Agreement in Modern Greek Coordinate Noun Phrases
Agreement in Modern Greek Coordinate Noun Phrases
Agreement in Modern Greek Coordinate Noun Phrases
Create successful ePaper yourself
Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.
<strong>Agreement</strong> <strong>in</strong> <strong>Modern</strong> <strong>Greek</strong> Coord<strong>in</strong>ate<br />
<strong>Noun</strong> <strong>Phrases</strong><br />
Desp<strong>in</strong>a Kazana<br />
A thesis submitted for the degree of Doctor of Philosophy<br />
Department of Language and L<strong>in</strong>guistics<br />
University of Essex<br />
June, 2011
Abstract<br />
A central issue posited by those who have studied agreement with coord<strong>in</strong>ate nouns is<br />
that the <strong>in</strong>dividual conjuncts have different agreement features from the coord<strong>in</strong>ate phrase<br />
(Dalrymple and Kaplan, 2000). This thesis explores agreement <strong>in</strong> MG coord<strong>in</strong>ate nouns<br />
with<strong>in</strong> the constra<strong>in</strong>t-based framework of Lexical-Functional Grammar. It <strong>in</strong>vestigates<br />
predicate-argument and head-modifier agreement with coord<strong>in</strong>ate nouns and agreement<br />
with disjunctively conjo<strong>in</strong>ed nouns.<br />
In predicate-argument agreement, the syntactic and semantic (Corbett, 1991; Wechsler,<br />
2009; Sadler, 2006) agreement pr<strong>in</strong>ciples occur crossl<strong>in</strong>guistically. I argue that a referential<br />
pr<strong>in</strong>ciple also exists, motivated by a Contextually Introduced Referent implied <strong>in</strong> the<br />
sentence. CIR is the superord<strong>in</strong>ate term of the coord<strong>in</strong>ate nouns and the predicative adjective/participle<br />
copies its agreement features. Thus, I propose an LFG syntactic account for<br />
referential agreement. In head-modifier agreement, MG modifiers behave different from the<br />
def<strong>in</strong>ite determ<strong>in</strong>er. I argue that a shared modifier shows concord agreement, it scopes<br />
over coord<strong>in</strong>ate nouns and occurs with the jo<strong>in</strong>t and split <strong>in</strong>terpretations. A shared determ<strong>in</strong>er<br />
shows <strong>in</strong>dex/concord agreement, it scopes over coord<strong>in</strong>ate nouns and occurs only<br />
with the jo<strong>in</strong>t <strong>in</strong>terpretation. The def<strong>in</strong>ite determ<strong>in</strong>er allows some exceptional patterns <strong>in</strong><br />
plural and occurs with a split read<strong>in</strong>g when nouns are a natural coord<strong>in</strong>ation. I capture natural<br />
coord<strong>in</strong>ation patterns follow<strong>in</strong>g the standard analysis on coord<strong>in</strong>ation. In agreement<br />
with disjunctive coord<strong>in</strong>ate nouns, I argue that native speakers choose an agree<strong>in</strong>g verb<br />
based on the <strong>in</strong>terpretation of disjunction as ‘and-coord<strong>in</strong>ate’ and as ‘exclusive-or’. I capture<br />
this follow<strong>in</strong>g Eggert’s analysis which treats or as a subset function, us<strong>in</strong>g lamda-DRT,<br />
and I <strong>in</strong>troduce a special lexical entry for or which can be <strong>in</strong>terpreted either way. However,<br />
I show that verb agreement is also determ<strong>in</strong>ed by other factors, such as the features of<br />
<strong>in</strong>dividual conjuncts, the type of sentence and the subject position <strong>in</strong> the sentence.
Acknowledgements<br />
This thesis would never have been completed without the help and support of a number of<br />
people. The first person I am deeply grateful to is my supervisor Professor Louisa Sadler.<br />
Her encouragement and support was really valuable to me and kept me go<strong>in</strong>g through the<br />
hard times all these years. Louisa has been an endless source of <strong>in</strong>spiration from many<br />
different aspects. Her deep knowledge <strong>in</strong> l<strong>in</strong>guistics and her ability to come across her<br />
ideas is unique. She taught me so many th<strong>in</strong>gs and answered countless questions through<br />
the time we worked together and she was always there to listen and advise me when I<br />
encountered difficulties. She supported me <strong>in</strong> evey way she could, not only on l<strong>in</strong>guistic<br />
issues but also on personal issues. Even towards the end when th<strong>in</strong>gs became really hard,<br />
she believed <strong>in</strong> me and motivated me to stick to my goal and not give up.<br />
I would also like to thank Doug Arnold, who was there to help me any time of the<br />
day, especially with my LaTex problems. Also, many thanks go to Andrew Spencer, Bob<br />
Borsley, who advised me and guided me <strong>in</strong> the supervisory board meet<strong>in</strong>gs, and Mary<br />
Dalrymple and Ash Asudeh who provided me with some <strong>in</strong>sightful comments <strong>in</strong> LFG09.<br />
My special thanks go to Maria Flouraki for be<strong>in</strong>g a good friend, for shar<strong>in</strong>g with me her<br />
passion <strong>in</strong> l<strong>in</strong>guistics, for her encouragement and for the great collaboration we had, which<br />
I hope will cont<strong>in</strong>ue.<br />
Many thanks go to a number of people <strong>in</strong> the department of Language and L<strong>in</strong>guistics<br />
at Essex University. In particular, I thank the staff and the members of the Constra<strong>in</strong>tbased<br />
L<strong>in</strong>guistics Discussion Group: Ryo Otoguro, Takafumi Maekawa, Antonis Polentas,<br />
Miriam Urgelles and Kakia Chatsiou. Also, my deep appreciation goes to Wyn Johnson<br />
for giv<strong>in</strong>g me the opportunity to work as a Graduate Teach<strong>in</strong>g Assistant and improve my<br />
teach<strong>in</strong>g skills <strong>in</strong> the field of l<strong>in</strong>guistics.<br />
My PhD research has been supported f<strong>in</strong>ancially from the Economic & Social Research<br />
Council, British Government. I am grateful to them s<strong>in</strong>ce they funded my work all these<br />
years and they also gave me the opportunity to participate <strong>in</strong> a number of conferences<br />
abroad.<br />
Last but not least, I would like to thank my family. Many thanks go to my parents<br />
and my sister for their support especially towards the end. My deepest thanks go to my<br />
partner and husband Apostolos P<strong>in</strong>akidis who believed <strong>in</strong> me from the very beg<strong>in</strong>n<strong>in</strong>g,<br />
urged me to start this difficult ‘journey’ and supported me <strong>in</strong> every aspect from the start.<br />
His motivat<strong>in</strong>g words and strong encouragement brought me back when I felt lost and gave<br />
me the courage to f<strong>in</strong>ish what I started. Though, I am mostly grateful to him for be<strong>in</strong>g a<br />
‘mother’ apart from a father to our baby son. I also need to thank him for his technical<br />
support and feedback and for patiently answer<strong>in</strong>g data questions, even though he found it<br />
tir<strong>in</strong>g at times. ..<br />
F<strong>in</strong>ally, I would like to say my deepest thanks to my baby son Manthos who ‘occasionally’<br />
showed some patience when see<strong>in</strong>g me work<strong>in</strong>g long hours <strong>in</strong> front of the computer.<br />
I would like to say a big sorry to him for not be<strong>in</strong>g there most of the time to look after<br />
him as I should. His smile and his hugs were the most important reason to f<strong>in</strong>ish what I<br />
started and look forward to what is com<strong>in</strong>g up next...
To my partner and son,<br />
Apostolos and Manthos P<strong>in</strong>akidis<br />
iii
List of Abbreviations<br />
MG = <strong>Modern</strong> <strong>Greek</strong><br />
pers = person<br />
gend = gender<br />
semgend = semantic gender<br />
num = number<br />
1 = first person<br />
2 = second person<br />
3 = third person<br />
masc = mascul<strong>in</strong>e<br />
fem = fem<strong>in</strong><strong>in</strong>e<br />
neut = neuter<br />
sg = s<strong>in</strong>gular<br />
pl = plural<br />
nom = nom<strong>in</strong>ative<br />
gen = genitive<br />
acc = accusative<br />
anim = animate<br />
CCA = Closest Conjunct <strong>Agreement</strong><br />
CIR = Contextually Introduced Referent
Contents<br />
1 Introduction 1<br />
1.1 Overview of the thesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3<br />
2 LFG: An overview 5<br />
2.1 Constituent-structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6<br />
2.2 Functional-Structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8<br />
2.2.1 C-structure and f-structure correspondence . . . . . . . . . . . . . . 10<br />
2.3 Semantic-Structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14<br />
2.4 <strong>Agreement</strong> <strong>in</strong> LFG . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17<br />
2.4.1 Coord<strong>in</strong>ation <strong>in</strong> LFG . . . . . . . . . . . . . . . . . . . . . . . . . . . 19<br />
2.4.1.1 Distributive vs Nondistributive Features . . . . . . . . . . . 20<br />
2.4.1.2 Predicate Coord<strong>in</strong>ation . . . . . . . . . . . . . . . . . . . . 21<br />
2.4.1.3 <strong>Noun</strong> Phrase Coord<strong>in</strong>ation . . . . . . . . . . . . . . . . . . 23<br />
2.4.2 The Semantics of Coord<strong>in</strong>ation . . . . . . . . . . . . . . . . . . . . . 25<br />
2.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28<br />
3 Theories of <strong>Agreement</strong> 29<br />
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29<br />
3.2 Theoretical Approaches to Predicate-Argument <strong>Agreement</strong> . . . . . . . . . 29<br />
3.2.1 Corbett’s Descriptive Approach . . . . . . . . . . . . . . . . . . . . . 29<br />
3.2.2 Dalrymple and Kaplan’s Set-Based Theory . . . . . . . . . . . . . . 32<br />
3.2.3 Wechsler’s Theory of Gender Resolution . . . . . . . . . . . . . . . . 36<br />
3.2.4 Sadler’s Theory of Resolution <strong>in</strong> Rumanian . . . . . . . . . . . . . . 41<br />
3.2.5 Badecker’s Optimality Theory . . . . . . . . . . . . . . . . . . . . . . 45<br />
3.3 Theoretical Approaches to Head-Modifier <strong>Agreement</strong> . . . . . . . . . . . . . 51<br />
3.3.1 <strong>Agreement</strong> <strong>in</strong> HPSG . . . . . . . . . . . . . . . . . . . . . . . . . . . 52<br />
3.3.2 K<strong>in</strong>g and Dalrymple’s Theory . . . . . . . . . . . . . . . . . . . . . . 54<br />
3.3.3 Dalrymple and Nikolaeva’s theory of natural and accidental coord<strong>in</strong>ation<br />
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61<br />
3.4 Discuss<strong>in</strong>g the theoretical approaches to agreement . . . . . . . . . . . . . . 66<br />
4 Predicate Argument <strong>Agreement</strong> <strong>in</strong> <strong>Modern</strong> <strong>Greek</strong> 70<br />
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70<br />
4.2 The <strong>Modern</strong> <strong>Greek</strong> Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70<br />
4.2.1 Resolution <strong>in</strong> MG Animate <strong>Noun</strong>s . . . . . . . . . . . . . . . . . . . 72<br />
4.2.2 Resolution <strong>in</strong> MG Inanimate <strong>Noun</strong>s . . . . . . . . . . . . . . . . . . 79<br />
4.2.3 Mix<strong>in</strong>g Animate and Inanimate <strong>Noun</strong>s <strong>in</strong> Coord<strong>in</strong>ate Constructions 87<br />
4.3 Resolution Pr<strong>in</strong>ciples <strong>in</strong> MG . . . . . . . . . . . . . . . . . . . . . . . . . . . 88<br />
4.3.1 Semantic Resolution <strong>in</strong> Animate <strong>Noun</strong>s . . . . . . . . . . . . . . . . 88<br />
4.3.2 Syntactic Resolution <strong>in</strong> Inanimate <strong>Noun</strong>s . . . . . . . . . . . . . . . 94<br />
4.4 Reference Transfer <strong>in</strong> Noncoord<strong>in</strong>ate Constructions . . . . . . . . . . . . . . 97
CONTENTS vi<br />
4.4.1 Reference Transfer <strong>in</strong> <strong>Modern</strong> <strong>Greek</strong> . . . . . . . . . . . . . . . . . . 100<br />
4.5 An Approach to Gender <strong>Agreement</strong> <strong>in</strong> MG . . . . . . . . . . . . . . . . . . 102<br />
4.5.1 Problems of the Current Theories for the MG data . . . . . . . . . . 102<br />
4.5.2 An Analysis of the MG Expected Patterns . . . . . . . . . . . . . . . 106<br />
4.5.3 An Approach to the MG Referential <strong>Agreement</strong> Patterns . . . . . . 109<br />
4.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115<br />
5 Head Modifier <strong>Agreement</strong> <strong>in</strong> <strong>Modern</strong> <strong>Greek</strong> 117<br />
5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117<br />
5.2 NP Internal <strong>Agreement</strong> and Interpretation of the Coord<strong>in</strong>ate <strong>Noun</strong>s . . . . 118<br />
5.3 NP Internal <strong>Agreement</strong> <strong>in</strong> <strong>Modern</strong> <strong>Greek</strong> . . . . . . . . . . . . . . . . . . . 121<br />
5.4 The <strong>Modern</strong> <strong>Greek</strong> Modifiers . . . . . . . . . . . . . . . . . . . . . . . . . . 121<br />
5.4.1 concord agreement with the MG Modifiers . . . . . . . . . . . . . 127<br />
5.5 The <strong>Modern</strong> <strong>Greek</strong> Def<strong>in</strong>ite Determ<strong>in</strong>er . . . . . . . . . . . . . . . . . . . . 135<br />
5.5.1 Uniform Number and Gender <strong>in</strong> Coord<strong>in</strong>ate <strong>Noun</strong>s . . . . . . . . . . 135<br />
5.5.2 Uniform Number and Different Gender <strong>in</strong> Coord<strong>in</strong>ate <strong>Noun</strong>s . . . . 140<br />
5.5.2.1 Motivation for the Exceptional Coord<strong>in</strong>ate Constructions . 142<br />
5.5.3 <strong>in</strong>dex/concord agreement with the MG Def<strong>in</strong>ite Determ<strong>in</strong>er . . . 145<br />
5.5.4 Natural Coord<strong>in</strong>ation <strong>in</strong> <strong>Modern</strong> <strong>Greek</strong> . . . . . . . . . . . . . . . . 148<br />
5.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153<br />
6 Disjunctively Conjo<strong>in</strong>ed <strong>Noun</strong> <strong>Phrases</strong> 155<br />
6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155<br />
6.2 Inclusive vs Exclusive or . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 156<br />
6.2.1 Natural Language or . . . . . . . . . . . . . . . . . . . . . . . . . . . 161<br />
6.3 Approaches to Disjunctive Coord<strong>in</strong>ated NPs . . . . . . . . . . . . . . . . . . 162<br />
6.3.1 Morgan’s Analysis of Disjunctive NPs . . . . . . . . . . . . . . . . . 162<br />
6.3.2 Peterson’s Analysis of Disjunctive NPs . . . . . . . . . . . . . . . . . 164<br />
6.3.3 An Intermodular Theory of Disjunction . . . . . . . . . . . . . . . . 168<br />
6.3.4 Discuss<strong>in</strong>g the three approaches . . . . . . . . . . . . . . . . . . . . . 175<br />
6.4 Verb <strong>Agreement</strong> <strong>in</strong> MG Disjunctive Coord<strong>in</strong>ate <strong>Noun</strong>s . . . . . . . . . . . . 176<br />
6.4.1 Interpretation and Verb <strong>Agreement</strong> <strong>in</strong> Disjunctive Coord<strong>in</strong>ate <strong>Noun</strong>s 177<br />
6.4.2 Syntax and Verb <strong>Agreement</strong> <strong>in</strong> Disjunctive Coord<strong>in</strong>ate <strong>Noun</strong>s . . . . 181<br />
6.4.3 Speakers’ Strategies and Verb <strong>Agreement</strong> <strong>in</strong> MG Disjunctive <strong>Noun</strong>s 186<br />
6.5 Issues on the MG positive & negative predisjunction-disjunction . . . . . . 191<br />
6.6 Analys<strong>in</strong>g Verb <strong>Agreement</strong> <strong>in</strong> Disjunctive Coord<strong>in</strong>ate <strong>Noun</strong>s . . . . . . . . . 196<br />
6.7 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 202<br />
7 Conclud<strong>in</strong>g Remarks 204<br />
A Questionnaires on Predicate Argument <strong>Agreement</strong> 207<br />
A.1 Discussion of questionnnaires . . . . . . . . . . . . . . . . . . . . . . . . . . 207<br />
A.2 Predicate Argument <strong>Agreement</strong> sets of examples . . . . . . . . . . . . . . . 208<br />
A.3 Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 209<br />
B Questionnaires on agreement with Disjunctively Conjo<strong>in</strong>ed <strong>Noun</strong>s 213<br />
B.1 Discussion of questionnnaires . . . . . . . . . . . . . . . . . . . . . . . . . . 213<br />
B.2 Disjunctive Coord<strong>in</strong>ate <strong>Noun</strong>s sets of examples . . . . . . . . . . . . . . . . 214<br />
B.3 Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 215<br />
C Orig<strong>in</strong>al Questionnaires <strong>in</strong> <strong>Greek</strong> 220
Chapter 1<br />
Introduction<br />
The objective of this thesis is to provide an analysis of agreement phenomena <strong>in</strong> <strong>Modern</strong><br />
<strong>Greek</strong> 1 coord<strong>in</strong>ate noun phrases jo<strong>in</strong>ed by the conjunction ke ‘and’ and the disjunction i ‘or’.<br />
We will focus on predicate-argument agreement, which is agreement of the coord<strong>in</strong>ate noun<br />
phrase with the predicate of the sentence (i.e. verb or predicative adjective/participle), and<br />
head-modifier agreement, which is agreement of modifiers (i.e. determ<strong>in</strong>ers, attributive<br />
adjectives) with the coord<strong>in</strong>ate noun phrase. Also, we will discuss verb agreement with<br />
disjunctively conjo<strong>in</strong>ed nouns.<br />
A number of features are <strong>in</strong>volved <strong>in</strong> agreement, which vary from language to language.<br />
MG predicate-argument agreement <strong>in</strong>volves the features gender, number and person. A<br />
s<strong>in</strong>gle or a coord<strong>in</strong>ate noun phrase shows agreement with the verb <strong>in</strong> number and person<br />
while it shows agreement with the predicative adjective/participle <strong>in</strong> gender and number.<br />
The follow<strong>in</strong>g example illustrates the different types of agreement with a noncoord<strong>in</strong>ate<br />
noun:<br />
(1) O<br />
Kostas<br />
the.masc.sg.3 Kostas.masc.sg.3<br />
‘Kostas is very active’<br />
<strong>in</strong>e<br />
is.sg.3<br />
poli<br />
very<br />
drastirios<br />
active.masc.sg<br />
In most languages, agreement with coord<strong>in</strong>ate noun phrases is <strong>in</strong>terest<strong>in</strong>g s<strong>in</strong>ce the<br />
coord<strong>in</strong>ate nouns usually have different agreement features from the features of the agree<strong>in</strong>g<br />
verb or predicative adjective/participle. For example, <strong>in</strong> MG coord<strong>in</strong>ate phrases conjuncts<br />
with a masc and a fem gender show masc agreement, conjuncts with a 1st and a 2nd<br />
person show 1st person agreement, while two s<strong>in</strong>g conjuncts show pl agreement. Consider<br />
the follow<strong>in</strong>g examples:<br />
(2) O antras ke<br />
the.masc.sg<br />
xarumeni<br />
happy.masc.pl<br />
man.masc.sg and<br />
‘The man and woman are happy’<br />
(3) Ego ke esi tha<br />
I.1.sg and you.2.sg will<br />
‘Me and you will meet at the house’<br />
i<br />
the.fem.sg<br />
vrethume<br />
meet.2.pl<br />
j<strong>in</strong>eka<br />
woman.fem.sg<br />
sto spiti<br />
at-the house<br />
<strong>in</strong>e<br />
are.pl<br />
In MG predicate-argument agreement with coord<strong>in</strong>ate nouns, we focus on the gender<br />
feature. Most l<strong>in</strong>guists acknowledge the existence of two different pr<strong>in</strong>ciples <strong>in</strong> gender<br />
agreement with coord<strong>in</strong>ate nouns, the syntactic and the semantic (Corbett, 1991; Wechsler<br />
and Zlatić, 2003; Dalrymple and Kaplan, 2000; Sadler, 2006). We will argue that MG<br />
1 We will refer to <strong>Modern</strong> <strong>Greek</strong> with the <strong>in</strong>itials MG from now on.
gender agreement with coord<strong>in</strong>ate noun phrases shows great variation due to an extra<br />
type of pr<strong>in</strong>ciple. MG admits syntactic agreement, semantic agreement, and a third type,<br />
which we will call referential agreement. Therefore, we will attempt to propose a syntacticsemantic<br />
analysis for the first two types of agreement and a syntactic analysis for referential<br />
agreement, although it is contextually determ<strong>in</strong>ed.<br />
MG head-modifier agreement <strong>in</strong>volves the features number, gender and case. A<br />
determ<strong>in</strong>er or attributive adjective agrees with a s<strong>in</strong>gle noun <strong>in</strong> all three features:<br />
(4) O<br />
Janis<br />
the.masc.sg.nom<br />
antras<br />
man.masc.sg.nom<br />
John.masc.sg.nom<br />
‘John is an active man’<br />
<strong>in</strong>e<br />
is<br />
drastirios<br />
active.masc.sg.nom<br />
In coord<strong>in</strong>ate noun phrases similar requirements hold. A shared attributive adjective<br />
always agrees with the closest conjunct <strong>in</strong> gender, irrespective of the fact that the two<br />
nouns may have the same or different gender features, and with both conjuncts <strong>in</strong> case and<br />
number. A shared determ<strong>in</strong>er agrees with both conjuncts <strong>in</strong> gender, case and number.<br />
These possibilities are shown below:<br />
(5) Oreo<br />
nice.neut.sg.nom<br />
‘Nice house and car’<br />
spiti<br />
house.neut.sg.nom<br />
(6) I<br />
fili<br />
the.masc.pl.nom<br />
mu<br />
my<br />
friends.masc.pl.nom<br />
‘My friends and colleagues’<br />
ke<br />
and<br />
ke<br />
and<br />
aftok<strong>in</strong>ito<br />
car.neut.sg.nom<br />
s<strong>in</strong>adelfi<br />
colleagues.masc.pl.nom<br />
In MG head-modifier agreement with coord<strong>in</strong>ate nouns, we discuss the number feature<br />
and the gender feature. K<strong>in</strong>g and Dalrymple (2004) proposed three different agreement<br />
systems (<strong>in</strong> concord, <strong>in</strong>dex and concord/<strong>in</strong>dex) that determ<strong>in</strong>e the number feature<br />
<strong>in</strong> head-modifier agreement <strong>in</strong> a number of languages. In MG, we will argue that the<br />
def<strong>in</strong>ite determ<strong>in</strong>er is captured by the concord/<strong>in</strong>dex system and it also occurs with a<br />
restricted set of structures that have the characteristics of natural coord<strong>in</strong>ation (Dalrymple<br />
and Nikolaeva, 2006). The rest of the modifiers are captured by the concord system. We<br />
will assume different analyses for the determ<strong>in</strong>er and the rest of the modifiers, while we will<br />
assume the standard analysis of accidental coord<strong>in</strong>ation for the restricted set of structures<br />
of natural coord<strong>in</strong>ation admitted by the determ<strong>in</strong>er.<br />
MG disjunctive coord<strong>in</strong>ate noun phrases will also be discussed <strong>in</strong> the current thesis.<br />
The ma<strong>in</strong> issues posited by those who have studied disjunctive coord<strong>in</strong>ate nouns (Morgan,<br />
1972, 1984, 1985; Peterson, 1986; Jenn<strong>in</strong>gs, 1994; Eggert, 2002; Alonso-Ovalle, 2006) are<br />
the semantics of disjunction as a truth-functional <strong>in</strong>clusive or and a truth-functional exclusive<br />
or, number verb agreement, which is either sg or pl, and the factors that determ<strong>in</strong>e<br />
it. Our focus is on number verb agreement. In most languages <strong>in</strong>clud<strong>in</strong>g MG, number<br />
predicate agreement with disjunctive conjo<strong>in</strong>ed nouns varies greatly. Follow<strong>in</strong>g Morgan<br />
(1985), we will argue that <strong>in</strong> MG the most crucial factor that determ<strong>in</strong>es verb agreement is<br />
the <strong>in</strong>terpretation of the disjunctive coord<strong>in</strong>ate phrase either as an ‘and-coord<strong>in</strong>ate’ result<strong>in</strong>g<br />
<strong>in</strong> plural verb agreement, or as an ‘exclusive-or’ result<strong>in</strong>g <strong>in</strong> s<strong>in</strong>gular verb agreement.<br />
Native speakers assign either the one or the other <strong>in</strong>terpretation to the coord<strong>in</strong>ate phrase<br />
and this <strong>in</strong>terpretation plays a crucial role <strong>in</strong> the choice of the agree<strong>in</strong>g verb. In the follow<strong>in</strong>g<br />
example, speakers choose either a sg verb or a pl verb. The first form is seen as a<br />
result of hav<strong>in</strong>g to choose only one of the two dr<strong>in</strong>ks that are available and therefore the<br />
2
1.1 Overview of the thesis 3<br />
emphasis is given on the choice of one of the two dr<strong>in</strong>ks. The second form is the result of<br />
the fact that two different dr<strong>in</strong>ks are available after the meal and therefore the emphasis<br />
is on the availability of two different dr<strong>in</strong>ks:<br />
(7) Kafes i tsai servir-ete/-onte<br />
coffee or tea is.sg/are.pl-served<br />
‘Coffee or tea is/are served after the d<strong>in</strong>ner’<br />
meta<br />
after<br />
to<br />
the<br />
gevma<br />
d<strong>in</strong>ner<br />
We will argue, however, that agreement ‘by read<strong>in</strong>g’ is one strategy alongside a number<br />
of others <strong>in</strong> MG disjunctive coord<strong>in</strong>ate phrases. Verb agreement is alternatively determ<strong>in</strong>ed<br />
by syntactic factors, such as the presence or absence of a predicative adjective/participle <strong>in</strong><br />
the coord<strong>in</strong>ate phrase, the type of sentence (i.e declarative, <strong>in</strong>terrogative), or the speaker’s<br />
strategies, such as pl w<strong>in</strong>s, prox and fc w<strong>in</strong>s (Peterson, 1986). Our proposal adopts<br />
the λ-DRT framework and captures successfully the two different <strong>in</strong>terpretations assigned<br />
to disjunctive coord<strong>in</strong>ate nouns. However, it captures only phenomena of <strong>in</strong>terpretation<br />
whereas a complete account should extend to all the aforementioned factors <strong>in</strong> order to<br />
cover the wide range of data.<br />
F<strong>in</strong>ally, it is <strong>in</strong>terest<strong>in</strong>g to note that agreement <strong>in</strong> noun phrase coord<strong>in</strong>ate structures <strong>in</strong><br />
MG has not been discussed <strong>in</strong> traditional handbooks and descriptive grammars. Most handbooks<br />
and grammars (Clairis et al. (2004), Holton et al. (1997), Triantaphyllidis (1994),<br />
Triantaphyllidis (2005), Joseph and Philippaki-Warburton (1987) and Mackridge (1985))<br />
present basic issues of the conjunction ke ‘and’ and the disjunction i such as the type of<br />
phrases they coord<strong>in</strong>ate (Mackridge (1985), Joseph and Philippaki-Warburton (1987)) or<br />
the fact that agreement is necessarilly plural <strong>in</strong> coord<strong>in</strong>ate structures (Holton et al. (1997),<br />
Triantaphyllidis (1994), Triantaphyllidis (2005)). The central phenomena of referential<br />
agreement, natural coord<strong>in</strong>ation and <strong>in</strong>terpretation <strong>in</strong> disjunctive coord<strong>in</strong>ate nouns <strong>in</strong> MG<br />
are not found <strong>in</strong> any of the traditional grammars so our research aims to shed some light<br />
<strong>in</strong>to new l<strong>in</strong>guistic areas <strong>in</strong> MG.<br />
1.1 Overview of the thesis<br />
The organisation of the thesis is as follows. The first chapter (i.e. Chapter 2) presents<br />
an overview of Lexical Functional Grammar summaris<strong>in</strong>g the basic ideas on constituentstructure,<br />
functional-structure, the correspondence between the two, and semantic-structure.<br />
We will particularly focus on how agreement and coord<strong>in</strong>ation phenomena are treated<br />
with<strong>in</strong> LFG. We present the syntactic and the semantic treatment of sentential coord<strong>in</strong>ation,<br />
predicate coord<strong>in</strong>ation and noun phrase coord<strong>in</strong>ation.<br />
Chapter 3 presents an overview of the most central theories of agreement. We will review<br />
theoretical approaches to predicate-argument agreement that discuss the gender feature<br />
<strong>in</strong> coord<strong>in</strong>ate nouns. These are Corbett (1983b, 1991), a descriptive approach to gender<br />
resolution, and Dalrymple and Kaplan (2000), a syntactic approach to gender resolution.<br />
Wechsler and Zlatić (2000) and Sadler (2006), who propose a syntactic-semantic analysis<br />
for gender resolution with<strong>in</strong> LFG, and Badecker (2008) a theory of gender agreement <strong>in</strong><br />
Optimality Theory. Next, we revisit the theories <strong>in</strong> head-modifier agreement that discuss<br />
the number feature <strong>in</strong> coord<strong>in</strong>ate nouns. These <strong>in</strong>clude Kathol (1999) and Wechsler and<br />
Zlatić (2000) with<strong>in</strong> HPSG; K<strong>in</strong>g and Dalrymple (2004), which presents three different<br />
agreement systems <strong>in</strong> NP <strong>in</strong>ternal agreement, and Dalrymple and Nikolaeva (2006), which<br />
deals with natural coord<strong>in</strong>ation phenomena <strong>in</strong> a number of languages with<strong>in</strong> LFG.<br />
Chapter 4 describes gender resolution <strong>in</strong> MG coord<strong>in</strong>ate noun phrases. We will discuss<br />
animate and <strong>in</strong>animate nouns and we will show that three ma<strong>in</strong> pr<strong>in</strong>ciples control<br />
gender agreement <strong>in</strong> MG animate and <strong>in</strong>animate nouns, the syntactic pr<strong>in</strong>ciple <strong>in</strong> <strong>in</strong>animate<br />
nouns, the semantic pr<strong>in</strong>ciple <strong>in</strong> animate nouns and the referential pr<strong>in</strong>ciple <strong>in</strong> both
1.1 Overview of the thesis 4<br />
groups of nouns. For the syntactic and semantic resolution pr<strong>in</strong>ciples, we will propose an<br />
analysis that will comb<strong>in</strong>e both syntax and semantics <strong>in</strong> the resolution process. For the<br />
referential agreement pr<strong>in</strong>ciple, we propose a prelim<strong>in</strong>ary analysis with<strong>in</strong> the framework of<br />
LFG by assum<strong>in</strong>g an extra level of representation, the Contextual level C, mapped to the<br />
syntactic level through the <strong>in</strong>verse mapp<strong>in</strong>g function d −1 , and we <strong>in</strong>troduce the Contextually<br />
Introduced Referent that accounts effectively for the unexpected gender agreement<br />
form found <strong>in</strong> certa<strong>in</strong> noun phrase coord<strong>in</strong>ations.<br />
Chapter 5 discusses ma<strong>in</strong>ly number agreement and gender agreement <strong>in</strong> head-modifier<br />
coord<strong>in</strong>ate noun phrases. A shared MG def<strong>in</strong>ite determ<strong>in</strong>er allows only a jo<strong>in</strong>t (i.e. when<br />
the coord<strong>in</strong>ate nouns refer to the same <strong>in</strong>dividual/entity) <strong>in</strong>terpretation and it is accounted<br />
for by the <strong>in</strong>dex/concord system. A shared MG modifier allows a jo<strong>in</strong>t and a split (i.e.<br />
when the coord<strong>in</strong>ate nouns refer to different <strong>in</strong>dividuals/entities) <strong>in</strong>terpretation and it is<br />
accounted for by the concord system. A number of data, however, show an unexpected<br />
behaviour, such as the case where a split <strong>in</strong>terpretation is admitted when a shared MG<br />
plural def<strong>in</strong>ite determ<strong>in</strong>er modifies two plural conjuncts. We will argue that these patterns<br />
must be treated as special cases of natural coord<strong>in</strong>ation but we will analyse them assum<strong>in</strong>g<br />
the f-structure of ord<strong>in</strong>ary coord<strong>in</strong>ate nouns s<strong>in</strong>ce MG behaves different and the f-structure<br />
of natural coord<strong>in</strong>ation (Dalrymple and Nikolaeva, 2006) cannot capture the MG patterns.<br />
Chapter 6 presents disjunctive coord<strong>in</strong>ate nouns <strong>in</strong> MG. Verb agreement varies greatly<br />
<strong>in</strong> these constructions s<strong>in</strong>ce it can be either sg or pl. Our data will show that different<br />
factors determ<strong>in</strong>e verb agreement, such as syntactic, semantic, pragmatic (i.e. contextual<br />
<strong>in</strong>terpretation of the phrase) or speakers’ <strong>in</strong>dividual strategies. Our analysis, however,<br />
focuses on the <strong>in</strong>terpretation issue of the disjunctive coord<strong>in</strong>ate phrase and it is formalised<br />
with<strong>in</strong> the framework of λ-DRT, which is context dependent. Thus, we provide an account<br />
tak<strong>in</strong>g <strong>in</strong>to consideration that the disjunctively conjo<strong>in</strong>ed noun phrase can be <strong>in</strong>terpreted<br />
either with an ‘exclusive’ sense, where a s<strong>in</strong>gular verb form is preferred, or with an ‘andcoord<strong>in</strong>ate’<br />
sense, where a plural verb form is preferred.<br />
The last chapter summarises the whole thesis and discusses some rema<strong>in</strong><strong>in</strong>g issues for<br />
future research.
Chapter 2<br />
LFG: An overview<br />
This chapter is an overview of the formalism of Lexical Functional Grammar (LFG) as<br />
presented <strong>in</strong> Nordl<strong>in</strong>ger and Bresnan (To appear), Dalrymple (2001), Falk (2001), Bresnan<br />
(2001) and Aust<strong>in</strong> (2001). LFG was orig<strong>in</strong>ally developed <strong>in</strong> the 1970s by Joan Bresnan and<br />
Ronald Kaplan. The actual name of the theory reveals its <strong>in</strong>ternal structure. It is a lexical<br />
theory, s<strong>in</strong>ce the lexicon plays a major role and most of the work is done with<strong>in</strong> the lexicon.<br />
The lexicon captures the syntactic or semantic <strong>in</strong>formation of the different word classes,<br />
<strong>in</strong>clud<strong>in</strong>g any exceptional <strong>in</strong>formation. LFG’s syntactic structures are built of words and<br />
the lexicon is the place where words are created. As a lexicalist theory, LFG respects the<br />
Lexical Integrity Pr<strong>in</strong>ciple. It is also “nonderivational” or “nontransformational” s<strong>in</strong>ce verbal<br />
diatheses relations are stated <strong>in</strong> the lexicon and not through transformations as <strong>in</strong> other<br />
theories. F<strong>in</strong>ally, syntactic structures are built monotonically, mean<strong>in</strong>g that <strong>in</strong>formation<br />
can be added but no deletion or change of <strong>in</strong>formation is allowed.<br />
The theory is also functional and not configurational. This means that grammatical<br />
functions, like subject or object, are central <strong>in</strong> the theory and they are used to represent<br />
the syntactic structures of the various languages. This type of syntactic <strong>in</strong>formation is represented<br />
not <strong>in</strong> terms of a phrase structure tree but <strong>in</strong> terms of feature-structures. The<br />
basic idea is that grammatical functions appear as features and other elements that have<br />
a specific function appear as the values to those features. F-structures exist <strong>in</strong> parallel to<br />
c-structures.<br />
As opposed to generative grammar, the theory is also constra<strong>in</strong>t-based. The architecture<br />
of LFG uses parallel structures <strong>in</strong> order to represent the different aspects of l<strong>in</strong>guistic<br />
<strong>in</strong>formation. These structures or levels have their own architecture, vocabulary and obey a<br />
series of constra<strong>in</strong>ts <strong>in</strong> order to function properly. The central levels of representation are<br />
the c(onstituent)-structure, the f(unctional)-structure, the a(rgument)-structure, the semantic<br />
(σ)- structure, the i(nformation)-structure, the m(orphological) and p(honological)structures.<br />
These parallel levels are l<strong>in</strong>ked to each other via correspondence pr<strong>in</strong>ciples,<br />
us<strong>in</strong>g correspondence functions or “projection” functions. C-structure, f-structure and astructure<br />
focus on the syntactic aspects of a language, s-structure deals with semantics,<br />
i-structure explores discourse functions and m-structure and p-structure focus on the morphological<br />
and phonological properties of a language. The different levels of representation<br />
and their <strong>in</strong>dependent constra<strong>in</strong>ts make the theory both universal and applicable to a<br />
number of variable languages. In the current thesis, we focus on the syntactic levels (the<br />
c-structure and f-structure) and the semantic level (the s-structure) and we will describe<br />
how LFG accounts for agreement, focus<strong>in</strong>g on agreement <strong>in</strong> coord<strong>in</strong>ate noun phrases.
