merged as DPs/Ds in the 
complement position of the VP. The clitic’s phi–features are visible to appropriate 
probes located higher in the clause structure, therefore the clitic is then attracted 
by V12 (or some other verbal v head) to its specifier position. In this way, an A–
chain is formed. Simultaneously, an A!–chain is formed with the clitic and an 
optional EPP (Extended Projection Principle) feature at v*, which bears a 
familiarity/–focus/old information D feature13. The clitic incorporates14 into the 
edge of v*, and eventually moves to T incorporated in the v*–V complex. The 
above assumptions are illustrated in (27), based on Mavrogiorgos’ tree structure 
in (34) (2009:99).   
 
(27) [TP [T! DP/Dcl–V–v*–T [v*P (DP/Dcl) [v*! [(V–v*) [VP (V) (DP/Dcl)]]]]] 
 
Returning to Revithiadou’s (2006) analysis, the idea advanced in line with 
Franks (1998) and Franks and Bo!kovi" (2001) is that a lower copy of the clitic is 
available in PF. On the basis of the clause structure adopted for CG (see (26) in 
chapter 1), repeated below for ease of reference, it is expected that the copy of the 
DP/Dcl within v*P is available in PF. 
 
(28) [C pu [COp oti/an/na/as [Neg dhen/min [CM tha/na/as [I cl + V...]]]]] 
 
Thus, the structure I assume for examples like (29) is illustrated in (30). 
 
(29) *CHI: (n)a  to  (p)ki  to […]. 
M     CL V      CL 
(30)  [CM (n)a [TP [T! to pki [v*P to [v*! [pki [VP pki to]]]]] 
 
                                                
12 Mavrogiorgos (2009) takes V to be a phase head by inheritance of phi–features from v*. 
13 With regard to the phrasal status of the clitic, Mavrogiorgos assumes that it “moves as an XP, 
but lands as an X by incorporating into the edge of v*, which is a minimal morpho–syntactic phase 
transparent at the edge” (2009:98). 
14 See section 3.3 of chapter 3 in Mavrogiorgos (2009) for a detailed discussion of how 
incorporation takes place. 
 
 
 
 
211 
The above proposal accounts for the realisation of the clitic pronoun in both 
positions in child data, unlike the adult data, where the PF regulates which of the 
two copies should be realised. 
Based on these facts, it seems that clitic placement in CG is the result of 
operations that take place at the syntax–phonology interface. The proposal I adopt 
follows the spirit of proposals in Bo!kovi" (2000, 2001) (for cliticisation in 
Serbo#Croatian) and the basic tenet is that the placement requirement imposed on 
CG clitics “can be captured in its entirety through a filtering effect of the 
phonology on the syntax” (2000:105). 
Such a proposal differs radically from the Prosodic Inversion approach, 
put forward by Halpern (1995) and adopted by Condoravdi and Kiparsky (2001) 
to account for clitic placement in Greek dialects. PI is a radical PF operation. 
Specifically, it is a PF reordering mechanism whose implementation results in a 
rightward movement of the clitic to the right edge of its prosodic word (see 
Halpern 1995, Halpern & Zwicky 1996). Hence, for CG, it is assumed that the 
clitic is placed pre#verbally by syntax and enclisis derives from the manifestation 
of PI (Condoravdi & Kiparsky 2001). If we were to explain the attested 
overgeneralisation of enclisis within such an approach, we should attribute it to 
the unconstrained application of PI, even in contexts in which this is banned in 
adult language. 
However, there are important drawbacks for such an assumption that have 
been already discussed in chapter 3 and are briefly recapitulated here. The basic 
tenet of such an analysis is the existence of a very powerful PF which regulates 
the re–ordering of the verb and the clitic within the CL–V cluster. However, 
operations such as movement and linearisation of constituents are standardly 
perceived as syntactic, hence this is not a desirable assumption theoretically. 
Moreover, the claim that these operations are syntactic, but triggered by prosodic 
requirements, is not compatible with the standard assumption that syntax cannot 
look ahead to the final prosodic structure (Mavrogiorgos 2012). 
Within the PF#controlled spell#out of copies approach, the movement and 
copying of clitic pronouns takes place in syntax, while phonology deletes all but 
the highest copy, except when the realisation of the latter would lead to a PF 
 
 
 
 
212 
violation (Franks 1998, Franks & Bo!kovi" 2001). Specifically, syntax has the 
following output, upon the manifestation of the clitic movement from v*P to T#: 
 
(31) (function word) CL V CL 
 
Phonology then determines word order “by filtering out certain well–formed 
syntactic representations” (Bo!kovi" 2001:94). 
The backbone of such an analysis is Chomsky’s (1993) copy theory of 
movement. Under the original trace theory of movement, movement of an element 
leaves behind a trace and chains are pronounced in the head position, where 
phonological information is located. However, Chomsky (1993) proposes that 
movement leaves behind a copy of the moved element rather than a trace15. While 
the standard assumption is that LF may regulate the realisation of these copies, 
several authors, including Franks (1998) and Franks and Bo!kovi" (2001), 
proposed that deletion of copies is available in PF as well. 
I follow the proposals in Franks (1998) and Franks and Bo!kovi" (2001) 
and assume that a chain is pronounced in the head position and lower copies are 
deleted in PF, unless the pronunciation in the head position would lead to a PF 
violation, in which case the lower member of the chain is pronounced and the 
head is deleted. Moreover, I adopt Franks and Bo!kovi"’s (2001) perspective and 
attribute a derivational nature16 to the clitic–verb linearisation in CG as well. 
The next question is how PF regulates the realisation of either the lower or 
the higher copy and how the overproduction of enclisis at the initial stages of L1A 
can be captured. There are at least two ways in which the PF$optimisation 
procedure can be treated, each of which has different implications for children’s 
early production. 
                                                
15 This proposal conforms to the Inclusiveness Condition “which restricts syntactic operations to 
re–arrangements of elements introduced into the structure from the lexicon. The condition 
prohibits syntax from creating new elements, i.e. from introducing into the derivation elements 
that were not inserted from the lexicon” (Bo!kovi" 2001:98). 
16  Franks and Bo!kovi" (2001) suggest a phase–based approach to the multiple spell–out 
hypothesis on the basis of facts from Bulgarian. In particular, they assume that information is sent 
from syntax to phonology derivationally, as the structure is being built. Following Chomsky 
(2000), they assume that information is sent at discrete junctures, the phases, and that CP, but not 
IP, is a phase. 
 
 
 
 
213 
Firstly, the syntactic output may be filtered out through the constraints for 
clitic prosodisation in CG, ranked as proposed in Revithiadou (2006) in 
Optimality theoretic (OT) terms17. The constraint ranking is outlined in (14) in 
chapter 3 and repeated below as (32). Within Revithiadou’s approach, at the initial 
stages of clitic L1 acquisition, during which only enclitics emerge, the 
hierarchical order of the constraints in child grammar differs from that of the adult 
grammar. It may be assumed that the constraint requiring rightward directionality 
to the clitic element, WCON (R), is ranked higher than WCON (L), which requires 
leftward directionality. 
 