2.1 Constituent-structure 6<br />
2.1 Constituent-structure<br />
C-structure encodes the surface phrasal syntactic organisation of a language, such as l<strong>in</strong>ear<br />
order, hierarchical group<strong>in</strong>gs and syntactic categories of constituents, while it <strong>in</strong>terfaces<br />
with the phonological component of the grammar. Constituency <strong>in</strong>formation is expressed<br />
through phrase structure trees which depend on phrase structure rules. LFG c-structures<br />
adopt the X-bar model, which accounts effectively for the crossl<strong>in</strong>guistic variation traced<br />
<strong>in</strong> phrase structures.<br />
LFG’s c-structure level assumes two types of categories, lexical and functional categories.<br />
The lexical categories are N(oun), P(reposition), V(erb), A(djective) and A(dverb),<br />
which function as heads of phrases of the correspond<strong>in</strong>g phrasal categories NP, PP, VP,<br />
AdjP and AdvP. The functional categories are I, C 1 . The category I (orig<strong>in</strong>ally for INFL)<br />
(Falk, 1984) is the head of IP, it functions as the head of a f<strong>in</strong>ite clause and can be filled by<br />
tensed auxiliary verbs or other f<strong>in</strong>ite verbs. The category C (Fassi-Fehri, 1981) is the head<br />
of CP and it can be filled by a verbal element or other elements. In English, for example,<br />
the C position can be filled by a complementiser like that.<br />
S<strong>in</strong>ce the organisation of LFG c-structure is based on X-bar theory, it follows the basic<br />
pr<strong>in</strong>ciples of the latter. In X-bar theory any functional or lexical category X is related to<br />
its correspond<strong>in</strong>g projections, the nonmaximal projection X ′ (with “one bar level”) and<br />
the maximal projection X ′′ (with “two bar levels”).<br />
The organisation of <strong>in</strong>formation <strong>in</strong> the phrase structure tree assumes that a lexical<br />
item X is the head of a phrasal item X ′ and it is also the sister to a number of complement<br />
and adjunct phrases (YP...). In turn, X ′ is the head of the phrasal constituent XP and<br />
it is also the sister to a number of specifier phrases (ZP...), all of which form the phrasal<br />
constituent XP (Dalrymple, 2001, 57). In figure (2.1), the whole sentence corresponds to<br />
the IP category. The head of IP is the phrasal item I ′ which also has as its head the lexical<br />
item I. The tensed auxiliary verb fills <strong>in</strong> the I position and the ma<strong>in</strong> verb appears with<strong>in</strong><br />
the VP. The specifier of IP is the NP ‘David’ and the complement of V ′ is the NP beans:<br />
NP<br />
N<br />
David<br />
IP<br />
I<br />
is<br />
I ′<br />
V<br />
eat<strong>in</strong>g<br />
VP<br />
V ′<br />
NP<br />
N<br />
beans<br />
Figure 2.1: A simple phrase structure tree<br />
A central pr<strong>in</strong>ciple <strong>in</strong> LFG’s c-structure is the Pr<strong>in</strong>ciple of Endocentricity and the categories<br />
<strong>in</strong>volved are called endocentric categories (Bloomfield, 1962). This means that the<br />
phrase structure rules of a language are organised as hierarchical, c-structure configurations<br />
follow<strong>in</strong>g X ′ Theory. Thus, the maximal projection XP and the nonmaximal projection X ′<br />
are headed categories whose head is X.<br />
1 Two more functional categories are D and K but they will not be discussed here. For details see Bresnan<br />
(1997, 2001c), Sadler (1997) and Dalrymple (2001).
2.1 Constituent-structure 7<br />
A second pr<strong>in</strong>ciple that governs LFG’s c-structure is the Pr<strong>in</strong>ciple of Exocentricity<br />
(Nordl<strong>in</strong>ger and Bresnan, To appear; Bresnan, 2001; Dalrymple, 2001). Although English<br />
and a number of other languages are endocentric, there is also evidence that languages<br />
can make use of exocentric categories. Such languages allow for flatter structures where<br />
the subject and all of the arguments are sisters to the head and syntactic functions are<br />
morphologically determ<strong>in</strong>ed. For these languages, LFG <strong>in</strong>troduces the exocentric and nonprojective<br />
category S, which is not headed by a lexical head of the same category as itself<br />
and dom<strong>in</strong>ates any number of lexical or phrasal heads. S is a category that conta<strong>in</strong>s a<br />
predicate and any number of arguments, <strong>in</strong>clud<strong>in</strong>g the subject of the phrase. The example<br />
below is from Falk (2001):<br />
(1) aanaye kut.t.i kan.t.u<br />
elephant.acc child.nom saw<br />
‘The child saw the elephant’<br />
S<br />
NP<br />
aanaye<br />
NP<br />
kut.t.i<br />
V<br />
kan. t.u<br />
S<strong>in</strong>ce LFG uses both pr<strong>in</strong>ciples, it can accommodate data of both types of languages more<br />
effectively and it is a universally applicable theory.<br />
LFG c-structures are subject to the Lexical Integrity Pr<strong>in</strong>ciple, which enforces a dist<strong>in</strong>ction<br />
between the morphological and the syntactic components. The <strong>in</strong>ternal structure<br />
of words is part of the lexicon and the structure of phrases or sentences is part of syntax.<br />
Morphology determ<strong>in</strong>es the <strong>in</strong>ternal word formation and syntax deals with phrasal or sentential<br />
composition. Thus, the formation of the <strong>in</strong>ternal structure of words is <strong>in</strong>visible to<br />
syntax. The pr<strong>in</strong>ciple is stated as follows:<br />
Lexical Integrity Pr<strong>in</strong>ciple<br />
Morphologically complete words are leaves of the c-structure tree and each leaf<br />
corresponds to one and only one c-structure node.<br />
(Bresnan, 2001, 92)<br />
Follow<strong>in</strong>g the above, LFG allows only fully <strong>in</strong>flected words to occupy the position of cstructure<br />
term<strong>in</strong>al nodes, and not parts of words or empty categories. Despite the dist<strong>in</strong>ction<br />
of morphology and syntax, it is possible for both components to contribute the same<br />
k<strong>in</strong>d of <strong>in</strong>formation, be<strong>in</strong>g functionally equivalent (Nordl<strong>in</strong>ger and Bresnan, To appear;<br />
Bresnan, 2001).<br />
For a theory that represents <strong>in</strong>formation <strong>in</strong> the form of phrase structure trees, phrase<br />
structure rules are also required to admit or describe the correspond<strong>in</strong>g phrase structure<br />
trees. Thus, PS rules function as conditions that will license the different nodes <strong>in</strong> a tree.<br />
The follow<strong>in</strong>g example illustrates a simple phrase structure rule:<br />
(2) S −→ NP VP<br />
The categories on the right of the arrow are the c-structure daughters of the categories<br />
on the left.<br />
LFG’s phrase structure rules are more expressive than the phrase structure rules of<br />
other theories s<strong>in</strong>ce they consist of a regular expression (Kaplan and Bresnan, 1982) <strong>in</strong><br />
the right-hand side. Thus, optionality, recursion and disjunction are some of the processes<br />
that are represented <strong>in</strong> the PS rules admitted by LFG. For example, the rule <strong>in</strong> (3) states<br />
that either an NP or a PP can appear as the specifier of IP. The disjunction of two
2.2 Functional-Structure 8<br />
different phrase structure categories is represented by the curly brackets and the different<br />
possibilities are stated with a vertical bar |:<br />
(3) IP −→ {NP|PP} I ′<br />
(Dalrymple, 2001, 93)<br />
Similarly, <strong>in</strong> (4) the parentheses around the NP denotes optionality and the Kleene star<br />
annotation on the PP <strong>in</strong>dicates that any number of PPs can appear after the optional NP:<br />
(4) VP −→ V (NP) PP*<br />
(Dalrymple, 2001, 94)<br />
PS rules allow two types of relations, dom<strong>in</strong>ance relations and precedence relations.<br />
Dom<strong>in</strong>ance relations are expressed through Immediate Dom<strong>in</strong>ance statements, stated with<br />
a comma between the daughter nodes. The rule below states that the VP dom<strong>in</strong>ates a V<br />
and an NP node:<br />
(5) VP −→ V, NP<br />
(Dalrymple, 2001, 96)<br />
Precedence relations maybe expressed through L<strong>in</strong>ear Precedence statements. This is<br />
stated by the <strong>in</strong>troduction of a separate constra<strong>in</strong>t with the symbol
2.2 Functional-Structure 9<br />
(8)<br />
⎡<br />
⎤<br />
pred ‘cry〈subj〉’<br />
⎢<br />
⎢tense<br />
past<br />
⎥<br />
⎢ ⎡ ⎤⎥<br />
g: ⎢ pred ‘mary’ ⎥<br />
⎢ ⎢ ⎥<br />
⎣subj<br />
f: ⎣num<br />
sg<br />
⎥<br />
⎦<br />
pers 3<br />
The label g corresponds to the f-structure of the whole sentence. The three attributes are:<br />
pred, tense and subj. The value of the pred attribute is the semantic form ‘cry〈subj〉’,<br />
which subcategorises for another grammatical function, a subj, and has a unique <strong>in</strong>stantiation,<br />
<strong>in</strong>dicated by s<strong>in</strong>gle quotes. The attribute tense has the atomic value past, whereas<br />
the attribute subj has a whole f-structure as its value, which is the f-structure f and it<br />
is the subject of the sentence. In turn, the subj f-structure with label f conta<strong>in</strong>s the basic<br />
syntactic <strong>in</strong>formation that mary contributes. The value of the pred attribute is also<br />
the semantic form ‘mary’, which gives rise to a uniquely <strong>in</strong>stantiated occurrence. The<br />
attribute num has a sg value and the attribute pers has a 3 value (Dalrymple, 2001, 31).<br />
In the same way that c-structures are descibed by a set of phrase structure rules,<br />
f-structures are described by a set of equations known as a “functional description” (Dalrymple,<br />
2001, 101). Thus, the equation <strong>in</strong> example (9) is satisfied by the f-structure <strong>in</strong><br />
example (10) s<strong>in</strong>ce it conta<strong>in</strong>s the attribute num with the value sg:<br />
(9) (f num)= sg<br />
(10) f [num sg]<br />
The relation between an f-structure and the correspond<strong>in</strong>g equation is formally presented<br />
as follows:<br />
(f α)= υ holds if and only if 〈α, υ〉 ∈ f, where f is an f-structure, α is a symbol,<br />
υ is a value<br />
(Dalrymple, 2001, 101)<br />
LFG requires that an f-structure must be the m<strong>in</strong>imal solution that satisfies all the<br />
constra<strong>in</strong>ts that the words and phrases of an utterance <strong>in</strong>troduce (Dalrymple, 2001, 101).<br />
In the case of the constra<strong>in</strong>t (9), the f-structure <strong>in</strong> (10) is the m<strong>in</strong>imal solution as required.<br />
The organisation of an f-structure is restricted by the well-formedness conditions of<br />
completeness, coherence and uniqueness. Completeness ensures that all of the arguments<br />
that a predicate subcategorises for are present <strong>in</strong> the f-structure. This is formally<br />
stated as follows:<br />
Completeness:<br />
An f-structure is locally complete if and only if it conta<strong>in</strong>s all the governable<br />
grammatical functions that its predicate governs. An f-structure is complete if<br />
and only if it and all its subsidiary f-structures are locally complete. (Dalrymple,<br />
2001, 37)<br />
The sentence below is <strong>in</strong>complete because there is an argument miss<strong>in</strong>g:<br />
(11) *David saw<br />
The f-structure <strong>in</strong> (12) of the pred ‘see’ requires the grammatical functions subj and<br />
obj. Example (11) conta<strong>in</strong>s only a subj and not an obj which makes it <strong>in</strong>complete and<br />
therefore unacceptable:<br />
(12) <br />
<br />
pred ‘see〈subj,obj〉’
2.2 Functional-Structure 10<br />
The pr<strong>in</strong>ciple of coherence can be seen as the opposite of completeness. It ensures that<br />
only the arguments subcategorised by the predicate are present <strong>in</strong> the f-structure and not<br />
any extra ones. The formal def<strong>in</strong>ition is below:<br />
Coherence<br />
An f-structure is locally coherent if and only if all the governable grammatical<br />
functions that it conta<strong>in</strong>s are governed by a local predicate. An f-structure is<br />
coherent if and only if it and all its subsidiary f-structures are locally coherent.<br />
(Dalrymple, 2001, 39)<br />
Sentence (13) is ungrammatical s<strong>in</strong>ce there is an extra grammatical function, the obj noun<br />
phrase the boy, not subcategorised by the predicate:<br />
(13) *Mary cried the boy<br />
Thus, the f-structure is <strong>in</strong>coherent s<strong>in</strong>ce the sentence conta<strong>in</strong>s the extra grammatical function<br />
obj, which the predicate cry does not require:<br />
(14) ⎡<br />
⎤<br />
pred ‘cry〈subj〉’<br />
⎢ ⎥<br />
⎢<br />
⎢subj<br />
pred ‘mary’ ⎥<br />
⎣ ⎦<br />
obj pred ‘boy’<br />
F<strong>in</strong>ally, f-structure postulates the uniqueness or consistency condition which ensures<br />
that an attribute must have only one value and not more than one:<br />
uniqueness or consistency<br />
In a given f-structure a particular attribute may have at most one value<br />
(Dalrymple, 2001, p.39)<br />
The sentence below is ungrammatical s<strong>in</strong>ce the subj noun phrase the girl is s<strong>in</strong>gular but<br />
the verb cry requires its subject to be plural.<br />
(15) *The girl cry<br />
The f-structure is ill-formed because the attribute num has two different values at the<br />
same time. To be well-formed, it needs to have either a sg or pl value and not both<br />
simultaneously:<br />
(16) ⎡<br />
⎤<br />
pred ‘cry〈subj〉’<br />
⎢ ⎥<br />
⎢<br />
⎣ pred ‘girl’ ⎥<br />
subj ⎦<br />
num sg/pl<br />
The most important aspect about LFG’s f-structure is that it follows the pr<strong>in</strong>ciple of<br />
universality which states “that <strong>in</strong>ternal structures are largerly <strong>in</strong>variant across languages”<br />
(Bresnan, 2001, 45). Thus, sentences <strong>in</strong> different languages, which are translational equivalents,<br />
will have the same f-structures despite possible differences <strong>in</strong> their c-structures.<br />
2.2.1 C-structure and f-structure correspondence<br />
This section provides a short description of the correspondence between the two levels<br />
of representation: c-structure and f-structure. LFG uses a formal way to represent the<br />
relation between the two structures, a function 5 φ (phi). This is stated as follows:<br />
5 A function is “a special type of relation which assigns a unique value to its argument” (Dalrymple,<br />
2001, 30).
2.2 Functional-Structure 11<br />
(17) φ: N → F<br />
(Kaplan, 1995)<br />
The correspondence between the c-structure nodes and the f-structure is many-to-one and<br />
each c-structure node is related to a specific and only one f-structure through the φ function.<br />
The examples below display simple diagrams. In example (18), a s<strong>in</strong>gle node corresponds<br />
to an f-structure, whereas <strong>in</strong> example (19) more than one node corresponds to the same<br />
f-structure. The correspondence from the c-structure to the f-structure is represented by<br />
an arrow:<br />
(18) V<br />
cried<br />
(19) NP<br />
N ′<br />
N<br />
Mary<br />
<br />
pred ‘cry〈subj〉’<br />
tense past<br />
⎡ ⎤<br />
pred ‘mary’<br />
⎢ ⎥<br />
⎣num<br />
sg ⎦<br />
pers 3<br />
<br />
The mapp<strong>in</strong>g function from c-structure to f-structure is not arbitrary but it obeys certa<strong>in</strong><br />
pr<strong>in</strong>ciples that are <strong>in</strong> accordance with the c-structural and f-structural configurations.<br />
In general, a c-structure head and its phrasal projections correspond to the same f-structure,<br />
as <strong>in</strong> example (19) 6 . Specifiers of the functional categories IP or CP are mapped either<br />
to the grammatical function subj or to the discourse functions topic or focus. Thus, <strong>in</strong><br />
English and many other languages the specifier position of IP can be filled by the subj<br />
function or by the topic/focus functions, whereas the specifier position of CP corresponds<br />
to the focus function.<br />
Complements of functional categories are f-structure co-heads and will be mapped to<br />
the same f-structure as their heads, and complements of lexical categories are mapped to<br />
any of the grammatical functions, exclud<strong>in</strong>g the subj and the discourse functions. These<br />
notions are shown below:<br />
(20) Mary is call<strong>in</strong>g Chris<br />
NP<br />
N<br />
Mary<br />
IP<br />
I<br />
is<br />
I ′<br />
V<br />
call<strong>in</strong>g<br />
VP<br />
NP<br />
N<br />
Chris<br />
⎡<br />
⎤<br />
pred ‘call〈subj,obj〉’<br />
⎢ ⎥<br />
⎢<br />
⎢subj<br />
pred ‘mary’ ⎥<br />
⎣ ⎦<br />
obj pred ‘chris’<br />
The specifier NP Mary is mapped to the subj function. The complement of the lexical<br />
category V, which is the NP Chris, is mapped to the obj function. The auxiliary and the<br />
ma<strong>in</strong> verb are co-heads and they correspond to the same f-structure whose pred value is<br />
the verb call. When more than one complement appears usually the first one is mapped<br />
to the obj and the second to the objθ, as <strong>in</strong> the English example <strong>in</strong> (21):<br />
6 An exception is Japanese which is a “pro-drop” language with no verbal agreement morphology and<br />
the s<strong>in</strong>gle word kowareta ‘broke’ appears without an overt subject noun phrase. Thus, the subj f-structure<br />
does not corrrespond to any node <strong>in</strong> the c-structure (Kameyama (1985) cited <strong>in</strong> Dalrymple (2001)). For a<br />
schematic representation see Dalrymple (2001, 71).
2.2 Functional-Structure 12<br />
(21) Mary gave Chris the keys<br />
NP<br />
N<br />
Mary<br />
V<br />
gave<br />
IP<br />
NP<br />
N<br />
I ′<br />
VP<br />
V ′<br />
Chris<br />
Det<br />
the<br />
NP<br />
N ′<br />
N<br />
keys<br />
⎡<br />
⎤<br />
pred ‘give〈subj,obj,objθ〉’<br />
⎢ ⎥<br />
⎢<br />
⎢subj<br />
g: pred ‘mary’<br />
⎥<br />
⎢<br />
⎥<br />
f: ⎢obj<br />
h: pred ‘chris ⎥<br />
⎢ <br />
⎥<br />
⎢<br />
⎣ pred ‘keys’ ⎥<br />
objθ i:<br />
⎦<br />
def +<br />
In such a mathematical framework, it is possible to express the <strong>in</strong>verse f-structure to cstructure<br />
mapp<strong>in</strong>g through the use of the <strong>in</strong>verse mapp<strong>in</strong>g function φ −1 . The mapp<strong>in</strong>g<br />
function from c-structure to f-structure and from f-structure to c-structure is possible s<strong>in</strong>ce<br />
the various levels of representation exist simultaneously <strong>in</strong> different dimensions and no<br />
derivational process is carried out.<br />
LFG postulates a formal and powerful way of express<strong>in</strong>g the above schematic correspondence<br />
between a c-structure and an f-structure by apply<strong>in</strong>g a series of functionaldescriptions<br />
or equations to the correspond<strong>in</strong>g phrase structure rules. At the core of these<br />
functional equations are two symbols: ∗, ˆ∗, together with the c-structure to f-structure<br />
function φ. The symbol ∗ denotes the f-structure of the current c-structure node on which<br />
the constra<strong>in</strong>t is placed and the symbol ˆ∗ denotes the f-structure corrrespond<strong>in</strong>g to the<br />
mother node. A more simplified way to state these notations is the use of arrows, the<br />
up-arrow (↑) and the down-arrow (↓). Based on Dalrymple (2001), these are restated as<br />
follows:<br />
(22) ˆ∗ mother’s f-structure = ↑<br />
∗ self’s f-structure = ↓<br />
The use of such notation relates to the way trees are represented. The upward arrow ↑<br />
denotes the mother node, while the downward ↓ denotes the current node itself, and the<br />
need to show any relevant <strong>in</strong>formation is locally passed up the tree from the daughter to<br />
the mother nodes.<br />
Consider the simple example, Mary saw Tom. To show the process clearly, we assign a<br />
variable to each node and relative f-structure, such as fip, fnp, fi ′ etc, and we assume the<br />
follow<strong>in</strong>g c-structure tree correspond<strong>in</strong>g to the f-structure below:<br />
(23) IPfip<br />
NPfnp I ′ fi ′<br />
⎡<br />
⎤<br />
pred ‘see〈subj,obj〉’<br />
⎢<br />
⎥<br />
⎢<br />
tense past ⎥<br />
⎢<br />
⎥<br />
Nfn<br />
Mary<br />
Vfv<br />
saw<br />
VPfvp<br />
V ′ fv ′<br />
NPfnp<br />
Nfn<br />
Tom<br />
fip,fi ′,fvp,fv ′,fv<br />
The f-descriptions for the above f-structure are as follows:<br />
⎢ pred ‘mary’ ⎥<br />
⎢subj<br />
fnp,fn<br />
⎥<br />
⎢<br />
num sg ⎥<br />
⎢ ⎥<br />
⎢<br />
⎣<br />
pred ‘tom’ ⎥<br />
obj fnp,fn<br />
⎦<br />
num sg
2.2 Functional-Structure 13<br />
(24) (fipsubj) = fnp<br />
fnp = fn<br />
(fnpred) = ‘mary’<br />
fip = fi ′<br />
fi ′ = fvp<br />
fvp = fv ′<br />
fv ′ = fv<br />
(fvpred) = ‘see’<br />
(fvtense) = past<br />
(fv ′obj) = fnp<br />
fnp = fn<br />
(fnpred) = ‘tom’<br />
The phrase structure rules with the relevant annotations for example (23) are the follow<strong>in</strong>g:<br />
(25) IP → NP I ′<br />
(fipsubj)= fnp<br />
NP → N<br />
fnp = fn<br />
fip = fi ′<br />
I ′ → ( I ) VP<br />
VP → V ′<br />
fvp = fv ′<br />
fi ′ = fvp<br />
V ′ → V NP<br />
fv ′ = fv (fv ′obj) = fnp<br />
The f-descriptions that correspond to a particular c-structure tree can be replaced with<br />
the ↑ and ↓ arrows, as follows:<br />
(26) IP → NP I ′<br />
NP → N<br />
(↑subj)= ↓ ↑ = ↓<br />
↑ = ↓<br />
I ′ → ( I ) VP<br />
VP → V ′<br />
↑ = ↓<br />
↑ = ↓<br />
V ′ → V NP<br />
↑ = ↓ (↑obj) = ↓<br />
Generally, <strong>in</strong> any annotated phrase structure rule, any daughter node, which appears<br />
on the right hand side of a rule, can be annotated with constra<strong>in</strong>ts that capture the relation<br />
between its f-structure and the f-structure of the mother node. If the daughter node is the<br />
head, they map to the same f-structure; if the daughter node is a nonhead, the daughter’s<br />
f-structure will have some relation (say, the obj relation) to the mother’s f-structure. The<br />
two arrows used <strong>in</strong> the rules assign a stronger locality, s<strong>in</strong>ce only the functional relation
2.3 Semantic-Structure 14<br />
between a daughter and its mother can be stated. Relations among the f-structures of<br />
a daughter node of an annotated rule with its grandmother node are not allowed to be<br />
expressed (Dalrymple, 2001, 117-120).<br />
The same metavariables are used <strong>in</strong> lexical entries. Information that relates to each<br />
lexical item will be captured under each lexical entry and that <strong>in</strong>formation corresponds to<br />
the relevant f-structure. The lexical entries for example (23) are illustrated below:<br />
(27) saw V (↑ pred) = ‘see〈subj,obj〉’<br />
(↑ tense) = past<br />
Mary N (↑ pred) = ‘mary’<br />
(↑ pers) = 3<br />
(↑ num) = sg<br />
Tom N (↑ pred) = ‘tom’<br />
(↑ pers) = 3<br />
(↑ num) = sg<br />
The ↑ and ↓ arrows have exactly the same use <strong>in</strong> a lexical entry as the one <strong>in</strong> the rules s<strong>in</strong>ce<br />
↑ arrow refers to the node that dom<strong>in</strong>ates the lexical item, and ↓ refers to the f-structure<br />
that corresponds to the word itself (Dalrymple, 2001, 120). The same equations can be<br />
annotated to c-structure nodes with which they are associated to make the connection<br />
clearer.<br />
The correspondence between the c-structure to f-structure is most effectively captured<br />
with<strong>in</strong> the pr<strong>in</strong>ciple of Economy of Expression which is stated as follows:<br />
economy of expression<br />
All syntactic phrase structure nodes are optional and are not used unless required<br />
by <strong>in</strong>dependent pr<strong>in</strong>ciples (completeness, coherence, semantic expressivity).<br />
(Bresnan, 2001, 91)<br />
This pr<strong>in</strong>ciple ensures that only the syntactic phrase structure nodes that contribute to<br />
important <strong>in</strong>formation at the f-structure level are licensed <strong>in</strong> the phrase structure tree while<br />
it disallows any empty c-structure category, which dom<strong>in</strong>ates a non term<strong>in</strong>al element, to<br />
be expressed <strong>in</strong> the c-structure tree. Thus, it is immediately related to the f-structure pr<strong>in</strong>ciples<br />
of completeness, coherence and semantic expressivity and admits only the necessary<br />
and m<strong>in</strong>imal <strong>in</strong>formation <strong>in</strong>to the c-structure, which will map to the relevant f-structure.<br />
2.3 Semantic-Structure<br />
LFG is a theoretical framework that does not follow the rule-to-rule hypothesis of the<br />
Pr<strong>in</strong>ciple of Compositionality, which states that the rules of syntax are closely related to<br />
the rules of semantics (Bach, 1989), while it does not assume that the rules that comb<strong>in</strong>e<br />
the mean<strong>in</strong>g of a sentence should depend on phrasal dom<strong>in</strong>ance or l<strong>in</strong>ear order<strong>in</strong>g relations<br />
(Dalrymple, 2001, 218). In LFG, semantic composition is associated to the f-structure and<br />
not to the c-structure level. An important advantage of encod<strong>in</strong>g semantic composition<br />
based on the f-structures is that certa<strong>in</strong> crossl<strong>in</strong>guistic facts are expressed <strong>in</strong> pr<strong>in</strong>ciples<br />
of semantic composition that would otherwise rema<strong>in</strong> obscure if we assumed an analysis<br />
follow<strong>in</strong>g c-structure. Thus, semantic composition relies on the <strong>in</strong>formation encoded at the<br />
f-structure level to derive the <strong>in</strong>formation at the separate level of semantic-structure.<br />
In LFG, each lexical entry contributes a mean<strong>in</strong>g constructor also known as a premise to<br />
a logical deduction. A mean<strong>in</strong>g constructor is a set of logical <strong>in</strong>structions applied to obta<strong>in</strong>
2.3 Semantic-Structure 15<br />
the mean<strong>in</strong>g of the f-structure headed by the lexical entry, by comb<strong>in</strong><strong>in</strong>g the mean<strong>in</strong>gs of the<br />
lexical entry and its syntactic arguments. Then, it uses logical ‘glue’ for mean<strong>in</strong>g assembly<br />
(Dalrymple et al., 2002; Dalrymple, 2001). Therefore, once all the required constra<strong>in</strong>ts<br />
are gathered, deduction <strong>in</strong> the logic is used to obta<strong>in</strong> a mean<strong>in</strong>g for the entire structure.<br />
Mean<strong>in</strong>g constructors consist of a left-hand side known as the mean<strong>in</strong>g side and the righthand<br />
side, which is the glue side, hav<strong>in</strong>g the form P : L where a mean<strong>in</strong>g P is paired with<br />
a l<strong>in</strong>ear logic formula L.<br />
The left-hand side or the mean<strong>in</strong>g side is often represented as a λ-expression 7 which<br />
is a convenient formalism to represent the mean<strong>in</strong>gs of sentences. The lambda operator λ<br />
constructs a function by abstract<strong>in</strong>g on a variable conta<strong>in</strong>ed <strong>in</strong> a proposition. The function<br />
is applied to its argument as follows:<br />
(28) Function application: [λX.P](a)<br />
The function λX.P is applied to the argument a.<br />
(Dalrymple, 2001, 225)<br />
The expression above states that the lambda operator b<strong>in</strong>ds the variable X, which occurs<br />
at least once <strong>in</strong> the proposition P. Thus, <strong>in</strong> the example Mary cried represented as<br />
[λX.cry(X)]Mary us<strong>in</strong>g the λ-calculus, if we replace all occurrences of X <strong>in</strong> P with Mary,<br />
we can derive the expression cry(Mary), which is the semantic equivalent. Although the λcalculus<br />
is a useful formalism for assembl<strong>in</strong>g the semantics of a sentence provided by surface<br />
constituent structure trees, it is not an effective tool to comb<strong>in</strong>e mean<strong>in</strong>gs of f-structure<br />
constituents. This is due to the fact that f-structures are unordered and λ-calculus follows<br />
a fixed order of comb<strong>in</strong><strong>in</strong>g a functor with its arguments (Dalrymple et al., 2002).<br />
To avoid the problems posed by the λ-calculus, LFG uses the glue approach a formula<br />
<strong>in</strong> l<strong>in</strong>ear logic <strong>in</strong> the right-hand side. This is a deductive approach to mean<strong>in</strong>g assembly<br />
which comb<strong>in</strong>es the mean<strong>in</strong>gs associated with attribute-value structures, through a series<br />
of constra<strong>in</strong>ts, so as to deduce the mean<strong>in</strong>g of the whole structure. Thus, once the mean<strong>in</strong>g<br />
constructor or premise of each f-structure has been derived then we use the glue approach<br />
to comb<strong>in</strong>e the mean<strong>in</strong>g constructor of the lexical entry with its syntactic arguments <strong>in</strong><br />
order to form the mean<strong>in</strong>g of the whole utterance.<br />
LFG also postulates a semantic-structure, which is a separate level of l<strong>in</strong>guistic representation.<br />
The f-structure is associated with the semantic-structure by the correspondence<br />
function σ8 , represented by an arrow from the f-structure to the s-structure. The semantic<br />
projection of an f-structure represents its mean<strong>in</strong>g and it is denoted with the subscript<br />
σ. Thus, the notation fσ is the semantic projection of the f f-structure. An example with<br />
a one-place predicate, such as Mary cried, <strong>in</strong>troduces the follow<strong>in</strong>g f-structure and the<br />
correspond<strong>in</strong>g s-structure:<br />
⎡<br />
⎤ <br />
(29) pred ‘cry〈subj〉’ fσ : (f subj)σ :<br />
⎢ ⎡ ⎤⎥<br />
⎢<br />
f: ⎢<br />
pred ‘mary’ ⎥<br />
⎢ ⎢ ⎥<br />
⎥<br />
⎣subj<br />
g: ⎣pers<br />
3 ⎦<br />
⎥<br />
⎦<br />
num sg<br />
For the simple sentence Mary cried, the follow<strong>in</strong>g augmented lexical entries are <strong>in</strong>troduced:<br />
7 Other formalisms are DRT and Predicate Calculus. For a def<strong>in</strong>ition and details on the lambda calculus<br />
see Partee et al. (1993).<br />
8 The <strong>in</strong>verse correpondence function σ −1 also holds of an s-structure to an f-structure.
2.3 Semantic-Structure 16<br />
(30) Mary N (g pred) = ‘mary’<br />
Mary : gσ<br />
cry V (f pred)= ‘cry’<br />
λX.cry(X) : (f subj)σ ⊸ fσ<br />
In the first lexical entry, the expression Mary : gσ is the mean<strong>in</strong>g constructor. It does<br />
not really construct any mean<strong>in</strong>g here s<strong>in</strong>ce it is an <strong>in</strong>dividual-denot<strong>in</strong>g expression. Mary<br />
is the mean<strong>in</strong>g and it is associated with the s-structure gσ which is the semantic projection<br />
of the f-structure g. The second lexical entry with the expression λX.cry(X) : (f subj)σ<br />
⊸ f σ shows how we construct mean<strong>in</strong>gs. The left part λX.cry(X) displays a one-place<br />
predicate us<strong>in</strong>g the λ-calculus. The right part is a formula which conta<strong>in</strong>s the connective ⊸<br />
known as the l<strong>in</strong>ear implication symbol <strong>in</strong> l<strong>in</strong>ear logic. This symbol expresses a mean<strong>in</strong>g<br />
like if...then...: and it carries a requirement of production and consumption. Thus, the<br />
formula <strong>in</strong> the second l<strong>in</strong>e of the lexical entry of the verb cry, (f subj)σ ⊸ fσ, <strong>in</strong>dicates<br />
that if a semantic resource (f subj)σ is found, it is consumed and the semantic resource<br />
fσ is produced, which is the semantics of the whole structure. Thus, the mean<strong>in</strong>g for the<br />
whole structure is produced once the mean<strong>in</strong>g of the arguments of the verb are obta<strong>in</strong>ed.<br />
The mean<strong>in</strong>g constructors for the argument Mary and the predicate cry are as follows:<br />
(31) [Mary] Mary : gσ<br />
[cry] λX.cry(X) : (f subj)σ ⊸ fσ<br />
If we <strong>in</strong>stantiate the f-structure for Mary with label g and the f-structure for cry with label<br />
f, we can derive the follow<strong>in</strong>g:<br />
(32) [Mary] Mary : gσ<br />
[cry] λX.cry(X) : gσ ⊸ fσ<br />
Discharg<strong>in</strong>g the antecedent of an implication on the right hand side of the formula<br />
labeled [cry] means apply<strong>in</strong>g a function to its argument on the left-hand side. The formula<br />
[Mary-cry] is def<strong>in</strong>ed as follows:<br />
(33) [Mary-cry] : cry(Mary) : fσ<br />
If we comb<strong>in</strong>e the premises [Mary] and [cry], we derive the follow<strong>in</strong>g 9 :<br />
(34) [Mary], [cry]<br />
⊢ [Mary-cry]<br />
⊢ cry(Mary) : fσ<br />
Glue semantics postulates specific rules which associate the mean<strong>in</strong>g side (on the left)<br />
with the glue side (on the right side), shown below:<br />
(35) X : gσ<br />
P : gσ ⊸ fσ<br />
P(X) : fσ<br />
(Dalrymple, 2001, 235)<br />
The rule states that the glue side requires as its argument a semantic structure gσ and the<br />
mean<strong>in</strong>g side also requires an argument for the one-place predicate. Once this argument is<br />
9 The symbol ⊢ stands for the l<strong>in</strong>ear logic entailment relation.
2.4 <strong>Agreement</strong> <strong>in</strong> LFG 17<br />
available <strong>in</strong> both cases then it is possible to derive the complete semantic resource <strong>in</strong> both<br />
the mean<strong>in</strong>g side and the glue sides.<br />
To summarise, l<strong>in</strong>ear logic allows the various mean<strong>in</strong>gs <strong>in</strong> a sentence to be expressed<br />
as premises <strong>in</strong> a logical deduction and derives the mean<strong>in</strong>g of the sentence by comb<strong>in</strong><strong>in</strong>g<br />
these mean<strong>in</strong>gs. Thus, through logical deduction mean<strong>in</strong>g contributions are perceived<br />
as resources that are produced and consumed while no semantic composition is assumed,<br />
follow<strong>in</strong>g the rules of phrasal composition. Accord<strong>in</strong>g to Dalrymple et al. (2002), this entails<br />
a resource-sensitivity of l<strong>in</strong>ear logic that disallows any arbitrary duplication, deletion or<br />
addition of formulas. Therefore, each word or phrase makes a unique contribution, which<br />
is shown <strong>in</strong> the f<strong>in</strong>al mean<strong>in</strong>g of the sentence.<br />
2.4 <strong>Agreement</strong> <strong>in</strong> LFG<br />
This section provides an <strong>in</strong>troduction to the basic notions of agreement and coord<strong>in</strong>ation<br />
with<strong>in</strong> the architecture of Lexical Functional Grammar. Most constra<strong>in</strong>t-based theories,<br />
which analyse agreement, assert that predicates, such as verbs, place constra<strong>in</strong>ts on the<br />
features of their arguments and more precisely on the central grammatical features person,<br />
number, gender and case.<br />
In LFG, agreement between a f<strong>in</strong>ite verb and its subject is stated through a series<br />
of equations <strong>in</strong> the lexical entries of the agree<strong>in</strong>g elements and it is represented at the fstructure<br />
level. For example, the verb below is morphologically marked to <strong>in</strong>dicate that its<br />
subject must be third person s<strong>in</strong>gular while the subject must carry the same feature-values<br />
to agree with the verb:<br />
(36) Mary arrives<br />
In the example above, the verb arrives agrees with the subject Mary. The f-structure<br />
<strong>in</strong>cludes the attributes person and number, and <strong>in</strong>stantiates the values for these attributes<br />
<strong>in</strong> the follow<strong>in</strong>g way:<br />
⎡<br />
⎤<br />
(37) pred ‘arrive〈subj〉’<br />
⎢ ⎡ ⎤⎥<br />
⎢<br />
f: ⎢<br />
pred ‘Mary’ ⎥<br />
⎢ ⎢ ⎥<br />
⎥<br />
⎣subj<br />
g: ⎣person<br />
3 ⎦<br />
⎥<br />
⎦<br />
num sg<br />
The agreement of the two elements is successfully captured by the equations contributed<br />
by the lexical entries for Mary and arrives, <strong>in</strong>stantiated as follows:<br />
(38) arrives: (f subj person) = 3<br />
(f subj num) = sg<br />
Mary: (g person) = 3<br />
(g num) = sg<br />
The annotations on the English phrase structure rules require that the subject of the verb<br />
be the f-structure g follow<strong>in</strong>g the equation (f subj) = g:<br />
(39) IP → NP VP<br />
(↑subj)=↓ ↑=↓<br />
S<strong>in</strong>ce the agreement features specified by the noun phrase match the agreement features<br />
required by the verb, we can conclude that the agreement features of the whole phrase are<br />
consistent and the sentence is well-formed.
2.4 <strong>Agreement</strong> <strong>in</strong> LFG 18<br />
In NP <strong>in</strong>ternal agreement, LFG assumes the dist<strong>in</strong>ction between concord and <strong>in</strong>dex<br />
agreement features associated with the nouns and posited at the level of functional<br />
structure, follow<strong>in</strong>g work <strong>in</strong> HPSG (Pollard and Sag, 1994; Kathol, 1999; Wechsler and<br />
Zlatić, 2000, 2003). concord features are associated to the declension class of a noun and<br />
control agreement between a noun and its determ<strong>in</strong>er or adjective, and <strong>in</strong>dex features are<br />
associated to the semantics of the noun and control agreement between a noun phrase and<br />
a bound pronoun or a verb (K<strong>in</strong>g and Dalrymple, 2004). The follow<strong>in</strong>g simplified phrase<br />
structure rule is <strong>in</strong>troduced:<br />
(40) NP → Det N<br />
↑=↓ ↑=↓<br />
(K<strong>in</strong>g and Dalrymple, 2004, 71)<br />
The annotation ↑=↓ <strong>in</strong> the phrase structure rule for both the det and the np requires<br />
that they are co-heads and therefore they share the same f-structure. K<strong>in</strong>g and Dalrymple<br />
(2004) propose the follow<strong>in</strong>g phrase structure tree and f-structure for the simple noun<br />
phrase ‘this boy’:<br />
(41) This boy<br />
DET<br />
this<br />
NP<br />
N<br />
boy<br />
⎡<br />
⎤<br />
spec ‘this’<br />
⎢<br />
⎥<br />
⎢<br />
pred ‘boy’ ⎥<br />
⎢<br />
⎥<br />
⎢concord<br />
num sg ⎥<br />
⎢<br />
⎣ <br />
⎥<br />
⎦<br />
<strong>in</strong>dex num sg<br />
(K<strong>in</strong>g and Dalrymple, 2004, 72)<br />
The lexical entries for the determ<strong>in</strong>er this and the noun boy are illustrated below:<br />
(42) this (↑spec) = ‘this’<br />
(↑concord num)= sg<br />
boy (↑pred) = ‘boy’<br />
(↑concord num)= sg<br />
(↑<strong>in</strong>dex num)= sg<br />
(K<strong>in</strong>g and Dalrymple, 2004, 72)<br />
The lexical entry of the determ<strong>in</strong>er this requires <strong>in</strong> its f-structure a noun with sg value<br />
for its concord num feature specification. This is imposed by the equation (↑concord<br />
num)=sg. The lexical entry of the noun boy contributes two types of agreement <strong>in</strong>formation,<br />
concord and <strong>in</strong>dex. In both specifications, it has sg value for the num feature.<br />
The former is related to the agreement of the noun with its determ<strong>in</strong>er and the latter is<br />
related to the agreement of the noun with the verb. The noun’s agreement <strong>in</strong>formation<br />
satisfies the requirements of the determ<strong>in</strong>er s<strong>in</strong>ce the noun contributes the same concord<br />
feature value as the one required by the determ<strong>in</strong>er, and they form a grammatical phrase.<br />
Therefore, noun phrase <strong>in</strong>ternal agreement depends on the concord feature only (Wechsler<br />
and Zlatić, 2000). If the determ<strong>in</strong>er specifies <strong>in</strong>formation that conflicts with that of<br />
the noun, such as <strong>in</strong> the phrase *this boys, the f-structure is unacceptable s<strong>in</strong>ce the value<br />
of the num feature is required to be simultaneously sg by the determ<strong>in</strong>er and pl by the<br />
noun, and therefore it does not obey the uniqueness condition.<br />
Thus, agreement <strong>in</strong> all languages <strong>in</strong>volves the <strong>in</strong>teraction of multiple elements and LFG<br />
has the mechanism to encode effectively this <strong>in</strong>teraction, us<strong>in</strong>g constra<strong>in</strong><strong>in</strong>g equations <strong>in</strong>
2.4 <strong>Agreement</strong> <strong>in</strong> LFG 19<br />
the lexical entries of the agree<strong>in</strong>g elements and represent<strong>in</strong>g these equations <strong>in</strong> well-formed<br />
f-structures.<br />
2.4.1 Coord<strong>in</strong>ation <strong>in</strong> LFG<br />
LFG architecture has the mechanism to capture the phenomenon of coord<strong>in</strong>ation, which<br />
will concern us <strong>in</strong> the rest of the thesis. We will briefly discuss all types of coord<strong>in</strong>ation<br />
but the ma<strong>in</strong> focus will be on noun phrase coord<strong>in</strong>ation.<br />
Coord<strong>in</strong>ation was orig<strong>in</strong>ally discussed with<strong>in</strong> the framework of LFG by Bresnan et al.<br />
(1985b) and was more formally explored by Kaplan and Maxwell (1988) and Dalrymple<br />
et al. (1995). Coord<strong>in</strong>ation phenomena are usually divided <strong>in</strong>to two different types, constituent<br />
coord<strong>in</strong>ation <strong>in</strong> which the coord<strong>in</strong>ated elements are phrasal or lexical constituents,<br />
and nonconstituent coord<strong>in</strong>ation where the coord<strong>in</strong>ated elements are fragments of phrasal<br />
constituents (Dalrymple et al., 1995). The follow<strong>in</strong>g sentences represent examples of constituent<br />
and nonconstituent coord<strong>in</strong>ation:<br />
(43) a. Constituent Coord<strong>in</strong>ation<br />
Bill cried and Mary laughed<br />
Mary cooked and ate the cake<br />
b. Nonconstituent Coord<strong>in</strong>ation<br />
Mary flew to Dallas on Wednesday and New York on Friday<br />
David <strong>in</strong>troduced Chris to Tracy and Mary to Tom<br />
Although constituent coord<strong>in</strong>ation seems more straightforward than nonconstituent<br />
coord<strong>in</strong>ation, it still raises important issues for the more traditional approaches, such as<br />
Conjunction Reduction (CR), proposed by Chomsky (1957) and ref<strong>in</strong>ed by Dougherty<br />
(1970, 1971). A ma<strong>in</strong> problem for those approaches was that it is not always possible to<br />
derive coord<strong>in</strong>ate phrases from full sentences. For example, the two sentences below cannot<br />
be conjo<strong>in</strong>ed and form the coord<strong>in</strong>ate structure <strong>in</strong> (44c):<br />
(44) a. The girl promised John to go<br />
b. The girl persuaded John to go<br />
c. *The girl promised and persuaded John to go<br />
(Kaplan and Maxwell, 1988, 303)<br />
A second problem relates to semantics and that it is possible to have a different semantic<br />
<strong>in</strong>terpretation <strong>in</strong> a coord<strong>in</strong>ate construction, which derives from two separate propositions.<br />
Thus, <strong>in</strong> example (45a), the noun man may refer to two different persons, whereas <strong>in</strong><br />
example (45b), man refers to a s<strong>in</strong>gle person:<br />
(45) a. A man saw Tom and a man greeted Tom<br />
b. A man saw and greeted Tom.<br />
In LFG, coord<strong>in</strong>ate structures, either constituent or nonconsituent, are treated as a<br />
set conta<strong>in</strong><strong>in</strong>g the f-structures that correspond to the two coord<strong>in</strong>ated elements (Bresnan<br />
et al., 1985b; Kaplan and Maxwell, 1988). In other words, each conjunct has its own fstructure<br />
which is part of a set. The major advantage of a set representation is that it has<br />
the ability to hold an unbounded number of elements <strong>in</strong> a s<strong>in</strong>gle construction and none of<br />
these elements dom<strong>in</strong>ates or scopes over the others. The follow<strong>in</strong>g rule from Dalrymple<br />
(2001, 362) captures coord<strong>in</strong>ate structures 10 :<br />
10 The Kleene plus operator ( + ) allows any number of IPs to appear with<strong>in</strong> the phrase.
2.4 <strong>Agreement</strong> <strong>in</strong> LFG 20<br />
(46) IP −→ IP + Conj IP<br />
↓∈↑ ↑=↓ ↓∈↑<br />
This phrase structure rule states that the coord<strong>in</strong>ate sentence consists of a sentence (or more<br />
than one), which is the first conjunct, a conjunction and another sentence follow<strong>in</strong>g the<br />
latter. The functional annotations <strong>in</strong> the phrase structure rule specify that the f-structure<br />
of each conjunct is an element of the f-structure that corresponds to the whole coord<strong>in</strong>ate<br />
phrase. A characteristic example of sentential coord<strong>in</strong>ation is the sentence Tom cried and<br />
Mary laughed where the two sentences Tom cried and Mary laughed are conjo<strong>in</strong>ed by the<br />
conjunction and. The correspondence of the c-structure to the relevant f-structure is seen<br />
below:<br />
(47) Tom cried and Mary laughed<br />
NP<br />
N<br />
Tom<br />
IP<br />
I ′<br />
VP<br />
V<br />
cried<br />
IP<br />
Conj<br />
and<br />
NP<br />
N<br />
Mary<br />
IP<br />
I ′<br />
VP<br />
V<br />
laughed<br />
⎧ ⎡<br />
⎤ ⎫<br />
pred ‘cry〈subj〉’<br />
⎣ ⎦<br />
⎪⎨ subj pred ‘tom’ ⎪⎬<br />
⎡<br />
⎤<br />
pred ‘laugh’<br />
⎣ ⎦<br />
⎪⎩ subj pred ‘mary’ ⎪⎭<br />
The whole structure appears as a set which conta<strong>in</strong>s the f-structures that correspond to<br />
the conjunct sentences. Sentential coord<strong>in</strong>ation is rather straightforward. Cases of nonsentential<br />
coord<strong>in</strong>ation are more complicated and LFG has the mechanism to capture them.<br />
Before turn<strong>in</strong>g to these, we <strong>in</strong>troduce the important dist<strong>in</strong>ction between distributive and<br />
nonditributive features.<br />
2.4.1.1 Distributive vs Nondistributive Features<br />
Coord<strong>in</strong>ate structures are special s<strong>in</strong>ce they have features that are different from the features<br />
of their elements, motivat<strong>in</strong>g a dist<strong>in</strong>ction between distributive and nondistributive<br />
features.<br />
A distributive feature is a feature which is associated with each member of the set and<br />
not with the set as a whole. For example, <strong>in</strong> a coord<strong>in</strong>ate noun phrase with nom<strong>in</strong>ative case<br />
case is a distributive feature s<strong>in</strong>ce each conjunct noun needs to specify <strong>in</strong> its f-structure<br />
the attribute case with nom value. Thus, each member of the set needs to have the pair<br />
case nom. An example from <strong>Greek</strong> illustrates the above:<br />
(48) To<br />
the<br />
agori<br />
boy.nom.sg.3<br />
ke<br />
and<br />
‘The boy and the girl are happy’<br />
to<br />
the<br />
koritsi<br />
girl.nom.sg.3<br />
<strong>in</strong>e<br />
are.pl<br />
xarumena<br />
happy
2.4 <strong>Agreement</strong> <strong>in</strong> LFG 21<br />
NP<br />
to-agori<br />
NP<br />
Conj<br />
ke<br />
NP<br />
to-koritsi<br />
⎡<br />
⎤<br />
num pl<br />
⎢⎧<br />
⎡ ⎤ ⎫⎥<br />
⎢ pred ‘agori’<br />
⎥<br />
⎢ ⎢ ⎥ ⎥<br />
⎢ ⎢case<br />
nom ⎥ ⎥<br />
⎢ ⎢ ⎥ ⎥<br />
⎢ d: ⎢<br />
⎢num<br />
sg ⎥ ⎥<br />
⎥ ⎥<br />
⎢ ⎢<br />
⎢ ⎣pers<br />
3<br />
⎥ ⎥<br />
⎦ ⎥<br />
⎢<br />
f: ⎢<br />
⎪⎨<br />
⎪⎬<br />
⎥<br />
def + ⎥<br />
⎢ ⎡<br />
⎤ ⎥<br />
⎢<br />
pred ‘koritsi’ ⎥<br />
⎢ ⎢<br />
⎥ ⎥<br />
⎢ ⎢case<br />
nom ⎥ ⎥<br />
⎢ ⎢<br />
⎥ ⎥<br />
⎢ b: ⎢<br />
⎢ ⎢num<br />
sg ⎥ ⎥<br />
⎥ ⎥<br />
⎢ ⎢<br />
⎣ ⎣pers<br />
3<br />
⎥ ⎥<br />
⎦ ⎥<br />
⎪⎩<br />
⎪⎭<br />
⎦<br />
def +<br />
Bresnan et al. (1985b) and Dalrymple and Kaplan (2000) formally state the above by<br />
extend<strong>in</strong>g the def<strong>in</strong>ition of function-application (f α) = υ to apply to sets as well. This is<br />
shown below:<br />
If α is a distributive feature and s is a set of f-structures, then (s α)= υ holds<br />
iff (f α) = υ for all f-structures f that are members of the set s.<br />
(Dalrymple, 2001, 365)<br />
A nondistributive feature, on the other hand, is one which appears as a property of the<br />
coord<strong>in</strong>ate phrase as a whole. In example (48) above, the num feature with value pl holds<br />
of the whole coord<strong>in</strong>ate phrase and not of the <strong>in</strong>dividual conjuncts and it is nondistributive.<br />
Dalrymple and Kaplan (2000, 72) propose the follow<strong>in</strong>g formal def<strong>in</strong>ition for nondistributive<br />
features:<br />
If α is a nondistributive feature, then (f α) = υ holds iff the pair 〈α,υ〉 ∈ f.<br />
(Dalrymple, 2001, 158)<br />
Apart from num and pers <strong>in</strong> the above example, preconjunctions, such as both and<br />
either, and conjunctions such as and and or are classified as nondistributive features s<strong>in</strong>ce<br />
they are attributes of the coord<strong>in</strong>ate phrase as a whole. Thus, the lexical entry for and or<br />
or is represented as follows:<br />
(49) and conj (↑conj) = and (Dalrymple, 2001, 267)<br />
The dist<strong>in</strong>ction of distributive and nondistributive features is central <strong>in</strong> the sections<br />
that follow.<br />
2.4.1.2 Predicate Coord<strong>in</strong>ation<br />
LFG also captures non-sentential coord<strong>in</strong>ation. Predicate coord<strong>in</strong>ation and more specifically<br />
verb coord<strong>in</strong>ation is a type of non-sentential coord<strong>in</strong>ation <strong>in</strong> which some parts of the<br />
sentence (i.e. subj and/or obj) are shared by the coord<strong>in</strong>ated constituents. The phrase<br />
structure rule for coord<strong>in</strong>ate predicates uses the same functional annotations as the ones<br />
proposed for sentential coord<strong>in</strong>ation but different categories are used s<strong>in</strong>ce predicates are<br />
the coord<strong>in</strong>ate elements. Thus, it is rewritten as follows:<br />
(50) V −→ V + Cnj V<br />
↓ ∈ ↑ ↓ ∈ ↑<br />
(Dalrymple, 2001, 364)
2.4 <strong>Agreement</strong> <strong>in</strong> LFG 22<br />
Based on the rule above, the c-structure and f-structure of the coord<strong>in</strong>ated predicates saw<br />
and bought are shown below:<br />
(51) saw and bought<br />
V<br />
saw<br />
V<br />
Cnj<br />
and<br />
V<br />
bought<br />
⎧<br />
⎫<br />
⎪⎨ pred ‘see〈subj,obj〉’<br />
⎪⎬<br />
f: <br />
<br />
⎪⎩ pred ‘buy〈subj,obj〉’ ⎪⎭<br />
When verbs are coord<strong>in</strong>ated they need to comply with the restrictions imposed by the<br />
Coherence and Completeness conditions for the sentence to be grammatically acceptable.<br />
For example, the two predicates <strong>in</strong> (51) are transitive and each of them needs to have<br />
a subj and an obj for the phrase to be grammatical. This restriction is satisfied if both<br />
predicates share the same subj and obj, or alternatively if the two features distribute to the<br />
<strong>in</strong>dividual conjuncts. In the f-structure and c-structure <strong>in</strong> (52) distributivity requirements<br />
are satisfied. The NP Tom functions as the subject of the f-structure set through the equation<br />
(↑subj)=↓ and the NP oranges functions as the object of the f-structure set through<br />
the equation (↑obj)=↓. Hence both subj and obj features distribute to the <strong>in</strong>dividual<br />
conjuncts and therefore the Coherence and Completeness conditions are satisfied:<br />
(52) Tom saw and bought oranges<br />
NP<br />
N<br />
Tom<br />
V<br />
saw<br />
IP<br />
V ′<br />
V<br />
Conj<br />
and<br />
I ′<br />
VP<br />
V<br />
bought<br />
NP<br />
N<br />
oranges<br />
⎧⎡<br />
⎤⎫<br />
pred ‘see〈subj,obj〉’<br />
⎢ ⎥<br />
⎢<br />
⎢subj<br />
pred ‘〈tom〉’ ⎥<br />
⎪⎨<br />
⎣ <br />
⎦⎪⎬<br />
obj pred ‘〈oranges〉’<br />
⎡<br />
⎤<br />
pred ‘buy’<br />
⎢<br />
⎥<br />
⎣subj<br />
⎦<br />
⎪⎩<br />
⎪⎭<br />
obj<br />
The major advantage of the notion of distributivity is that it accounts for both grammatical<br />
and ungrammatical cases. In the case above, the subj and obj are distributed<br />
to each member of the set and this is why the two verbs share the same values for the<br />
subj and obj attributes and the sentence is grammatical. The second advantage is that<br />
distributivity accounts for ungrammatical cases as the ones below:<br />
(53) *Mary promised and persuaded John to go<br />
V<br />
NP<br />
N<br />
Mary<br />
promised<br />
V ′<br />
V<br />
Conj<br />
and<br />
IP<br />
V<br />
I ′<br />
VP<br />
persuaded<br />
NP<br />
N<br />
John<br />
Part<br />
to<br />
VP<br />
The f-structure of the above sentence is shown below:<br />
V ′<br />
V<br />
go
2.4 <strong>Agreement</strong> <strong>in</strong> LFG 23<br />
(54)<br />
⎧⎡<br />
⎤ ⎫<br />
pred ‘promise〈subj,xcomp〉’<br />
⎢ <br />
⎥<br />
⎢<br />
⎢subj<br />
pred ‘〈Mary〉’<br />
⎥<br />
⎢<br />
⎥<br />
⎢obj<br />
pred ‘〈John〉’ ⎥<br />
⎢ <br />
<br />
⎥<br />
⎢<br />
⎪⎨ ⎣ pred ‘go〈Mary〉’ ⎥<br />
xcomp<br />
⎦ ⎪⎬<br />
f:<br />
subj<br />
⎡<br />
⎤<br />
pred ‘persuade〈subj,obj,xcomp〉’<br />
⎢<br />
⎢subj<br />
⎥<br />
⎢<br />
⎥<br />
⎢obj<br />
⎥<br />
⎢ <br />
⎥<br />
⎢<br />
⎣ pred ‘go〈John〉’<br />
⎥<br />
⎦<br />
⎪⎩<br />
xcomp<br />
⎪⎭<br />
subj<br />
The two xcomps are required to be token identical but they are not s<strong>in</strong>ce they do not<br />
share the same subject. Thus, the phrase is ungrammatical.<br />
2.4.1.3 <strong>Noun</strong> Phrase Coord<strong>in</strong>ation<br />
<strong>Noun</strong> phrase coord<strong>in</strong>ation is the last type of constituent coord<strong>in</strong>ation we will present and<br />
which will concern us <strong>in</strong> the rest of the thesis. The central agreement features <strong>in</strong> NP<br />
coord<strong>in</strong>ation are number, person, gender, case and <strong>in</strong> some languages noun class 11 .<br />
As <strong>in</strong> the other types of coord<strong>in</strong>ation, noun phrase coord<strong>in</strong>ation uses the standard LFG<br />
rule:<br />
(55) NP −→ NP + Cnj NP<br />
↓ ∈ ↑ ↑=↓ ↓ ∈ ↑<br />
(Dalrymple, 2001, 381)<br />
<strong>Noun</strong> phrase coord<strong>in</strong>ation also assumes the dist<strong>in</strong>ction between concord and <strong>in</strong>dex 12<br />
agreement features, as non-coord<strong>in</strong>ate nouns. <strong>in</strong>dex agreement is formally stated by K<strong>in</strong>g<br />
and Dalrymple (2004) as follows:<br />
<strong>in</strong>dex agreement with a coord<strong>in</strong>ate phrase depends on the <strong>in</strong>dex features of<br />
the coord<strong>in</strong>ate phrase as a whole. (K<strong>in</strong>g and Dalrymple, 2004, 74)<br />
In predicate-argument agreement, the <strong>in</strong>dex features are number, person, gender<br />
and optionally case 13 . A very simple example of a coord<strong>in</strong>ate noun phrase is given below<br />
from Dalrymple and Kaplan (2000, 778):<br />
(56) Jose y yo hablamos/*habláis/*hablan<br />
Jose and I speak.1.pl/*speak.2.pl/*speak.3.pl<br />
‘Jose and I speak’<br />
11 The most characteristic languages with noun class are Bantu languages.<br />
12 Dalrymple and Kaplan (2000) and Dalrymple (2001) when discuss<strong>in</strong>g agreement <strong>in</strong> coord<strong>in</strong>ate noun<br />
phrases do not assume the dist<strong>in</strong>ction of the <strong>in</strong>dex and concord features with<strong>in</strong> the f-structure. This<br />
dist<strong>in</strong>ction was later <strong>in</strong>troduced by K<strong>in</strong>g and Dalrymple (2004).<br />
13 case is considered to be either distributive or non-distributive, which means that it is one of the <strong>in</strong>dex<br />
or concord features, depend<strong>in</strong>g on the language <strong>in</strong>volved. Dalrymple and Kaplan (2000) seem to suggest<br />
that case and noun class are distributive but Sadler (2003) and McCloskey (1986) argue this is too strong<br />
a position when consider<strong>in</strong>g case mismatches <strong>in</strong> coord<strong>in</strong>ate structures <strong>in</strong> a number of languages <strong>in</strong>clud<strong>in</strong>g<br />
English such as the example She and him/he will drive to the movies (Sadler, 2003, 74).
2.4 <strong>Agreement</strong> <strong>in</strong> LFG 24<br />
The coord<strong>in</strong>ate phrase conta<strong>in</strong>s a first person s<strong>in</strong>gular conjunct and a third person s<strong>in</strong>gular<br />
conjunct. However, verb agreement is as for a first person plural phrase. Thus, the <strong>in</strong>dex<br />
person and <strong>in</strong>dex number agreement features of the coord<strong>in</strong>ate structure differ from the<br />
features of <strong>in</strong>dividual conjuncts. A simple functional structure which shows clearly the<br />
representation of the <strong>in</strong>dex features and the features of the conjuncts for the coord<strong>in</strong>ate<br />
phrase <strong>in</strong> (56) is shown below from Dalrymple and Kaplan (2000):<br />
⎡⎧<br />
⎡ ⎤⎫⎤<br />
(57)<br />
⎢<br />
pred ‘Jose’<br />
⎢ ⎢ ⎥<br />
⎥<br />
⎢ f: ⎣person<br />
3 ⎦<br />
⎥<br />
⎢⎪⎨<br />
⎪⎬ ⎥<br />
⎢ num sg ⎥<br />
⎢ ⎡ ⎤ ⎥<br />
⎢ pred ‘pro’ ⎥<br />
⎢<br />
fi: ⎢ ⎢ ⎥ ⎥<br />
⎢<br />
i: ⎣person<br />
1 ⎦ ⎥<br />
⎢⎪⎩<br />
⎪⎭<br />
⎥<br />
⎢ num sg<br />
⎥<br />
⎢<br />
⎥<br />
⎢ person 1 ⎥<br />
⎢<strong>in</strong>dex<br />
⎥<br />
⎢<br />
⎣<br />
num pl ⎥<br />
⎦<br />
pred conj<br />
K<strong>in</strong>g and Dalrymple (2004) propose that <strong>in</strong>dex features are nondistributive features.<br />
Thus, the set represent<strong>in</strong>g a coord<strong>in</strong>ate structure is allowed to have features that represent<br />
the agreement features for the coord<strong>in</strong>ate set. These are the <strong>in</strong>dex features.<br />
concord agreement is also def<strong>in</strong>ed by K<strong>in</strong>g and Dalrymple (2004) as follows:<br />
concord agreement with a coord<strong>in</strong>ate phrase depends on the concord features<br />
of each conjunct. (K<strong>in</strong>g and Dalrymple, 2004, 77)<br />
In head-modifier agreement, the concord features are usually number, gender and<br />
optionally case. In the example below from MG, the adjective needs to agree with each<br />
conjunct <strong>in</strong> case and num, otherwise the phrase is ungrammatical:<br />
(58) Eksipno<br />
clever.sg.nom<br />
‘Clever boy and girl’<br />
agori<br />
boy.sg.nom<br />
ke<br />
and<br />
koritsi<br />
girl.sg.nom<br />
The adjective is nom<strong>in</strong>ative s<strong>in</strong>gular and specifies <strong>in</strong> its lexical entry the feature values<br />
concord num sg and concord case nom. Therefore, each conjunct also needs to be<br />
nom<strong>in</strong>ative s<strong>in</strong>gular. The f-structure of the coord<strong>in</strong>ate phrase eksipno agori ke koritsi ‘clever<br />
boy and girl’ below shows the basic concepts of the concord feature and the features of<br />
the conjuncts:<br />
(59) eksipno agori ke koritsi<br />
eksipno: (↑concord num) = sg<br />
(↑concord case) = nom
2.4 <strong>Agreement</strong> <strong>in</strong> LFG 25<br />
⎡ <br />
<br />
⎤<br />
adj [pred ‘clever’]<br />
⎢<br />
⎥<br />
⎢<br />
⎥<br />
⎢<strong>in</strong>dex<br />
[num pl](nondistrib. feat. of c) ⎥<br />
⎢⎧<br />
⎢ ⎡<br />
⎤⎫<br />
⎥<br />
⎢ pred ‘boy’<br />
⎥<br />
⎢ ⎢ ⎥<br />
⎥<br />
⎢ ⎢<br />
⎢ b: ⎢<br />
num sg ⎥ ⎥<br />
⎢ ⎢concord<br />
⎥ ⎥<br />
c: ⎢ ⎣ case nom ⎥ ⎥<br />
⎦ ⎥<br />
⎢⎪⎨<br />
⎪⎬ ⎥<br />
⎢ <strong>in</strong>dex [num sg]<br />
⎥<br />
⎢ ⎡<br />
⎤ ⎥<br />
⎢ pred ‘girl’<br />
⎥<br />
⎢ ⎢ ⎥<br />
⎥<br />
⎢ ⎢<br />
⎢ g: ⎢<br />
num sg ⎥ ⎥<br />
⎢ ⎢concord<br />
⎥ ⎥<br />
⎣ ⎣ case nom ⎥ ⎥<br />
⎦ ⎥<br />
⎦<br />
⎪⎩ <strong>in</strong>dex [num sg]<br />
⎪⎭<br />
Thus, the agreement features of the adjective distribute to each conjunct and require<br />
feature match<strong>in</strong>g <strong>in</strong> order for the phrase to be grammatical. K<strong>in</strong>g and Dalrymple (2004)<br />
propose that concord features are distributive features and they are associated only with<br />
the <strong>in</strong>dividual members of the set represent<strong>in</strong>g the coord<strong>in</strong>ate phrase and not the set as a<br />
whole. These features appear as features of the <strong>in</strong>dividual conjuncts only.<br />
To summarise, <strong>in</strong> noun phrase coord<strong>in</strong>ation LFG assumes the dist<strong>in</strong>ction between two<br />
central features <strong>in</strong>dex and concord. concord features are distributive and are associated<br />
only with the <strong>in</strong>dividual conjuncts of a set of a coord<strong>in</strong>ate phrase. These are usually<br />
number, gender and case. <strong>in</strong>dex features are nondistributive mean<strong>in</strong>g that the set as<br />
a whole represent<strong>in</strong>g the coord<strong>in</strong>ate phrase, can have <strong>in</strong>dex features. These are usually<br />
person, number, gender and case (depend<strong>in</strong>g on the language) as well as any type<br />
of conjunction (Dalrymple and Kaplan, 2000). Thus, a coord<strong>in</strong>ate phrase as a whole is a<br />
hybrid object that apart from the elements it conta<strong>in</strong>s, it also has its own attributes and<br />
values.<br />
2.4.2 The Semantics of Coord<strong>in</strong>ation<br />
In this section, we will summarise semantics <strong>in</strong> coord<strong>in</strong>ate structures. Both sentential<br />
and subsentential coord<strong>in</strong>ate structures derive their mean<strong>in</strong>gs from the mean<strong>in</strong>gs of their<br />
conjuncts. A simple coord<strong>in</strong>ate structure and the mean<strong>in</strong>g representation is shown below:<br />
(60) Tom cried and Mary laughed<br />
cry(Tom) ∧ laugh(Mary)<br />
For the whole sentence to be true, it must be the case that both conjuncts are true. The<br />
follow<strong>in</strong>g mean<strong>in</strong>g constructor and f-structure are assumed for the above sentence, follow<strong>in</strong>g<br />
Dalrymple (2001):<br />
(61) cry(Tom) ∧ laugh(Mary): fσ<br />
⎧ ⎡<br />
⎤⎫<br />
pred ‘cry〈subj〉’<br />
⎪⎨<br />
p: ⎣ ⎦<br />
subj pred ‘tom’ ⎪⎬<br />
f: ⎡<br />
⎤<br />
pred ‘laugh〈subj〉’<br />
q: ⎣ ⎦<br />
⎪⎩ subj pred ‘mary’ ⎪⎭<br />
The conjunction ∧ above imposes the restriction that both propositions must be true for<br />
the whole proposition to be true. The mean<strong>in</strong>g constructor is as follows:
2.4 <strong>Agreement</strong> <strong>in</strong> LFG 26<br />
(62) [and] λX.λY.X∧Y: (↑ ∈)σ ⊸ [(↑ ∈)σ ⊸ ↑σ]<br />
(Dalrymple, 2001, 374)<br />
The glue side states that two semantic resources are required of type t represented by the<br />
(↑ ∈)σ symbol, which corresponds to each proposition. These semantic resources are<br />
elements of the set which is stated by the ∈ symbol. Once these resources are found and<br />
consumed only then we are able to create a semantic resource for the whole set, represented<br />
as ↑σ.<br />
The detailed mean<strong>in</strong>g constructors for the separate conjuncts and for the conjunction<br />
and are shown below as <strong>in</strong> Dalrymple (2001):<br />
(63) [cry] λX.cry(X) : hσ ⊸ pσ<br />
[Tom] Tom : hσ<br />
[laugh] λX.laugh(X) : iσ ⊸ qσ<br />
[Mary] Mary : iσ<br />
[and] λX.λY.X∧Y: pσ ⊸ [qσ ⊸ fσ]<br />
Thus, we first comb<strong>in</strong>e the mean<strong>in</strong>g constructors of the first conjunct [cry] and [Tom]<br />
then of the second conjunct [laugh] and [Mary] and f<strong>in</strong>ally the result of the two will<br />
be comb<strong>in</strong>ed with the conjunction [and] <strong>in</strong> order to give the whole coord<strong>in</strong>ate structure.<br />
Follow<strong>in</strong>g Dalrymple (2001), this is shown below:<br />
(64) [cry], [Tom] ⊢ cry(Tom) : pσ<br />
[laugh], [Mary] ⊢ laugh(Mary) : qσ<br />
[and], [Tom-cry], [Mary-laugh] ⊢ cry(Tom) ∧ laugh(Mary): fσ<br />
<strong>Noun</strong> phrase coord<strong>in</strong>ate structures follow a similar process with the ma<strong>in</strong> difference that<br />
the conjunction and is a group-form<strong>in</strong>g ‘and’. This type of and requires arguments of type<br />
e, which is the type used for <strong>in</strong>dividuals. The mean<strong>in</strong>g constructor of the group-form<strong>in</strong>g<br />
‘and’ is illustrated below:<br />
(65) [and] λX.λY.X,Y: (↑ ∈)σ ⊸ [(↑ ∈)σ ⊸ ↑σ]<br />
(Dalrymple, 2001, 384)<br />
The glue side states that two semantic resources are required of type e, which are<br />
members of the set and they are represented as (↑ ∈)σ which corresponds to each<br />
conjunct. Once these are found, they are consumed and therefore we can deduce the<br />
semantic resource of the whole coord<strong>in</strong>ate phrase, represented as ↑σ.<br />
A simple coord<strong>in</strong>ate noun phrase, such as Tom and Mary met, has the correspond<strong>in</strong>g<br />
mean<strong>in</strong>g constructor and f-structure:<br />
(66) meet({Tom,Mary}) : fσ<br />
⎡<br />
⎤<br />
pred ‘meet〈subj〉’<br />
⎢ ⎡<br />
⎤⎥<br />
⎢<br />
num pl ⎥<br />
⎢ ⎢<br />
⎥<br />
⎥<br />
⎢ ⎢conj<br />
and ⎥<br />
⎥<br />
f: ⎢ ⎢⎧<br />
⎢subj<br />
g: ⎢<br />
⎫⎥<br />
⎥<br />
⎢ ⎢⎪⎨<br />
h: pred ‘tom’<br />
⎪⎬<br />
⎥<br />
⎥<br />
⎥<br />
⎥<br />
⎢ ⎢<br />
⎣ ⎣ <br />
⎥<br />
⎥<br />
⎪⎩ i: pred ‘mary’ ⎪⎭<br />
⎦<br />
⎥<br />
⎦<br />
(Dalrymple, 2001, 384)<br />
The mean<strong>in</strong>g deduction for the above coord<strong>in</strong>ate phrase is as follows:
2.4 <strong>Agreement</strong> <strong>in</strong> LFG 27<br />
(67) [meet] λX.meet(X) : gσ ⊸ fσ<br />
[Tom] Tom : hσ<br />
[Mary] Mary : iσ<br />
[g-and] λX.λY.X,Y: hσ ⊸ [iσ ⊸ gσ]<br />
Once we comb<strong>in</strong>e the two nouns [Tom], [Mary], and the [g-and], we derive the premise<br />
[Tom-and-Mary] as follows:<br />
(68) [Tom-and-Mary] {Tom,Mary} : gσ<br />
The semantic resource gσ is required by the mean<strong>in</strong>g constructor of the verb [meet].<br />
Once the latter is comb<strong>in</strong>ed with the coord<strong>in</strong>ate phrase [Tom-and-Mary], we can derive<br />
the mean<strong>in</strong>g constructor for the whole sentence:<br />
(69) [Tom-and-Mary], [meet] ⊢ meet({Tom, Mary}) : fσ<br />
A similar process holds if one of the conjuncts is a quantifier. The mean<strong>in</strong>g for the<br />
simple sentence Tom and a student met is the follow<strong>in</strong>g accord<strong>in</strong>g to Dalrymple (2001):<br />
(70) Tom and a student met<br />
a(Y, student(Y), meet({Tom, Y}))<br />
The relevant mean<strong>in</strong>g constructor and f-structure for the sentence is shown below, based<br />
on Dalrymple (2001):<br />
(71) a(Y, student(Y), meet({Tom, Y})) : fσ<br />
⎡<br />
pred ‘meet〈subj〉’<br />
⎢ ⎡<br />
⎢<br />
num pl<br />
⎢ ⎢<br />
⎢ ⎢conj<br />
and<br />
⎢ ⎢⎧<br />
f: ⎢ ⎢<br />
<br />
⎢ ⎢<br />
⎢subj<br />
g: ⎢⎪⎨<br />
h: pred ‘tom’<br />
⎢ ⎢ ⎡<br />
⎢ ⎢<br />
⎢ ⎢<br />
⎣ ⎣ i: ⎣<br />
⎪⎩<br />
spec<br />
⎤<br />
⎤ ⎥<br />
⎥<br />
⎥<br />
⎥<br />
⎥<br />
⎫⎥<br />
⎥<br />
⎥<br />
⎥<br />
⎥<br />
⎥<br />
⎤<br />
⎪⎬ ⎥<br />
⎥<br />
⎥<br />
⎥<br />
pred ‘a’<br />
⎥<br />
⎥<br />
⎦<br />
⎥<br />
⎥<br />
⎪⎭<br />
⎦<br />
⎥<br />
⎦<br />
pred ‘student’<br />
(Dalrymple, 2001, 384)<br />
The mean<strong>in</strong>g deduction for the above coord<strong>in</strong>ate phrase is as follows 14 :<br />
(72) [meet] λX.meet(X) : gσ ⊸ fσ<br />
[Tom] Tom : hσ<br />
[a-student] λS.a(Y, student(Y),S(Y)) : ∀H.[iσ ⊸ H] ⊸ H<br />
[g-and] λX.λY.X,Y: hσ ⊸ [iσ ⊸ gσ]<br />
First, we comb<strong>in</strong>e the mean<strong>in</strong>g [Tom] and [g-and] <strong>in</strong> order to produce the mean<strong>in</strong>g constructor<br />
[Tom-and]:<br />
(73) [Tom-and] λY.{Tom, Y} : iσ ⊸ gσ<br />
14 The universal quantifier ∀ means all or every. It b<strong>in</strong>ds the variable H and ranges over semantic structures<br />
which correspond to different scopes of the quantifier. Also, the expression [iσ ⊸ H] ⊸ H asserts that if a<br />
resource of any H is found that satisfies [iσ ⊸ H] then H can be obta<strong>in</strong>ed (Partee et al., 1993, Chapter 7).
2.5 Summary 28<br />
Next, the follow<strong>in</strong>g abstraction rule is used which allows us to <strong>in</strong>troduce and then discharge<br />
a hypothetical premise <strong>in</strong> the deduction:<br />
(74) X : [iσ]<br />
.<br />
P(X) : fσ<br />
λX.P(X) : iσ ⊸ fσ<br />
(Dalrymple, 2001, 236)<br />
The hypothetical premise X : [iσ ] is <strong>in</strong>troduced which can be comb<strong>in</strong>ed with the premises<br />
[Tom-and] and [meet] <strong>in</strong> order to give the mean<strong>in</strong>g constructor [Tom-and-X-meet].<br />
This is shown below:<br />
(75) X : [iσ ] [Tom-and]<br />
{Tom, X} : gσ<br />
[meet]<br />
meet({Tom, X}) : fσ<br />
[Tom-and-X-meet] λX.meet({Tom, X}) : iσ ⊸ fσ<br />
(Dalrymple, 2001, 386)<br />
F<strong>in</strong>ally, if we comb<strong>in</strong>e the premise [Tom-and-X-meet] with the premise [a-student] we<br />
derive the whole phrase:<br />
(76) [Tom-and-X-meet], [a-student] ⊢ a(Y, student(Y), meet({Tom,Y})): fσ<br />
(Dalrymple, 2001, 386)<br />
The above section presented a brief summary of how semantic composition works <strong>in</strong><br />
coord<strong>in</strong>ate structures us<strong>in</strong>g l<strong>in</strong>ear logic and <strong>in</strong> particular ‘glue’ semantics. The basic idea<br />
is to derive the mean<strong>in</strong>g of a sentence by comb<strong>in</strong><strong>in</strong>g the mean<strong>in</strong>gs of its parts, which are<br />
represented as premises. Hence, through a series of logical deductions we can derive the<br />
mean<strong>in</strong>g of the sentence.<br />
2.5 Summary<br />
This chapter briefly presented the central concepts <strong>in</strong> the formalism of Lexical-Functional<br />
Grammar. We focused on the syntactic and semantic levels of representation, analys<strong>in</strong>g<br />
agreement <strong>in</strong> coord<strong>in</strong>ate constructions. We tried to describe the central notions of sentential<br />
and non-sentential coord<strong>in</strong>ation, which generally follow the same mechanism. The rest of<br />
the thesis deals with non-sentential noun phrase coord<strong>in</strong>ation. The next chapter focuses on<br />
a review of the most important theoretical approaches to predicate-argument agreement<br />
and head-modifier agreement that capture various phenomena <strong>in</strong> noun phrase coord<strong>in</strong>ation.