(32) FAITH (Acc), EXH, WCON (L), NON REC >> PCON, WCON (R) 
 
Alternatively, clitic placement may be regulated by some PF filters such as 
those proposed by Bo!kovi" (2000/2001) for cliticisation in Serbo–Croatian. 
Bo!kovi" proposes that Serbo–Croatian clitics undergo syntactic movement and 
phonology places them in the second position of their I–phrase (2000:114–115). 
In particular, he suggests that Serbo–Croatian clitics are suffixes18 within their I–
phrase (Selkirk 1995) and that they are specified as such in the lexicon. This 
specification is represented by the following PF filters: a. #_ and b. suffix 
(Bo!kovi" 2000:107). The former indicates that clitics appear in the initial part 
within their domain of cliticisation (I–phrase) and the latter indicates that they are 
suffixes, hence right–adjacent to their host. Syntactic outputs that violate this 
lexical requirement of Serbo–Croatian clitics are filtered out in PF. 
Extending Bo!kovi"’s (2000, 2001) analysis, one can account for 
cliticisation in CG as follows: the Tobler–Mussafia effect is assumed to be a 
second position effect whose domain of application is the I–phrase (Revithiadou 
& Spyropoulos 200819). The overgeneralisation of enclitics in early CG can then 
                                                
17 The use of OT constraints within the context of generative approaches is not such a novelty. 
Franks and King (2000) give an account of cliticisation in Serbo–Croatian in which syntax is 
generative, but PF is regulated by OT constraints.  
18 Crucially, Bo!kovi" does not take clitics to be affixes. Instead, he uses the term suffix “to 
indicate a phonologically weak element that follows its host” (2000:104, footnote 29). 
19 Revithiadou and Spyropoulos (2008) consider CG clitics second position (2P) elements in the 
sense that they appear in the second position of their I–phrase. 
 
 
 
 
214 
be attributed to a misanalysis of the PF filters that regulate clitic placement. In 
particular, children assume that clitics are verbal suffixes rather than suffixes to 
the initial constituent within their I–phrase and they always place them post–
verbally. What needs to be stressed is that like Bo!kovi", I do not assume that 
clitics are affixes (2000:104, footnote 29). Instead, I take the term suffix to 
indicate a phonologically weak element that prosodifies in the preceding element 
within the same I–phrase.  
The acquisition data do not provide evidence in favour of one or the other 
approach. However, corroborative evidence for the second position status of CG 
clitics (within their I–phrase) comes from the diachronic development of clitic 
pronouns in Greek (Pappas 2001; 2004, Revithiadou & Spyropoulos 2008). 
Revithiadou and Spyropoulos indicate that Greek clitics in the Late Classical and 
early Post–Classical period (4th–2nd c. BC) had to adhere to the Wackernagel 
Law and occupy the 2nd position in the clause. The relaxation of this Law led to 
the emergence of two subsystems, one of which retained the 2P status of clitic 
pronouns. However, the domain of application of second position restrictions was 
no longer the clause but instead the phonological phrase. This subsystem was the 
ancestor of the CG grammar. A standard assumption that I adopt is that the 
Tobler–Mussafia effect is pertinent to the syntax–phonology interface, and hence 
the domain of application of phonological operations that regulate clitic placement 
in contemporary CG could be none other than the I–phrase, as suggested in 
Revithiadou and Spyropoulos (2008). Within such an analysis, phonology 
determines the 2P effect through filtering out an over–generating syntax. 
 
 
 
 
 
 
 
 
 
 
 
 
 
215 
CONCLUDING REMARKS 
 
The current study has provided evidence for an interesting phenomenon attested 
in early CG, namely the over–production of the enclisis pattern, on the basis of 
experimental as well as spontaneous data, both cross–sectional and longitudinal. 
This target–deviant pattern is attested in children aged 2;6 to 3;0 with single 
clitics or clitic clusters, and in derivationally more complex structures, such as CD 
and CLLD. Three developmental stages have been identified in clitic L1A in CG: 
the first is characterised by the use of the enclisis pattern alone in all syntactic 
contexts, the second by the manifestation of post–verbal clitic placement in 
enclisis contexts, pre– and post–verbal placement in proclisis contexts, as well as 
sparse occurrences of double realisation of the clitic pronoun (both pre– and post–
verbally), and the last by target–like clitic placement in both enclisis and proclisis 
contexts. 
From a theoretical perspective, the over–production of the enclisis pattern 
is explained within Revithiadou’s (2006) account of clitic placement in CG. The 
movement of CG clitics leaves behind copies, which are available in PF. 
Phonology deletes all but the highest, unless its realisation would lead to a PF 
violation (Franks 1998, Franks & Bo!kovi" 2001). Following ideas in Bo!kovi" 
(2000, 2001), I attribute the Tobler–Mussafia effect in CG to the second position 
status of CG clitics within their I–phrase, while I locate the domain of cliticisation 
in CG at the syntax–phonology interface.  
The outcome of the current study has an important implication for the 
categorisation of clitic languages from the perspective of L1 acquisition. 
Following Mavrogiorgos’ (2012) tri–partition with respect to the types of clitic 
languages, clitic placement in finiteness–sensitive languages, including Catalan, 
Italian, Romanian, Spanish and Standard Modern Greek among others, is target–
like from the onset of L1A. With regard to languages exhibiting the Tobler–
Mussafia effect, like CG and European Portuguese, a different picture emerges. 
The current study has offered ample evidence for the over–production of the 
enclisis pattern in all syntactic contexts at the onset of L1A in CG, as well as for 
the double realisation of the clitic pronoun both pre– and post–verbally in 
 
 
 
 
216 
proclisis contexts. It has been shown that similar target–deviant structures are 
produced by young EP–speaking children as well (Lobo & Costa 2012). On the 
basis of this divergence, I attribute the placement errors in the clitic production of 
CG– and EP–speaking children to the type of the clitic language they acquire. In 
other words, while children acquiring Tobler–Mussafia type languages manifest 
clitic misplacement at the onset of L1A, children acquiring finiteness–sensitive 
languages exhibit adult–like clitic placement from the onset. With respect to 
languages exhibiting second position restrictions Serbo–Croatian is a 
representative example. Here there is some evidence for unraised clitic pronouns 
in 3–year–olds (Ilic & Ud Deen 2004); no data are available for earlier stages of 
clitic L1A to test whether young children fail to produce clitic pronouns in their 
raised positions.  
The existence or absence of clitic misplacement in early languages is 
contingent on the type of clitic language being acquired. A direct comparison of 
the acquisition process in CG and SMG, two varieties sharing the morphological 
paradigm of pronominal clitics, while representing a different type of clitic 
language, is indicative in this respect. Early CG exhibits target–deviant patterns 
including misplacements and realisations of multiple copies of the clitic pronoun, 
while early SMG patterns with adult SMG with respect to clitic placement. The 
target–deviant productions in early CG are attributed to the Tobler–Mussafia 
properties of the variety being acquired.  
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
217 
LIMITATIONS OF THE CURRENT STUDY – 
IMPLICATIONS FOR FUTURE RESEARCH 
 
The present study has used different methodologies to investigate the L1A of 
clitic placement in CG on the basis of data from 58 children. This is the largest 
database for the study of clitic placement in enclisis and proclisis contexts in CG–
speaking children so far. This study aimed to define the age range within which 
non–adult clitic placement is attested in early CG, and thus children aged 2;6 to 4 
were recruited. One important outcome of the present study is that it has identified 
that target–deviant patterns occur in the data of children whose age falls within 
the age range 2;6 to 3;0 and in proclisis contexts alone. This outcome points both 
to limitations of this study and to avenues for further research. 
a) First, the overall number of children falling within the critical age 
was, obviously, a subset of the total tested: 18 out of the 50 participants in the 
experiment and 6 out of the 8 participants of the spontaneous speech study belong 
to this age group. A follow–up study should now focus on this particular age 
group and enlarge the database for it. 
b) Having gained an overall picture of both enclisis and proclisis 
contexts, a follow–up study could now concentrate on exploring in detail the 
different proclisis contexts. This should include an experimental design especially 
developed for the elicitation of negative utterances, very few of which were 
obtained with the method used.  
If the conclusion of the current study is on the right track and the Tobler–
Mussafia properties of the language being acquired do indeed result in target–
deviant productions, then similar phenomena are expected in other languages 
exhibiting Tobler–Mussafia properties. Apart from EP, in which clitic 
misplacement has already been attested (Lobo & Costa 2012), such a 
phenomenon is expected to occur in early Galician and Berber, another language 
in which clitics are subject to position restrictions similar to those of CG and EP 
(Petinou & Terzi 2002:23, footnote 19). It would certainly be fruitful to extend the 
analysis to the L1 acquisition of languages of this type.  
! 218 
REFERENCES 
 