Chapter 3<br />
Theories of <strong>Agreement</strong><br />
3.1 Introduction<br />
Several theories have been developed concern<strong>in</strong>g the phenomenon of agreement <strong>in</strong> noun<br />
phrase coord<strong>in</strong>ate constructions. Most theories focus on the gender, number and person<br />
features. The current section presents an overview of the theories that analyse the gender<br />
and number agreement features <strong>in</strong> coord<strong>in</strong>ate nouns. Predicate-argument agreement and<br />
head-modifier agreement will be presented with reference to the features of gender and<br />
number, respectively.<br />
In predicate-argument agreement, gender ‘resolution’ (Givon, 1970) will be discussed<br />
s<strong>in</strong>ce it presents crossl<strong>in</strong>guistic and language-specific variation. The central notion concern<strong>in</strong>g<br />
gender agreement resolution, as proposed by a number of l<strong>in</strong>guists (Corbett, 1991;<br />
Wechsler and Zlatić, 2003; K<strong>in</strong>g and Dalrymple, 2004), is whether the resolved form is<br />
based on a syntactic pr<strong>in</strong>ciple, a semantic pr<strong>in</strong>ciple or both pr<strong>in</strong>ciples.<br />
In head-modifier agreement, the number feature will be the ma<strong>in</strong> focus. number<br />
agreement of determ<strong>in</strong>ers with coord<strong>in</strong>ate nouns shows unexpected variation s<strong>in</strong>ce there are<br />
languages, such as English, where number agreement is determ<strong>in</strong>ed syntactically and other<br />
languages, such as Russian, where syntax or semantics determ<strong>in</strong>e number agreement. The<br />
person feature does not display any special characteristics at least with<strong>in</strong> the languages<br />
presented here and it will not be discussed <strong>in</strong> any detail.<br />
3.2 Theoretical Approaches to Predicate-Argument <strong>Agreement</strong><br />
gender resolution has been mostly discussed <strong>in</strong> theoretical approaches to predicate-argument<br />
agreement. The term ‘resolution’ (Givon, 1970) refers to any rule that specifies the form<br />
of an agree<strong>in</strong>g element, known as the target, when the controller, the element which determ<strong>in</strong>es<br />
the agreement, consists of conjo<strong>in</strong>ed nouns. A characteristic approach to gender<br />
resolution is that of Corbett (1991) which uses a list of statements <strong>in</strong> the form of sets of resolution<br />
rules for the agreement patterns of different languages. Other approaches <strong>in</strong>clude<br />
those of Dalrymple and Kaplan (2000), Wechsler and Zlatić (2003, Chap. 8) and Sadler<br />
(2006), which use the set-based approach, but they differ significantly <strong>in</strong> the operation they<br />
use s<strong>in</strong>ce they resort either to union or to <strong>in</strong>tersection. F<strong>in</strong>ally, a fairly recent proposal<br />
on predicate-argument agreement <strong>in</strong> coord<strong>in</strong>ate constructions is that of Badecker (2007,<br />
2008), based on the <strong>in</strong>teract<strong>in</strong>g effects of ranked constra<strong>in</strong>ts with<strong>in</strong> Optimality Theoretic<br />
Syntax. Each of these theories will be presented and discussed <strong>in</strong> this chapter.<br />
3.2.1 Corbett’s Descriptive Approach<br />
Corbett (1983b, 1991) <strong>in</strong>troduces a set of resolution rules for the features person, number<br />
and gender. Concern<strong>in</strong>g the person and number features, Corbett (1983b) argues that
3.2 Theoretical Approaches to Predicate-Argument <strong>Agreement</strong> 30<br />
resolution is straightforward s<strong>in</strong>ce these features show little variation crossl<strong>in</strong>guistically<br />
<strong>in</strong> their resolved forms. person resolution, for example, follows the precedence hierarchy<br />
where “the first person takes precedence over the second, and the second over the<br />
third”(Corbett, 1991, 262), (Siewierska, 2004). person resolution rules are universal s<strong>in</strong>ce<br />
they occur frequently across languages and they also match the hierarchy of reference,<br />
which is able to constra<strong>in</strong> the pronom<strong>in</strong>al system, as proposed by Zwicky (1977, 718, 725).<br />
The person resolution rules occur <strong>in</strong> a specific order and apply <strong>in</strong> that same order:<br />
Person Resolution Rules<br />
1. If the conjuncts <strong>in</strong>clude a first person, first person agreement forms will be<br />
used;<br />
2. If the conjuncts <strong>in</strong>clude a second person, second person agreement forms will<br />
be used;<br />
(The default condition is that third person agreement forms are used.)<br />
(Corbett, 1991, 262)<br />
Also, accord<strong>in</strong>g to Corbett (1983b, 1991, 2000), <strong>in</strong> number resolution conjo<strong>in</strong>ed s<strong>in</strong>gular<br />
elements need plural number. His number resolution rules cover languages with and<br />
without a dual number:<br />
Number Resolution Rules<br />
1. If there are two conjuncts only, both of which are <strong>in</strong> the s<strong>in</strong>gular, then dual<br />
agreement form will be used;<br />
2. In all other cases, provid<strong>in</strong>g there is at least one non-plural conjunct or three<br />
conjuncts, plural agreement forms will be used;<br />
(Corbett, 1991, 263)<br />
Languages without a dual number must follow only the second resolution rule but<br />
languages with a dual number need to follow the first one too. Thus, for a language like<br />
Slovene, a South Slavonic language with s<strong>in</strong>gular, dual and plural number <strong>in</strong> its system,<br />
if two s<strong>in</strong>gulars are conjo<strong>in</strong>ed, the verb appears <strong>in</strong> the dual. If there are more than two<br />
nouns or if one of them is dual or plural, then the predicate is plural.<br />
Gender, on the other hand, is a much more complex feature s<strong>in</strong>ce it shows great variation<br />
across languages and poses problems of whether its resolved form is semantically<br />
or syntactically driven. The central notion of Corbett’s theory related to gender is the<br />
existence of three different strategies <strong>in</strong> gender resolution, a syntactic, a semantic and a<br />
mixed strategy, comb<strong>in</strong><strong>in</strong>g the two former.<br />
Syntactic resolution means “the gender of the conjo<strong>in</strong>ed nouns <strong>in</strong>volved is what counts,<br />
rather than their mean<strong>in</strong>g” (Corbett, 1991, p.279). Corbett (1991) claims that French,<br />
Spanish, Latvian, H<strong>in</strong>di, Panjabi and <strong>Modern</strong> Hebrew display syntactic gender resolution.<br />
French, for example, is a language with two genders, mascul<strong>in</strong>e and fem<strong>in</strong><strong>in</strong>e. If a noun<br />
phrase consists of a mascul<strong>in</strong>e noun, the mascul<strong>in</strong>e form is used, otherwise the fem<strong>in</strong><strong>in</strong>e is<br />
used. The fact that resolution rules are the same <strong>in</strong> animate and <strong>in</strong>animate nouns leads<br />
Corbett (1991) to the conclusion that French 1 follows the syntactic pr<strong>in</strong>ciple of resolution.<br />
The follow<strong>in</strong>g examples display the above:<br />
(1) le<br />
a<br />
livre et le cahier<br />
sont neufs<br />
book.masc and an excercise-book.masc are new.masc.pl<br />
‘the book and excercise book are new’<br />
1 We shall see that this conclusion is disputed by Wechsler and Zlatić (2003) and Wechsler (2009), who<br />
present a set-based semantic theory of agreement and assert that French shows mixed gender resolution<br />
accord<strong>in</strong>g to which animate nouns follow the semantic pr<strong>in</strong>ciple and <strong>in</strong>animate nouns follow the syntactic<br />
pr<strong>in</strong>ciple.
3.2 Theoretical Approaches to Predicate-Argument <strong>Agreement</strong> 31<br />
(2) le<br />
a<br />
père et la mère sont excellents<br />
father.masc and a mother.fem are excellent.masc<br />
‘the father and mother are excellent’<br />
(Corbett, 1991, p.279)<br />
The resolution rules <strong>in</strong> French are stated as follows:<br />
Gender Resolution Rules for French<br />
1. If at least one conjunct is mascul<strong>in</strong>e, the mascul<strong>in</strong>e form is used;<br />
2. Otherwise the fem<strong>in</strong><strong>in</strong>e is used;<br />
(Corbett, 1991, 279)<br />
Semantic gender resolution “<strong>in</strong>volves reference to the mean<strong>in</strong>g of the conjo<strong>in</strong>ed elements<br />
even if this implies disregard for their syntactic gender”(Corbett, 1991, p.269). In other<br />
words the <strong>in</strong>herent mean<strong>in</strong>g of the conjuncts determ<strong>in</strong>es the resolution form to be used and<br />
not their grammatical gender. Corbett (1991) claims that this type is found <strong>in</strong> Dravidian<br />
and Bantu languages. Tamil, for <strong>in</strong>stance, is a language with three genders, mascul<strong>in</strong>e<br />
(for nouns denot<strong>in</strong>g male rationals), fem<strong>in</strong><strong>in</strong>e (for female rationals) and neuter (for nonrationals)<br />
<strong>in</strong> the s<strong>in</strong>gular and two agreement forms <strong>in</strong> the plural, the rational (for mascul<strong>in</strong>es<br />
and fem<strong>in</strong><strong>in</strong>es) and the neuter. When all conjuncts denote rationals (whether of the same<br />
or mixed gender) the rational form is used, otherwise if the conjuncts are neuter, the neuter<br />
agreement is used. Thus, Corbett (1991) characterises this type of resolution as semantic<br />
<strong>in</strong> nature and presents the follow<strong>in</strong>g examples:<br />
(3) raaman-um murukan-um va-nt-aaÒka<br />
Raman-and Murugan-and come-past-3rd.pl.rational<br />
‘Raman and Murugan came’<br />
(4) naay-um<br />
dog-and<br />
puune-yum<br />
cat-and<br />
‘The dog and the cat came’<br />
(Corbett, 1991, 269)<br />
va-nt-atuÒka<br />
come-past-3rd.pl.neut<br />
The resolution rules are presented as follows:<br />
Gender Resolution Rules for Tamil<br />
1. If all conjuncts denote rationals, the rational form is used;<br />
2. If all conjuncts denote non-rationals, the neuter form is used;<br />
3. Otherwise the rational plural may be used<br />
(Corbett, 1991, 285)<br />
The third type, accord<strong>in</strong>g to Corbett (1991), comb<strong>in</strong>es both the syntactic and semantic<br />
resolution pr<strong>in</strong>ciples. He claims that <strong>in</strong> languages like Polish, Lat<strong>in</strong> and Rumanian these<br />
two strategies coexist. He argues that persons display semantic resolution and the rest of<br />
the nouns show syntactic resolution. In Lat<strong>in</strong>, for <strong>in</strong>stance, when the coord<strong>in</strong>ate nouns<br />
have a different gender feature then resolution is determ<strong>in</strong>ed based on whether the nouns<br />
are animate or not. If they are animate the mascul<strong>in</strong>e form is used, otherwise neuter is the<br />
resolved gender. These are shown below:<br />
(5) quam pridem pater mihi<br />
how long-ago father.masc me.dat<br />
mortu-i essent<br />
dead-masc.pl were<br />
‘How long ago my father and mother had died’<br />
et<br />
and<br />
mater<br />
mother.fem
3.2 Theoretical Approaches to Predicate-Argument <strong>Agreement</strong> 32<br />
(6) murus<br />
wall.masc<br />
et<br />
and<br />
porta<br />
gate.fem<br />
de<br />
from<br />
caelo<br />
sky<br />
‘The wall and the gate had been struck by lightn<strong>in</strong>g’<br />
(Corbett, 1991, 287)<br />
The resolution rules are presented as below:<br />
tact-a<br />
struck-neut.pl<br />
Gender Resolution Rules for Lat<strong>in</strong><br />
1. If all conjuncts are mascul<strong>in</strong>e, the mascul<strong>in</strong>e is used;<br />
2. If all conjuncts are fem<strong>in</strong><strong>in</strong>e, the fem<strong>in</strong><strong>in</strong>e is used;<br />
3. If all conjuncts denote humans, then the mascul<strong>in</strong>e is used;<br />
4. Otherwise the neuter is used<br />
(Corbett, 1991, 285)<br />
erant<br />
were<br />
The first two rules display feature-match<strong>in</strong>g or syntactic resolution, the fourth rule is a<br />
case of syntactic resolution, whereas the third rule follows the semantic pr<strong>in</strong>ciple s<strong>in</strong>ce the<br />
resolved form is determ<strong>in</strong>ed by the human/non-human dist<strong>in</strong>ction. Other languages such<br />
as Polish and Rumanian that fall with<strong>in</strong> the same category are analysed <strong>in</strong> more detail <strong>in</strong><br />
Corbett (1983b, 1991).<br />
A more formal analysis of the above descriptive approach is found <strong>in</strong> Corbett and<br />
Fraser (1993). Corbett and Fraser (1993) propose a prelim<strong>in</strong>ary formal analysis with<strong>in</strong><br />
Network Morphology where they present a DATR account of Russian nom<strong>in</strong>al <strong>in</strong>flections.<br />
To account for animacy <strong>in</strong> Russian nom<strong>in</strong>al morphology, Corbett and Fraser (1993) express<br />
the ‘prediction rules’ as ‘feature change rules’(130). The latter change the value of certa<strong>in</strong><br />
feature comb<strong>in</strong>ations and are extremely powerful. Corbett and Fraser (1993) reta<strong>in</strong> the idea<br />
of the copy<strong>in</strong>g rule <strong>in</strong> its most general form while the ma<strong>in</strong> work done by the feature-change<br />
rules is given to default statements under nom<strong>in</strong>al. Some of the defaults are overridden<br />
<strong>in</strong> particular parts of <strong>in</strong>dividual declension classes 2 .<br />
To summarise, Corbett (1991) provides a series of descriptive rules for all features<br />
person, number and gender and captures a range of languages. However, he does not<br />
attempt a formal approach to these descriptive rules until later <strong>in</strong> Corbett and Fraser<br />
(1993) where he <strong>in</strong>troduces a prelim<strong>in</strong>ary analysis.<br />
3.2.2 Dalrymple and Kaplan’s Set-Based Theory<br />
Dalrymple and Kaplan (2000) attempt a formal account of feature resolution. They propose<br />
a set-based theory of person and gender resolution <strong>in</strong> coord<strong>in</strong>ate constructions with<strong>in</strong><br />
the constra<strong>in</strong>t-based LFG framework. The central problem beh<strong>in</strong>d resolution is that a<br />
coord<strong>in</strong>ate noun phrase has its own feature-values for person, number and gender, as<br />
any other noun or pronoun, dist<strong>in</strong>ct from the conjuncts’ feature-values. In the follow<strong>in</strong>g<br />
example, the first conjunct is third(3) person s<strong>in</strong>gular(sg) and the second conjunct is<br />
first(1) person s<strong>in</strong>gular(sg). The predicate resolves to features not found <strong>in</strong> any of the two<br />
conjuncts, the first person plural(1.pl):<br />
(7) Jose y yo hablamos/*habláis/*hablan<br />
Jose and I speak-1pl/*speak-2pl/*speak-3pl<br />
(Dalrymple and Kaplan, 2000, 778)<br />
In the f-structure below, the first conjunct has different feature-values from the second<br />
conjunct while the whole coord<strong>in</strong>ate phrase has its own feature-values which are different<br />
from the two conjuncts:<br />
2 Details of the defaults are shown <strong>in</strong> Corbett and Fraser (1993, 131).
3.2 Theoretical Approaches to Predicate-Argument <strong>Agreement</strong> 33<br />
⎡⎧<br />
⎡ ⎤⎫⎤<br />
(8)<br />
⎢<br />
pred ‘Jose’<br />
⎢ ⎢ ⎥ ⎥<br />
⎢ f: ⎣person<br />
3 ⎦ ⎥<br />
⎢⎪⎨<br />
⎪⎬ ⎥<br />
⎢ num sg ⎥<br />
⎢ ⎡ ⎤ ⎥<br />
⎢ pred ‘pro’ ⎥<br />
⎢<br />
fi⎢<br />
⎢ ⎥ ⎥<br />
⎢<br />
i: ⎣person<br />
1 ⎦ ⎥<br />
⎢⎪⎩<br />
⎪⎭<br />
⎥<br />
⎢ num sg ⎥<br />
⎢<br />
⎥<br />
⎢<br />
⎢person<br />
1 ⎥<br />
⎢<br />
⎥<br />
⎣num<br />
pl ⎦<br />
conj and<br />
(Dalrymple and Kaplan, 2000, 785)<br />
To account for resolution phenomena <strong>in</strong> coord<strong>in</strong>ate nouns and the dist<strong>in</strong>ct feature-values<br />
of the conjuncts from the coord<strong>in</strong>ate phrase as a whole, Dalrymple and Kaplan (2000) use<br />
set-valued features, assert<strong>in</strong>g that “sets encode complex values” (p.780). They assign set<br />
values to the agreement attributes person and gender of the <strong>in</strong>dividual conjuncts. The<br />
first, second and third persons are represented with any of the follow<strong>in</strong>g marker sets, such<br />
as the empty set {}, the s<strong>in</strong>gleton sets {s} (for speakers), {h} (for hearer) 3 or the two<br />
member set {s,h}. Similarly, the mascul<strong>in</strong>e, fem<strong>in</strong><strong>in</strong>e and neuter genders are represented<br />
with any of the marker sets, such as the empty set {}, the s<strong>in</strong>gleton sets {m}, {f} or<br />
{n} and the two member sets {m,f}, {f,n} or {m,n}. We need to note that the use and<br />
variation of the set designators depends on the language itself and whether it <strong>in</strong>cludes all<br />
the person and gender values mentioned above.<br />
Once the sets are assigned to the features of <strong>in</strong>dividual conjuncts, the resolved feature<br />
value of the coord<strong>in</strong>ate phrase is the set of the union of the values of the conjunct daughters.<br />
The union operator constructs the smallest set and ensures that all the members of the<br />
<strong>in</strong>put sets are <strong>in</strong>cluded with<strong>in</strong> that. Dalrymple and Kaplan (2000) propose a def<strong>in</strong>ition of<br />
set union:<br />
(9) x ∪ y is the smallest set z such that x ⊆ z and y ⊆ z<br />
(Dalrymple and Kaplan, 2000, 785)<br />
This proposition states that “the union will result if we take the smallest set that satisfies<br />
a collection of separately stated mother-daughter subset assertions” (Dalrymple and<br />
Kaplan, 2000, p.785). Dalrymple and Kaplan (2000) argue that LFG can accommodate<br />
both the condition of the smallest set, us<strong>in</strong>g the m<strong>in</strong>imal f-structures for the functional<br />
description of an utterance, and the mother-daughter subset relation, us<strong>in</strong>g the ↑ and<br />
↓ metavariables. Thus, they <strong>in</strong>troduce the follow<strong>in</strong>g PS rule with the relevant annotations:<br />
(10) NP → NP CONJ NP<br />
↓∈↑ ↓∈↑<br />
(↓person) ⊆ (↑person) (↓person) ⊆ (↑person)<br />
(↓gender) ⊆ (↑gender) (↓gender) ⊆ (↑gender)<br />
(Dalrymple and Kaplan, 2000, 788)<br />
The first l<strong>in</strong>e, already known from coord<strong>in</strong>ation of categories, states that each NP<br />
conjunct daughter needs to be a member of the conjunct set of the mother node. The<br />
second l<strong>in</strong>e states that the person value of the coord<strong>in</strong>ate phrase will form the smallest<br />
set conta<strong>in</strong><strong>in</strong>g the values of the <strong>in</strong>dividual conjuncts, and the same statement is applied to<br />
gender.<br />
3 For languages with more person features, such as the <strong>in</strong>clusive or exclusive, more marker sets are<br />
required. For details see Dalrymple and Kaplan (2000, 780-83).
3.2 Theoretical Approaches to Predicate-Argument <strong>Agreement</strong> 34<br />
Dalrymple and Kaplan (2000) apply their theory to different languages by present<strong>in</strong>g<br />
person resolution <strong>in</strong> Fula, English, Spanish and Slovak, and gender resolution <strong>in</strong> H<strong>in</strong>di.<br />
In person resolution, they focus on the syntactic dist<strong>in</strong>ctions <strong>in</strong> pronouns and not on the<br />
referential ones. For languages like English, they assume that the person feature has sets<br />
whose elements are drawn from the markers s and h. Thus, 1st person is represented as<br />
the set {s,h}, 2nd person as the set {h} and 3rd person as the empty set {}:<br />
(11) English, Spanish and Slovak<br />
{s,h}: first person<br />
{h}: second person<br />
{}: third person<br />
(Dalrymple and Kaplan, 2000, 784)<br />
The agreement patterns for the above languages are shown below:<br />
(12) English, Spanish and Slovak<br />
1 & 2 = 1<br />
1 & 3 = 1<br />
2 & 3 = 2<br />
3 & 3 = 3<br />
(Dalrymple and Kaplan, 2000, 784)<br />
These agreement patterns state that when a 1st person is coord<strong>in</strong>ated with either a<br />
second person or a third person resolution is 1st person, when a 2nd person is coord<strong>in</strong>ated<br />
with a 3rd person pronoun resolution is 2nd person, and when both conjuncts are 3rd<br />
person resolution is 3rd person.<br />
The table below displays how the union of the conjunct sets produces the set of the<br />
coord<strong>in</strong>ate phrase:<br />
(13) {s,h}(1) ∪ {h}(2) = {s,h}(1)<br />
{s,h}(1) ∪ {}(3) = {s,h}(1)<br />
{h}(2) ∪ {}(3) = {h}(2)<br />
{}(3) ∪ {}(3) = {}(3)<br />
(Dalrymple and Kaplan, 2000, 784)<br />
In example (14) from Spanish, the third person s<strong>in</strong>gular noun comb<strong>in</strong>es with the second<br />
person s<strong>in</strong>gular pronoun. As predicted by the union of the two conjuncts, the resolution<br />
must be second person plural.<br />
(14) Jose y tu hablais<br />
Jose-3.sg and you-2.sg speak-2.pl<br />
The relevant f-structure is presented below:<br />
(15) Jose y tu hablais
3.2 Theoretical Approaches to Predicate-Argument <strong>Agreement</strong> 35<br />
⎡<br />
⎤<br />
pred ‘speak’<br />
⎢ ⎡<br />
⎢<br />
⎧ ⎡ ⎤⎫⎤<br />
⎥<br />
⎢ ⎢<br />
pred ‘Jose’<br />
⎢ ⎢ ⎢ ⎥ ⎥<br />
⎥<br />
⎢ ⎢ j: ⎣person<br />
{} ⎦<br />
⎥<br />
⎥<br />
⎥<br />
⎥<br />
⎢ ⎢<br />
⎢ ⎢⎪⎨<br />
num sg ⎪⎬<br />
⎥<br />
⎥<br />
⎥<br />
⎥<br />
⎢ ⎢<br />
⎢ ⎢ ⎡ ⎤ ⎥<br />
⎥<br />
⎥<br />
⎥<br />
f ⎢ ⎢ pred ‘pro’ ⎥<br />
⎥<br />
⎢ ⎢<br />
⎢subj<br />
c: ⎢ ⎢ ⎥<br />
⎥<br />
⎥<br />
⎢ ⎢ i: ⎢<br />
⎢ ⎢ ⎣person<br />
h ⎥ ⎥<br />
⎥<br />
⎦ ⎥<br />
⎥<br />
⎢ ⎢⎪⎩<br />
⎪⎭<br />
⎥<br />
⎥<br />
⎢ ⎢ num sg<br />
⎥<br />
⎥<br />
⎥<br />
⎥<br />
⎢ ⎢<br />
⎢ ⎢ ⎥<br />
⎥<br />
⎥<br />
⎥<br />
⎢ ⎢<br />
⎣ ⎣<br />
person h ⎥<br />
⎥<br />
⎦<br />
⎥<br />
⎦<br />
num pl<br />
(Dalrymple and Kaplan, 2000, 785)<br />
Thus, the person feature of the above coord<strong>in</strong>ate noun phrase is the union of the person<br />
features of the conjuncts. The relevant marker sets appear <strong>in</strong> the two conjuncts. Those<br />
sets are subsets of the smallest set that appears <strong>in</strong> the coord<strong>in</strong>ate NP.<br />
In gender resolution, Dalrymple and Kaplan (2000) acknowledge the existence of both<br />
a syntactic and a semantic resolved gender form, although they focus only on syntactic<br />
gender resolution. A case of semantic resolution where the two conjuncts are fem but the<br />
resolved form is masc, is shown below:<br />
(16) [La personne avec la barbe]<br />
[the person with the beard].fem<br />
idiots/?*idiotes<br />
idiots.masc/?*fem<br />
‘The person with the beard and Marie are idiots’<br />
(Wechsler, 2009)<br />
et<br />
and<br />
Marie<br />
Marie.fem<br />
sont<br />
are<br />
For syntactic gender resolution, they present H<strong>in</strong>di as an example, which is a two gender<br />
system, consist<strong>in</strong>g of mascul<strong>in</strong>e and fem<strong>in</strong><strong>in</strong>e nouns. In H<strong>in</strong>di, when a mascul<strong>in</strong>e and a<br />
fem<strong>in</strong><strong>in</strong>e noun are coord<strong>in</strong>ated, the coord<strong>in</strong>ate phrase has a mascul<strong>in</strong>e gender, as <strong>in</strong> (17).<br />
When the two nouns are fem<strong>in</strong><strong>in</strong>e the coord<strong>in</strong>ate phrase is fem<strong>in</strong><strong>in</strong>e, as <strong>in</strong> (18):<br />
(17) H<strong>in</strong>di<br />
meraa<br />
my<br />
m˜e<br />
loc<br />
kuttaa aur merii<br />
dog.masc and my<br />
rahte<br />
hãĩ<br />
live.subj gend=masc.pl<br />
billii<br />
cat.fem<br />
‘My dog and my cat live with me <strong>in</strong> the house’<br />
(18) yah larki aur uski mãã<br />
this girl.fem and her mother.fem<br />
rahtii<br />
hãĩ<br />
live.subj gend=fem<br />
‘This girl and her mother live <strong>in</strong> Delhi’<br />
(Dalrymple and Kaplan, 2000, 788-9)<br />
mere<br />
with<br />
dilli<br />
Delhi<br />
saath<br />
me<br />
m˜e<br />
loc<br />
g h ar<br />
house<br />
The above observation is <strong>in</strong>formally presented as a set of rules for gender resolution<br />
below:<br />
(19) H<strong>in</strong>di<br />
· If the conjuncts <strong>in</strong>clude at least one mascul<strong>in</strong>e noun phrase, the mascul<strong>in</strong>e form
3.2 Theoretical Approaches to Predicate-Argument <strong>Agreement</strong> 36<br />
is used.<br />
· Otherwise the fem<strong>in</strong><strong>in</strong>e form is used.<br />
(Dalrymple and Kaplan, 2000, 789)<br />
The agreement patterns are shown below:<br />
(20) H<strong>in</strong>di<br />
masc & masc= masc<br />
masc & fem= masc<br />
fem & fem= fem<br />
(Dalrymple and Kaplan, 2000, 789)<br />
Due to the two-way gender system, the marker sets are only two, us<strong>in</strong>g a s<strong>in</strong>gle primitive<br />
marker {m}:<br />
(21) H<strong>in</strong>di<br />
masc{m}<br />
fem{}<br />
(Dalrymple and Kaplan, 2000, 789)<br />
Apply<strong>in</strong>g the union operation, they draw the appropriate resolution results for H<strong>in</strong>di:<br />
(22) H<strong>in</strong>di<br />
{m}(masc) ∪ {m}(masc) = {m}(masc)<br />
{m}(masc) ∪ {}(fem) = {m}(masc)<br />
{}(fem) ∪ {}(fem) = {}(fem)<br />
(Dalrymple and Kaplan, 2000, 789)<br />
Dalrymple and Kaplan (2000), us<strong>in</strong>g the feature set representation account for agreement<br />
phenomena <strong>in</strong> a number of other languages, <strong>in</strong>clud<strong>in</strong>g Icelandic a language with a<br />
three gender system. In the last section, we will discuss some disadvantages of their theory,<br />
such as that feature assignment is not cross-l<strong>in</strong>guistic but different languages assume<br />
different set-valued features (V<strong>in</strong>cent and Börjars, 2000) and that Slovene and MG, which<br />
have more complicated gender systems, cannot be captured properly by the current theory<br />
without avoid<strong>in</strong>g the addition of further constra<strong>in</strong>ts.<br />
3.2.3 Wechsler’s Theory of Gender Resolution<br />
Wechsler and Zlatić (2000) and Wechsler and Zlatić (2003, Chap. 8) present an account for<br />
gender resolution which <strong>in</strong>cludes both the semantic and syntactic pr<strong>in</strong>ciples. They <strong>in</strong>vestigate<br />
three languages French, Serbian/Croatian and Icelandic and conclude that <strong>in</strong>animate<br />
coord<strong>in</strong>ate nouns are subject to syntactic resolution and animate coord<strong>in</strong>ate nouns are<br />
subject to semantic resolution. They observe that semantic resolution <strong>in</strong> animate coord<strong>in</strong>ate<br />
NPs is a crossl<strong>in</strong>guistic phenomenon, even though non-coord<strong>in</strong>ate animate/<strong>in</strong>animate<br />
nouns follow grammatical gender.<br />
Evidence for the existence of a semantic resolution pr<strong>in</strong>ciple <strong>in</strong> animate coord<strong>in</strong>ate<br />
nouns is derived from all three languages. French, for <strong>in</strong>stance, shows fem<strong>in</strong><strong>in</strong>e resolution<br />
when the conjuncts are fem<strong>in</strong><strong>in</strong>e, and mascul<strong>in</strong>e is the default gender <strong>in</strong> all other cases, <strong>in</strong><br />
both animate and <strong>in</strong>animate coord<strong>in</strong>ate nouns:<br />
(23) French rule (M, F):<br />
F.PL: F + F<br />
M.PL: elsewhere<br />
(Wechsler and Zlatić, 2003, 173)
3.2 Theoretical Approaches to Predicate-Argument <strong>Agreement</strong> 37<br />
Accord<strong>in</strong>g to Wechsler and Zlatić (2003), the follow<strong>in</strong>g examples are a clear evidence<br />
that resolution <strong>in</strong> animate nouns obeys the semantic pr<strong>in</strong>ciple. In example (24), the noun<br />
la sent<strong>in</strong>elle ‘sentry’ is grammatically fem<strong>in</strong><strong>in</strong>e and it can refer to both males and females.<br />
When that noun is coord<strong>in</strong>ated with the grammatically fem<strong>in</strong><strong>in</strong>e noun la femme ‘woman’,<br />
the result is nonetheless mascul<strong>in</strong>e resolution:<br />
(24) La sent<strong>in</strong>elle et sa femme<br />
the.fem.sg sentry and his wife.fem.sg<br />
pris/*prises en otage<br />
taken.masc/*fem.pl as hostage<br />
‘The sentry and his wife were taken hostage’<br />
(Wechsler and Zlatić, 2003, 177)<br />
ont été<br />
were<br />
Similarly, <strong>in</strong> example (25), the noun mannequ<strong>in</strong> ‘fashion model’ is grammatically mascul<strong>in</strong>e<br />
but it also refers to a female. When its referent is fem<strong>in</strong><strong>in</strong>e and it is conjo<strong>in</strong>ed with the<br />
grammatically fem<strong>in</strong><strong>in</strong>e noun maquilleuse ‘make-up artist’, the result triggers fem<strong>in</strong><strong>in</strong>e<br />
plural agreement:<br />
(25) Le mannequ<strong>in</strong> et sa maquilleuse<br />
the.masc fashion-model and her make-up-artist.fem.sg<br />
assises dans le co<strong>in</strong><br />
seated.fem.pl <strong>in</strong> the corner<br />
‘The fashion model and her make-up artist are seated <strong>in</strong> the corner’<br />
(Wechsler and Zlatić, 2003, 178)<br />
sont<br />
are<br />
The above patterns show that French animate coord<strong>in</strong>ate nouns show semantic resolution.<br />
Inanimate nouns follow the syntactic pr<strong>in</strong>ciple of resolution with the patterns shown <strong>in</strong><br />
(23).<br />
Serbian/Croatian has similar resolution rules to French but it is a three gender system,<br />
<strong>in</strong>clud<strong>in</strong>g a neuter gender. Animate and <strong>in</strong>animate nouns follow the resolution rules below:<br />
(26) Serbian/Croatian rule (M, F, NT):<br />
F.PL: F + F<br />
M.PL: elsewhere<br />
(Wechsler and Zlatić, 2003, 174)<br />
Evidence that resolution is sex based and not grammatically based <strong>in</strong> animate nouns<br />
is drawn from examples where the grammatically neuter noun devojče ‘little girl’ is coord<strong>in</strong>ated<br />
with the grammatically fem<strong>in</strong><strong>in</strong>e noun devojka ‘girl’ and they resolve to fem<strong>in</strong><strong>in</strong>e<br />
and not to mascul<strong>in</strong>e:<br />
(27) Ova velika devojka i moje malo<br />
this.fem.sg big girl and my little<br />
se lepo igrale/?igrali<br />
aux.pl refl well played.fem.pl/?masc.pl<br />
‘This big girl and my little girl played well’<br />
(Wechsler and Zlatić, 2003, 179)<br />
devojče su<br />
girl.neut.sg<br />
Inanimate nouns <strong>in</strong> Serbian/Croatian are syntactically resolved as <strong>in</strong> (26).<br />
F<strong>in</strong>ally, Icelandic, which displays the follow<strong>in</strong>g resolution patterns, shows semantic<br />
resolution <strong>in</strong> animate nouns and syntactic resolution <strong>in</strong> <strong>in</strong>animate nouns:<br />
(28) Icelandic rule (M, F, NT):<br />
F.PL: F + F
3.2 Theoretical Approaches to Predicate-Argument <strong>Agreement</strong> 38<br />
M.PL: M + M<br />
NT.PL: elsewhere (<strong>in</strong>clud<strong>in</strong>g NT and all mixtures)<br />
(Wechsler and Zlatić, 2003, 174)<br />
In the example below, the noun skáld ‘poet’ is grammatically neuter, <strong>in</strong>dependent of<br />
the sex of the poet. When it is coord<strong>in</strong>ated with another male denot<strong>in</strong>g noun and it refers<br />
to a male poet then resolution is mascul<strong>in</strong>e:<br />
(29) Skaldiþ<br />
the-poet.neut<br />
og<br />
and<br />
Jón<br />
Jon<br />
‘The poet and Jon are famous’<br />
(Wechsler and Zlatić, 2003, 179)<br />
eru<br />
are<br />
frægir/*fræg<br />
famous.masc/*neut.pl<br />
Inanimate nouns <strong>in</strong> Icelandic are syntactically resolved as <strong>in</strong> (28).<br />
Thus, the above patterns support Wechsler and Zlatić’ s claim that animate coord<strong>in</strong>ate<br />
NPs are subject to semantic resolution while <strong>in</strong> coord<strong>in</strong>ate constructions when grammatical<br />
and natural gender diverge, resolution follows the natural and not the grammatical<br />
gender as opposed to noncoord<strong>in</strong>ate structures, which always follow grammatical gender.<br />
Inanimate coord<strong>in</strong>ate NPs are subject to syntactic resolution.<br />
In animate nouns, Wechsler and Zlatić (2003, Chapt.8) associate grammatical genders<br />
with semantic correlates: mascul<strong>in</strong>e gender correlates with male sex, and fem<strong>in</strong><strong>in</strong>e gender<br />
with female sex. An NP is assigned either an <strong>in</strong>herent grammatical gender or a semantic<br />
gender. Grammatical gender is assigned to a non-coord<strong>in</strong>ate NP, which <strong>in</strong>herits the head<br />
features accord<strong>in</strong>g to a feature-shar<strong>in</strong>g mechanism between the head noun and the phrase<br />
it heads. Otherwise, semantic gender is assigned to the NP. Semantic feature assignment<br />
occurs <strong>in</strong> two cases; first, when the head noun is lexically genderless, as <strong>in</strong> non-coord<strong>in</strong>ated<br />
proper nouns (Dupont) or <strong>in</strong> sex-neutral nouns (journaliste). Second, when the phrase lacks<br />
a head noun, as <strong>in</strong> coord<strong>in</strong>ate NPs, which lack a head noun and therefore a grammatical<br />
gender feature. Thus, non-coord<strong>in</strong>ated NPs are assigned grammatical gender unless they<br />
are lexically genderless, <strong>in</strong> which case they acquire their gender feature based on semantic<br />
<strong>in</strong>terpretations. Coord<strong>in</strong>ated NPs are assigned semantic gender, s<strong>in</strong>ce they are perceived<br />
as headless constructions. Wechsler and Zlatić (2003) state the above succ<strong>in</strong>ctly with the<br />
follow<strong>in</strong>g pr<strong>in</strong>ciple:<br />
Gender agreement with an animate NP that lacks <strong>in</strong>herent gender is always<br />
<strong>in</strong>terpreted semantically.<br />
(Wechsler and Zlatić, 2003, p.181)<br />
In French and Serbian/Croatian, which have similar resolution patterns, they propose<br />
two types of genders to derive the resolution facts. The s(emantic)-genders are those with<br />
semantic correlates, such as mascul<strong>in</strong>e and fem<strong>in</strong><strong>in</strong>e; those without a semantic correlate,<br />
such as neuter are referred to as e(mpty/expletive)-genders. For the s-genders they <strong>in</strong>troduce<br />
the follow<strong>in</strong>g semantic <strong>in</strong>terpretations:<br />
(30) Semantic <strong>in</strong>terpretations of French or Serbian/Croatian s-genders (for NPs without<br />
<strong>in</strong>herent gender)<br />
a. fem<strong>in</strong><strong>in</strong>e: ‘female’<br />
b. mascul<strong>in</strong>e: ‘non-female’<br />
(Wechsler and Zlatić, 2003, 182)<br />
The semantic <strong>in</strong>terpretations for s-genders follow the crossl<strong>in</strong>guistic generalisation accord<strong>in</strong>g<br />
to which the gender used for mixed-sex groups <strong>in</strong> most languages is also the gender
3.2 Theoretical Approaches to Predicate-Argument <strong>Agreement</strong> 39<br />
used for coord<strong>in</strong>ate structures. In French, this gender is mascul<strong>in</strong>e, found <strong>in</strong> plural pronouns<br />
(ils), non-sex-differentiated plural proper nouns (les Dupont), non-sex-differentiated<br />
plural common nouns (les America<strong>in</strong>s) and <strong>in</strong> coord<strong>in</strong>ate nouns (Corbett (1991); Farkas<br />
and Zec (1995) all cited <strong>in</strong> Wechsler and Zlatić (2003, Chap.8, 181)). In Serbian/Croatian<br />
mascul<strong>in</strong>e is the gender used for mixed-sex groups, also found <strong>in</strong> plural pronouns (oni) and<br />
<strong>in</strong> coord<strong>in</strong>ate nouns. In Icelandic, neuter is the gender used for mixed-sex groups, and it<br />
is found <strong>in</strong> coord<strong>in</strong>ate constructions too. Thus, <strong>in</strong> Icelandic Wechsler and Zlatić (2003)<br />
propose different semantic representations to derive the desired results, such as mascul<strong>in</strong>e:<br />
‘male’ and fem<strong>in</strong><strong>in</strong>e: ‘female’.<br />
In French and Serbian/Croatian animate nouns, Wechsler and Zlatić (2003) assume<br />
that the positively def<strong>in</strong>ed semantic feature ‘female’ is a distributive property “where for<br />
any distributive property P and set s, P(s) iff ∀ f ∈ s.P(f)” (Dalrymple and Kaplan, 2000,<br />
769). Thus, a group of animate entities is a female group if and only if all of its members<br />
are female. On the other hand, the negatively def<strong>in</strong>ed semantic feature ‘non-female’ is not<br />
distributive s<strong>in</strong>ce a ‘non-female’ group is any group that refers either to a mixed group<br />
where at least one member is male and the rest are females, or to a group where all of the<br />
members are males. Thus, <strong>in</strong> (30) ‘female’ is distributive s<strong>in</strong>ce it refers to female members<br />
only, whereas ‘non-female’ is not distributive (i.e. ¬‘female’ is true of at least one member<br />
of the group) s<strong>in</strong>ce it may refer to a mixed group where at least one member is male or<br />
to a male group where all of the members are male. In the two languages, the semantic<br />
<strong>in</strong>terpretations assigned to the correspond<strong>in</strong>g grammatical genders entail that mascul<strong>in</strong>e is<br />
the resolved gender for plural animates.<br />
In <strong>in</strong>animate nouns agreement is grammatical. Wechsler and Zlatić (2003) assume the<br />
<strong>in</strong>tersection operation based on the notion of distributivity, mean<strong>in</strong>g that the set represent<strong>in</strong>g<br />
a property of a group is the <strong>in</strong>tersection of the sets represent<strong>in</strong>g the properties of<br />
the members of the group. They represent fem<strong>in</strong><strong>in</strong>e and mascul<strong>in</strong>e genders as positively<br />
or negatively specified features to produce the required results. For example, <strong>in</strong> French<br />
a mascul<strong>in</strong>e and a fem<strong>in</strong><strong>in</strong>e noun resolve <strong>in</strong>to mascul<strong>in</strong>e. Thus, each positively specified<br />
grammatical gender, such as fem<strong>in</strong><strong>in</strong>e, is represented as the s<strong>in</strong>gleton set {f}, and each<br />
negatively specified gender, such as non-fem<strong>in</strong><strong>in</strong>e (i.e. mascul<strong>in</strong>e), is the empty set {}.<br />
The feature assignment is shown below:<br />
(31) a. fem<strong>in</strong><strong>in</strong>e = {f}<br />
b. non-fem<strong>in</strong><strong>in</strong>e (mascul<strong>in</strong>e) = {}<br />
(Wechsler and Zlatić, 2003, 184)<br />
The <strong>in</strong>tersection of two members if one is {f} and the other {} results <strong>in</strong> a mascul<strong>in</strong>e<br />
grammatical gender as required {f} ∩ {} = {}. Otherwise the <strong>in</strong>tersection of two members<br />
which are both fem<strong>in</strong><strong>in</strong>e results <strong>in</strong> fem<strong>in</strong><strong>in</strong>e {f} ∩ {f} = {f}.<br />
E-gender features, like neuter, which lack any semantic correlates and do not follow<br />
distributivity, are accounted for by remov<strong>in</strong>g them from the computation of gender features<br />
for NPs denot<strong>in</strong>g groups. They achieve that by <strong>in</strong>tersect<strong>in</strong>g the calculated set with the<br />
set Gs, which represents the s-genders <strong>in</strong> a language. The follow<strong>in</strong>g universal constra<strong>in</strong>t is<br />
<strong>in</strong>troduced:<br />
Constra<strong>in</strong>t on <strong>in</strong>animate aggregate discourse referents<br />
Let γ1...γn be null or unary sets, represent<strong>in</strong>g the respective genders of the<br />
<strong>in</strong>animate discourse referents κ1...κn; let γx be a null or unary set represent<strong>in</strong>g<br />
the gender of the aggregate (group-denot<strong>in</strong>g) discourse referent X, where the<br />
members of X are κ1...κn; and let Gs be the set of s-gender features <strong>in</strong> the<br />
grammar. Then γx= γ1 ∩...∩ γn ∩ Gs.<br />
(Wechsler and Zlatić, 2003, 184)
3.2 Theoretical Approaches to Predicate-Argument <strong>Agreement</strong> 40<br />
This constra<strong>in</strong>t states that the gender value for an aggregate discourse referent is the<br />
<strong>in</strong>tersection of the gender values of the discourse referent’s elements (conjuncts), m<strong>in</strong>us<br />
any features that are not s-gender features (Wechsler and Zlatić, 2003, 184). E-genders<br />
can be represented as either unary {n} or null {} sets.<br />
For example <strong>in</strong> French, the representation of genders is {f} for fem<strong>in</strong><strong>in</strong>e, with a semantic<br />
correlate ‘female’ and {} for mascul<strong>in</strong>e with a semantic correlate ‘non-female’. The set of<br />
s-genders will be the unary set {f} only and this will be <strong>in</strong>tersected with the <strong>in</strong>animate<br />
discourse referents:<br />
(32) French<br />
a. Set of s-genders: Gs = {f}<br />
b. Set representations of the genders:<br />
fem: {f} (< ‘female’)<br />
masc: {} (< ‘non-female’)<br />
c. Calculat<strong>in</strong>g gender for plural referents<br />
masc & masc = masc {} ∩ {} ∩ Gs = {}<br />
fem & fem = fem {f} ∩ {f} ∩ Gs = {f}<br />
masc & fem = masc {} ∩ {f} ∩ Gs = {}<br />
(Wechsler and Zlatić, 2003, 185)<br />
The above rule and derivation of resolution applies not only to coord<strong>in</strong>ate phrases but<br />
generally to aggregate discourse referents.<br />
To account for semantic or grammatical gender assignment, they use the LFG formalism,<br />
<strong>in</strong> which the features of a grammatical object can be specified by two dist<strong>in</strong>ct<br />
elements appear<strong>in</strong>g <strong>in</strong> the f-structure. They use def<strong>in</strong><strong>in</strong>g equations for the semantic gender<br />
and constra<strong>in</strong><strong>in</strong>g equations for the grammatical gender. The semantic values are ‘female’<br />
and ‘non-female’ and they refer to any <strong>in</strong>dividual that is either female or non-female, respectively.<br />
The grammatical values are {f} and {} for the fem<strong>in</strong><strong>in</strong>e or mascul<strong>in</strong>e genders,<br />
respectively.<br />
Their proposal works as follows. The adjective below specifies disjunctively two different<br />
values for the feature gend. The ‘female’ value is the semantic value whereas the {f} value<br />
is semantically vacuous and therefore it is the grammatical gender value:<br />
(33) a. competente, A<br />
(↑pred) = ‘competent〈(↑subj)〉’<br />
(↑subj gend) =c{f} ∨ (↑subj gend) = ‘female’<br />
(↑subj num) = sg<br />
(Wechsler and Zlatić, 2003, 191)<br />
In example (34), the noun sent<strong>in</strong>elle specifies [gend {f}], <strong>in</strong>herited by the NP [La sent<strong>in</strong>elle<br />
á la barbe]. The semantic form of the predicative adjective [gend ‘female’] would conflict<br />
with the grammatical form specified by the noun and therefore the disjunct [gend {f}] is<br />
selected for the subject. Thus, <strong>in</strong> this case agreement is syntactic:<br />
(34) [La sent<strong>in</strong>elle à la barbe] est compétente<br />
(↑gend)= {f} (↑subj gend)=c{f}<br />
The bearded sentry is competent<br />
(Wechsler and Zlatić, 2003, 192)<br />
Next, we show how their theory accounts for animate coord<strong>in</strong>ate nouns. In the lexical<br />
entry below, the French predicative adjective competents specifies disjunctively a grammatical<br />
gender [gend {}] and a semantic gender [gend ‘non-female’]:
3.2 Theoretical Approaches to Predicate-Argument <strong>Agreement</strong> 41<br />
(35) a. competents, A<br />
(↑pred) = ‘compétent〈(↑subj)〉’<br />
(↑subj gend) =c{} ∨ (↑subj gend) = ‘non-female’<br />
(↑subj num) = pl<br />
In example (36), the coord<strong>in</strong>ate construction is headless and therefore it lacks a grammatical<br />
gender. The grammatical gender of the adjective [gend {}] cannot f<strong>in</strong>d the same<br />
feature <strong>in</strong> the coord<strong>in</strong>ate noun phrase s<strong>in</strong>ce the latter lacks a grammatical gender. The semantic<br />
gender value is the only available choice for the noun phrase. Thus, the coord<strong>in</strong>ate<br />
noun phrase is <strong>in</strong>terpreted semantically:<br />
(36) Jean<br />
Jean.masc.sg<br />
et<br />
and<br />
‘Jean and Marie are competent’<br />
Marie<br />
Marie.fem.sg<br />
sont<br />
sont<br />
competents<br />
competent.masc.pl<br />
To summarise, Wechsler and Zlatic’s (2003) theory assumes that non-coord<strong>in</strong>ate NPs<br />
(animate/<strong>in</strong>animate) will be assigned grammatical agreement unless they lack an <strong>in</strong>herent(grammatical)<br />
gender <strong>in</strong> which case they should be <strong>in</strong>terpreted semantically. Coord<strong>in</strong>ate<br />
animate nouns will be assigned semantic agreement, s<strong>in</strong>ce they lack an <strong>in</strong>herent grammatical<br />
gender, while coord<strong>in</strong>ate <strong>in</strong>animate nouns will be assigned grammatical agreement if<br />
they are mascul<strong>in</strong>e or fem<strong>in</strong><strong>in</strong>e; otherwise, agreement is derived by the constra<strong>in</strong>t on <strong>in</strong>animate<br />
aggregate discourse referents if they are neuter. Thus, Wechsler and Zlatić (2003) use<br />
both semantic and grammatical (syntactic) resolution, where the features are represented<br />
semantically and grammatically, respectively. Some unfortunate aspects of this approach<br />
are that it does not consider languages with different resolution patterns <strong>in</strong> animate and<br />
<strong>in</strong>animate nouns or the problem of undergeneration of some agreement patterns. These<br />
issues will be analysed <strong>in</strong> the last section.<br />
3.2.4 Sadler’s Theory of Resolution <strong>in</strong> Rumanian<br />
Sadler (2006) also proposes an analysis of gender resolution <strong>in</strong> Rumanian and she captures<br />
the dist<strong>in</strong>ction between the syntactic and semantic resolution <strong>in</strong> <strong>in</strong>animate and animate<br />
nouns, respectively.<br />
Accord<strong>in</strong>g to Farkas (1990), Farkas and Zec (1995), Lumsden (1992) and Wechsler and<br />
Zlatić (2003), Rumanian is a language with three genders: mascul<strong>in</strong>e, fem<strong>in</strong><strong>in</strong>e and neuter.<br />
Generally, animate nouns are masc when they refer to males, and fem when they refer<br />
to females. In coord<strong>in</strong>ate constructions, animate nouns follow the semantic pr<strong>in</strong>ciple of<br />
resolution. Thus, if any of the conjuncts are male-denot<strong>in</strong>g then resolution is mascul<strong>in</strong>e,<br />
otherwise resolution is fem<strong>in</strong><strong>in</strong>e. The follow<strong>in</strong>g examples, which show the above general<br />
resolution rules, are from Farkas and Zec (1995) and Moosally (1998), respectively:<br />
(37) Maria<br />
Maria.fem.sg<br />
¸si<br />
and<br />
‘Maria and mother were seen’<br />
(Farkas and Zec, 1995, 94)<br />
(38) Maria<br />
Maria.fem.sg<br />
¸si<br />
and<br />
‘Maria and father were seen’<br />
(Moosally, 1998, 112)<br />
mama<br />
mother.fem.sg<br />
tata<br />
father.masc.sg<br />
au fost vǎzute<br />
were seen.fem.pl<br />
au fost vǎzuti<br />
were seen.masc.sg<br />
The relevant table for the above patterns is from Sadler (2006, 306):
3.2 Theoretical Approaches to Predicate-Argument <strong>Agreement</strong> 42<br />
Table 3.1: <strong>Agreement</strong> of Rumanian Animate Coord<strong>in</strong>ate <strong>Noun</strong>s<br />
NP1 NP2 Target<br />
female male male.pl<br />
female female female.pl<br />
male male male.pl<br />
Inanimate nouns may belong to any of the three genders: mascul<strong>in</strong>e, fem<strong>in</strong><strong>in</strong>e and<br />
neuter. However, an <strong>in</strong>terest<strong>in</strong>g pattern emerges accord<strong>in</strong>g to Corbett (1991), noncoord<strong>in</strong>ate<br />
<strong>in</strong>animates dist<strong>in</strong>guish three agreement classes but they have only two target genders:<br />
mascul<strong>in</strong>e and fem<strong>in</strong><strong>in</strong>e. Thus, a mascul<strong>in</strong>e noun admits mascul<strong>in</strong>e agreement <strong>in</strong> its dependents,<br />
a fem<strong>in</strong><strong>in</strong>e noun admits fem<strong>in</strong><strong>in</strong>e agreement but a neuter noun admits mascul<strong>in</strong>e<br />
agreement <strong>in</strong> the s<strong>in</strong>gular and fem<strong>in</strong><strong>in</strong>e agreement <strong>in</strong> the plural, as seen below:<br />
(39) un<br />
a.masc<br />
‘a beautiful chair’<br />
(40) douǎ<br />
two.fem<br />
scaun<br />
chair.neut.sg<br />
scaune<br />
chairs.neut.pl<br />
‘two beautiful chairs’<br />
(Sadler, 2006, 303)<br />
frumos<br />
beautiful.masc.sg<br />
frumoase<br />
beautiful.fem.pl<br />
Corbett (1991), despite oppos<strong>in</strong>g views (Wechsler and Zlatić, 2003; Bateman and Pol<strong>in</strong>sky,<br />
2005), claims that the three-gender agreement dist<strong>in</strong>ction suppports the existence of<br />
three genders especially when one considers <strong>in</strong>animate coord<strong>in</strong>ate nouns, which follow syntactic<br />
resolution. In the latter, if all conjuncts are mascul<strong>in</strong>e, resolution is mascul<strong>in</strong>e,<br />
otherwise resolution is fem<strong>in</strong><strong>in</strong>e even if the conjuncts are neuter. The follow<strong>in</strong>g table is<br />
from Sadler (2006, 304):<br />
Table 3.2: <strong>Agreement</strong> of Rumanian Inanimate Coord<strong>in</strong>ate <strong>Noun</strong>s<br />
NP1 NP2 AP NP1 NP2 AP<br />
nsg nsg fpl fsg msg fpl<br />
fsg fsg fpl fsg nsg fpl<br />
msg msg mpl msg nsg fpl<br />
Sadler’s (2006) proposal treats targets of agreement as underspecify<strong>in</strong>g “the agreement<br />
features of their controllers” <strong>in</strong>stead of treat<strong>in</strong>g “neuter nouns as lexically underspecified<br />
for gender” (Sadler, 2006, 310). In other words, targets are not assigned a specific value<br />
based on their controllers but they place their own constra<strong>in</strong>ts. <strong>Noun</strong>s are specified as<br />
belong<strong>in</strong>g to one of the three gender classes, mascul<strong>in</strong>e, fem<strong>in</strong><strong>in</strong>e or neuter. The lexical<br />
entries of nouns normally appear as follows:<br />
(41) copac (↑ pred)=‘tree’ rochie (↑ pred)=‘dress’<br />
(↑ gend)=masc (↑ gend)=fem<br />
(↑ num)=sg (↑ num)=sg<br />
scaun (↑ pred)=‘chair’<br />
(↑ gend)=neut<br />
(↑ num)=sg<br />
(Sadler, 2006, 310)
3.2 Theoretical Approaches to Predicate-Argument <strong>Agreement</strong> 43<br />
Adjectives and determ<strong>in</strong>ers are represented as follows:<br />
(42) frumoasă frumos › i<br />
((adj∈↑)gend)=fem ((adj∈↑)gend)=masc<br />
((adj∈↑)num)= sg ((adj∈↑)num)= pl<br />
(↑pred)=‘beautiful’ (↑pred)=‘beautiful’<br />
frumos frumoase<br />
((adj∈↑)gend)¬=fem ((adj∈↑)gend)¬=masc<br />
((adj∈↑)num)= sg ((adj∈↑)num)= pl<br />
(↑pred)=‘beautiful’ (↑pred)=‘beautiful’<br />
(Sadler, 2006, 311)<br />
The lexical entries of the first two adjectives can comb<strong>in</strong>e with fsg and mpl nouns<br />
respectively, s<strong>in</strong>ce the adjectives share the gender value of their head. The next two lexical<br />
entries, for msg and fpl forms of nom<strong>in</strong>al modifiers, are underspecified, mean<strong>in</strong>g that the<br />
mascul<strong>in</strong>e s<strong>in</strong>gular form cannot comb<strong>in</strong>e with a fem<strong>in</strong><strong>in</strong>e noun but it will comb<strong>in</strong>e freely<br />
with a mascul<strong>in</strong>e or neuter s<strong>in</strong>gular, and similarly the fem<strong>in</strong><strong>in</strong>e plural will comb<strong>in</strong>e with the<br />
fem<strong>in</strong><strong>in</strong>e or neuter plural. Participles and predicative adjectives are specified <strong>in</strong> a similar<br />
way (Sadler, 2006, 311).<br />
Sadler (2006, 312) adopts the follow<strong>in</strong>g marker sets to represent the values of the three<br />
genders <strong>in</strong> Rumanian:<br />
(43) Marker Sets for Rumanian <strong>Noun</strong>s<br />
masc{m}<br />
fem{m,n}<br />
neut{n}<br />
In the example below, the fem<strong>in</strong><strong>in</strong>e noun requires a fem<strong>in</strong><strong>in</strong>e predicative adjective:<br />
(44) O<br />
a.fem.sg<br />
garoafǎ<br />
carnation.fem.sg<br />
‘A white carnation is expensive’<br />
(Farkas, 1990, 539)<br />
alba<br />
white.fem.sg<br />
e<br />
is<br />
scumpǎ<br />
expensive.fem.sg<br />
The lexical entry of the adjective must allow a fem<strong>in</strong><strong>in</strong>e {m,n} value for the gender of its<br />
subject s<strong>in</strong>ce the noun is fem<strong>in</strong><strong>in</strong>e. A constra<strong>in</strong><strong>in</strong>g equation must be used to constra<strong>in</strong> the<br />
gender value. This is represented as follows:<br />
(45) scumpǎ (subj gend must be fem)<br />
(↑subj gend)=c{m n}<br />
(↑subj num)= sg<br />
(↑pred)=‘expensive’<br />
(Sadler, 2006, 313)<br />
In the follow<strong>in</strong>g example, the noun sc › aune ‘chair’ is lexically specified (↑ gend)={n}.<br />
The agreement features of the predicative adjective need to be underspecified s<strong>in</strong>ce the<br />
noun itself is neuter(plural) but the agree<strong>in</strong>g elements are fem<strong>in</strong><strong>in</strong>e(plural):
3.2 Theoretical Approaches to Predicate-Argument <strong>Agreement</strong> 44<br />
(46) Nis › te<br />
some.fem.pl<br />
sc › aune<br />
chair.neut.pl<br />
‘Some comfortable chairs are useful’<br />
(Farkas, 1990, 540)<br />
confortabile<br />
comfortable.fem.pl<br />
e<br />
are<br />
folositoare<br />
useful.fem.pl<br />
The lexical entry of the adjective cannot be the closed set conta<strong>in</strong><strong>in</strong>g the s<strong>in</strong>gle element {m}<br />
(mean<strong>in</strong>g that it has to be either neuter {n} or fem<strong>in</strong><strong>in</strong>e {m,n}). The negative constra<strong>in</strong>t<br />
is used to exclude the possibility of a mascul<strong>in</strong>e value <strong>in</strong> the gender feature:<br />
(47) folositoare (subj gend can’t be masc)<br />
(↑subj gend)¬={m} (f.pl)<br />
(↑subj num)= pl<br />
(↑pred)=‘useful’<br />
(Sadler, 2006, 313)<br />
After plac<strong>in</strong>g the above constra<strong>in</strong>ts, Sadler (2006) turns to coord<strong>in</strong>ate nouns, and follow<strong>in</strong>g<br />
Dalrymple and Kaplan (2000), uses set union to derive resolution. The standard<br />
PS rule of coord<strong>in</strong>ation, add<strong>in</strong>g the extra feature anim with values + or −, specify<strong>in</strong>g that<br />
nouns are animate or <strong>in</strong>animate, accounts for <strong>in</strong>animate coord<strong>in</strong>ate nouns:<br />
(48) NP → NP CONJ NP<br />
↓∈↑ ↓∈↑<br />
(↓gend) ⊆ (↑gend) (↓gend) ⊆ (↑gend)<br />
(↓anim)= − (↓anim)= −<br />
(Sadler, 2006, 315)<br />
The above rule accounts for all the coord<strong>in</strong>ation cases of <strong>in</strong>animate nouns, as shown below:<br />
Table 3.3: Coord<strong>in</strong>ation with set values <strong>in</strong> Inanimate Rumanian <strong>Noun</strong>s<br />
NP1 NP2 NPcoord Target Morph<br />
{m}(masc) {m}(masc) {m}(masc) mpl<br />
{m,n}(fem) {m,n}(fem) {m,n}(fem) fpl<br />
{n}(neut) {n}(neut) {n}(neut) fpl<br />
{m,n}(fem) {n}(neut) {m,n}(fem) fpl<br />
{m}(masc) {n}(neut) {m,n}(fem) fpl<br />
{m,n}(fem) {m}(masc) {m,n}(fem) fpl<br />
(Sadler, 2006, 314)<br />
The pattern where two neuter s<strong>in</strong>gular nouns resolve to fem<strong>in</strong><strong>in</strong>e plural is also accounted<br />
s<strong>in</strong>ce the lexical entry of the predicative adjective <strong>in</strong>cludes the constra<strong>in</strong>t (↑subj gend)¬={m},<br />
which disallows {m} as a gend value but allows either the {n} or the {m,n} sets. This is<br />
shown below:<br />
(49) Scaunul ¸si dulapul<br />
chair.def.nsg and cupboard.def.nsg<br />
‘The chair and the cupboard are white’<br />
(50) albe (subj gend can’t be masc)<br />
(↑subj gend)¬= {m} (f.pl)<br />
(↑subj num)= pl<br />
sînt<br />
are<br />
albe<br />
white.fpl
3.2 Theoretical Approaches to Predicate-Argument <strong>Agreement</strong> 45<br />
(↑pred)=‘white’<br />
(Sadler, 2006, 4)<br />
Corbett (1991) argues that animate nouns <strong>in</strong> Rumanian follow semantic resolution. He<br />
<strong>in</strong>troduces the follow<strong>in</strong>g generalisation:<br />
Gender Resolution Rules for Rumanian<br />
1. If one conjunct denotes a male animate then resolution is masc;<br />
2. If all conjuncts are masc, then masc is used;<br />
3. otherwise fem is used<br />
(Corbett, 1991, 289)<br />
To capture the notion of a male referent, Sadler (2006) uses an extra f-structure feature<br />
semgend with values male and female. Thus, lexical entries of nouns which denote<br />
male <strong>in</strong>dividuals (<strong>in</strong>clud<strong>in</strong>g nouns that are syntactically fem) have the (↑semgend)= male<br />
feature-value and nouns that denote female <strong>in</strong>dividuals are specified as (↑semgend)= female.<br />
She <strong>in</strong>troduces the follow<strong>in</strong>g rule so as to derive resolution <strong>in</strong> animate nouns:<br />
(51) NP → NP CONJ NP<br />
↓∈↑ ↓∈↑<br />
(↓anim)= + (↓anim)= +<br />
(Sadler, 2006, 317)<br />
[(↑∈ semgend)¬= male<br />
(↑ gend)={m,n}<br />
| (↑ gend)={m}]<br />
The first two l<strong>in</strong>es above are the standard rule of coord<strong>in</strong>ation. The second l<strong>in</strong>e states that<br />
both conjuncts need to be animate s<strong>in</strong>ce this rule is animate specific. In the next equation,<br />
Sadler (2006) uses negation, which has a universal <strong>in</strong>terpretation, so as to state that none<br />
of the members of the set should specify a feature-value male <strong>in</strong> which case the coord<strong>in</strong>ate<br />
phrase should resolve <strong>in</strong>to fem ((↑ gend={m,n})), otherwise the gender of the coord<strong>in</strong>ate<br />
phrase should be masc (| (↑ gend={m}) 4 .<br />
The present proposal is rather powerful s<strong>in</strong>ce it accounts for both the semantic and<br />
syntactic resolution cases found <strong>in</strong> animate and <strong>in</strong>animate nouns <strong>in</strong> Rumanian, accord<strong>in</strong>gly.<br />
It also takes <strong>in</strong>to account the central dist<strong>in</strong>ction that animate nouns resolve accord<strong>in</strong>g to<br />
the semantic pr<strong>in</strong>ciple and that <strong>in</strong>animate nouns resolve accord<strong>in</strong>g to the syntactic one. A<br />
problematic aspect is the addition of the extra feature semgend that will be discussed <strong>in</strong><br />
the last section.<br />
3.2.5 Badecker’s Optimality Theory<br />
Badecker (2008) proposes a theory of agreement with<strong>in</strong> the framework of Optimality Theory,<br />
which is based on the <strong>in</strong>teraction of a series of ranked constra<strong>in</strong>ts. His focuses on the<br />
gender feature for languages that preserve gender dist<strong>in</strong>ctions across the s<strong>in</strong>gular-plural<br />
divide. In particular, he argues that gender resolution patterns arise from the <strong>in</strong>teraction of<br />
a series of universal but violable markedness and faithfulness constra<strong>in</strong>ts, conceived with<strong>in</strong><br />
4 This is also compatible with (↑∈ semgend)¬=male s<strong>in</strong>ce it needs to make the two pairs of the dis-<br />
junction mutually exclusive.
3.2 Theoretical Approaches to Predicate-Argument <strong>Agreement</strong> 46<br />
OT (Pr<strong>in</strong>ce and Smolensky, 2002). He assumes the dist<strong>in</strong>ction <strong>in</strong> noun phrase agreement<br />
between concord and <strong>in</strong>dex features (Kathol, 1999; Sadler, 1999; Wechsler and Zlatić,<br />
2000, 2003). He analyzes languages that reta<strong>in</strong> the same gender dist<strong>in</strong>ction with both s<strong>in</strong>gular<br />
and plural nouns, such as <strong>Modern</strong> <strong>Greek</strong>, Icelandic, Serbian/Croatian, Slovene and<br />
Lat<strong>in</strong>, which are three-gender systems, and the two gender systems, Italian and French.<br />
The central concept of his theory is based on six markedness and faithfulness constra<strong>in</strong>ts<br />
for gender resolution, whose rank<strong>in</strong>g and importance vary across the languages he discusses.<br />
The first three are markedness constra<strong>in</strong>ts and govern the expression of the three genders<br />
(fem<strong>in</strong><strong>in</strong>e, mascul<strong>in</strong>e and neuter), presented below:<br />
Constra<strong>in</strong>t 1, 2, 3<br />
*F, *M, *N: Do not express gender feature f (for each of Fem<strong>in</strong><strong>in</strong>e, Mascul<strong>in</strong>e<br />
and Neuter). (Badecker, 2008, 9)<br />
The ma<strong>in</strong> function of the above constra<strong>in</strong>ts is to disallow the occurrence of a particular<br />
gender feature <strong>in</strong> the output and impose economy of structure. Their order varies depend<strong>in</strong>g<br />
on the language. These markedness constra<strong>in</strong>ts are dom<strong>in</strong>ated by a number of other<br />
highly ranked faithfulness constra<strong>in</strong>ts, which require “an identity relation between the <strong>in</strong>dex<br />
features of conjo<strong>in</strong>ed NPs and those of their constituent conjuncts” (Badecker, 2008,<br />
9). The first faithfulness constra<strong>in</strong>t requires the same <strong>in</strong>dex between the coord<strong>in</strong>ate phrase<br />
and at least one of the conjuncts for the number and gender features, and tries to capture<br />
the effect of conjunct number on gender resolution with mixed gender value conjuncts.<br />
Thus, conjo<strong>in</strong>ed mascul<strong>in</strong>e and fem<strong>in</strong><strong>in</strong>e s<strong>in</strong>gular number nouns can resolve to neuter, and<br />
conjo<strong>in</strong>ed mascul<strong>in</strong>e and fem<strong>in</strong><strong>in</strong>e plural number nouns can resolve to mascul<strong>in</strong>e:<br />
Constra<strong>in</strong>t 4<br />
There is a conjunct phrase whose number and gender feature values are identical<br />
to the resolution values for the conjo<strong>in</strong>ed phrase as a whole (∃c[identN ,G(c,ph 5 )]);<br />
violated if the <strong>in</strong>dex number and/or gender of a conjo<strong>in</strong>ed phrase differs from<br />
that of all its conjuncts. (Badecker, 2008, 10)<br />
The above constra<strong>in</strong>t is violated when the number resolved value is plural and the<br />
conjucts are s<strong>in</strong>gular and when the gender resolved value is different from the gender value<br />
of the <strong>in</strong>dividual conjuncts. A relevant example is the <strong>Modern</strong> <strong>Greek</strong> coord<strong>in</strong>ate phrase<br />
below, from Badecker (2008, 3):<br />
(52) Laspomena itan o<br />
diadromos<br />
muddy.neut.pl was the.masc.sg corridor.masc.sg<br />
i skala<br />
the.fem.sg flight-of-stairs.fem.sg<br />
‘The corridor and the flight of stairs were muddy’<br />
The next faithfulness constra<strong>in</strong>t imposes the requirement that the gender feature value<br />
of a coord<strong>in</strong>ate phrase is identical to the gender feature value of each conjunct. It does not<br />
place any restrictions on the number value. Badecker (2008) states this as follows:<br />
ke<br />
and<br />
(53) Constra<strong>in</strong>t 5<br />
Every phrasal conjunct is identical <strong>in</strong> its gender feature values to the resolution<br />
values for the conjo<strong>in</strong>ed phrase as a whole (∀c[identG(c,ph)]); violated if the<br />
gender of a conjo<strong>in</strong>ed phrase differs from that of any of its conjuncts.<br />
(Badecker, 2008, 10)<br />
5 The symbol c stands for conjunct and the symbol ph stands for the coord<strong>in</strong>ate phrase.
3.2 Theoretical Approaches to Predicate-Argument <strong>Agreement</strong> 47<br />
The present constra<strong>in</strong>t is satisfied when the resolved gender value is the same as the<br />
gender value of the conjuncts, and it is violated if at least one of the gender values of the<br />
conjuncts differ from the gender value of the coord<strong>in</strong>ate phrase. The current constra<strong>in</strong>t is<br />
satisfied <strong>in</strong> example (54), but it is violated <strong>in</strong> (55):<br />
(54) La donna e la<br />
the.fem.sg woman and the.fem.sg<br />
preoccupate/*preoccupati<br />
worried.fem.pl/*masc.pl<br />
‘The woman and the girl are very worried’<br />
(55) L’ uomo e la<br />
the.masc.sg man and the.fem.sg<br />
preoccupati/*preoccupate<br />
worried.masc.pl/*fem.pl<br />
‘The man and the woman are very worried’<br />
(Badecker, 2008, 6)<br />
ragazza<br />
girl<br />
donna<br />
woman<br />
sono<br />
are.3.pl<br />
sono<br />
are.3.pl<br />
molto<br />
very<br />
molto<br />
very<br />
The last constra<strong>in</strong>t, which relates to the number feature, is known as SemNr and<br />
requires the resolved number value of a conjo<strong>in</strong>ed phrase to be plural (or dual). This<br />
constra<strong>in</strong>t ensures that the morpho-syntactic representation of number (s<strong>in</strong>gular, dual,<br />
plural) obeys the semantic number. Accord<strong>in</strong>g to Badecker (2008), this constra<strong>in</strong>t is highly<br />
ranked compared to the others <strong>in</strong> most languages analysed:<br />
(56) Constra<strong>in</strong>t 6<br />
SemNr: the <strong>in</strong>dex number of a conjo<strong>in</strong>ed NP (if it has one) cannot differ from the<br />
phrase’s semantic number.<br />
(Badecker, 2008, 10)<br />
In his analysis of <strong>Modern</strong> <strong>Greek</strong> <strong>in</strong>animate nouns 6 , he claims that conjo<strong>in</strong>ed mixed<br />
gender nouns with mascul<strong>in</strong>e and fem<strong>in</strong><strong>in</strong>e conjuncts resolve to neut when conjuncts are<br />
both s<strong>in</strong>gular and to masc when they are both plural. The resolution patterns are as<br />
follows:<br />
Table 3.4: <strong>Agreement</strong> of Inanimate Coord<strong>in</strong>ate <strong>Noun</strong>s<br />
UNIFORM GENDER MIXED GENDER<br />
NP1 & NP2 AP NP1 & NP2 AP<br />
m & m mpl msg & fsg npl<br />
mpl & fpl mpl<br />
f & f fpl m & n npl<br />
n & n npl f & n npl<br />
(Badecker, 2008, 9)<br />
Accord<strong>in</strong>g to the above table, he <strong>in</strong>troduces the follow<strong>in</strong>g rank<strong>in</strong>g for <strong>Modern</strong> <strong>Greek</strong>:<br />
(57) <strong>Modern</strong> <strong>Greek</strong><br />
SemNr ≫ ∃c[identN ,G(c,ph)] ≫ ∀c[identG(c,ph)] ≫ *F ≫ *M ≫ *N<br />
(Badecker, 2008, 10)<br />
The rank<strong>in</strong>g above treats neut as the default gender and fem as the least preferred one.<br />
S<strong>in</strong>ce the faithfulness constra<strong>in</strong>ts ident are placed higher <strong>in</strong> the rank<strong>in</strong>g, the markedness<br />
6 Full discussion of what patterns are actually found <strong>in</strong> MG will follow <strong>in</strong> the next chapters.