Agouraki, Y. (1991). A Modern Greek complementizer and its significance for 
universal grammar. UCL Working Papers in Linguistics 3: 1–24. 
Agouraki, Y. (1993). Spec–Head Licensing: The Scope of the Theory. 
Unpublished PhD Dissertation, University College London. 
Agouraki, Y. (2001). The position of clitics in Cypriot Greek. In: Ralli, A., 
Joseph, B. & M. Janse (eds.), Proceedings of the First International 
Conference of Modern Greek Dialects and Linguistic Theory, 1–18. 
University of Patras. 
Agouraki, Y. (2010). It–clefts and stressed operators in the preverbal field of 
Cypriot Greek. Lingua 120: 527–554. 
Aguado, G. (1988). Valoracion de la competencia morfosintactica en el niño de 
dos años y medio. Infancia y Aprendizaje 43: 73–95. 
Alexiadou, A. (1994). Issues in the syntax of Adverbs. Unpublished PhD 
Dissertation, University of Potsdam. 
Alexiadou, A. (1997). Adverb Placement: a Case Study in Antisymmetric Syntax. 
Amsterdam: John Benjamins. 
Alexiadou, A. & E. Anagnostopoulou (1998). Parametrizing Agr: word order, V–
movement and EPP–checking. Natural Language and Linguistic Theory 
16: 491–539. 
Amery, H. & S. Cartwright (2009). First Hundred Words in English. London: 
Usborne Publishing. 
Anagnostopoulou, E. (1994). Clitic Dependencies in Modern Greek. Unpublished 
PhD Dissertation, University of Salzburg. 
Anagnostopoulou, E. (2006). Clitic Doubling. In: Everaert, M. & H. van 
Riemsdijk (eds.), The Blackwell Companion of Syntax, 519–581. Oxford: 
Blackwell. 
Antelmi, D. (1992). Uipotesi Maturazionale Nel Vacquisizione del Linguaggio. 
Indagine Longitudinale su una Bambina Italiana. Unpublished PhD 
Dissertation, University of Venice/ Padua. 
! 219 
Arvaniti, A. (2010). Linguistic practices in Cyprus and the emergence of Cypriot 
Standard Greek. Mediterranean Language Review 17: 15–45. 
Avram, A. (1986). Sandhi phenomena in Romanian. In: Henning, A. (ed.), Sandhi 
Phenomena in the Languages of Europe, 551–574. Berlin/New 
York/Amsterdam: Mouton de Gruyter. 
Avram, L. (1999). Clitic Omission in Child Romanian. Ms. University of 
Bucharest/University of Potsdam. 
Avram, L. (2000). From possessive clitics to object clitics: a unifying analysis. In: 
Tasmowski L. (coord.), The Expression of Possession in Romance and 
Germanic Languages, 83–100. Cluj–Napoca: Clusium. 
Avram, L. & M. Coene (2007). Object Clitics as Last Resort. Implications for 
language acquisition. In: Baauw, S., van Kampen J. & M. Pinto (eds.), The 
Acquisition of Romance Languages. Selected papers from The Romance 
Turn II, 7–26. Utrecht: LOT. 
Babyonyshev, M. & S. Marin (2006). Acquisition of Pronominal Clitics in 
Romanian. Catalan Journal of Linguistics 5: 17–44. 
Benincà, P. & G. Cinque (1993). Su alcune differenze tra enclisi e proclisi. In 
Omaggio a Gianfranco Folena, 2313–2326. Editoriale Programma, 
Padova. 
Bermúdez–Otero, R. & A. R. Luís. (2009). Cyclic domains and prosodic spans in 
the phonology of European Portuguese functional morphs [Handout]. 
Paper presented at the “Old World Conference in Phonology 6” 
(Edinburgh, 24 January 2009). (downloadable at: http://www.bermudez-
otero.com/bermudez-otero&luis.pdf) 
Beukema, F. & P. Coopmans (1989). A Government–Binding Perspective on the 
Imperative in English. Journal of Linguistics 25: 417–436. 
Bonet, E. (1994). The Person–Case constraint: A morphological approach. The 
Morphology–Syntax Connection, MIT Working Papers in Linguistics 22, 
33–52. 
Bo!kovi", #. (2000). Second Position Cliticization. Syntax and / or Phonology? 
In: Beukema, F. & M. den Dikken (eds.), Clitic Phenomena in European 
Languages, 71–119. Amsterdam / Philadelphia: John Benjamins. 
! 220 
Bo!kovi", #. (2001). On the Nature of the Syntax–Phonology Interface: 
Cliticization and Related Phenomena. Amsterdam, London, New York: 
Elsevier Science. 
Cardinaletti, A. & M. Starke (1999). The typology of structural deficiency: A case 
study of the three classes of pronouns. In: van Riemsdijk H. (ed.), Clitics 
in the languages of Europe, 145–233. Berlin: Mouton de Gruyter. 
Carnie, A., Harley, H. & E. Pyatt (2000). VSO Order as Raising Out of IP? Some 
Evidence from Old Irish. In Carnie, A. & E. Guilfoyle (eds.), The Syntax 
of Verb Initial Languages, 39–59. Oxford University Press. 
Chatzikyriakidis, S. (2010). Clitics in 4 Dialects of Modern Greek: A Dynamic 
Account. Unpublished PhD Dissertation, King’s College, London. 
Chatzikyriakidis, S. (2012). A dynamic account of clitic positioning in Cypriot 
Greek. Lingua 112: 642–672. 
Chomsky, N. (1991). Some notes on economy of derivation and representation. 
In: Freidin, R. (ed.), Principles and parameters in comparative grammar, 
417–454. Cambridge, MA: MIT Press. 
Chomsky, N. (1993). A Minimalist Program for Linguistic Theory. In: Hale, K. & 
S. J. Keyser (eds.), The View from Building 20: Essays in Linguistics in 
Honor of Sylvain Bromberger, 1–52. Cambridge, MA: MIT Press. 
Chomsky, N. (1995). The Minimalist Program. Cambridge, MA: MIT Press. 
Chomsky, N. (1998). Minimalist inquiries: the framework. Ms. MIT. 
Chomsky, N. (2000). Minimalist inquiries: The framework. In: Martin, R., 
Michaels, D. & J. Uriagereka (eds.), Step by step: essays on minimalist 
syntax in honor of Howard Lasnik, 89–155. Cambridge, MA: MIT Press. 
Chondrogianni, V. (2008). The acquisition of determiners and clitic pronouns by 
child and adult L2 learners of Greek. Unpublished PhD Dissertation, 
University of Cambridge. 
Cinque, G. (1990). Types of A–Dependencies. Cambridge, MA: The MIT Press. 
Condoravdi, C. & P. Kiparsky (2001). Clitics and clause structure. Journal of 
Greek Linguistics 2, 1–40. 
Costa, J. & M. Lobo (2007a). Clitic omission, null objects or both in the 
acquisition of European Portuguese? In: Baauw, S., Drijkoningen, F. & M. 
! 221 
Pinto (eds.), Romance Languages and Linguistic Theory 2005, 59–72. 
Amsterdam: John Benjamins. 
Costa, J. & M. Lobo (2007b). Complexidade e omissão de clíticos: o caso dos 
reflexos. XXII Encontro Nacional da Associação Portuguesa de 
Linguística. Textos Seleccionados, 303–313. Lisboa: APL. 
Costa, J., Lobo, M., Carmona, J. & C. Silva (2008). Clitic omission in European 
Portuguese: correlation with null objects? In: Gavarró A. & M. J. Freitas 
(eds.), Language acquisition and development: Proceedings of GALA 
2007, 133–143. Newcastle–upon–Tyne: Cambridge Scholars Publishing. 
Costa, J. & M. Lobo (2009). Clitic omission in the acquisition of European 
Portuguese: Data from comprehension. In: Pires, A. & J. Rothman (eds.), 
Minimalist inquiries into child and adult language acquisition: case 
studies across Portuguese, 63–84. Berlin/ New York: Mouton de Gruyter. 
Duarte, I., Matos, G. & I. H. Faria (1995). Specificity of European Portuguese 
Clitics in Romance. In: Faria, I. H. & M. J. Freitas (eds.), Studies on the 
Acquisition of Portuguese, 129–154. Lisboa: APL. 
Duarte, I. & G. Matos (2000). Romance Clitics and the Minimalist Program. In 
Costa, J. (ed.), Portuguese Syntax. New Comparative Studies, 116–142. 
Oxford University Press.  
Eisenbeiss, S. (2009). Contrast is the Name of the Game. Contrast–Based Semi–
Structured Elicitation Techniques for Studies on Children’s Language 
Acquisition. Essex Research Reports in Linguistics 57 (7). Department of 
Language and Linguistics, University of Essex, Colchester, UK. 
(downloadable at: 
http://www.essex.ac.uk/linguistics/publications/errl/errl57-7.pdf) 
Ferguson, C.A. (1959). Diglossia. Word 15: 325–340. [Reprinted in Paulston, 
C.B. & G.R. Tucker (eds.). 2003. Sociolinguistics: The Essential 
Readings, 345–358. Oxford: Blackwell] 
Ferguson, S. & E. Groat (1995). Defining Shortest Move. Ms. Harvard 
University. 
Fontana, J. (1996). Phonology and Syntax in the Interpretation of the Tobler– 
Mussafia Law. In: Halpern, A. & A. Zwicky (eds.), Approaching Second. 
! 222 
Second Position Clitics and Related Phenomena, 41–83. Stanford, CA: 
CSLI Publications. 
Franks, S. (1998). Clitics in Slavic. Position paper presented at the “Comparative 
Slavic Morphosyntax” Workshop, Indiana University (Spencer, Indiana, 
5–7 June 1998). (downloadable at 
http://www.indiana.edu/~slavconf/linguistics/frnks.pdf) 
Franks, S. (2000). Clitics at the interface: An introduction to clitic phenomena in 
European languages. In: Beukema, F. & M. den Dikken (eds.), Clitic 
Phenomena in European Languages, 1–46. Amsterdam: John Benjamins. 
Franks, S. & T. King (2000). A handbook of Slavic clitics. New York: Oxford 
University Press. 
Franks, S. & !. Bo"kovi# (2001). An Argument for Multiple Spell–Out. 
Linguistic Inquiry 32 (1): 174–183. 
Grohmann, K.K., Panagiotidis, P. & S. Tsiplakou (2006). Properties of Wh–
question formation in Cypriot Greek. In: Janse, M., Joseph, B. and A. Ralli 
(eds.), Proceedings of the Second International Conference of Modern 
Greek Dialects and Linguistic Theory, 83–98. University of Patras. 
Grohmann, K.K. (2011). Some directions for the systematic investigation of the 
acquisition of Cypriot Greek: A new perspective on production abilities 
from object clitic placement. In: Rinke E. & T. Kupisch (eds.), The 
Development of Grammar: Language acquisition and diachronic change. 
In honour of Jürgen M. Meisel (Hamburg Studies on Multilingualism 11), 
179–203. Amsterdam: John Benjamins. 
Grohmann, K.K. & E. Papadopoulou (2011). Question(able) Issues in Cypriot 
Greek. Linguistic Analysis 37 (1–2): 8–38. 
Grohmann, K.K., Theodorou, E., Pavlou, N., Leivada, E., Papadopoulou, E. & S. 
Martínez–Ferreiro (2012). The Development of Object Clitic Placement in 
Cypriot Greek and the Romance Connection. In: Ferré, S., Prévost, P., 
Tuller, L. & R. Zebib (eds.), Selected Proceedings of the Romance Turn IV 
Workshop on the Acquisition of Romance Languages, 128–152. 
Newcastle–upon–Tyne: Cambridge Scholars Publishing. 
Grüter, T. (2006). Object clitics and null objects in the acquisition of French. 
! 223 
Unpublished PhD Dissertation, McGill University, Montreal. 
Guasti, M.T. (1993/1994). Verb Syntax in Italian Child Grammar: Finite and 
Nonfinite Verbs. Language Acquisition 3 (1): 1–40. 
Haegeman, L. (1996). Root Infinitives, Clitics and Truncated Structures. In: 
Clahsen H. (ed.), Generative Perspectives on Language Acquisition, 271–
308. Amsterdam/Philadelphia: John Benjamins. 
Halle, M. & A. Marantz. (1993). Distributed Morphology and the Pieces of 
Inflection. In: Hale, K. & S. J. Keyser (eds.), The View from Building 20: 
Essays in Linguistics in Honor of Sylvain Bromberger, 111–176. 
Cambridge, MA: MIT Press. 
Halpern, A. (1995). On the Placement and Morphology of Clitics. Stanford, 
California: CSLI Publications. 
Halpern, A. & A. Zwicky (1996). Approaching second: second position clitics 
and related phenomena. Stanford, California: CSLI Publications. 
Hamann, C., Rizzi, L. & U. Frauenfelder (1996). On the Acquisition of Subject 
and Object Clitics in French. In: Clahsen H. (ed.), Generative Perspectives 
on Language Acquisition, 309–334. Amsterdam/Philadelphia: John 
Benjamins. 
Heim, I. & A. Kratzer (1998). Semantics in generative grammar. Oxford: 
Blackwell. 
Hickey, T. (1991). Mean length of utterance and the acquisition of Irish. Journal 
of Child Language 18: 553–569. 
Hiramatsu, K. (1997). A production/judgment asymmetry in children’s negative 
questions. Ms. University of Connecticut, Storrs. 
Hiramatsu, K. (2000a). What did Move didn’t erase? In: Braze, D., Kudo, Y. & K. 
Hiramatsu (eds.), UConn Working Papers in Linguistics 10, 75–110. 
Department of Linguistics, University of Connecticut, Storrs. 
Hiramatsu, K. (2000b). Accessing linguistic competence: Evidence from 
children’s and adults’ acceptability judgments. Unpublished PhD 
Dissertation, University of Connecticut, Storrs. 
Holton, D. Mackridge, P. & I. Philippaki–Warburton (1997). Greek: A 
Comprehensive Grammar of the Modern Language. London/New York: 
! 224 
Routledge. 
Horrocks, G.C. (1997). Greek: a history of the language and its speakers. 
London: Longman. 
Hyams, N. (2002). Clausal Structure in Child Greek: A reply to Varlokosta, 
Vainikka and Rohrbacher and a reanalysis. The Linguistic Review 19: 
225–269. (also downloadable at 
http://www.linguistics.ucla.edu/people/hyams/hyams.htm) 
Iatridou, S. (1994). Comments on the Paper by Roberts. In: Lightfoot, D. & N. 
Hornstein (eds.), Verb Movement, 243–60. Cambridge: CUP. 
Ilic, T. & K. Ud Deen (2004). Object raising and cliticization in Serbo–Croatian 
child language. In: van Kampen, J. and S. Baauw (eds.), Proceedings of 
GALA 2003 (Generative Approaches to Language Acquisition) Volume 1, 
235–243. Utrecht: LOT. 
Ioannidou, E. (2009). Using the ‘improper’ language in the classroom: the conflict 
between language use and legitimate varieties in education. Evidence from 
a Greek Cypriot classroom. Language and Education 23 (3): 263–278. 
Ioannidou, E. & A. Sophocleous (2010). “Now, is this how we are going to say 
it?” Comparing teachers’ language practices in primary and secondary 
state education in Cyprus. Linguistics and Education 21: 298–313. 
Janse, M. (1998). Cappadocian clitics and the syntax-morphology interface. In 
Joseph, B., Horrocks, G.C. and I. Philippaki–Warburton (eds.), Themes in 
Greek Linguistics II, 257–281. Philadelphia: Benjamins. 