3.2 Theoretical Approaches to Predicate-Argument <strong>Agreement</strong> 48<br />
constra<strong>in</strong>ts alone do not determ<strong>in</strong>e the resolution value but rather the ident constra<strong>in</strong>ts<br />
will play a determ<strong>in</strong><strong>in</strong>g factor.<br />
In coord<strong>in</strong>ate phrases with identical gender conjuncts as <strong>in</strong> fem<strong>in</strong><strong>in</strong>e s<strong>in</strong>gular nouns, the<br />
resolution value is fem plural, which violates the ident constra<strong>in</strong>t ∃c[identN ,G(c,ph)].<br />
However, the fem<strong>in</strong><strong>in</strong>e plural value is the optimal one s<strong>in</strong>ce it does not violate the SemNr,<br />
which is placed higher <strong>in</strong> the rank<strong>in</strong>g, and the ∀c[identG(c,ph)], which outranks the<br />
markedness constra<strong>in</strong>ts.<br />
(58) I kuz<strong>in</strong>a ke i<br />
the.fem.sg<br />
kathares<br />
clean.fem.pl<br />
kitchen.fem.sg and the.fem.sg<br />
‘The kitchen and the bathroom are both clean’<br />
(Badecker, 2008, 11)<br />
The table below shows the optimal form of the relevant example:<br />
tualeta<br />
bathroom.fem.sg<br />
Table 3.5: MG Coord<strong>in</strong>ate <strong>Noun</strong>s<br />
Identical Gender Conjuncts<br />
[f.sg & f.sg]<strong>in</strong>dex a SemNr ∃c[identN ,G(c,ph)] ∀c[identG(c,ph)] *F *M *N<br />
α= n.sg *! * * *<br />
α= m.sg *! * * *<br />
α= f.sg *! *<br />
α= n.pl * *! *<br />
α= m.pl * *! *<br />
Zα= f.pl * *<br />
(Badecker, 2008, 12)<br />
In mixed gender cases, he argues that <strong>in</strong> mascul<strong>in</strong>e and fem<strong>in</strong><strong>in</strong>e s<strong>in</strong>gular conjuncts,<br />
resolution is neut plural, as <strong>in</strong> example (59). In mascul<strong>in</strong>e and fem<strong>in</strong><strong>in</strong>e plural conjuncts,<br />
resolution is masc plural, as <strong>in</strong> example (60):<br />
(59) O dromos ke i<br />
the.masc.sg road.masc.sg and the.fem.sg<br />
jemata kosmo<br />
full.neut.pl of-people<br />
‘The road and the square were full of people.’<br />
platia<br />
square.fem.sg<br />
(60) I<br />
dromi ke i<br />
the.masc/fem.pl roads.masc.pl and the.masc/fem.pl<br />
platies itan jemati kosmo<br />
squares.fem.pl were full.mpl of-people<br />
‘The roads and the squares were full of people’<br />
(Badecker, 2008, 14)<br />
itan<br />
were<br />
Accord<strong>in</strong>g to the rank<strong>in</strong>g of the constra<strong>in</strong>ts proposed by Badecker (2008), the above<br />
resolution forms depend mostly on the gender markedness constra<strong>in</strong>ts and on the identity<br />
constra<strong>in</strong>t ∃c[identN ,G(c,ph)], which determ<strong>in</strong>es number and gender simultaneously and<br />
is ranked above the markedness constra<strong>in</strong>ts for the three genders. In table (3.6), which<br />
corresponds to example (59), we see that conjo<strong>in</strong>ed s<strong>in</strong>gular nouns violate both identity<br />
constra<strong>in</strong>ts, thus markedness constra<strong>in</strong>ts will determ<strong>in</strong>e the gender option. S<strong>in</strong>ce neuter is<br />
the least marked gender, the resolved form is neut plural (Badecker, 2008, 14):<br />
<strong>in</strong>e<br />
are
3.2 Theoretical Approaches to Predicate-Argument <strong>Agreement</strong> 49<br />
Table 3.6: MG Coord<strong>in</strong>ate <strong>Noun</strong>s<br />
Conjo<strong>in</strong>ed Mascul<strong>in</strong>e and Fem<strong>in</strong><strong>in</strong>e s<strong>in</strong>gular conjuncts<br />
[m.sg & f.sg]<strong>in</strong>dex a SemNr ∃c[identN ,G(c,ph)] ∀c[identG(c,ph)] *F *M *N<br />
α= n.sg *! * * *<br />
α= m.sg *! * *<br />
α= f.sg *! * *<br />
Zα= n.pl * * *<br />
α= m.pl * * *!<br />
α= f.pl * * *!<br />
(Badecker, 2008, 14)<br />
In table (3.7), which refers to example (60), neuter plural resolution violates the identity<br />
constra<strong>in</strong>t for gender and number (∃c[identN ,G(c,ph)]), whereas neither mascul<strong>in</strong>e plural<br />
nor fem<strong>in</strong><strong>in</strong>e plural violate the same constra<strong>in</strong>t. Therefore, there are only two possible<br />
resolved forms between mascul<strong>in</strong>e and fem<strong>in</strong><strong>in</strong>e, which makes masc plural the resolution<br />
value:<br />
Table 3.7: MG Coord<strong>in</strong>ate <strong>Noun</strong>s<br />
Conjo<strong>in</strong>ed Mascul<strong>in</strong>e and Fem<strong>in</strong><strong>in</strong>e plural conjuncts<br />
[m.pl & f.pl]<strong>in</strong>dex a SemNr ∃c[identN ,G(c,ph)] ∀c[identG(c,ph)] *F *M *N<br />
α= n.sg *! * * *<br />
α= m.sg *! * * *<br />
α= f.sg *! * * *<br />
α= n.pl *! * *<br />
Zα= m.pl * *<br />
α= f.pl * *!<br />
(Badecker, 2008, 14)<br />
The same constra<strong>in</strong>ts hold for a number of other attested resolution patterns concern<strong>in</strong>g<br />
languages such as French and Italian, Icelandic, Serbian and Croatian, but the rank<strong>in</strong>g is<br />
different conform<strong>in</strong>g to the resolution patterns of each language. Badecker (2008) presents<br />
the resolution patterns of Italian and French <strong>in</strong> a tabular form as follows:<br />
Table 3.8: Syntactic resolution patterns for Italian and French<br />
UNIFORM GENDER MIXED GENDER<br />
Italian, French: m & m = m m & f = m<br />
f & f = f<br />
(Badecker, 2008, 9)<br />
Accord<strong>in</strong>g to Badecker (2008) the follow<strong>in</strong>g constra<strong>in</strong>t rank<strong>in</strong>g works for languages such<br />
as Italian and French:<br />
(61) Italian, French etc.<br />
*N ≫ ... ≫ SemNr ≫ ∀c[identG(c,ph)] ≫ *F ≫ *M ≫∃c[identN ,G(c,ph)]<br />
(Badecker, 2008, 10)<br />
To account for mixed gender conjuncts the identity constra<strong>in</strong>t ∀c[identG(c,ph)] is placed<br />
higher than the two gender markedness constra<strong>in</strong>ts *F and *M while *F is also placed higher<br />
than *M. For cases of identical gender conjuncts, Badecker (2008) places the markedness<br />
constra<strong>in</strong>ts below the ident constra<strong>in</strong>t while number is accounted aga<strong>in</strong> by the high position<br />
of the SemNr constra<strong>in</strong>t. The table below shows the optimal form of a mixed gender<br />
comb<strong>in</strong>ation:
3.2 Theoretical Approaches to Predicate-Argument <strong>Agreement</strong> 50<br />
Table 3.9: Italian/French Coord<strong>in</strong>ate <strong>Noun</strong>s<br />
Mixed gender comb<strong>in</strong>ations<br />
[m.pl & f.pl]<strong>in</strong>dex a *N SemNr ∀c[identG(c,ph)] *F *M ∃c[identN ,G(c,ph)]<br />
α= n.sg *! * * *<br />
α= m.sg *! * * *<br />
α= f.sg *! * * *<br />
α= n.pl *! * *<br />
Zα= m.pl * *<br />
α= f.pl * *!<br />
(Badecker, 2008, 14)<br />
Other laguages, such as Serbian/Croatian, Slovene and Icelandic 7 are analysed but we<br />
will not discuss the analysis here due to space limitations.<br />
Accord<strong>in</strong>g to Badecker (2008), the above gender resolution patterns obey the syntactic<br />
resolution pr<strong>in</strong>ciple (Corbett, 1991; Wechsler and Zlatić, 2003) s<strong>in</strong>ce they are determ<strong>in</strong>ed by<br />
morphological properties like gender and number, which are syntactic features. He proposes<br />
a similar analysis for languages whose resolved gender is controlled by semantic factors,<br />
such as the animacy (or human/non-human) of the conjuncts. In Lat<strong>in</strong>, for example, he<br />
accounts for semantic resolution by assum<strong>in</strong>g that the semantic dist<strong>in</strong>ctions correspond<br />
to morpho-syntactic features like [± Animate] (or [±Human]). The resolution patterns <strong>in</strong><br />
Lat<strong>in</strong> for identical or mixed gender conjuncts are presented below:<br />
Table 3.10: Semantically motivated resolution for Lat<strong>in</strong><br />
MIXED GENDER<br />
uniform gender uniformly [+human] uniformly [-human]<br />
m & m = m m & f = m m & f = n<br />
f & f = f m & n = m m & n = n<br />
n & n = n f & n = m f & n = n<br />
(Badecker, 2008, 9)<br />
In example (62), the coord<strong>in</strong>ate phrase consists of a mascul<strong>in</strong>e and a fem<strong>in</strong><strong>in</strong>e [-human]<br />
conjunct which resolve to neuter plural. In example (63), the conjuncts are mascul<strong>in</strong>e and<br />
fem<strong>in</strong><strong>in</strong>e but s<strong>in</strong>ce they are [+human], they resolve to mascul<strong>in</strong>e:<br />
(62) Murus et porta de caelo<br />
wall.masc.sg<br />
errant<br />
were<br />
and gate.fem.sg from sky<br />
‘The wall and gate have been struck by lighten<strong>in</strong>g’<br />
(63) Quam pridem pater mehi<br />
how long-ago father.masc.sg me.dat<br />
mortu-i essen<br />
dead.masc.pl were<br />
‘How long ago my father and mother had died’<br />
(Corbett, 1991; Badecker, 2008)<br />
et<br />
and<br />
tact-a<br />
struck.neut.pl<br />
mater<br />
mother.fem.sg<br />
Badecker (2008) <strong>in</strong>troduces an extra constra<strong>in</strong>t to account for the [±Human] dist<strong>in</strong>ction,<br />
which is a comb<strong>in</strong>ation of the markedness constra<strong>in</strong>t *N with the *[+Human] feature. Its<br />
function is to exclude neuter as a resolution value for conjo<strong>in</strong>ed [+Human] NPs. The<br />
7 For a complete discussion on the rank<strong>in</strong>g of constra<strong>in</strong>ts and relevant examples <strong>in</strong> Serbian/Croatian,<br />
Slovene and Icelandic see Badecker (2008, 20-26).
3.3 Theoretical Approaches to Head-Modifier <strong>Agreement</strong> 51<br />
problem with this constra<strong>in</strong>t is that it also disallows neuter as the resolved gender for<br />
conjo<strong>in</strong>ed nouns that are both neuter. To avoid that, Badecker (2008) places this constra<strong>in</strong>t<br />
lower <strong>in</strong> the rank<strong>in</strong>g, below the ident constra<strong>in</strong>t for gender, but above the markedness<br />
constra<strong>in</strong>ts. F<strong>in</strong>ally, the feature [Human] is assigned to the conjo<strong>in</strong>ed phrase, when the<br />
conjuncts are either [+Human] or [-Human], and this will only affect gender resolution<br />
<strong>in</strong> mixed gender phrases because the ident constra<strong>in</strong>t ∀c[identG(c,ph)] is placed higher<br />
than the rest of the constra<strong>in</strong>ts which relate to gender.<br />
(64) Lat<strong>in</strong><br />
SemNr ≫ ∀c[identG(c,ph)] ≫ *N&*[+Human] ≫ *F ≫ *M ≫ *N<br />
(Badecker, 2008, 25)<br />
The tables below display how rank<strong>in</strong>g works. For mixed gender cases every resolution<br />
value will necessarily violate the constra<strong>in</strong>t ∀c[identG(c,ph)] and therefore the markedness<br />
constra<strong>in</strong>ts will determ<strong>in</strong>e the gender value from the least marked to the most marked <strong>in</strong><br />
[+H] nouns. In table (3.11), gender resolution for [-H] will be determ<strong>in</strong>ed by the rank<strong>in</strong>g<br />
of the simple markedness constra<strong>in</strong>ts (*F > *M > *N) s<strong>in</strong>ce *N&[+H] is not violated by<br />
any resolution options and the resolved gender is always neut:<br />
Table 3.11: Lat<strong>in</strong> Coord<strong>in</strong>ate <strong>Noun</strong>s<br />
Gender resolution as modulated by [-Human]<br />
[m.sg,[-h] & f.sg,[-h]]<strong>in</strong>dex a ∀c[identG(c,ph)] *N&*[+Human] *F *M *N ∃c[identN ,G(c,ph)]<br />
Zα= n.pl,[-h] * * *<br />
α= m.pl,[-h] * *!<br />
α= f.pl,[-h] * *!<br />
(Badecker, 2008, 22)<br />
In table (3.12), gender resolution for [+H] will exclude neuter as the resolved gender due<br />
to the locally conjo<strong>in</strong>ed markedness constra<strong>in</strong>t *N&[+H] while the rema<strong>in</strong><strong>in</strong>g options are<br />
controlled by the rest of the constra<strong>in</strong>ts and the result<strong>in</strong>g value is masc for any comb<strong>in</strong>ation<br />
of conjunct genders which are [+Human]:<br />
Table 3.12: Lat<strong>in</strong> Coord<strong>in</strong>ate <strong>Noun</strong>s<br />
Gender resolution as modulated by [+Human]<br />
[m.sg,[+h] & f.sg,[+h]]<strong>in</strong>dex a ∀c[identG(c,ph)] *N&*[+Human] *F *M *N ∃c[identN ,G(c,ph)]<br />
α= n.pl,[+h] * *! * *<br />
Zα= m.pl,[+h] * *<br />
α= f.pl,[+h] * *!<br />
(Badecker, 2008, 22)<br />
To conclude, the above constra<strong>in</strong>t-based approach which depends on the rank<strong>in</strong>g of<br />
those constra<strong>in</strong>ts captures the multiple patterns and it is rather flexible s<strong>in</strong>ce each language,<br />
depend<strong>in</strong>g on the data requirements, adopts a different rank<strong>in</strong>g of the constra<strong>in</strong>ts and<br />
allows the admission of new constra<strong>in</strong>ts. Two problems are that it becomes non-economical<br />
through the addition of different constra<strong>in</strong>ts for different languages, and the rank<strong>in</strong>g of the<br />
various forms lead<strong>in</strong>g to an optimal form is not favoured <strong>in</strong> agreement cases where more<br />
than one agreement form is considered grammatically correct, such as the ones we will<br />
discuss <strong>in</strong> MG. These issues are discussed <strong>in</strong> the last section.<br />
3.3 Theoretical Approaches to Head-Modifier <strong>Agreement</strong><br />
The theoretical approaches to head-modifier agreement <strong>in</strong> coord<strong>in</strong>ate structures analyse<br />
NP-<strong>in</strong>ternal coord<strong>in</strong>ation follow<strong>in</strong>g the two-fold nature of agreement features <strong>in</strong>dex and
3.3 Theoretical Approaches to Head-Modifier <strong>Agreement</strong> 52<br />
concord with<strong>in</strong> the noun phrase and the dist<strong>in</strong>ction between distributive and nondistributive<br />
features. The theories that will be presented here are those proposed by Kathol (1999)<br />
and Wechsler and Zlatić (2003) <strong>in</strong> HPSG, K<strong>in</strong>g and Dalrymple (2004) and Dalrymple and<br />
Nikolaeva (2006) <strong>in</strong> LFG, focus<strong>in</strong>g on number.<br />
3.3.1 <strong>Agreement</strong> <strong>in</strong> HPSG<br />
A number of recent proposals <strong>in</strong> NP agreement <strong>in</strong> HPSG have postulated the existence of<br />
a head feature agr (Kathol, 1999) or concord (Wechsler and Zlatić, 2000) responsible<br />
for morphosyntactic agreement, together with the <strong>in</strong>dex feature responsible for semantic<br />
agreement.<br />
Kathol (1999) extends the treatment of agreement <strong>in</strong>troduced by Pollard and Sag (1994)<br />
and treats agreement as a phenomenon which “<strong>in</strong>volves merg<strong>in</strong>g of <strong>in</strong>formation contributed<br />
by various resources <strong>in</strong> the sentence” (Kathol, 1999, 234). He proposes that selector categories,<br />
such as verbs, record their own morphosyntactic agreement features and also select<br />
the agreement features of their argument. This results <strong>in</strong> structure shar<strong>in</strong>g with<strong>in</strong> the<br />
selector category. For example, <strong>in</strong> English the verb walks has the follow<strong>in</strong>g lexical entry:<br />
(65) ⎡<br />
⎡ ⎤⎤<br />
f<strong>in</strong>ite<br />
⎢<br />
⎢ ⎥⎥<br />
⎢cat|hd|morsyn|<br />
agr ⎣pers<br />
1 3rd⎦<br />
⎥<br />
⎢<br />
num 2 sg<br />
⎥<br />
⎢ ⎡<br />
⎤ ⎥<br />
⎢ walk<strong>in</strong>g<br />
⎥<br />
⎢ ⎢ ⎡ ⎤⎥<br />
⎥<br />
⎢ ⎢ <strong>in</strong>dex ⎥ ⎥<br />
⎢ ⎢<br />
⎢cont⎢<br />
⎢<br />
⎢ ⎢ ⎢pers<br />
⎥<br />
⎥ ⎥<br />
1⎥<br />
⎥ ⎥<br />
⎢ ⎢walker<br />
3 ⎢ ⎥<br />
⎥ ⎥<br />
⎣ ⎣ ⎣num<br />
2 ⎦<br />
⎥ ⎥<br />
⎦ ⎥<br />
⎦<br />
gend<br />
(Kathol, 1999, 236)<br />
The above entry shows that only pers and num are <strong>in</strong>volved <strong>in</strong> subject-verb agreement<br />
s<strong>in</strong>ce these two features are appropriate for the subsort f<strong>in</strong>ite of agr. Also, the value of<br />
these features is structure-shared with the value of the same features found on the <strong>in</strong>dex of<br />
subject of the verb. F<strong>in</strong>ally, agr, which is the value of morsyn, is part of a larger group<br />
of l<strong>in</strong>guistic <strong>in</strong>formation 8 .<br />
Kathol (1999) posits two types of agreement, morphosyntactic and semantic as follows:<br />
(66) morphosyntactic agr(selector) ≈ 9 agr(arg)<br />
semantic agr(selector) ≈ <strong>in</strong>dex(arg)<br />
For <strong>in</strong>stance, the follow<strong>in</strong>g example <strong>in</strong> French is a hybrid case of agreement, comb<strong>in</strong><strong>in</strong>g<br />
subject-verb morphosyntactic num agreement, subject-verb semantic pers agreement, and<br />
semantic gend and num agreement between the predicative adjective and the noun:<br />
(67) Vous êtes belle<br />
you.2.pl are.pl beautiful.sg.fem<br />
‘You are beautiful’ (Kathol, 1999, 239)<br />
Morphosyntactic num agreement between the subject and the verb is given by the agr<br />
feature. Semantic pers agreement between the verb and the subject and gend and num<br />
agreement between the predicative adjective and the noun are given by the <strong>in</strong>dex feature.<br />
8 For a more elaborate f-structure and detailed explanations of it see Kathol (1999, 236-9).<br />
9 The symbol “≈” means “is structured-shared <strong>in</strong> its relevant parts with” (Kathol, 1999, 241).
3.3 Theoretical Approaches to Head-Modifier <strong>Agreement</strong> 53<br />
In French, <strong>in</strong>dex also reflects the anchor<strong>in</strong>g conditions, mean<strong>in</strong>g that if a phrase refers to a<br />
female <strong>in</strong>dividual the <strong>in</strong>dex <strong>in</strong> its gend feature will have the value fem. However, the analysis<br />
proposed for French does not hold for all languages but it varies s<strong>in</strong>ce morphosyntactic<br />
and semantic agreement differs from language to language 10 . Kathol’s proposal, who posits<br />
a dist<strong>in</strong>ction between morphosyntactic and semantic agreement, captures a wide range of<br />
data that cannot be accounted for by the analysis proposed by Pollard and Sag (1994).<br />
Pollard and Sag (1994, 61) assume a dist<strong>in</strong>ction between <strong>in</strong>dex agreement, which arises<br />
when <strong>in</strong>dices are required to be token identical, concord or syntactic agreement, which<br />
<strong>in</strong>volves structure-shar<strong>in</strong>g between features (referr<strong>in</strong>g ma<strong>in</strong>ly to the features of case and<br />
declension class <strong>in</strong> German) on “selector” and selected category, and pragmatic agreement,<br />
which arises when contextual background assumptions need to be consistent.<br />
Wechsler and Zlatić (2000) also treat agreement as hybrid or mixed and present a theory<br />
of agreement <strong>in</strong> Serbo-Croatian. They view agreement <strong>in</strong> terms of “two elements specify<strong>in</strong>g<br />
partial <strong>in</strong>formation about a s<strong>in</strong>gle l<strong>in</strong>guistic object” (Wechsler and Zlatić, 2003, 7) and this<br />
<strong>in</strong>formation com<strong>in</strong>g from the two elements must be compatible to result <strong>in</strong> agreement. They<br />
assume that any “<strong>in</strong>flected noun has two different feature sets that determ<strong>in</strong>e the agreement<br />
values it triggers” (Wechsler and Zlatić, 2000, 7pp). These are concord, which <strong>in</strong>volves<br />
the features case, num and gender, and <strong>in</strong>dex which <strong>in</strong>volves the features person, num<br />
and gender.<br />
The concord features of a noun are related to the noun’s <strong>in</strong>flected form. Thus,<br />
concord|gender is related to the declension class of a noun and concord|case and<br />
concord|number determ<strong>in</strong>e its <strong>in</strong>flected form. The relation between morphology and<br />
concord is shown as follows:<br />
(68) morphology ⇐⇒ concord<br />
(Wechsler and Zlatić, 2000, 800)<br />
On the other hand, <strong>in</strong>dex|gender and <strong>in</strong>dex|number are associated to the noun’s<br />
semantics, i.e. whether the noun denotes a male or female or an aggregate or non-aggregate<br />
entity. This is shown as follows:<br />
(69) <strong>in</strong>dex ⇐⇒ semantics<br />
(Wechsler and Zlatić, 2000, 800)<br />
The concord and <strong>in</strong>dex features are related to each other as follows:<br />
(70) concord ⇐⇒ <strong>in</strong>dex<br />
(Wechsler and Zlatić, 2000, 800)<br />
Wechsler and Zlatić (2000) summarise <strong>in</strong> a schema the four types of <strong>in</strong>formation concern<strong>in</strong>g<br />
a noun as follows:<br />
(71) morphology ⇐⇒ concord ⇐⇒ <strong>in</strong>dex ⇐⇒ semantics<br />
(Wechsler and Zlatić, 2000, 800)<br />
The ma<strong>in</strong> focus of their work is cases of mismatch agreement and they predict three<br />
types of mismatches <strong>in</strong> languages such as Serbo-Croatian. The first mismatch occurs between<br />
declension and concord. For example, some nouns have mascul<strong>in</strong>e grammatical properties<br />
and they refer to males but they decl<strong>in</strong>e follow<strong>in</strong>g the declension used for fem<strong>in</strong><strong>in</strong>e<br />
nouns such as Serbian/Croatian. A second mismatch is between <strong>in</strong>dex and the semantics<br />
of a noun. For example, a noun may refer to a female but governs mascul<strong>in</strong>e agreement.<br />
10 For the discussion of languages such as Spanish, German and English see Kathol (1999, 234-250).
3.3 Theoretical Approaches to Head-Modifier <strong>Agreement</strong> 54<br />
F<strong>in</strong>ally, the mismatch we are <strong>in</strong>terested <strong>in</strong> is that between concord and <strong>in</strong>dex. They give as<br />
a characteristic example the Serbo-Croatian noun deca ‘children’. This noun triggers fem<strong>in</strong><strong>in</strong>e<br />
s<strong>in</strong>gular agreement on attributive modifiers and non-f<strong>in</strong>ite predicate phrases, which<br />
is syntactic agreement, but it triggers neuter plural agreement on coreferential pronouns,<br />
which is semantic agreement. This is illustrated below:<br />
(72) Posmatrali smo ovu dobru decu.<br />
watched.1pl aux this.f.sg good.f.sg children.acc.<br />
su se lepo igrala<br />
aux.3pl refl nicely played.nt.pl<br />
‘We watched these good childreni. Theyi played well.’<br />
(Wechsler and Zlatić, 2000, 51)<br />
Ona<br />
they.nt.pl<br />
Wechsler and Zlatić (2000) <strong>in</strong>troduce the follow<strong>in</strong>g lexical <strong>in</strong>formation for the noun<br />
deca:<br />
(73) ⎡ ⎤<br />
num s<strong>in</strong>g<br />
⎢<br />
concord ⎥<br />
⎢ gend fem ⎥<br />
⎢ <br />
⎥<br />
⎢<br />
⎣ num plur ⎥<br />
<strong>in</strong>dex<br />
⎦<br />
gend neut<br />
(Wechsler and Zlatić, 2003, 51)<br />
The noun specifies two different features concord and <strong>in</strong>dex. concord refers to the<br />
morphosyntactic agreement and <strong>in</strong>dex to the semantic agreement. For Serbo-Croatian, determ<strong>in</strong>ers,<br />
attributive adjectives and secondary predicates show concord agreement whereas<br />
verbs, primary predicates and bound anaphors show <strong>in</strong>dex agreement.<br />
Both approaches proposed by Kathol (1999) and Wechsler and Zlatić (2000) are consistent<br />
with Corbett’s dist<strong>in</strong>ction between “agreement ad formam” (syntactic agreement)<br />
and “agreement ad sensum” (semantic agreement) and his “agreement hierarchy” (Corbett,<br />
1979, 1983a, 1991), presented below:<br />
(74) The <strong>Agreement</strong> Hierarchy<br />
attributive < predicate < relative pronoun < personal pronoun<br />
Corbett (1979) also suggests that the follow<strong>in</strong>g constra<strong>in</strong>t on agreement systems hold:<br />
(75) As we move rightwards along the hierarchy, the likelihood of semantic agreement<br />
will <strong>in</strong>crease monotonically (Corbett, 1979, 210)<br />
Both approaches capture agreement mismatches and also form the basis for the <strong>in</strong>troduction<br />
of the same features with a similar function <strong>in</strong> Lexical Functional Grammar.<br />
3.3.2 K<strong>in</strong>g and Dalrymple’s Theory<br />
Determ<strong>in</strong>er-noun agreement is straightforward with noncoord<strong>in</strong>ate constructions but more<br />
complicated with coord<strong>in</strong>ate nouns. In general, <strong>in</strong> noncoord<strong>in</strong>ate nouns, a s<strong>in</strong>gular determ<strong>in</strong>er<br />
goes with a s<strong>in</strong>gular noun and a plural determ<strong>in</strong>er goes with a plural noun (K<strong>in</strong>g<br />
and Dalrymple, 2004). In coord<strong>in</strong>ate phrases, however, agreement is not always straightforward.<br />
In English, for <strong>in</strong>stance, <strong>in</strong> example (76) the s<strong>in</strong>gular determ<strong>in</strong>er will modify two<br />
s<strong>in</strong>gular conjuncts. The phrase as a whole, however, behaves as a plural phrase and forces<br />
plural agreement to the predicate, as <strong>in</strong> example (77):<br />
(76) This boy and girl/ a boy and girl
3.3 Theoretical Approaches to Head-Modifier <strong>Agreement</strong> 55<br />
(77) This boy and girl are happy<br />
(K<strong>in</strong>g and Dalrymple, 2004, 69)<br />
In plural number, the plural determ<strong>in</strong>er will modify two plural conjuncts:<br />
(78) These boys and girls (K<strong>in</strong>g and Dalrymple, 2004, 69)<br />
Mixed coord<strong>in</strong>ations where one conjunct is s<strong>in</strong>gular and the other is plural require<br />
separate determ<strong>in</strong>ers:<br />
(79) These boys and girl*/These boys and this girl<br />
(K<strong>in</strong>g and Dalrymple, 2004, 69)<br />
K<strong>in</strong>g and Dalrymple (2004) adopt the concord and <strong>in</strong>dex feature dist<strong>in</strong>ction with<strong>in</strong><br />
the noun phrase (Chapter 2, section 2.4.1.3). They show that a phrase such as this boy and<br />
girl has a s<strong>in</strong>gular concord value and a plural <strong>in</strong>dex value, account<strong>in</strong>g for the s<strong>in</strong>gular<br />
number <strong>in</strong> the determ<strong>in</strong>er and the two conjuncts, and the plural number <strong>in</strong> the verb,<br />
respectively. The concord and <strong>in</strong>dex dist<strong>in</strong>ction is straightforwardly related to the LFG<br />
dist<strong>in</strong>ction between distributive and nondistributive features.<br />
Unlike concord, <strong>in</strong>dex is a non-distributive feature. The follow<strong>in</strong>g def<strong>in</strong>ition for<br />
<strong>in</strong>dex agreement is repeated from above:<br />
(80) <strong>in</strong>dex agreement with a coord<strong>in</strong>ate phrase depends on the <strong>in</strong>dex features of the<br />
coord<strong>in</strong>ate phrase as a whole. (K<strong>in</strong>g and Dalrymple, 2004, 74)<br />
In particular, a coord<strong>in</strong>ate phrase is treated as plural due to the presence of a pl value<br />
<strong>in</strong> its <strong>in</strong>dex feature regardless of whether its conjuncts are s<strong>in</strong>gular or plural. Even a<br />
phrase with s<strong>in</strong>gular nouns behaves as a plural phrase:<br />
(81) Joe and Fred are happy<br />
K<strong>in</strong>g and Dalrymple (2004, 74) propose a dist<strong>in</strong>ction between semantic number and<br />
syntactic (<strong>in</strong>dex) number. They argue that semantically the number feature of a phrase<br />
represents the number of <strong>in</strong>dividuals it refers to. When one referent is denoted the num<br />
has a s<strong>in</strong>gular value but when more than one referent is denoted the value is plural. As<br />
mentioned above, coord<strong>in</strong>ate phrases usually refer to more than one <strong>in</strong>dividual; thus, the<br />
phrase this boy and girl behaves as a plural phrase and has a plural value <strong>in</strong> its num<br />
feature, as shown <strong>in</strong> S-V agreement:<br />
(82) This boy and girl are happy<br />
There are cases, however, when a coord<strong>in</strong>ate phrase refers to a s<strong>in</strong>gle <strong>in</strong>dividual, as <strong>in</strong><br />
my friend and colleague. In this case, the phrase behaves like a s<strong>in</strong>gular phrase and the<br />
value of the num feature is s<strong>in</strong>gular:<br />
(83) My friend and colleague is eat<strong>in</strong>g a pizza<br />
K<strong>in</strong>g and Dalrymple (2004) associate this semantic dist<strong>in</strong>ction with a syntactic dist<strong>in</strong>ction<br />
<strong>in</strong> <strong>in</strong>dex num requirements. They suggest that a phrase like this boy and girl <strong>in</strong>volves<br />
a group-form<strong>in</strong>g <strong>in</strong>terpretation of the conjunction and, requir<strong>in</strong>g a pl value for the <strong>in</strong>dex<br />
num of the noun phrase s<strong>in</strong>ce the group is composed of ‘a boy and a girl’. They propose<br />
the follow<strong>in</strong>g LFG annotation:<br />
(84) Group-form<strong>in</strong>g and<br />
(↑<strong>in</strong>dex num) = pl<br />
(K<strong>in</strong>g and Dalrymple, 2004, 76)
3.3 Theoretical Approaches to Head-Modifier <strong>Agreement</strong> 56<br />
Heycock and Zamparelli (2000, 2005) use the term split <strong>in</strong>terpretation/read<strong>in</strong>g to refer to<br />
the group-form<strong>in</strong>g <strong>in</strong>terpretation of the conjunction and.<br />
On the other hand, a coord<strong>in</strong>ate phrase like my friend and colleague denotes a Boolean<br />
<strong>in</strong>terpretation of and, mean<strong>in</strong>g that the <strong>in</strong>dividual or <strong>in</strong>dividuals referred to have both<br />
properties simultaneously. Thus, <strong>in</strong> example (83) each <strong>in</strong>dividual must be a friend and a<br />
colleague at the same time. In this case, syntactically the <strong>in</strong>dex num of the coord<strong>in</strong>ate<br />
phrase must be the same as the <strong>in</strong>dex num of one of the conjuncts, annotated as follows:<br />
(85) Boolean and<br />
(↑<strong>in</strong>dex num) = (↑∈ <strong>in</strong>dex num)<br />
(K<strong>in</strong>g and Dalrymple, 2004, 76)<br />
The notation (↑∈<strong>in</strong>dex num) refers to a member of the conjunct set where the set<br />
membership symbol is an attribute. At the same time, they assume that semantically<br />
all of the conjuncts have the same number. <strong>Phrases</strong> like my friend and colleaques have<br />
no boolean <strong>in</strong>terpretation s<strong>in</strong>ce the s<strong>in</strong>gular noun friend and the plural noun colleaques<br />
cannot corefer to the same entity. Heycock and Zamparelli (2000, 2005) use the term jo<strong>in</strong>t<br />
<strong>in</strong>terpretation/read<strong>in</strong>g for the Boolean and.<br />
The concord feature is a distributive feature, associated only with the <strong>in</strong>dividual<br />
members of the set represent<strong>in</strong>g the coord<strong>in</strong>ate phrase and not the set as a whole. The<br />
follow<strong>in</strong>g def<strong>in</strong>ition for concord agreement is repeated from above:<br />
(86) concord agreement with a coord<strong>in</strong>ate phrase depends on the concord features<br />
of each conjunct. (K<strong>in</strong>g and Dalrymple, 2004, 77)<br />
A s<strong>in</strong>gular determ<strong>in</strong>er will require s<strong>in</strong>gular conjuncts and a plural determ<strong>in</strong>er plural conjuncts<br />
s<strong>in</strong>ce the concord features of the conjuncts must match the concord features of<br />
the determ<strong>in</strong>er <strong>in</strong> order to be licenced.<br />
The f-structure of the coord<strong>in</strong>ate phrase this boy and girl illustrates the above:<br />
⎡<br />
⎤<br />
(87) spec ‘this’(nondistrib. feat. of c)<br />
⎢<br />
⎥<br />
⎢<strong>in</strong>dex<br />
[num pl](nondistrib. feat. of c) ⎥<br />
⎢<br />
⎢⎧<br />
⎡<br />
⎤⎫<br />
⎥<br />
⎢ pred ‘boy’<br />
⎥<br />
⎢ ⎢<br />
⎥<br />
⎥<br />
⎢ b: ⎣concord<br />
[num sg] ⎦<br />
⎥<br />
c: ⎢⎪⎨<br />
⎪⎬<br />
⎢ <strong>in</strong>dex [num sg]<br />
⎥<br />
⎢<br />
⎡<br />
⎤<br />
⎥<br />
⎢ pred ‘girl’<br />
⎥<br />
⎢<br />
⎢<br />
⎥<br />
⎥<br />
g:<br />
⎣<br />
⎣concord<br />
[num sg] ⎦<br />
⎥<br />
⎪⎩<br />
⎪⎭<br />
⎦<br />
<strong>in</strong>dex [num sg]<br />
(K<strong>in</strong>g and Dalrymple, 2004, 77)<br />
In the f-structure above, the set c has both nondistributive and distributive features. The<br />
spec feature with its value ‘this’ and the <strong>in</strong>dex feature with its value [num pl] are the<br />
nondistributive features. These are associated with the whole set c while the <strong>in</strong>dex pl<br />
feature-value means that the verb is required to show plural agreement. The coord<strong>in</strong>ate<br />
structure also has features which are associated with each of the members b and g of the set<br />
c. These are concord with value [num sg] and <strong>in</strong>dex with value [num sg]. The concord<br />
feature is a distributive feature and it can never be associated with the coord<strong>in</strong>ate phrase<br />
as a whole.<br />