Kayne, R.S. (1975). French Syntax. The Transformational Cycle. Cambridge, 
MA: MIT Press. 
Kayne, R.S. (1994). The Antisymmetry of Syntax. Cambridge, MA: The MIT 
Press. 
Laka, I. (1990). Negation in syntax: On the nature of functional categories and 
projections. Unpublished PhD Dissertation, MIT. 
Lasnik, H. (1981). Restricting the theory of transformations. In: Hornstein, N. and 
D. Lightfoot (eds.), Explanation in linguistics: the logical problem of 
language acquisition, 152–73. London: Longman.  
Leivada, E., Mavroudi, P. & A. Epistithiou (2010). (Meta)language(s) or 
! 225 
bidialectism? Acquisition of clitic placement by mainland Greeks, Greek 
Cypriots and biethnics in the diglossic context of Cyprus. In: Botinis, A. 
(ed.), Proceedings of the ISCA Tutorial and Research Workshop on 
Experimental Linguistics 2010, 97–100. Athens: University of Athens. 
Liberman, M & A. Prince (1977). On Stress and Linguistic Rhythm. Linguistic 
Inquiry 8 (2): 249–336. 
Lobo, M. & J. Costa (2012). On the acquisition of clitic placement in European 
Portuguese. Paper presented at the CYCLIA Workshop on the Acquisition 
of Clitics (Nicosia, Cyprus, May 25–26 2012). 
MacWhinney, B. (2000). The CHILDES Project: Tools for Analyzing Talk. 3rd 
Edition. Mahwah, NJ: Lawrence Erlbaum Associates. 
MacWhinney, B. (2012a). The CHILDES Project: Tools for Analyzing Talk – 
Electronic Edition. Part 1: The CHAT Transcription Format. 
(downloadable at http://childes.psy.cmu.edu/manuals/chat.pdf) 
MacWhinney, B. (2012b). The CHILDES Project: Tools for Analyzing Talk – 
Electronic Edition. Part 2: The CLAN Programs. (downloadable at  
http://childes.psy.cmu.edu/manuals/clan.pdf) 
Malakoff, M. E., Mayes, L. C., Schottenfeld, R. & S. Howell (1999). Language 
Production in 24–Month–Old Inner–City Children of Cocaine–and–Other–
Drug–Using Mothers. Journal of Applied Developmental Psychology 20 
(1): 159–180. 
Manzini, M.R. & L. Savoia (1999). The Syntax of Middle–reflexive and Object 
Clitics: A Case of Parametrization in Arberërsh Dialects. In: Mandalà, M. 
(ed.), Studi in Onore di Luigi Marlekaj, 328–383. Bari: Adriatica. 
Marin, S. (2004). Pronominal Clitic Positioning is Morpho–Phonological: 
Evidence from Romanian Adult and Acquisition Data. Ms. Yale 
University. 
Marinis, T. (2000). The acquisition of clitic objects in Modern Greek: Single 
Clitics, Clitic Doubling, Clitic Left Disclocation. In ZAS Papers in 
Linguistics 15, 259–281. Berlin: ZAS. 
Mastropavlou, M., Petinou, K. & I.M. Tsimpli (2014). Morpho–phonology and 
compensation: tense marking and object clitics in Greek and Cypriot–
! 226 
Greek children with SLI. Ms. University of Ioannina, Cyprus University of 
Technology and Aristotle University of Thessaloniki. 
Mavrogiorgos, M. (2009). Proclisis and Enclisis in Greek. Unpublished PhD 
Dissertation, University of Cambridge. 
Mavrogiorgos, M. (2010). Clitics in Greek: a minimalist account of proclisis and 
enclisis. Amsterdam: John Benjamins. 
Mavrogiorgos, M. (2012). Enclisis at the syntax–PF interface. Ms. University of 
Ulster. 
Mussafia, A. (1886). Una particolarità sintattica della lingua italiana dei primi 
secoli. In: Ascoli, G. I. et al.  Miscellanea di filologia e linguistica. In 
memoria di Napoleone Caix e Ugo Angelo Canello, 255–261. Firenze: 
Successori Le Monnier.  
Mussafia, A. (1898). Enclisi o proclisi del pronome personale atono qual oggeto. 
Romania 27 (195): 145–146. 
Pancheva, R. (2005). The Rise and Fall of Second–Position Clitics. Natural 
Language and Linguistic Theory 23 (1): 103–167. 
Pappas, P. A. (2001). Weak Object Pronoun Placement in Later Medieval and 
Early Modern Greek. Unpublished PhD Dissertation, The Ohio State 
University. 
Pappas, P. A. (2004). Variation and Morphosyntactic Change in Greek: From 
clitics to affixes. Hampshire: Palgrave Macmillan. 
Pappas, P. A. (2011). A magnitude estimation test for Cypriot Greek clitics. Paper 
presented at GeSuS (Prague, Czech Republic, May 26–28 2011). 
Parodi, T. & I. M. Tsimpli (2005). ‘Real’ and apparent optionality in second 
language grammars: finiteness and pronouns in null operator structures. 
Second Language Research 21 (3): 250–285. 
Pavlou, N. (2009). Flavours of mbu in Cypriot Greek. Paper presented at the 9th 
International Conference of Greek Linguistics, University of Chicago 
(October 29–31, 2009). 
Pavlou, P. & A. Papapavlou (2004). Issues of dialect use in education from the 
Greek Cypriot perspective. International Journal of Applied Linguistics, 
14(2): 243–258. 
! 227 
Petinou, K. & A. Terzi (2002). Clitic Misplacement Among Normally Developing 
Children and Children With Specific Language Impairment and the Status 
of Infl Heads. Language Acquisition 10 (1): 1–28. 
(Philippaki–)Warburton, I. (1977): Modern Greek clitic pronouns and the 
‘surface–structure constraints’ hypothesis. Journal of Linguistics 13: 259–
281. 
Philippaki–Warburton, I. (1987). The theory of empty categories and the pro–drop 
parameter in Modern Greek. Journal of Linguistics 23: 289–318. 
Philippaki–Warburton, I. (1998). Functional categories and Modern Greek syntax. 
The Linguistic Review 15: 159–186. 
Philippaki–Warburton, I. & V. Spyropoulos (1999): On the boundaries of 
inflection and syntax: Greek pronominal clitics and particles. In: Booij, G. 
& J. Van Marle (eds.), The Yearbook of Morphology 1998, 45–72. 
Dordrecht: Kluwer. 
Philippaki–Warburton, I. & V. Spyropoulos (2004). A change of mood: The 
development of the Greek mood system. Linguistics 42, 523–549. 
Piñón, C. (1993). SigmaP and Hungarian. WCCFL 11:388–404. 
Prince, A. & P. Smolensky (1993). Optimality Theory: constraint interaction in 
Generative Grammar. Report no. RuCCS–TR–2, Rutgers University 
Center for Cognitive Science, New Brunswick, NJ. 
Ralli, A. (2006). Syntactic and morphosyntactic phenomena in Modern Greek 
dialects: The state of the art. Journal of Greek Linguistics 7: 121–159. 
Reinhart, T. (1999). The processing cost of reference set computations: guess 
patterns in acquisition. OTS Working papers in Linguistics. 
Revithiadou, A. (2006). Prosodic filters on syntax: an interface account of second 
position clitics. Lingua 116: 79–111. 
Revithiadou, A. (2008). A cross–dialectal study of cliticization in Greek. Lingua 
118: 1393–1415. 
Revithiadou, A. & V. Spyropoulos (2008). Greek object clitic pronouns: A 
typological survey of their grammatical properties. STUF 61, 39–53. 
Rivero, M. L. & A. Terzi (1995). Imperatives, V–Movement and Logical Mood. 
Journal of Linguistics 31 (2), 301–332. 
! 228 
Rizzi, L. (1993/94). Some notes on linguistic theory and language development: 
The case of root infinitives. Language Acquisition 3 (4), 371–393. 
Rizzi, L. (1997). The fine structure of the left periphery. In: Haegeman, L. (ed.), 
Elements of grammar: Handbook in generative syntax, 281–337. 
Dordrecht: Kluwer Academic Publishers. 
Roberts, I. (2010). Agreement and Head Movement: Clitics, Incorporation, and 
Defective Goals. Cambridge, MA: MIT Press.  
Roussou, A. (1991). Nominalized clauses in the syntax of Modern Greek. UCL 
Working Papers 3.  
Roussou, A. (1994). The Syntax of Complementisers. Unpublished PhD 
Dissertation, University College London.  
Roussou, A. (2000). On the left periphery. Modal particles and complementisers. 
Journal of Greek Linguistics 1: 65–94.  
Schaeffer, J. (1997). Direct object scrambling in Dutch and Italian child language. 
UCLA Dissertations in Linguistics 17. 
Schaeffer, J. C. (2000). The acquisition of direct object scrambling and clitic 
placement. Syntax and Pragmatics. Amsterdam/Philadelphia: John 
Benjamins. 
Selkirk, E.O. (1995). The prosodic structure of function words. In: Beckman, J., 
Walsh Dickey, L. and S. Urbanczyk (eds.), University of Massachusetts 
Occasional Papers in Linguistics 18: Papers on Optimality Theory, 439–
469. Amherst: University of Massachusetts. 
Silva, C. (2008). Assimetrias na aquisição de clíticos diferenciados em português 
europeu. Unpublished MA Dissertation. Universidade Nova de Lisboa. 
Simeonidis, C. P. (2006). History of the Cypriot Greek Dialect [in Greek]. 
Nicosia: Holy Monastery of Kykkos Centre of Research. 
Soares, C. (2006). La syntaxe de la peripherie gauche en Portuguais Europeen et 
son acquisition. PhD Dissertation, University of Paris 8.  
Sportiche, D. (1996). Clitic Constructions. In: Rooryck, J. & L. Zaring (eds.) 
Phrase Structure and the Lexicon, 213–276. Dordrecht: Kluwer Academic 
Publishers. 
Stephany, U. (1997). The Acquisition of Greek. In: Slobin, D.I. (ed.), The 
! 229 
crosslinguistic study of language acquisition. Volume 4, 183–334. 
Mahwah, New Jersey: Lawrence Erlbaum Associates.  
Terzi, A. (1999a). Clitic combinations, their hosts and their ordering. Natural 
Language and Linguistic Theory 17: 85–121. 
Terzi, A. (1999b). Cypriot Greek clitics and their positioning restrictions. In: 
Alexiadou, A., Horrocks, G. & M. Stavrou (eds.), Studies in Greek Syntax, 
227–240. Dordrecht: Kluwer Academic Publishers. 
Theophanopoulou–Kontou, D. (1986/1987). Empty categories and clitics in 
Modem Greek: The object case. Glossologia 5/6:41–68. 
Tobler, A. (1875/1912). [Review of] Jules le Coultre, De l’ordre des mots dans 
Crestien de Troyes. Extrait du programme de Pâques 1875 du Collége 
Vitzthum. Dresde 1875. Göttingische gelehrte Anzeigen 2 (34): 1057–
1082. (Republished in: Tobler, A. (ed.) Vermischte Beiträge zur 
französischen Grammatik 5, 395– 415. Leipzig: S. Hirzel.) 
Tsakali, V. & K. Wexler (2003). The Acquisition of Person Case Constraint. 
Paper presented at the 19th International Symposium on Theoretical & 
Applied Linguistics, Aristotle University of Thessaloniki (Thessaloniki, 
Greece, 3–5 April 2009). 
Tsakali, V. & K. Wexler (2004). Why Children Omit Clitics in Some Languages 
but not in Others: New Evidence from Greek. In: van Kampen, J. and S. 
Baauw (eds.), Proceedings of GALA 2003 (Generative Approaches to 
Language Acquisition) Volume 2, 493–504. Utrecht: LOT.   
Tsakali, P. (2006). The Syntax and Acquisition of Pronominal Clitics: a 
Crosslinguistic Study with Special Reference to Modern Greek. 
Unpublished PhD Dissertation, University College London. 
Tsimpli, I. M. (1992/1996). The Prefunctional Stage of First Language 
Acquisition: A crosslinguistic Study. PhD Dissertation, University College 
London. New York / London: Garland. 
Tsimpli, I. M. (1995). Focusing in Modern Greek. In: Kiss, K. É. (ed.), Discourse 
Configurational Languages, 176–206. Oxford: OUP. 
Tsimpli, I. M. (1999). Null operators, clitics and identification: a comparison 
between Greek and English. In: Alexiadou, A., Horrocks, G. & M. Stavrou 
! 230 
(eds.), Studies in Greek Syntax, 241–262. Dordrecht: Kluwer Academic 
Publishers. 
Tsimpli, I. M. & S. Stavrakaki (1999). The effects of a morpho–syntactic deficit 
in the determiner system: The case of a Greek SLI child. Lingua 108 (1): 
31–85. 
Tsiplakou, S., Panagiotidis, P. & K.K. Grohmann (2005). Properties of Cypriot 
Greek wh–question formation. Paper presented at the “7th International 
Conference on Greek Linguistics” (York, UK, 8–10 September 2005). 
Tsoulas, G. (1993). Remarks on the structure and interpretation of na–clauses. 
Studies in Greek Linguistics 14: 191–206. 
Tzakosta, M. (2003). The acquisition of clitics in Greek: A phonological 
perspective. In: Agbayani, B., Koskinen, P. and V. Samiian (eds.), 
Proceedings of the 31st West Coast Conference of Linguistics. Vol. 14. 
Department of Linguistics, California State University at Fresno. 
Tzakosta, M. (2004a). The acquisition of the clitic group in Greek. Proceedings of 
the 24th Annual Meeting of Greek Linguistics, 693–704. Thessaloniki: 
Aristotle University of Thessaloniki. 
Tzakosta, M. (2004b). The enclisis–proclisis asymmetry in the acquisition of 
clitics in Greek. In: Katsimali, G., Kalokairinos, A., Anagnostopoulou, E. 
and I. Kappa (eds.) Proceedings of the 6th International Conference in 
Greek Linguistics. Vol. 1, 91–98. Rethymno: University of Crete. 
Uriagereka, J. (1995). Aspects of the Syntax of Clitic Placement in Western 
Romance. Linguistic Inquiry 26 (1): 79–123. 
Varlokosta, S., Vainikka, A. & B. Rohrbacher (1998). Functional projections, 
markedness, and “root infinitives” in early child Greek. The Linguistic 
Review 15: 187–207. 
Veloudis, I. & I. Philippaki–Warburton (1983). The subjunctive in Modern Greek 
[in Greek]. Studies in Greek Linguistics 4: 151–168. 
Vlachos, C. (2010). Wh–in–situ: The Case of Greek. In Torrens, V., Escobar, L., 
Gavarró, A. and J. Gutiérrez (eds.), Movement and Clitics, 85–113. 
Newcastle–upon–Tyne: Cambridge Scholars Publishing. 
Wexler, K. (1998). Very early parameter setting and the unique checking 
! 231 
constraint: A new explanation of the optional infinitive stage. Lingua 106: 
23–79. 
Wexler, K., Gavarró, A. & V. Torrens (2004). Feature checking and object clitic 
omission in child Catalan and Spanish. In: Bok–Bennema, R., 
Hollebrandse, B., Kampers–Manhe, B. & P. Sleeman (eds.), Romance 
Languages and Linguistic Theory 2002, 253–268. 
Amsterdam/Philadelphia: John Benjamins. 
Yiakoumetti, A., Evans, M. & E. Esch (2005). Language Awareness in a 
Bidialectal Setting: The Oral Performance and Language Attitudes of 
Urban and Rural Students in Cyprus. Language Awareness 14 (4): 
254!260. 
 