The complete representation of the s<strong>in</strong>gular coord<strong>in</strong>ate phrase this boy and girl, <strong>in</strong>clud<strong>in</strong>g<br />
the lexical entry of the determ<strong>in</strong>er, is presented below:
3.3 Theoretical Approaches to Head-Modifier <strong>Agreement</strong> 57<br />
(88) This boy and girl<br />
this: (↑concord num) = sg<br />
DET<br />
this<br />
NP<br />
N<br />
boy<br />
N<br />
Conj<br />
and<br />
N<br />
girl<br />
(K<strong>in</strong>g and Dalrymple, 2004, 78)<br />
⎡<br />
⎤<br />
spec ‘this’(nondis. feat. )<br />
⎢<br />
⎥<br />
⎢<strong>in</strong>dex<br />
[num pl](nondis. feat.) ⎥<br />
⎢<br />
⎢⎧<br />
⎡<br />
⎤⎫<br />
⎥<br />
⎢ pred ‘boy’<br />
⎥<br />
⎢ ⎢<br />
⎥ ⎥<br />
⎢ b: ⎣concord<br />
[num sg] ⎦ ⎥<br />
c: ⎢⎪⎨<br />
⎪⎬<br />
⎢ <strong>in</strong>dex [num sg]<br />
⎥<br />
⎢<br />
⎡<br />
⎤ ⎥<br />
⎢ pred ‘girl’<br />
⎥<br />
⎢<br />
⎢<br />
⎥ ⎥<br />
g:<br />
⎣<br />
⎣concord<br />
[num sg] ⎦ ⎥<br />
⎪⎩<br />
⎪⎭<br />
⎦<br />
<strong>in</strong>dex [num sg]<br />
The follow<strong>in</strong>g example shows the correspond<strong>in</strong>g plural coord<strong>in</strong>ate phrase these boys<br />
and girls, <strong>in</strong> which the demonstrative determ<strong>in</strong>er specifies <strong>in</strong> its lexical entry a concord<br />
feature with a [num pl] value and would require conjuncts with the same value:<br />
(89) These boys and girls<br />
these: (↑concord num) = pl<br />
DET<br />
these<br />
NP<br />
N<br />
boys<br />
N<br />
Conj<br />
and<br />
N<br />
girls<br />
(K<strong>in</strong>g and Dalrymple, 2004, 80)<br />
⎡<br />
⎤<br />
spec ‘these’<br />
⎢<br />
⎥<br />
⎢<strong>in</strong>dex<br />
[num pl] ⎥<br />
⎢<br />
⎢⎧<br />
⎡<br />
⎤⎫<br />
⎥<br />
⎢ pred ‘boys’ ⎥<br />
⎢ ⎢<br />
⎥ ⎥<br />
⎢ b: ⎣concord<br />
[num pl] ⎦ ⎥<br />
c: ⎢⎪⎨<br />
⎪⎬<br />
⎢ <strong>in</strong>dex [num pl]<br />
⎥<br />
⎢<br />
⎡<br />
⎤ ⎥<br />
⎢ pred ‘girls’<br />
⎥<br />
⎢<br />
⎢<br />
⎥ ⎥<br />
g:<br />
⎣<br />
⎣concord<br />
[num pl] ⎦ ⎥<br />
⎪⎩<br />
⎪⎭<br />
⎦<br />
<strong>in</strong>dex [num pl]<br />
Coord<strong>in</strong>ations modified by either a s<strong>in</strong>gular or a plural determ<strong>in</strong>er <strong>in</strong> which one noun is<br />
s<strong>in</strong>gular and the other is plural are disallowed. The reason is that the two nouns will have<br />
f-structures with concord and <strong>in</strong>dex [num sg] feature-values and concord and <strong>in</strong>dex<br />
[num pl] feature-values, respectively. When these coord<strong>in</strong>ated nouns are modified by a<br />
determ<strong>in</strong>er that requires concord [num sg], the sg value is <strong>in</strong>compatible with the value<br />
of the pl conjunct and when they are modified by a determ<strong>in</strong>er that requires concord<br />
[num pl], the pl value is <strong>in</strong>compatible with the value of the sg conjunct, result<strong>in</strong>g <strong>in</strong><br />
ungrammaticallity:<br />
(90) *This boy and girls<br />
DET<br />
this<br />
NP<br />
N<br />
boy<br />
N<br />
Conj<br />
and<br />
N<br />
girls<br />
⎡<br />
⎤<br />
spec ‘this’<br />
⎢<br />
⎥<br />
⎢<strong>in</strong>dex<br />
[num sg]<br />
⎥<br />
⎢<br />
⎢⎧<br />
⎡<br />
⎤ ⎫ ⎥<br />
⎢ pred ‘boy’ ⎥<br />
⎢ ⎢<br />
⎥ ⎥<br />
⎢ b: ⎣concord<br />
[num sg] ⎦ ⎥<br />
c: ⎢⎪⎨<br />
⎪⎬ ⎥<br />
⎢ <strong>in</strong>dex [num sg] ⎥<br />
⎢<br />
⎡<br />
⎤ ⎥<br />
⎢ pred ‘girls’<br />
⎥<br />
⎢<br />
⎢<br />
⎥ ⎥<br />
g:<br />
⎣<br />
⎣concord<br />
[num pl/sg] ⎦ ⎥<br />
⎪⎩<br />
⎪⎭<br />
⎦<br />
<strong>in</strong>dex [num pl]
3.3 Theoretical Approaches to Head-Modifier <strong>Agreement</strong> 58<br />
(K<strong>in</strong>g and Dalrymple, 2004, 81)<br />
K<strong>in</strong>g and Dalrymple (2004) consider both the concord and <strong>in</strong>dex features as central<br />
to NP-<strong>in</strong>ternal agreement. Some languages may require both features or just one of the<br />
two features to participate <strong>in</strong> agreement. Based on this assumption, they propose four<br />
possible agreement patterns <strong>in</strong> NP-<strong>in</strong>ternal agreement <strong>in</strong> terms of the number feature: an<br />
unrestricted one, the concord/<strong>in</strong>dex one, the concord one and the <strong>in</strong>dex one.<br />
The first one is the least restricted accord<strong>in</strong>g to which the determ<strong>in</strong>er places no constra<strong>in</strong>ts<br />
on the number of the nouns it comb<strong>in</strong>es with. A characteristic example <strong>in</strong> English<br />
is the def<strong>in</strong>ite determ<strong>in</strong>er the, which can comb<strong>in</strong>e with coord<strong>in</strong>ated s<strong>in</strong>gular nouns, plural<br />
nouns and even mixed number plural and s<strong>in</strong>gular nouns. The lexical entry of the<br />
determ<strong>in</strong>er is as follows:<br />
(91) the: (↑spec) = ‘the’<br />
All the examples below are well formed:<br />
(92) a. the dog and cat<br />
b. the dogs and cats<br />
c. the dogs and cat<br />
The second system is the most restrictive and <strong>in</strong>volves both <strong>in</strong>dex agreement and redundantly<br />
concord agreement <strong>in</strong> the s<strong>in</strong>gular, and concord agreement and redundantly<br />
<strong>in</strong>dex agreement <strong>in</strong> the plural. The s<strong>in</strong>gular determ<strong>in</strong>er can modify s<strong>in</strong>gular nouns or s<strong>in</strong>gular<br />
coord<strong>in</strong>ated nouns, which refer to the same <strong>in</strong>dividual, as <strong>in</strong> my friend and colleague.<br />
Determ<strong>in</strong>er shar<strong>in</strong>g is not allowed with conjo<strong>in</strong>ed s<strong>in</strong>gular nouns referr<strong>in</strong>g to more than<br />
one <strong>in</strong>dividual or with coord<strong>in</strong>ated structures consist<strong>in</strong>g of mixed number conjuncts. The<br />
plural determ<strong>in</strong>er can modify plural nouns and coord<strong>in</strong>ated plural nouns, but it cannot<br />
modify coord<strong>in</strong>ations where one of the conjuncts is s<strong>in</strong>gular (K<strong>in</strong>g and Dalrymple, 2004,<br />
85). For this system, the follow<strong>in</strong>g annotations are proposed:<br />
(93) concord and <strong>in</strong>dex systems:<br />
a. S<strong>in</strong>gular determ<strong>in</strong>er<br />
(↑concord num) = sg<br />
(↑<strong>in</strong>dex num) = sg<br />
b. Plural determ<strong>in</strong>er<br />
(↑concord num) = pl<br />
(↑<strong>in</strong>dex num) = pl<br />
(K<strong>in</strong>g and Dalrymple, 2004, 85)<br />
Accord<strong>in</strong>g to K<strong>in</strong>g and Dalrymple (2004), languages fall<strong>in</strong>g with<strong>in</strong> such a system are<br />
Brazilian Portuguese and German. In Brazilian Portuguese, the s<strong>in</strong>gular determ<strong>in</strong>er o ‘the’<br />
can modify a s<strong>in</strong>gular noun, as <strong>in</strong> (94), and coord<strong>in</strong>ated s<strong>in</strong>gular nouns referr<strong>in</strong>g to the<br />
same <strong>in</strong>dividual (jo<strong>in</strong>t read<strong>in</strong>g), as <strong>in</strong> (95), but not nouns with different referents, as <strong>in</strong><br />
(96):<br />
(94) o<br />
the.m.sg<br />
‘The dog’<br />
cachorro<br />
dog.m.sg
3.3 Theoretical Approaches to Head-Modifier <strong>Agreement</strong> 59<br />
(95) o<br />
the.m.sg<br />
presidente<br />
president.m.sg<br />
e<br />
and<br />
‘The president and director of Air France’<br />
(96) *o<br />
the.m.sg<br />
cachorro<br />
dog.m.sg<br />
e<br />
and<br />
‘The dog and cat’<br />
(K<strong>in</strong>g and Dalrymple, 2004, 91)<br />
gato<br />
cat.m.sg<br />
The lexical entry of the determ<strong>in</strong>er o is as follows:<br />
(97) o: (↑spec) = ‘the’<br />
(↑concord num) = sg<br />
(↑<strong>in</strong>dex num) = sg<br />
(K<strong>in</strong>g and Dalrymple, 2004, 92)<br />
diretor<br />
director.m.sg<br />
da<br />
of<br />
Air France<br />
Air France<br />
S<strong>in</strong>ce the def<strong>in</strong>ite determ<strong>in</strong>er specifies a sg value for the <strong>in</strong>dex num feature, a coord<strong>in</strong>ate<br />
phrase with pl value <strong>in</strong> its <strong>in</strong>dex num feature, as the one <strong>in</strong> (96), would conflict the<br />
requirement of the determ<strong>in</strong>er and would result <strong>in</strong> an ungrammatical phrase.<br />
The third system <strong>in</strong>volves concord features only. A s<strong>in</strong>gular determ<strong>in</strong>er can modify<br />
s<strong>in</strong>gular nouns and coord<strong>in</strong>ated structures with s<strong>in</strong>gular conjuncts. A plural determ<strong>in</strong>er<br />
can modify a plural noun and coord<strong>in</strong>ate phrases with plural nouns. Determ<strong>in</strong>er shar<strong>in</strong>g<br />
is disallowed with mixed number coord<strong>in</strong>ate nouns (K<strong>in</strong>g and Dalrymple, 2004, 86). The<br />
annotations for this system are as follows:<br />
(98) concord systems:<br />
a. S<strong>in</strong>gular determ<strong>in</strong>er<br />
(↑concord num) = sg<br />
b. Plural determ<strong>in</strong>er<br />
(↑concord num) = pl<br />
(K<strong>in</strong>g and Dalrymple, 2004, 86)<br />
Apart form the English demonstrative determ<strong>in</strong>er this/these and that/ those, a language<br />
with such a system is F<strong>in</strong>nish and H<strong>in</strong>di/Urdu. The follow<strong>in</strong>g examples, are from<br />
F<strong>in</strong>nish (K<strong>in</strong>g and Dalrymple, 2004, 89). In (99), the determ<strong>in</strong>er tämä ‘this’ can modify a<br />
s<strong>in</strong>gular noun. In (100), it modifies two s<strong>in</strong>gular conjuncts but not a s<strong>in</strong>gular and a plural<br />
conjunct, as <strong>in</strong> (101):<br />
(99) tämä<br />
this.sg<br />
‘this cat’<br />
(100) tämä<br />
this.sg<br />
(101) *tämä<br />
this.sg<br />
kissa<br />
cat.sg<br />
kissa<br />
cat.sg<br />
‘this cat and dog’<br />
kissa<br />
cat.sg<br />
‘this cat and dog’<br />
ja<br />
and<br />
ja<br />
and<br />
koira<br />
dog.sg<br />
koirat<br />
dog.pl<br />
The lexical entry of the determ<strong>in</strong>er tämä ‘this’ is as follows:
3.3 Theoretical Approaches to Head-Modifier <strong>Agreement</strong> 60<br />
(102) tämä: (↑spec) = ‘tämä’<br />
(↑concord num) = sg<br />
(K<strong>in</strong>g and Dalrymple, 2004, 89)<br />
The sg value <strong>in</strong> the concord num feature distributes to each conjunct and requires the<br />
same sg value <strong>in</strong> the concord num feature of the <strong>in</strong>dividual conjuncts; otherwise the<br />
result is ungrammatical.<br />
The last system they propose is the <strong>in</strong>dex only which is not very common crossl<strong>in</strong>guistically<br />
and places constra<strong>in</strong>ts on the <strong>in</strong>dex feature only. This system allows a plural<br />
determ<strong>in</strong>er to modify either a plural noun or coord<strong>in</strong>ated s<strong>in</strong>gular nouns referr<strong>in</strong>g to different<br />
<strong>in</strong>dividuals, while a s<strong>in</strong>gular determ<strong>in</strong>er can modify a s<strong>in</strong>gular noun or s<strong>in</strong>gular<br />
coord<strong>in</strong>ated nouns only if they refer to the same <strong>in</strong>dividual (K<strong>in</strong>g and Dalrymple, 2004,<br />
86). Such a language is Russian. The annotations for the <strong>in</strong>dex system are as follows:<br />
(103) <strong>in</strong>dex system<br />
a. S<strong>in</strong>gular determ<strong>in</strong>er<br />
(↑<strong>in</strong>dex num) = sg<br />
b. Plural determ<strong>in</strong>er<br />
(↑<strong>in</strong>dex num) = pl<br />
(K<strong>in</strong>g and Dalrymple, 2004, 86)<br />
In example (104), the s<strong>in</strong>gular determ<strong>in</strong>er étot/éta/éto ‘this’ modifies a nonconjo<strong>in</strong>ed<br />
s<strong>in</strong>gular noun, while <strong>in</strong> (105) it modifies a coord<strong>in</strong>ate noun phrase with s<strong>in</strong>gular conjuncts,<br />
referr<strong>in</strong>g to the same <strong>in</strong>dividual:<br />
(104) éta<br />
this.f.sg<br />
‘this woman’<br />
(105) étot<br />
this.m.sg<br />
ˇzenˇsč<strong>in</strong>a<br />
woman.f.sg<br />
drug<br />
friend.m.sg<br />
i<br />
and<br />
‘this friend and colleague’<br />
(K<strong>in</strong>g and Dalrymple, 2004, 95)<br />
kollega<br />
colleague.m.sg<br />
The plural determ<strong>in</strong>er éti modifies a plural noun, as <strong>in</strong> (106), coord<strong>in</strong>ated plural nouns,<br />
as <strong>in</strong> (107), and s<strong>in</strong>gular coord<strong>in</strong>ated nouns, which refer to more than one <strong>in</strong>dividual, as <strong>in</strong><br />
(108):<br />
(106) éti<br />
these.pl<br />
‘these men’<br />
(107) éti<br />
these.pl<br />
muˇzčny<br />
man.m.pl<br />
muˇzč<strong>in</strong>y<br />
man.m.pl<br />
‘these men and women’<br />
(108) éti<br />
these.pl<br />
muˇzc<strong>in</strong>a<br />
man.m.sg<br />
i<br />
and<br />
i<br />
and<br />
‘this man and woman’<br />
(K<strong>in</strong>g and Dalrymple, 2004, 95)<br />
ˇzenˇsč<strong>in</strong>y<br />
woman.f.pl<br />
ˇzenˇsč<strong>in</strong>a<br />
woman.f.sg<br />
The lexical entry for this type of determ<strong>in</strong>ers is presented below:
3.3 Theoretical Approaches to Head-Modifier <strong>Agreement</strong> 61<br />
(109) éti: (↑spec) = ‘éti’<br />
(↑<strong>in</strong>dex num) = pl<br />
(K<strong>in</strong>g and Dalrymple, 2004, 96)<br />
Such a determ<strong>in</strong>er constra<strong>in</strong>s the <strong>in</strong>dex value of the coord<strong>in</strong>ate phrase as a whole and<br />
therefore it can be used even with s<strong>in</strong>gular nouns.<br />
To conclude, K<strong>in</strong>g and Dalrymple (2004) <strong>in</strong>troduce three different agreement systems<br />
based on the dist<strong>in</strong>ction between <strong>in</strong>dex and concord agreement features and capture a<br />
wide range of data concern<strong>in</strong>g number NP-<strong>in</strong>ternal agreement <strong>in</strong> coord<strong>in</strong>ated constructions.<br />
3.3.3 Dalrymple and Nikolaeva’s theory of natural and accidental coord<strong>in</strong>ation<br />
Another theory concern<strong>in</strong>g NP-<strong>in</strong>ternal agreement is that proposed by Dalrymple and<br />
Nikolaeva (2006). They exam<strong>in</strong>e agreement phenomena <strong>in</strong> adjectives modify<strong>in</strong>g coord<strong>in</strong>ated<br />
nouns <strong>in</strong> F<strong>in</strong>nish and propose a special f-structure for coord<strong>in</strong>ate nouns that are<br />
cases of “natural” coord<strong>in</strong>ation, which is different from the f-structure of nouns <strong>in</strong> “accidental”<br />
coord<strong>in</strong>ation. The dist<strong>in</strong>ction between “natural” and “accidental” noun phrase<br />
coord<strong>in</strong>ation turns out to be an important factor <strong>in</strong> expla<strong>in</strong><strong>in</strong>g some agreement patterns<br />
<strong>in</strong> MG.<br />
Adjectives <strong>in</strong> F<strong>in</strong>nish agree with the nouns they modify <strong>in</strong> number. S<strong>in</strong>gular nouns are<br />
modified by s<strong>in</strong>gular adjectives and plural nouns by plural adjectives, as <strong>in</strong> the follow<strong>in</strong>g<br />
examples:<br />
(110) ilo<strong>in</strong>en<br />
happy.sg<br />
poika<br />
boy.sg<br />
‘the/a happy boy’<br />
(111) iloiset<br />
happy.pl<br />
pojat<br />
boy.pl<br />
‘the happy boys’<br />
(Dalrymple and Nikolaeva, 2006, 824)<br />
In coord<strong>in</strong>ated constructions, plural nouns are modified by a plural adjective, which<br />
scopes over both conjuncts, as seen <strong>in</strong> example (112). There are cases, however, when<br />
s<strong>in</strong>gular nouns are modified by a plural adjective, as <strong>in</strong> (113):<br />
(112) nuoria<br />
young.part.pl<br />
‘young [girls and boys]’<br />
(113) Iloiset<br />
happy.pl<br />
kädessä<br />
hand.<strong>in</strong>es<br />
mies<br />
man.sg<br />
tyttöjä<br />
girls.part.pl<br />
ja<br />
and<br />
ja<br />
and<br />
poika<br />
boy.sg<br />
poikia<br />
boys.part.pl<br />
lähtivät<br />
left.3.pl<br />
‘The happy [man and boy] left together hand <strong>in</strong> hand’<br />
(Dalrymple and Nikolaeva, 2006, 825)<br />
yhdessä<br />
together<br />
käsi<br />
hand<br />
In other cases, a plural adjective with s<strong>in</strong>gular nouns is considered ungrammatical, as<br />
<strong>in</strong> (114):<br />
(114) *Han<br />
he<br />
osti<br />
bought.3.sg<br />
uudet<br />
new.acc.pl<br />
‘He bought a new [house and car]’<br />
(Dalrymple and Nikolaeva, 2006, 829)<br />
talon<br />
house.acc<br />
ja<br />
and<br />
auton<br />
car.acc
3.3 Theoretical Approaches to Head-Modifier <strong>Agreement</strong> 62<br />
Dalrymple and Nikolaeva (2006) show that the acceptance of a plural adjective <strong>in</strong><br />
s<strong>in</strong>gular coord<strong>in</strong>ated nouns is attributed to a semantic dist<strong>in</strong>ction between “natural” and<br />
“accidental” coord<strong>in</strong>ation (Haspelmath, 2004a; Wälchli, 2005). Thus, plural adjectives<br />
with s<strong>in</strong>gular conjuncts are allowed only with naturally coord<strong>in</strong>ated nouns while accidental<br />
coord<strong>in</strong>ation disallows them.<br />
In general, accidental coord<strong>in</strong>ation refers to “coord<strong>in</strong>ation of items that are not expected<br />
to co-occur and which do not have a close semantic relation”(Wälchli, 2005, 5), and natural<br />
coord<strong>in</strong>ation refers to “a semantic relation <strong>in</strong> which two entities are closely related <strong>in</strong><br />
mean<strong>in</strong>g and form a conceptual unit” (Haspelmath, 2004a; Wälchli, 2005). There are<br />
cases, however, when natural coord<strong>in</strong>ation is marked syntactically.<br />
Dalrymple and Nikolaeva (2006) show that, apart from the semantic relation between<br />
the two conjuncts <strong>in</strong> a natural coord<strong>in</strong>ation construction, there are various syntactic characteristics<br />
that are found <strong>in</strong> these structures. Natural coord<strong>in</strong>ation constructions often<br />
<strong>in</strong>clude special coord<strong>in</strong>ation markers. For example, Ud<strong>in</strong>e, which is a ‘with’ language<br />
(Stassen, 2000) 11 , uses the postposition zuÒe ‘with’ <strong>in</strong> accidental coord<strong>in</strong>ation but the<br />
postposition mule ‘with’ to denote natural coord<strong>in</strong>ation. This is illustrated below:<br />
(115) Natural Coord<strong>in</strong>ation<br />
bi<br />
I<br />
mamasa<br />
wife<br />
‘I and my wife’<br />
mule/*zuÒe<br />
with/with<br />
(116) Accidental Coord<strong>in</strong>ation<br />
bi<br />
I<br />
Sergej<br />
Sergej<br />
zuÒe/*mule<br />
with/with<br />
‘I and Sergej’<br />
(Stassen (2000), cited <strong>in</strong> Dalrymple and Nikolaeva (2006, 831))<br />
Also, natural coord<strong>in</strong>ation often <strong>in</strong>volves what Dalrymple and Nikolaeva (2006) refer<br />
to as phonological reduction of the coord<strong>in</strong>ation markers. They present an example from<br />
the Oceanic languages claim<strong>in</strong>g that <strong>in</strong> Lenakel the conjunction m is often found <strong>in</strong> natural<br />
coord<strong>in</strong>ation, which is a phonologically reduced form of m@ne found <strong>in</strong> accidental<br />
coord<strong>in</strong>ation (Moyse-Faurie and Lynch, 2004).<br />
Another characteristic of natural coord<strong>in</strong>ation is “tight coord<strong>in</strong>ation” (Dalrymple and<br />
Nikolaeva, 2006, 831). This means that features such as def<strong>in</strong>iteness, case, possession,<br />
number are marked only once or not at all <strong>in</strong> the coord<strong>in</strong>ate phrase. Thus, <strong>in</strong> German<br />
some cases of natural coord<strong>in</strong>ation lack determ<strong>in</strong>ers completely, a pattern which is generally<br />
disallowed under accidental coord<strong>in</strong>ation. This is presented below:<br />
(117) Natural Coord<strong>in</strong>ation<br />
Sonne<br />
sun<br />
und<br />
and<br />
Mond<br />
moon<br />
‘the sun and the moon’<br />
(118) Accidental Coord<strong>in</strong>ation<br />
der<br />
the<br />
Mond<br />
moon<br />
und<br />
and<br />
e<strong>in</strong><br />
a<br />
Sechser<br />
sixpence<br />
11 Accord<strong>in</strong>g to Stassen (2000), a ‘with’ language is one <strong>in</strong> which nom<strong>in</strong>al coord<strong>in</strong>ation is expressed by an<br />
adpositional phrase with the postposition ‘with’ so that the conjuncts have a different syntactic status.
3.3 Theoretical Approaches to Head-Modifier <strong>Agreement</strong> 63<br />
‘the moon and sixpence’<br />
(Lambrecht (1984), cited <strong>in</strong> Dalrymple and Nikolaeva (2006, 831))<br />
F<strong>in</strong>ally, <strong>in</strong> Eastern Armenian natural coord<strong>in</strong>ation <strong>in</strong>volves a s<strong>in</strong>gle <strong>in</strong>flectional mark<strong>in</strong>g,<br />
seen below:<br />
(119) Natural Coord<strong>in</strong>ation<br />
ał<br />
salt<br />
u<br />
and<br />
hac’-d<br />
bread.2sg<br />
‘your salt and bread’<br />
(120) Accidental Coord<strong>in</strong>ation<br />
dproc’-i-s<br />
school.dat.1sg<br />
ev<br />
and<br />
usuc’ič’-ner-i-s<br />
teacherpl.dat.1sg<br />
‘for my school and my teacher’<br />
(M<strong>in</strong>assian (1980), cited <strong>in</strong> Dalrymple and Nikolaeva (2006, 831))<br />
Although the above characteristics are associated with cases of natural coord<strong>in</strong>ation <strong>in</strong><br />
the above languages, they can occasionally occur <strong>in</strong> cases of accidental coord<strong>in</strong>ation too 12 .<br />
Dalrymple and Nikolaeva (2006) po<strong>in</strong>t out that not all languages use the above syntactic<br />
characteristics <strong>in</strong> natural coord<strong>in</strong>ation. There are languages that may not use any of these<br />
forms but still display cases of natural coord<strong>in</strong>ation, rely<strong>in</strong>g only on the relation between<br />
the nouns. Dalrymple and Nikolaeva (2006) assume the existence of two types of relations<br />
between the conjuncts with<strong>in</strong> a natural coord<strong>in</strong>ation construction. The first is an <strong>in</strong>tr<strong>in</strong>sic<br />
relation, such as that of ‘father and mother’ or ‘cup and saucer’ and may <strong>in</strong>volve cases of<br />
reduced syntax, or frozen word order (Benor and Levy, 2006). The second type of relation<br />
is “contextually specified” and depends on <strong>in</strong>formation derived from previous discourse or<br />
general world knowledge (Lambrecht, 1984; Heycock and Zamparelli, 2003; Dalrymple and<br />
Nikolaeva, 2006). For example, Wälchli (2005) presents the two nouns Igor-t Natashǎt<br />
‘Igor and Natasha’. He argues that these nouns are not <strong>in</strong>tr<strong>in</strong>sically but contextually<br />
related s<strong>in</strong>ce they are perceived as partners and they are coord<strong>in</strong>ated without the need for<br />
the conjunction di ‘and’, which is found <strong>in</strong> accidental coord<strong>in</strong>ation 13 . This second type is<br />
also found <strong>in</strong> natural coord<strong>in</strong>ation structures of the F<strong>in</strong>nish language.<br />
Dalrymple and Nikolaeva (2006) argue that <strong>in</strong> F<strong>in</strong>nish the semantic dist<strong>in</strong>ction beween<br />
natural and accidental coord<strong>in</strong>ation is expressed by different syntactic constructions without<br />
the need “for different coord<strong>in</strong>ation strategies or different morphosyntactic mark<strong>in</strong>gs<br />
on conjuncts”(825) <strong>in</strong> these constructions. In fact, F<strong>in</strong>nish natural coord<strong>in</strong>ation structures<br />
resemble those of accidental coord<strong>in</strong>ation but their difference lies <strong>in</strong> the syntactic features<br />
that are associated with each structure. The features of natural coord<strong>in</strong>ation have similarities<br />
with the features of plural nouns. Natural coord<strong>in</strong>ation admits a plural adjective and<br />
two s<strong>in</strong>gular nouns while the nouns are usually contextually related.<br />
In F<strong>in</strong>nish, two specific structures admit a plural adjective with s<strong>in</strong>gular nouns and<br />
exhibit natural coord<strong>in</strong>ation. The first one is when a plural “symmetric” adjective, such<br />
as similar, match<strong>in</strong>g or compatible modifies two s<strong>in</strong>gular coord<strong>in</strong>ate nouns. The follow<strong>in</strong>g<br />
example is acceptable accord<strong>in</strong>g to the speakers:<br />
12 Some exceptions are discussed <strong>in</strong> Dalrymple and Nikolaeva (2006).<br />
13 More arguments to support the same view are presented <strong>in</strong> Dalrymple and Nikolaeva (2006, 832),<br />
Lambrecht (1984, 794) and Heycock and Zamparelli (2003, 445).
3.3 Theoretical Approaches to Head-Modifier <strong>Agreement</strong> 64<br />
(121) On<br />
is<br />
ja<br />
and<br />
vaikeaa<br />
difficult<br />
teatteri<br />
theatre<br />
suunnitella<br />
design<br />
yhteensopivat/*yhteensopiva<br />
compatible.pl/*compatible.sg<br />
‘It is difficult to design a compatible [university and theatre]’<br />
(Dalrymple and Nikolaeva, 2006, 826)<br />
yliopisto<br />
university<br />
The second structure is when the nouns have any of the semantic features of def<strong>in</strong>iteness,<br />
animacy and humaness and are <strong>in</strong> subject position. In the follow<strong>in</strong>g example, repeated from<br />
above, the conjuncts are both considered as def<strong>in</strong>ite, human and occur <strong>in</strong> subject position:<br />
(122) Iloiset<br />
happy.pl<br />
mies<br />
man.sg<br />
ja<br />
and<br />
poika<br />
boy.sg<br />
lähtivät<br />
left.3.pl<br />
‘The happy [man and boy] left together hand <strong>in</strong> hand’<br />
(Dalrymple and Nikolaeva, 2006, 825)<br />
yhdessä<br />
together<br />
käsikädessä<br />
hand-hand.<strong>in</strong>es<br />
Also, <strong>in</strong> the example below, although the two nouns are def<strong>in</strong>ite <strong>in</strong>animate nouns, a<br />
plural adjective is still allowed accord<strong>in</strong>g to Dalrymple and Nikolaeva (2006), s<strong>in</strong>ce the two<br />
conjuncts refer to unique entities, they are subjects, and resemble somehow proper names:<br />
(123) ...wieniläiset<br />
...Viennese.pl<br />
kuoro<br />
choir<br />
ja<br />
and<br />
orkesteri<br />
orchestra<br />
hyökyvät<br />
surge.3.pl<br />
‘...Viennese choir and orchestra surge as they should’<br />
(Dalrymple and Nikolaeva, 2006, 828)<br />
ku<strong>in</strong><br />
as<br />
pitää<br />
needed<br />
F<strong>in</strong>ally, Dalrymple and Nikolaeva (2006) argue that when conjuncts are <strong>in</strong>def<strong>in</strong>ite then<br />
plural modification is less frequent and most of the times totally ungrammatical. Thus,<br />
the follow<strong>in</strong>g example cannot be accepted by native speakers of F<strong>in</strong>nish:<br />
(124) *Tässä<br />
this.<strong>in</strong>es<br />
kaupungissa<br />
city.<strong>in</strong>es<br />
on<br />
is<br />
hyvät<br />
good.pl<br />
yliopisto<br />
university<br />
‘There is a good university and theatre <strong>in</strong> this city’<br />
(Dalrymple and Nikolaeva, 2006, 829)<br />
ja<br />
and<br />
teatteri<br />
theatre<br />
To summarise the f<strong>in</strong>d<strong>in</strong>gs of Dalrymple and Nikolaeva (2006), all speakers accept a<br />
plural adjective with two s<strong>in</strong>gular conjuncts when the modifier is symmetric and when the<br />
conjuncts are either def<strong>in</strong>ite, animate, or human and <strong>in</strong> subject position. Speakers reject<br />
<strong>in</strong> general <strong>in</strong>def<strong>in</strong>ites and non-subject def<strong>in</strong>ites with plural adjectives 14 . The acceptance<br />
of a plural adjective with two s<strong>in</strong>gular nouns is attributed to the phenomenon of natural<br />
coord<strong>in</strong>ation. Thus, F<strong>in</strong>nish is one of those languages that allows both natural and accidental<br />
coord<strong>in</strong>ation. Accidental coord<strong>in</strong>ation is found with any k<strong>in</strong>d of nouns without<br />
any restrictions either syntactic or semantic <strong>in</strong> the coord<strong>in</strong>ate structure but modification of<br />
s<strong>in</strong>gular conjuncts by a plural adjective is disallowed. Natural coord<strong>in</strong>ation, on the other<br />
hand, <strong>in</strong>volves contextually related nouns which occur <strong>in</strong> s<strong>in</strong>gular number and are modified<br />
by a plural adjective, without us<strong>in</strong>g any specific syntactic constructions, such as tight<br />
coord<strong>in</strong>ation or <strong>in</strong>flectional mark<strong>in</strong>g.<br />
Dalrymple and Nikolaeva (2006) capture natural coord<strong>in</strong>ation phenomena with<strong>in</strong> LFG<br />
and assume the concord and <strong>in</strong>dex feature dist<strong>in</strong>ction. Accidentally coord<strong>in</strong>ated s<strong>in</strong>gular<br />
nouns talo ja auto ‘house and car’ have plural <strong>in</strong>dex s<strong>in</strong>ce the coord<strong>in</strong>ate phrase is plural as<br />
a whole, and s<strong>in</strong>gular concord because each conjunct is s<strong>in</strong>gular. The example is shown<br />
below:<br />
14 A detailed analysis and presentation of the F<strong>in</strong>nish data and some exceptions are found <strong>in</strong> Dalrymple<br />
and Nikolaeva (2006, 825-30).
3.3 Theoretical Approaches to Head-Modifier <strong>Agreement</strong> 65<br />
(125) Accidental coord<strong>in</strong>ation<br />
*vanhat/*vanha<br />
old.pl/old.sg<br />
‘old [house and car]’<br />
AdjP<br />
Adj<br />
*vanhat/*vanha<br />
NP<br />
N’<br />
talo<br />
house<br />
N<br />
talo<br />
N’<br />
N<br />
Conj<br />
ja<br />
ja<br />
and<br />
N<br />
auto<br />
(Dalrymple and Nikolaeva, 2006, 840)<br />
auto<br />
car<br />
⎡<br />
⎤<br />
adj [pred ‘old’]<br />
⎢<br />
⎥<br />
⎢<strong>in</strong>dex<br />
[num pl] ⎥<br />
⎢<br />
⎢⎧<br />
⎡<br />
⎤⎫<br />
⎥<br />
⎢ pred ‘house’ ⎥<br />
⎢ ⎢<br />
⎥ ⎥<br />
⎢ ⎣concord<br />
[num sg] ⎦ ⎥<br />
⎢⎪⎨<br />
⎪⎬ ⎥<br />
⎢ <strong>in</strong>dex [num sg] ⎥<br />
⎢<br />
⎡<br />
⎤ ⎥<br />
⎢ pred ‘car’<br />
⎥<br />
⎢<br />
⎢<br />
⎥ ⎥<br />
⎣<br />
⎣concord<br />
[num sg] ⎦ ⎥<br />
⎪⎩<br />
⎪⎭<br />
⎦<br />
<strong>in</strong>dex [num sg]<br />
F<strong>in</strong>nish adjectives constra<strong>in</strong> both the concord and <strong>in</strong>dex values of the conjuncts,<br />
mean<strong>in</strong>g that a s<strong>in</strong>gular adjective requires s<strong>in</strong>gular concord and <strong>in</strong>dex and a plural<br />
adjective requires plural concord and <strong>in</strong>dex. The example above cannot be modified<br />
by either a s<strong>in</strong>gular or a plural adjective. A s<strong>in</strong>gular adjective is disallowed with s<strong>in</strong>gular<br />
conjuncts s<strong>in</strong>ce the phrase ‘house and car’ has plural <strong>in</strong>dex, contradict<strong>in</strong>g the s<strong>in</strong>gular<br />
<strong>in</strong>dex required by the adjective, lead<strong>in</strong>g to agreement feature mismatch. Similarly, a<br />
plural adjective is also disallowed s<strong>in</strong>ce ‘house and car’ has s<strong>in</strong>gular concord but the<br />
adjective requires plural concord and aga<strong>in</strong> the agreement features do not match.<br />
In F<strong>in</strong>nish plural adjectives are allowed with s<strong>in</strong>gular conjuncts <strong>in</strong> natural coord<strong>in</strong>ation.<br />
In order to capture this structure, Dalrymple and Nikolaeva (2006) assume that F<strong>in</strong>nish<br />
specifies a different functional structure, which resembles somehow that of a simple plural<br />
noun. This structure is illustrated below:<br />
(126) Natural coord<strong>in</strong>ation<br />
uskolliset<br />
faithful.pl<br />
aviomies<br />
husband<br />
‘faithful [husband and wife]’<br />
AdjP<br />
Adj<br />
uskolliset<br />
NP<br />
N’<br />
N<br />
aviomies<br />
N’<br />
N<br />
Conj<br />
ja<br />
ja<br />
and<br />
N<br />
vaimo<br />
vaimo<br />
wife<br />
(Dalrymple and Nikolaeva, 2006, 842)<br />
⎡<br />
⎤<br />
pred ‘conj’<br />
⎢<br />
⎥<br />
⎢concord<br />
[num pl] ⎥<br />
⎢<br />
⎥<br />
⎢<br />
⎢<strong>in</strong>dex<br />
[num pl] ⎥<br />
⎢ ⎡<br />
⎤ ⎥<br />
⎢ pred ‘husband’ ⎥<br />
⎢ ⎢<br />
⎥⎥<br />
⎢conj1<br />
⎣concord<br />
[num sg] ⎦ ⎥<br />
⎢ <strong>in</strong>dex [num sg]<br />
⎥<br />
⎢<br />
⎡<br />
⎤⎥<br />
⎢ pred ‘wife’<br />
⎥<br />
⎢<br />
⎢<br />
⎥<br />
conj2<br />
⎢<br />
⎣concord<br />
[num sg] ⎦ ⎥<br />
⎢<br />
⎣ <strong>in</strong>dex [num sg] ⎥<br />
⎦<br />
adj [pred ‘faithful’]<br />
The plural adjective is allowed <strong>in</strong> natural coord<strong>in</strong>ation s<strong>in</strong>ce both concord and <strong>in</strong>dex<br />
have plural values which match the agreement requirements of a plural adjective.