 
 
 
 
 
  
 
 
 
 
 
 
 
 
 
 
 
Appendix I: Experimental Study !
!!
232 
Experimental Material: Picture–Based Task 
 
 
CONTENT REMOVED FOR COPYRIGHT REASONS 
 
 
 
 
(1) EXP: Ti theli na kami to koritsaki to kadhro? 
          “What does the girl want to do with the frame?” 
CHI: Na to   vali / kremasi (sto diho). 
 M CL.NEU.ACC  put / hang–3S on–the wall 
 
[Elicited Clause: na–clause]        !!!!!!!!!!!!!!!!!!!!!!!!!!!!!
 
Appendix I: Experimental Study !
!!
233 
CONTENT REMOVED FOR COPYRIGHT REASONS !!!!
(2) EXP: Ti ekamen o shilos tis koruas? 
          “What did the dog do to the girl?” 
CHI: Epkjasen tis   tin klatsa (tis). 
 Took–3S CL.FEM.DAT  the sock POSS 
[Elicited Clause: root clause] 
 
(3) EXP: Ti ekamen i korua tin alin klatsa? 
          “What did the girl do with the other sock?” 
CHI: Eforisen tin. 
         Put on–3S CL.FEM.ACC  
[Elicited Clause: root clause]        
    
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 !
Appendix I: Experimental Study !
!!
234 
CONTENT REMOVED FOR COPYRIGHT REASONS !!!!
(4) EXP: Ti enna kami o papas tu moru? 
          “What is dad going to do to the child?” 
CHI: Enna tu   forisi  to triko (tu). 
          M CL.NEU.DAT  put on–3S the jumper POSS 
[Elicited Clause: enna–clause]        !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
Appendix I: Experimental Study !
!!
235 
CONTENT REMOVED FOR COPYRIGHT REASONS !!!!
(5) EXP: Ti ekamen i mama to pkjato? 
          “What did mum do with the plate?” 
CHI: Evalen  to    pano. 
          Put–3S  CL.NEU.ACC  up 
[Elicited Clause: root clause]        !!
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
Appendix I: Experimental Study !
!!
236 
CONTENT REMOVED FOR COPYRIGHT REASONS !!!!
(6) EXP: Ti theli na kami to pedhaki tin agheladha? 
          “What does the boy want to do with the cow?” 
CHI: Na ti                   vali mesa. 
          M CL.FEM.ACC  put–3S in 
[Elicited Clause: na–clause]        !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
Appendix I: Experimental Study !
!!
237 
CONTENT REMOVED FOR COPYRIGHT REASONS !!!!
(7) EXP: Ti kamni i mama tu moru? 
          “What does mum do with the baby?” 
CHI: Thkjevazi tu                  paramithi. 
          Read–3S CL.NEU.DAT fairy–tale 
[Elicited Clause: root clause]        !!!!!!!!!!!!!!!!!!!!!
 
 
 
 
 
 
 
 
 
Appendix I: Experimental Study !
!!
238 
CONTENT REMOVED FOR COPYRIGHT REASONS 
 
 
 
(8) EXP: Ti enna kami i kiria tu mitsi? 
          “What is the lady going to do to the boy?” 
CHI: Enna tu                   vali / forisi  to paputsi 
M CL.MASC.DAT put on / wear–3S the shoe 
[Elicited Clause: enna–clause]        
 
 
 
 
 
 
 
 
 
 
 
 
Appendix I: Experimental Study !
!!
239 
Answer Sheet 
 
CL: Clitic 
OM: Omission 
NP: Noun Phrase 
N/A: No Answer 
OT: Other!
Nursery School:  School Reference Number:  
Child’s Name: Code:  
Date Of Birth:                                                                                        A: 2;6–3;0 
                                                                              Age Group:        B: 3;0–3;6 
Age: Years; Months. Days:                                                                   C: 3;6–4;0 
PRODUCTION 
 CL OM NP N/A OT 
1. Ti theli na kami to koritsaki to kadhro?      
2. Ti ekamen o shilos tis koruas?      
3. Ti ekamen i korua tin alin klatsa?      
4. Ti ena kami o papas tu moru?       
5. Ti ekamen i mama to pkjato?      
6. Ti theli na kami to pedhaki tin agheladha?      
7. Ti kamni i mama tu moru?      
8. Ti ena kami i kiria tu mitsi?      
PLACEMENT 
Root Clauses Na / Enna–clauses 
2. Correct / Incorrect 
3. Correct / Incorrect 
5. Correct / Incorrect 
7. Correct / Incorrect 
1. Correct / Incorrect 
4. Correct / Incorrect 
6. Correct / Incorrect 
8. Correct / Incorrect 
Overall 
Correct  = 
Incorrect = 
Overall 
Correct  = 
Incorrect = 
Appendix II: Clitic Production (Experimental Data) 240 
Subjects 
 
Age Sex ROOT 
 
MODAL CLAUSES 
TARGET NON–TARGET TARGET NON–TARGET 
CLITIC N/A OTHER OMISSION CLITIC N/A OTHER OMISSION 
AGE GROUP A (2;6–3;0) 
 
P1 2;6.7 = 
2;6 
0 3 1 0 0 1 3 0 0 
P2 2;6.8 = 
2;6 
0 3 0 1 0 3 1 0 0 
P3 2;6.15 = 
2;6 
0 4 0 0 0 2 2 0 0 
P4 2;6.25 = 
2;7 
0 3 0 1 0 4 0 0 0 
P5 2;7.17 = 
2;8 
0 1 0 3 0 0 1 3 0 
P6 2;8 = 
2;8 
1 1 0 3 0 2 1 1 0 
P7 2;8.17 = 
2;9 
0 4 0 0 0 3 0 1 0 
P8 2;8.25 = 
2;9 
0 1 3 0 0 0 3 1 0 
P9 2;9.27 = 
2;10 
1 3 1 0 0 2 1 1 0 
Appendix II: Clitic Production (Experimental Data) 241 
P10 2;10.11 
= 2;10 
1 4 0 0 0 3 1 0 0 
P11 2;10.29 
= 2;11 
0 3 0 1 0 3 1 0 0 
P12 2;11 = 
2;11 
1 1 0 1 1 1 2 0 1 
P13 2;11.7 = 
2;11 
1 4 0 0 0 3 0 0 1 
P14 2;11.9 = 
2;11 
0 3 0 1 0 3 1 0 0 
P15 2;11.23 
= 3;0 
0 2 1 1 0 1 2 0 1 
P16 3;0.4 = 
3;0 
1 2 0 0 2 2 0 2 0 
P17 3;0.11 = 
3;0 
0 2 0 1 1 2 2 0 0 
P18 
 
3;0.13 = 
3;0 
0 4 0 0 0 3 0 1 0 
Overall 2;6–3;0 – 48 6 13 4 38 21 10 3 
AGE GROUP B (3;0–3;6) 
 
P19 3;0.26 = 
3;1 
1 3 0 1 0 3 1 0 0 
Appendix II: Clitic Production (Experimental Data) 242 
P20 3;1.7 = 
3;1 
1 2 1 1 0 4 0 0 0 
P21 3;1.13 = 
3;1 
1 4 0 0 0 2 2 0 0 
P22 3;1.21 = 
3;2 
0 4 0 0 0 2 1 0 1 
P23 3;2.1 = 
3;2 
1 1 0 3 0 1 3 0 0 
P24 3;2.11 = 
3;2 
1 3 0 0 1 4 0 0 0 
P26 3;2.15 = 
3;2 
0 4 0 0 0 4 0 0 0 
P27 3;2.18 = 
3;3 
1 3 0 1 0 1 2 1 0 
P28 3;2.29 = 
3;3 
1 1 0 2 1 3 1 0 0 
P29 3;3.10 = 
3;3 
0 3 1 0 0 0 4 0 0 
P30 3;3.14 = 
3;3 
1 2 0 2 0 1 1 2 0 
P31 3;3.23 = 
3;4 
1 3 0 0 1 2 2 0 0 
P32 3;4.25 = 
3;5 
0 2 1 0 1 1 2 0 1 
Appendix II: Clitic Production (Experimental Data) 243 
P33 3;5.6 = 
3;5 
1 2 0 2 0 4 0 0 0 
P34 3;5.9 = 
3;5 
0 3 0 1 0 3 1 0 0 
P35 3;5.14 = 
3;5 
0 3 0 1 0 3 1 0 0 
P36 3;5.25 = 
3;6 
0 4 0 0 0 4 0 0 0 
P37 3;6.9 = 
3;6 
1 1 0 1 1 3 0 1 0 
P38 3;6.9 = 
3;6 
1 3 0 1 0 3 1 0 0 
P39 3;6.11 = 
3;6 
1 3 0 1 0 4 0 0 0 
P40 3;6.13 = 
3;6 
0 4 0 0 0 3 1 0 0 
P41 
 
3;6.15 = 
3;6 
1 4 0 0 0 3 0 0 1 
Overall 3;0–3;6 – 62 3 17 5 58 23 4 3 
AGE GROUP C (3;6–4;0) 
P42 3;7 1 3 1 0 0 3 1 0 0 
Appendix II: Clitic Production (Experimental Data) 244 
!!!!
 
P43 3;6.29 = 
3;7 
1 3 1 0 0 1 1 2 0 
P44 3;6.22 = 
3;7 
0 4 0 0 0 4 0 0 0 
P45 3;7.8 =  
;7 
0 2 1 1 0 2 2 0 0 
P46 3;7.24 = 
3;8 
1 2 2 0 0 1 3 0 0 
P47 3;8.12 = 
3;8 
0 3 0 1 0 4 0 0 0 
P48 3;8.17 = 
3;9 
0 3 0 1 0 3 0 1 0 
P49 3;8.18 = 
3;9 
0 3 0 1 0 3 0 0 1 
P50 3;9.9 = 
3;9 
1 2 0 1 1 3 1 0 0 
P51 3;10.27 
= 3;11 
0 3 0 0 1 3 0 0 1 
Overall 3;6–4;0 – 28 5 5 2 27 8 3 2 
Appendix III: Clitic Placement (Experimental Data) 
!
245 
Subjects Age Sex ROOT ROOT 
COR 
ROOT 
INCOR 
M M 
COR 
M 
INCOR 
P1 2;6.7 = 
2;6 
0 3 3 0 1 1 0 
P2 2;6.8 = 
2;6 
0 3 3 0 3 3 0 
P3 2;6.15 = 
2;6 
0 4 4 0 2 2 0 
P4 2;6.25 = 
2;7 
0 3 3 0 4 4 0 
P5 2;7.17 = 
2;8 
0 1 1 0 0 0 0 
P6 2;8 = 
2;8 
1 1 1 0 2 0 2 
P7 2;8.17 = 
2;9 
0 4 4 0 3 3 0 
P8 2;8.25 = 
2;9 
0 1 1 0 0 0 0 
P9 2;9.27 = 
2;10 
1 3 3 0 2 0 2 
P10 2;10.11 = 
2;10 
1 4 4 0 3 3 0 
P11 2;10.29 = 
2;11 
0 3 3 0 3 3 0 
P12 2;11 = 
2;11 
1 1 1 0 1 1 0 
Appendix III: Clitic Placement (Experimental Data) 
!
246 
P13 2;11.7 = 
2;11 
1 4 4 0 3 0 3 
P14 2;11.9 = 
2;11 
0 3 3 0 3 0 3 
P15 2;11.23 = 
3;0 
0 2 2 0 1 1 0 
P16 3;0.4 = 
3;0 
1 2 2 0 2 2 0 
P17 3;0.11 = 
3;0 
0 2 2 0 2 2 0 
P18 
 
3;0.13 = 
3;0 
0 4 4 0 3 0 3 
P19 3;0.26 = 
3;1 
1 2 2 0 3 3 0 
P20 3;1.7 = 
3;1 
1 2 2 0 4 4 0 
P21 3;1.13 = 
3;1 
1 4 4 0 2 2 0 
P22 3;1.21 = 
3;2 
0 4 4 0 2 0 2 
P23 3;2.1 = 
3;2 
1 1 1 0 1 1 0 
P24 3;2.11 = 
3;2 
1 3 3 0 4 4 0 
Appendix III: Clitic Placement (Experimental Data) 
!
247 
P26 3;2.15 = 
3;2 
0 4 4 0 4 4 0 
P27 3;2.18 = 
3;3 
1 3 3 0 1 0 1 
P28 3;2.29 = 
3;3 
1 1 1 0 3 3 0 
P29 3;3.10 = 
3;3 
0 3 2 1 0 0 0 
P30 3;3.14 = 
3;3 
1 2 2 0 1 1 0 
P31 3;3.23 = 
3;4 
1 3 3 0 2 2 0 
P32 3;4.25 = 
3;5 
0 2 2 0 1 1 0 
P33 3;5.6 = 
3;5 
1 3 3 0 4 4 0 
P34 3;5.9 = 
3;5 
0 3 3 0 3 3 0 
P35 3;5.14 = 
3;5 
0 3 3 0 3 3 0 
P36 3;5.25 = 
3;6 
0 4 4 0 4 4 0 
P37 3;6.9 = 
3;6 
1 1 1 0 3 3 0 
Appendix III: Clitic Placement (Experimental Data) 
!
248 
P38 3;6.9 = 
3;6 
1 3 3 0 3 3 0 
P39 3;6.11 = 
3;6 
1 3 3 0 4 4 0 
P40 3;6.13 = 
3;6 
0 4 4 0 3 3 0 
P41 
 
3;6.15 = 
3;6 
1 4 4 0 3 3 0 
P42 3;7 1 3 3 0 3 3 0 
P43 3;6.29 = 
3;7 
1 3 3 0 1 1 0 
P44 3;6.22 = 
3;7 
0 4 4 0 4 4 0 
P45 3;7.8 =  
;7 
0 2 2 0 2 2 0 
P46 3;7.24 = 
3;8 
1 2 2 0 1 1 0 
P47 3;8.12 = 
3;8 
0 3 3 0 4 4 0 
P48 3;8.17 = 
3;9 
0 3 3 0 3 3 0 
P49 3;8.18 = 
3;9 
0 3 3 0 3 3 0 
Appendix III: Clitic Placement (Experimental Data) 
!
249 
 P50 3;9.9 = 
3;9 
1 2 2 0 3 3 0 
P51 3;10.27 = 
3;11 
0 3 3 0 3 3 0