3.4 Discuss<strong>in</strong>g the theoretical approaches to agreement 66<br />
The two characteristics concern<strong>in</strong>g the f-structure of natural coord<strong>in</strong>ation are that these<br />
structures have a specific number of conjuncts, usually no more than two whereas accidental<br />
coord<strong>in</strong>ation can hold more than two conjuncts. Secondly, natural coord<strong>in</strong>ation f-structure<br />
has plural concord and plural <strong>in</strong>dex which makes it similar to the f-structure of a simple<br />
plural noun (Dalrymple and Nikolaeva, 2006).<br />
On the whole, the above analysis captures natural coord<strong>in</strong>ation <strong>in</strong> F<strong>in</strong>nish. Dalrymple<br />
and Nikolaeva (2006) claim that the architecture of LFG itself is important s<strong>in</strong>ce it allows<br />
the separation of c-structure and f-structure, facilitat<strong>in</strong>g the analysis of F<strong>in</strong>nish by us<strong>in</strong>g<br />
the same c-structure <strong>in</strong> both natural and accidental coord<strong>in</strong>ation but a different f-structure.<br />
Thus, morphology and phrase structure is identical <strong>in</strong> both types of coord<strong>in</strong>ation while the<br />
only dist<strong>in</strong>ction is made at the f-structure level. Dalrymple and Nikolaeva (2006, 843-<br />
47) present data from other languages, such as Tundra Nenets, Russian and Bahd<strong>in</strong>ani<br />
Kurdish, which are captured by the same approach.<br />
An important aspect of the current theory is that it deals with cases of natural coord<strong>in</strong>ation<br />
and not only with accidental coord<strong>in</strong>ation as most theories do. Also, it proposes<br />
a different f-structure for natural coord<strong>in</strong>ation patterns that captures a great number of<br />
data <strong>in</strong> various languages. The ma<strong>in</strong> drawback is that there are languages like MG whose<br />
natural coord<strong>in</strong>ation patterns cannot be captured by the same f-structure. We will discuss<br />
this issue, however, <strong>in</strong> more detail <strong>in</strong> the next section.<br />
3.4 Discuss<strong>in</strong>g the theoretical approaches to agreement<br />
The above section presented a summary of the theoretical approaches to predicate-argument<br />
and head-modifier agreement, which discuss the gender and number features <strong>in</strong> noun<br />
phrase coord<strong>in</strong>ate structures. In this last section, we will discuss and compare the aforementioned<br />
theories.<br />
In theoretical approaches to predicate-argument agreement, it is generally acknowledged<br />
that gender resolution should comb<strong>in</strong>e syntactic and semantic pr<strong>in</strong>ciples. Most of the<br />
analyses <strong>in</strong>corporate <strong>in</strong>to their account a proposal along these l<strong>in</strong>es apart from Dalrymple<br />
and Kaplan (2000). The latter present a syntactically motivated gender agreement theory<br />
but they admit that there are patterns <strong>in</strong> a number of languages that should be exam<strong>in</strong>ed<br />
from a semantic perspective s<strong>in</strong>ce they show semantic resolution. The follow<strong>in</strong>g example<br />
<strong>in</strong> French is repeated from above:<br />
(127) La personnne avec la barbe<br />
[The person with the beard].fem<br />
idiots/?*idiotes<br />
idiots.masc/?*fem<br />
‘The person with the beard and Marie are idiots’<br />
(Wechsler, 2009)<br />
et<br />
and<br />
Marie<br />
Marie.fem<br />
sont<br />
are<br />
Secondly, all of the theories capture the specific patterns they present <strong>in</strong> the languages<br />
they discuss, even though their actual analysis differs significantly. Dalrymple and Kaplan<br />
(2000) use set-union and Wechsler and Zlatić (2003) use set-<strong>in</strong>tersection. Badecker (2008)<br />
captures gender agreement <strong>in</strong> a large group of languages, rang<strong>in</strong>g from Romance, Slavic,<br />
Icelandic to MG, adopt<strong>in</strong>g a limited number of OT constra<strong>in</strong>ts whose rank<strong>in</strong>g depends<br />
on the language analysed. F<strong>in</strong>ally, Sadler (2006), who also uses set-union, accommodates<br />
mixed gender agreement <strong>in</strong> Rumanian assum<strong>in</strong>g two different PS rules for <strong>in</strong>animate and<br />
animate nouns.<br />
In all these theories, however, there are some important issues that should be discussed.<br />
Corbett (1991) uses a descriptive approach which <strong>in</strong>cludes a series of rules at a theoretical<br />
level. His approach is purely descriptive and does not assume a specific theoretical
3.4 Discuss<strong>in</strong>g the theoretical approaches to agreement 67<br />
framework while he does not attempt to expla<strong>in</strong> the l<strong>in</strong>guistic motivation for some agreement<br />
patterns and not for others; <strong>in</strong>stead he merely records them. Also, the rules that he<br />
proposes for the gender feature are language specific and not crossl<strong>in</strong>guistic as would be<br />
desirable. The ma<strong>in</strong> advantage, though, is that he emphasises the dist<strong>in</strong>ction between two<br />
separate pr<strong>in</strong>ciples <strong>in</strong> gender resolution, the semantic and the syntactic, and this idea was<br />
widely adopted by his followers. Only later on <strong>in</strong> Corbett and Fraser (1993), he proposes<br />
an analysis for animacy <strong>in</strong> Russian nom<strong>in</strong>al <strong>in</strong>flections with<strong>in</strong> Network Morphology.<br />
Dalrymple and Kaplan (2000) present a rather powerful theory comb<strong>in</strong><strong>in</strong>g set-valued<br />
features by set-union. The first problem occurs <strong>in</strong> Slovene where they need to impose an<br />
additional constra<strong>in</strong>t <strong>in</strong> the conjunction conj, an equation f ∈(↑gender), s<strong>in</strong>ce the sets<br />
used for the three genders and their union cannot capture the pattern where two neuter<br />
nouns resolve to mascul<strong>in</strong>e. Thus, the {f} set needs to participate also <strong>in</strong> the union process.<br />
Therefore, their approach cannot cover all cases of coord<strong>in</strong>ation but needs additional<br />
constra<strong>in</strong>ts for any extra or unexpected syntactic pattern that occurs <strong>in</strong> a language. Two<br />
other problems discussed by Wechsler and Zlatić (2003) are those of “undergeneration”<br />
and “overgeneration”. The former possibly refers to those cases of semantic resolution<br />
that Dalrymple and Kaplan (2000) do not capture at all, while the latter refers to some<br />
patterns that are unattested and are still generated by a set-based approach with the union<br />
operation. Wechsler and Zlatić (2003) claim that the follow<strong>in</strong>g unattested pattern can be<br />
generated by the current proposal:<br />
(128) *Unattested (m, f, nt)<br />
a. FEM & FEM = FEM<br />
b. Elsewhere NEUT (NEUT & NEUT; MASC & MASC; and mixes)<br />
(Wechsler and Zlatić, 2003, p.186)<br />
Their presupposition, however, is not expla<strong>in</strong>ed any further <strong>in</strong> terms of the marker sets that<br />
should be assigned to the different genders and s<strong>in</strong>ce there are not any such known gender<br />
systems so far, we cannot consider their argument valid. F<strong>in</strong>ally, V<strong>in</strong>cent and Börjars<br />
(2000) discuss a few more problems concern<strong>in</strong>g the analysis proposed by Dalrymple and<br />
Kaplan (2000), which are worth mentioned here. Regard<strong>in</strong>g first person, they argue that<br />
the different use of features across languages implies that personal identity is conceived<br />
differently while the primitives s and h do not ma<strong>in</strong>ta<strong>in</strong> their semantic content, mean<strong>in</strong>g<br />
that h cannot mean hearer if it appears <strong>in</strong> the set of the pronoun I. Also, V<strong>in</strong>cent and<br />
Börjars (2000) argue that English requires two types of first person based on semantic<br />
and pragmatic evidence. In a control context, you and I differs from John and I <strong>in</strong> the<br />
value of pro 15 . Regard<strong>in</strong>g gender, they argue that feature assignment <strong>in</strong> each language<br />
is determ<strong>in</strong>ed by the union operation, which is compatible with LFG’s characteristic of<br />
construct<strong>in</strong>g the smallest f-structure. An alternative operation could be set <strong>in</strong>tersection,<br />
which constructs the largest set. However, Mary Dalrymple po<strong>in</strong>ts out that <strong>in</strong>tersection is<br />
at odds with the basic architecture of LFG and the fact that the smallest set is consistent<br />
with a given f-description. This means that the <strong>in</strong>tersection of identical gender conjuncts,<br />
such as {m} and {m}, would result <strong>in</strong> {} s<strong>in</strong>ce the empty set is the smallest set which is<br />
15 V<strong>in</strong>cent and Börjars (2000, 7) argue that this is obvious <strong>in</strong> anaphoric control where split antecedence<br />
is allowed and the vale of PRO can refer to two dist<strong>in</strong>ct locations <strong>in</strong> the clause’s f-structure. Consider the<br />
follow<strong>in</strong>g examples:<br />
(1) a. Do you remember that I discussed with you offer<strong>in</strong>g ourselves as candidates for the election?<br />
b. Do you remember that I discussed with John offer<strong>in</strong>g ourselves as candidates for the election?<br />
In (1a), the subject of offer<strong>in</strong>g and antecedent of ourselves is the comb<strong>in</strong>ation of you and I, which is first<br />
person <strong>in</strong>clusive. In (1b), the subject of offer<strong>in</strong>g and antecedent of ourselves is either you and I, John and<br />
I or John, you and I.
3.4 Discuss<strong>in</strong>g the theoretical approaches to agreement 68<br />
a subset of both {m} and {m} (V<strong>in</strong>cent and Börjars, 2000, 16). This is an undesirable<br />
result. F<strong>in</strong>ally, V<strong>in</strong>cent and Börjars (2000) argue that the different set valued features<br />
adopted <strong>in</strong> each language are aga<strong>in</strong>st a crossl<strong>in</strong>guistic analysis. Although we agree with<br />
V<strong>in</strong>cent and Borjars’ arguments concern<strong>in</strong>g person, we strongly disagree with their claims<br />
on gender. The theories showed that gender varies greatly and even <strong>in</strong> a s<strong>in</strong>gle language<br />
different resolution results are drawn depend<strong>in</strong>g on the type of the conjuncts (i.e. animate<br />
or <strong>in</strong>animate). Thus, a crossl<strong>in</strong>gustic theory would be difficult to assume s<strong>in</strong>ce there will<br />
still be languages that would not be accounted for.<br />
The theory developed by Wechsler and Zlatić (2003) has a major advantage. It acknowledges<br />
the dist<strong>in</strong>ction of different resolution rules <strong>in</strong> animate and <strong>in</strong>animate nouns<br />
and provides an account based on that. However, the ma<strong>in</strong> problem is that they do not<br />
analyse languages where the resolution patterns are not common <strong>in</strong> animate and <strong>in</strong>animate<br />
nouns. They make a brief reference to Rumanian as one of those languages, claim<strong>in</strong>g that<br />
“the resolution mechanism for <strong>in</strong>animates is <strong>in</strong>dependent of the determ<strong>in</strong>ants of animate<br />
gender agreement”(Wechsler and Zlatić, 2003, 189) but they do not proceed to any analysis<br />
with<strong>in</strong> the context of their own theory. Moreover, their theory seems to be problematic<br />
for the Slovene extra pattern. The requirement for a masc resolved gender cannot be generated<br />
from the <strong>in</strong>tersection of two neuter conjuncts s<strong>in</strong>ce the <strong>in</strong>tersection of two neuter<br />
nouns will result <strong>in</strong> neuter and not mascul<strong>in</strong>e. Thus, we argue that the problem of undergeneration<br />
is also detected <strong>in</strong> this theory. F<strong>in</strong>ally, as Badecker (2008) claims, Wechsler<br />
and Zlatić (2003) do not provide any account for cases where number <strong>in</strong>teracts with gender<br />
<strong>in</strong> order to provide the required gender agreement form, which is a major drawback<br />
for languages like MG and Serbian/Croatian. This is <strong>in</strong>deed true and the same criticism<br />
could apply to Dalrymple and Kaplan’s (2000) approach, who do not take <strong>in</strong>to account the<br />
<strong>in</strong>teraction of other <strong>in</strong>dex features apart from the gender when they provide an account<br />
for the latter.<br />
Sadler (2006) proposes an account for languages with different resolution patterns <strong>in</strong><br />
animate and <strong>in</strong>animate nouns. The previous theories have addressed the issue but they<br />
have not proposed any concrete analysis, leav<strong>in</strong>g <strong>in</strong> the marg<strong>in</strong> a number of languages<br />
with this characteristic, such as Lat<strong>in</strong>, Polish, MG and many others. Also, Sadler (2006)<br />
developes a concise proposal s<strong>in</strong>ce <strong>in</strong> a simple rule us<strong>in</strong>g disjunction she captures both<br />
semantic and syntactic resolution found <strong>in</strong> animate and <strong>in</strong>animate Rumanian nouns, accord<strong>in</strong>gly.<br />
A second advantage is that she comb<strong>in</strong>es a crossl<strong>in</strong>guistic analysis, the set-union<br />
rule, which accounts for the expected patterns of resolution <strong>in</strong> <strong>in</strong>animate nouns found <strong>in</strong><br />
most languages, with a language specific one, the semantic rule of resolution proposed for<br />
Rumanian animate nouns. In fact, this approach could capture resolution <strong>in</strong> coord<strong>in</strong>ate<br />
nouns <strong>in</strong> a number of languages s<strong>in</strong>ce the only modifications required would be to adapt<br />
the equations <strong>in</strong> the semantic rule of resolution for the animate nouns, accord<strong>in</strong>g to the<br />
language analysed, and reta<strong>in</strong> the standard rule of coord<strong>in</strong>ation for the <strong>in</strong>animate nouns.<br />
Her attempt carries the same <strong>in</strong>tuitions as that of Wechsler and Zlatić (2003). The only<br />
disadvantage is that she assumes an extra f-structure feature, semgend, to account for<br />
the semantic patterns <strong>in</strong> Rumanian. It might be desirable to replace this feature by a<br />
constra<strong>in</strong>t which refers to the semantic structure on <strong>in</strong>terpretation. We will show, though,<br />
that this extra feature is required <strong>in</strong> the analysis of the MG animate nouns and possibly<br />
<strong>in</strong> other similar languages, mean<strong>in</strong>g that it is not really a redundant feature.<br />
Badecker (2008) proposes an analysis of gender resolution with<strong>in</strong> the Optimality Theory.<br />
He assumes a number of similar constra<strong>in</strong>ts which differ across languages only <strong>in</strong><br />
their rank<strong>in</strong>g and captures a great range of gender agreement patterns mak<strong>in</strong>g the proposal<br />
economic and crossl<strong>in</strong>guistic. He also takes <strong>in</strong>to consideration the different gender<br />
resolution patterns of animate and <strong>in</strong>animate nouns for some languages and other factors,<br />
such as the number factor <strong>in</strong> the conjuncts, which determ<strong>in</strong>e gender resolution <strong>in</strong> some
3.4 Discuss<strong>in</strong>g the theoretical approaches to agreement 69<br />
cases. A problem is that for any language with different features such as Tamil with the<br />
rational/non-rational dist<strong>in</strong>ction, new constra<strong>in</strong>ts need to be assumed which may result<br />
<strong>in</strong> a different group of constra<strong>in</strong>ts depend<strong>in</strong>g on the language. This would underm<strong>in</strong>e the<br />
economy and conciseness of the exist<strong>in</strong>g theory. Also, agreement is modelled by competition<br />
imply<strong>in</strong>g that many different forms may be possible but only one is the optimal. In<br />
some languages, however, like MG we will see that there are other possible agreement patterns<br />
and therefore agreement is not the result of only one correct form. This fact makes<br />
the theory rather undesirable.<br />
The head-modifier theories focus on number NP <strong>in</strong>ternal agreement. Kathol (1999) was<br />
the first to postulate a dist<strong>in</strong>ction between the two features and these were later adopted<br />
<strong>in</strong> LFG. Also, Wechsler and Zlatić (2000) <strong>in</strong>troduce two additional types of <strong>in</strong>formation<br />
declension and semantics, and capture further agreement mismatches, apart from those<br />
found between concord and <strong>in</strong>dex.<br />
The proposal developed by K<strong>in</strong>g and Dalrymple (2004) is powerful s<strong>in</strong>ce it extends<br />
to determ<strong>in</strong>ers and adjectives as modifiers and it is crossl<strong>in</strong>guistic, account<strong>in</strong>g for various<br />
phenomena <strong>in</strong> NP-<strong>in</strong>ternal agreement <strong>in</strong> a wide range of languages. MG, however,<br />
presents some restrictions that extend beyond the area of syntax. In particular, the <strong>in</strong>dex/concord<br />
system <strong>in</strong> the s<strong>in</strong>gular accounts for noun phrases such as my friend and<br />
colleague is where <strong>in</strong>dex has a sg value and the coord<strong>in</strong>ate phrase is semantically <strong>in</strong>terpreted<br />
as referr<strong>in</strong>g to one and the same person only (jo<strong>in</strong>t read<strong>in</strong>g). In the plural, the<br />
same system accounts for noun phrases such as my friends and colleagues are where <strong>in</strong>dex<br />
has a pl value and the coord<strong>in</strong>ate phrase is semantically <strong>in</strong>terpreted as denot<strong>in</strong>g the<br />
same referents (jo<strong>in</strong>t) or different referents (split) s<strong>in</strong>ce the resolved number is plural due<br />
to the plural number <strong>in</strong> both conjuncts. Although the <strong>in</strong>dex/concord system captures<br />
syntactically the MG data, the semantic dist<strong>in</strong>ction <strong>in</strong> plural number is problematic s<strong>in</strong>ce<br />
MG occurs only with the jo<strong>in</strong>t read<strong>in</strong>g when a shared determ<strong>in</strong>er scopes over coord<strong>in</strong>ate<br />
nouns either <strong>in</strong> the s<strong>in</strong>gular or <strong>in</strong> the plural number. The split read<strong>in</strong>g is allowed only <strong>in</strong><br />
specific constructions, as we shall see. Thus, MG plural coord<strong>in</strong>ate nouns behave different<br />
when compared to the languages discussed by K<strong>in</strong>g and Dalrymple (2004).<br />
F<strong>in</strong>ally, the theory proposed by Dalrymple and Nikolaeva (2006) is important s<strong>in</strong>ce it is<br />
the only one that discusses the impact of “natural” and “accidental” noun phrase coord<strong>in</strong>ation<br />
<strong>in</strong> agreement <strong>in</strong> constra<strong>in</strong>t-based frameworks, such as LFG and HPSG. The difference<br />
between the two types of coord<strong>in</strong>ation and the agreement requirements are captured by<br />
two different f-structures. The natural coord<strong>in</strong>ation f-structure does not represent its conjuncts<br />
us<strong>in</strong>g a set and it postulates plural concord and plural <strong>in</strong>dex requirements. The<br />
ma<strong>in</strong> drawback, however, is whether the f-structure proposed is found crossl<strong>in</strong>guistically.<br />
We will show that at least <strong>in</strong> MG the exact same f-structure <strong>in</strong>troduced by Dalrymple and<br />
Nikolaeva (2006) cannot capture the MG cases of natural coord<strong>in</strong>ation and the agreement<br />
requirements. Languages with similar behaviour may exist and the present f-structure<br />
would require modifications depend<strong>in</strong>g on the characteristics of each language.<br />
Despite the problems discussed <strong>in</strong> the various approaches, each of these theories contributes<br />
greatly and sheds light on the agreement patterns detected <strong>in</strong> the various languages.<br />
The follow<strong>in</strong>g chapters will focus on the analysis of the MG data and we will show that the<br />
ma<strong>in</strong> concepts of some of these theories are crucial for the explanation of some agreement<br />
patterns <strong>in</strong> MG.
Chapter 4<br />
Predicate Argument <strong>Agreement</strong> <strong>in</strong><br />
<strong>Modern</strong> <strong>Greek</strong><br />
4.1 Introduction<br />
<strong>Modern</strong> <strong>Greek</strong> predicate-argument agreement <strong>in</strong> coord<strong>in</strong>ate noun phrases <strong>in</strong>volves the features<br />
gender, number and person. In <strong>Modern</strong> <strong>Greek</strong>, number and person resolution is<br />
straightforward and resembles the patterns found <strong>in</strong> other languages. person resolution,<br />
follows the precedence hierarchy (see Corbett (1991); Siewierska (2004)) where the first<br />
person takes precedence over the second, and the second takes precedence over the third.<br />
In number resolution, coord<strong>in</strong>ation of s<strong>in</strong>gular elements requires plural number (outside<br />
of the jo<strong>in</strong>t read<strong>in</strong>g) <strong>in</strong> the outcome. gender resolution, however, is much more complex<br />
due to the variation it shows and the difficulty for the current theories <strong>in</strong> account<strong>in</strong>g for it.<br />
In this chapter, we will exam<strong>in</strong>e gender agreement <strong>in</strong> predicate argument relations<br />
<strong>in</strong> <strong>Modern</strong> <strong>Greek</strong>. The first section presents the MG data. The next section shows that<br />
MG coord<strong>in</strong>ate nouns follow different resolution pr<strong>in</strong>ciples; animate nouns follow semantic<br />
resolution and <strong>in</strong>animate nouns follow syntactic resolution. Our analysis will attempt<br />
to capture both. However, a number of unexpected resolution patterns occur which are<br />
evidence for a third resolution pr<strong>in</strong>ciple. This is referentially derived and depends on<br />
the assumption of a Contextually Introduced element <strong>in</strong> the discourse. The phenomenon<br />
of referential agreement has not been discussed so far <strong>in</strong> the present literature, such as<br />
descriptive grammars or traditional handbooks (Clairis et al. (2004), Holton et al. (1997),<br />
Triantaphyllidis (1994), Triantaphyllidis (2005), Joseph and Philippaki-Warburton (1987)<br />
and Mackridge (1985)). The ma<strong>in</strong> characteristics of referential agreement and an analysis<br />
of the data are presented <strong>in</strong> the last two sections.<br />
4.2 The <strong>Modern</strong> <strong>Greek</strong> Data<br />
MG animate and <strong>in</strong>animate nouns belong to one of the three gender classes: mascul<strong>in</strong>e,<br />
fem<strong>in</strong><strong>in</strong>e and neuter. In animate nouns, mascul<strong>in</strong>e refers to males whereas fem<strong>in</strong><strong>in</strong>e refers<br />
to females. Some exceptions <strong>in</strong> MG are nouns whose grammatical gender contradicts<br />
the natural/semantic one, such as agori ‘boy’ and koritsi ‘girl’, which are grammatically<br />
neuter but have a mascul<strong>in</strong>e and a fem<strong>in</strong><strong>in</strong>e natural/semantic gender, respectively, the<br />
noun anthropos ‘human be<strong>in</strong>g/person’, which is grammatically mascul<strong>in</strong>e but refers to both<br />
males and females, and nouns, which are grammatically neuter but lack a natural/semantic<br />
gender, such as pedi ‘child’, moro ‘baby’ and atomo ‘person’ (see Holton et al. (1997) and<br />
Mackridge (1985)). Inanimate nouns belong to any of the three gender classes, assigned<br />
on formal grounds depend<strong>in</strong>g on the declension class of the noun.<br />
Non-coord<strong>in</strong>ate noun phrases control their agreement “target” (Corbett, 1991), forc<strong>in</strong>g
4.2 The <strong>Modern</strong> <strong>Greek</strong> Data 71<br />
the latter to show grammatical agreement with the former. Thus, the predicative adjective<br />
or participle will agree <strong>in</strong> gender and number 1 with the noun it is predicated of 2 :<br />
(1) O Kostas<br />
the.masc.sg Kostas.masc.sg<br />
‘Kostas is handsome’<br />
(constructed)<br />
(2) I j<strong>in</strong>ekes<br />
the.fem.pl women.fem.pl<br />
‘The women are happy’<br />
(constructed)<br />
<strong>in</strong>e<br />
is.sg<br />
<strong>in</strong>e<br />
are.pl<br />
omorf-os<br />
handsome.masc.sg<br />
xarumen-es<br />
happy.fem.pl<br />
Even nouns whose grammatical gender contradicts their natural control predicative<br />
adjectives and force them to show grammatical gender agreement. For example, the noun<br />
to koritsi ‘the girl’ allows only a neuter predicative adjective despite the fact that it has a<br />
female referent:<br />
(3) To koritsi<br />
the.neut.sg girl.neut.sg<br />
‘The girl is beautiful’<br />
(constructed)<br />
<strong>in</strong>e<br />
is.sg<br />
omorf-o<br />
beautiful.neut.sg<br />
Semantic agreement is also possible once a pronoun occurs subsequently <strong>in</strong> the sentence<br />
and refers back to the “controller” (Corbett, 1991) noun, follow<strong>in</strong>g Corbett’s <strong>Agreement</strong><br />
Hierarchy 3 .<br />
MG <strong>in</strong>animate concrete and abstract nouns 4 also show grammatical agreement. Thus,<br />
the predicative adjective or participle shows agreement with the noun it is predicated of:<br />
(4) O kipos<br />
the.masc.sg garden.masc.sg<br />
‘The garden is big’<br />
(constructed)<br />
(5) I dimokratia<br />
the.fem.sg democracy.fem.sg<br />
‘Democracy is necessary’<br />
(constructed)<br />
<strong>in</strong>e<br />
is.sg<br />
<strong>in</strong>e<br />
is.sg<br />
megal-os<br />
big.masc.sg<br />
aparetit-i<br />
necessary.fem.sg<br />
Gender resolution <strong>in</strong> MG coord<strong>in</strong>ate nouns is <strong>in</strong>terest<strong>in</strong>g due to the existence of different<br />
resolution patterns. We impose a dist<strong>in</strong>ction of two major group patterns, the expected<br />
and unexpected ones.<br />
1 As previously mentioned, number is not important for the time be<strong>in</strong>g and therefore we will not discuss<br />
it further.<br />
2 The follow<strong>in</strong>g examples were collected from <strong>Greek</strong> newspapers and various books. I will use the abbre-<br />
viation G.N. to stand for <strong>Greek</strong> Newspaper and G.M. to stand for <strong>Greek</strong> Magaz<strong>in</strong>e followed by the specific<br />
newspaper or magaz<strong>in</strong>e from which I have collected each example. The examples that were constructed are<br />
<strong>in</strong>dicated with the word ‘constructed’.<br />
3 The agreement hierarchy is repeated below:<br />
(1) <strong>Agreement</strong> Hierarchy<br />
attributive < predicate < relative pronoun < personal pronoun<br />
Corbett (1979) argues that “as we move rightwards along the hierarchy, the likelihood of semantic agree-<br />
ment will <strong>in</strong>crease monotonically”(210).<br />
4 We subdivide <strong>in</strong>animate nouns <strong>in</strong>to concrete and abstract s<strong>in</strong>ce this turns out to be relevant <strong>in</strong> the<br />
sections that follow.
4.2 The <strong>Modern</strong> <strong>Greek</strong> Data 72<br />
The expected patterns refer to cases of either grammatical (syntactic) or semantic resolution.<br />
We assume that syntactic resolution occurs <strong>in</strong> MG <strong>in</strong>animate nouns and semantic<br />
resolution occurs <strong>in</strong> MG animate nouns, follow<strong>in</strong>g Corbett (1991) and Wechsler and Zlatić<br />
(2003). Thus, the expected cases of MG are those that show either syntactic or semantic<br />
gender resolution, accord<strong>in</strong>g to animacy dist<strong>in</strong>ctions.<br />
Alongside the expected gender patterns observed, an analysis of data collection by<br />
questionnaire showed the existence of a number of unexpected agreement patterns <strong>in</strong> both<br />
groups of nouns. We will argue that these unexpected patterns fall with<strong>in</strong> a third pr<strong>in</strong>ciple,<br />
which is motivated by “reference transfer” (Nunberg, 1977, 1995, 2005; Pollard and Sag,<br />
1994). Therefore, we postulate three pr<strong>in</strong>ciples of resolution <strong>in</strong> MG, the syntactic, the<br />
semantic and the referential one. As already mentioned above, this last pr<strong>in</strong>ciple has not<br />
been discussed anywhere <strong>in</strong> the literature on MG.<br />
Another unusual aspect of coord<strong>in</strong>ate nouns <strong>in</strong> MG is that <strong>in</strong> some cases only the number<br />
of the conjuncts, whether s<strong>in</strong>gular or plural, plays a role <strong>in</strong> determ<strong>in</strong><strong>in</strong>g the preference<br />
for a resolution pattern. Thus, <strong>in</strong> <strong>in</strong>animate nouns a specific resolved gender occurs with<br />
higher frequency <strong>in</strong> s<strong>in</strong>gular conjuncts than <strong>in</strong> plural conjuncts or <strong>in</strong> plural conjuncts than<br />
<strong>in</strong> s<strong>in</strong>gular conjuncts.<br />
The current discussion will start with the presentation of the patterns found <strong>in</strong> animate<br />
coord<strong>in</strong>ate nouns and will proceed to <strong>in</strong>animate coord<strong>in</strong>ate nouns while we will analyse coord<strong>in</strong>ate<br />
phrases with s<strong>in</strong>gular conjuncts separately from plural conjuncts. The method<br />
used to draw the results of the gender values <strong>in</strong> both groups of nouns is based on the responses<br />
of 26 <strong>Greek</strong> native speakers who answered certa<strong>in</strong> questionnaires. The participants<br />
are all university graduates from North and South Greece. The questionnaires appeared<br />
<strong>in</strong> two different forms, <strong>in</strong> the form of ‘multiple choice’ and <strong>in</strong> the form of ‘fill <strong>in</strong> the gaps’.<br />
The answers of the consultants were based on acceptability judgements and, even though<br />
the number of participants is relatively small, their answers are <strong>in</strong> accordance to my own<br />
orig<strong>in</strong>al <strong>in</strong>tuition about the various gender resolution patterns. F<strong>in</strong>ally, we need to note<br />
that the examples that will follow are all part of the questionnaires that were issued to the<br />
consultants 5 .<br />
4.2.1 Resolution <strong>in</strong> MG Animate <strong>Noun</strong>s<br />
In animate coord<strong>in</strong>ate nouns, the ma<strong>in</strong> f<strong>in</strong>d<strong>in</strong>gs showed that coord<strong>in</strong>ate phrases with s<strong>in</strong>gular<br />
and plural conjuncts display the same expected patterns. Additional unexpected<br />
patterns are found <strong>in</strong> the case of plural conjuncts.<br />
The expected resolved patterns of animate s<strong>in</strong>gular coord<strong>in</strong>ate nouns are discussed first.<br />
The responses of the 26 consultants are presented <strong>in</strong> the follow<strong>in</strong>g table 6 :<br />
Table 4.1: Animate S<strong>in</strong>gular Coord<strong>in</strong>ate <strong>Noun</strong>s<br />
Consultant Number: 26<br />
NP1 NP2 Resolution NP1 NP2 Resolution<br />
Mpl Fpl Npl Mpl Fpl Npl<br />
Msg Msg 26 - - Msg Fsg 26 - -<br />
Fsg Fsg - 26 - Msg Nm/f /0sg 26 - -<br />
Nsg Nsg - - 26 Fsg Nm/0sg 26 - -<br />
- - Fsg Nf sg 5 21 -<br />
5 For more details on the methodology and rationale on construct<strong>in</strong>g the questionnaires and on the<br />
responses of the consultants see Appendix A.<br />
6 The order of the conjuncts <strong>in</strong> the table does not correspond to the order of the conjuncts <strong>in</strong> the sentences<br />
that appear <strong>in</strong> the questionnaires. The questionnaires <strong>in</strong>cluded sentences with conjuncts <strong>in</strong> the reverse order<br />
too. We simply present the comb<strong>in</strong>ation of conjuncts <strong>in</strong> the table and show the responses.
4.2 The <strong>Modern</strong> <strong>Greek</strong> Data 73<br />
When a mascul<strong>in</strong>e animate noun is coord<strong>in</strong>ated with either a mascul<strong>in</strong>e, a fem<strong>in</strong><strong>in</strong>e<br />
or a neuter 7 the resolution form is always mascul<strong>in</strong>e, as <strong>in</strong> examples (6), (7) and (8),<br />
respectively:<br />
(6) O Petros ke o<br />
Janis<br />
the.masc.sg Peter.masc.sg and the.masc.sg John.masc.sg<br />
<strong>in</strong>e kurasmen-i. I dulia tus exi eksantlisi.<br />
are.pl tired.masc.pl. The job them.masc.pl has exhausted<br />
‘Peter and John are tired. The work has exhausted them’<br />
(constructed)<br />
(7) O pateras ke i mitera<br />
the.masc.sg father.masc.sg and the.fem.sg mother.fem.sg<br />
<strong>in</strong>e kurasmen-i. I zoi tus <strong>in</strong>e diskoli<br />
are.pl tired.masc.pl. The life their.masc.pl is difficult<br />
ke apetitiki<br />
and demand<strong>in</strong>g<br />
‘The father and mother are tired. Their life is difficult and demand<strong>in</strong>g’<br />
(constructed)<br />
(8) O Kostas ke<br />
the.masc.sg Kostas.masc.sg and<br />
<strong>in</strong>e poli exipn-i<br />
are.pl very clever.masc.pl<br />
‘Kostas and his child are very clever’<br />
(constructed)<br />
to<br />
the.neut.sg<br />
pedi<br />
child.neut.sg<br />
The responses of the consultants showed that a mascul<strong>in</strong>e gender is chosen by all (26 out<br />
of 26) consultants when a mascul<strong>in</strong>e noun cooccurs with another mascul<strong>in</strong>e conjunct, a<br />
fem<strong>in</strong><strong>in</strong>e conjunct or a neuter conjunct.<br />
Otherwise, resolution is fem<strong>in</strong><strong>in</strong>e if two fem<strong>in</strong><strong>in</strong>e nouns are coord<strong>in</strong>ated, as <strong>in</strong> (9), and<br />
neuter if the coord<strong>in</strong>ated nouns are neuter, as <strong>in</strong> (10):<br />
(9) I Maria<br />
the.fem.sg Maria.fem.sg<br />
<strong>in</strong>e xarumen-es<br />
are.pl happy.fem.pl<br />
‘Mary and Kathr<strong>in</strong>e are happy’<br />
(constructed)<br />
ke<br />
and<br />
i<br />
the.fem.sg<br />
(10) To agori ke to<br />
the.neut.sg boy.neut.sg and the.neut.sg<br />
vriskontan kathismen-a piso<br />
were.pl seated.neut.pl back<br />
‘The boy and the girl were seated at the back’<br />
(constructed)<br />
Kater<strong>in</strong>a<br />
Kater<strong>in</strong>a.fem.sg<br />
koritsi<br />
girl.neut.sg<br />
As shown <strong>in</strong> table (4.1), all consultants (26 of 26) chose a fem<strong>in</strong><strong>in</strong>e gender <strong>in</strong> example (9),<br />
and a neuter gender <strong>in</strong> example (10).<br />
7 As presented <strong>in</strong> section 1.2, the two grammatically neuter nouns agori ‘boy’ and and koritsi ‘girl’ refer<br />
to a male and a female,respectively, while the pedi ‘child’, moro ‘baby’ and atomo ‘person’ refer to either.<br />
For these nouns, we also <strong>in</strong>troduce the notation Nm/f /0. Thus, if a grammatically neuter noun refers to a<br />
male, it will be described as Nm, if it refers to a female, it will be described as Nf and if it refers to a noun<br />
without a natural/semantic gender, it will be described as N0.<br />
tu<br />
his
4.2 The <strong>Modern</strong> <strong>Greek</strong> Data 74<br />
When a grammatically mascul<strong>in</strong>e noun is coord<strong>in</strong>ated with a grammatically neuter<br />
noun, which is either a semantic neuter noun or a semantically-vacuous neuter noun 8 ,<br />
gender resolution is also mascul<strong>in</strong>e. In the follow<strong>in</strong>g examples, all consultants (26 out of<br />
26) chose a mascul<strong>in</strong>e gender irrespective of the gender denotation of the second conjunct:<br />
(11) O Kostas ke to<br />
the.masc.sg Kostas.masc.sg and the.neut.sg<br />
<strong>in</strong>e etimi ja anaxorisi<br />
are.pl ready.masc.pl to depart<br />
‘Kostas and the boy are ready to depart’<br />
(constructed)<br />
(12) O Kostas ke to<br />
the.masc.sg Kostas.masc.sg and the.neut.sg<br />
<strong>in</strong>e etimi ja anaxorisi<br />
are.pl ready.masc.pl to depart<br />
‘Kostas and the girl are ready to depart’<br />
(constructed)<br />
(13) O Kostas ke to<br />
the.masc.sg Kostas.masc.sg and the.neut.sg<br />
<strong>in</strong>e etimi ja anaxorisi<br />
are.pl ready.masc.pl to depart<br />
‘Kostas and the baby are ready to depart’<br />
(constructed)<br />
agori<br />
boy.neut.sg<br />
koritsi<br />
girl.neut.sg<br />
moro<br />
baby.neut.sg<br />
Also, when two nouns denote females then resolution is fem<strong>in</strong><strong>in</strong>e; for example, when<br />
the grammatically neuter noun to koritsi ‘the girl’, which refers to a female, is coord<strong>in</strong>ated<br />
with a grammatically fem<strong>in</strong><strong>in</strong>e noun 21 (out of 26) consultants chose the fem<strong>in</strong><strong>in</strong>e gender.<br />
However, the questionnaires showed that 5 (out of 26) consultants chose mascul<strong>in</strong>e resolution.<br />
This is due to the fact that mascul<strong>in</strong>e is the default semantic gender <strong>in</strong> MG. Thus,<br />
when referents denote females, resolution is fem<strong>in</strong><strong>in</strong>e and mascul<strong>in</strong>e occurs less often as a<br />
default gender form. The example that was given to the consultants is shown below:<br />
(14) I Maria ke to koritsi<br />
the.fem.sg Maria.fem.sg and the.neut.sg girl.neut.sg<br />
<strong>in</strong>e enthusiasmen-es/-i ja t<strong>in</strong> kenurja<br />
are.pl enthusiastic.fem.pl/masc.pl for the new<br />
tus<br />
zoi<br />
their.fem/masc.pl life.<br />
‘Mary and her girl are enthusiastic about their new life’<br />
(constructed)<br />
Any other case of coord<strong>in</strong>ation of a grammatically fem<strong>in</strong><strong>in</strong>e noun with a grammatically<br />
neuter noun that refers to a male or with a semantically-vacuous neuter noun yields<br />
mascul<strong>in</strong>e resolution. An example is shown below:<br />
(15) I Maria<br />
the.fem.sg<br />
agori/moro<br />
Maria.fem.sg<br />
boy.neut.sg/baby.neut.sg<br />
ke<br />
and<br />
<strong>in</strong>e<br />
are.pl<br />
to<br />
the.neut.sg<br />
xarumen-i<br />
happy.masc.pl<br />
8 We will <strong>in</strong>troduce the term semantic neuter noun to refer to the two grammatically neuter nouns that<br />
have a semantic gender, to agori ‘the boy’ and to koritsi ‘the girl’, and semantically-vacuous neuter noun<br />
to refer to the neuter nouns that do not have a semantic gender, such as to pedi ‘the child’, to moro ‘the<br />
baby’ and to atomo ‘the person’.<br />
tis<br />
her
4.2 The <strong>Modern</strong> <strong>Greek</strong> Data 75<br />
‘Mary and the boy/baby are happy’<br />
(constructed)<br />
Semantically-vacuous neuter nouns, such as to pedi ‘the child’ and to moro ‘the baby’,<br />
always display syntactic agreement <strong>in</strong> noncoord<strong>in</strong>ated constructions. In coord<strong>in</strong>ate phrases,<br />
these nouns can be referentially specified, referr<strong>in</strong>g to a specific person whose gender is<br />
known, or unspecified, denot<strong>in</strong>g a person whose gender is unknown. When these nouns are<br />
unspecified for a referent, the resolution is mascul<strong>in</strong>e whether they are coord<strong>in</strong>ated with a<br />
grammatically mascul<strong>in</strong>e or a grammatically fem<strong>in</strong><strong>in</strong>e noun. When they are referentially<br />
specified, however, the resolved gender depends on whether they cooccur with a male or<br />
female denot<strong>in</strong>g noun. For <strong>in</strong>stance, <strong>in</strong> example (16) when the neuter noun pedi ‘child’ has<br />
a male or a female referrent <strong>in</strong> the questionnaire all consultants (26 out of 26) choose a<br />
mascul<strong>in</strong>e gender. Thus, the grammatically mascul<strong>in</strong>e noun pateras ‘father’ is coord<strong>in</strong>ated<br />
with the grammatically neuter noun pedi ‘child’ and whether or not pedi ‘child’ has a male<br />
or female denotation the resolved gender is always mascul<strong>in</strong>e. In example (17), when the<br />
grammatically neuter noun to moro ‘baby’, which is coord<strong>in</strong>ated with a fem<strong>in</strong><strong>in</strong>e noun,<br />
refers to a male or an unspecified referent, all consultants chose a mascul<strong>in</strong>e gender (26 out<br />
of 26). Therefore, resolution is mascul<strong>in</strong>e. However, when the neuter noun to moro ‘baby’<br />
has a female referent, 21 (out of 26) consultants chose a fem<strong>in</strong><strong>in</strong>e gender, and therefore<br />
resolution is fem<strong>in</strong><strong>in</strong>e. In this last case, mascul<strong>in</strong>e also occurs as a resolved gender but only<br />
5 (out of 26) consultants chose the mascul<strong>in</strong>e form:<br />
(16) O pateras ke to<br />
the.masc.sg father.masc.sg and the.neut.sg<br />
vrethikan dolofonimen-i<br />
were-found.pl murdered.masc.pl<br />
‘The father and child were found murdered’<br />
(G.N.: To Vima)<br />
pedi<br />
child.neut.sg<br />
(17) I Maria ke to moro tis <strong>in</strong>e<br />
the Mary.fem.sg and the baby.neut.sg her are.pl<br />
kukl-i/-es.<br />
Tus/tis<br />
vlepis ke<br />
beautiful.masc/fem.pl. Them.masc/fem.pl you-see and<br />
tus/tis<br />
thavmazeis<br />
them.masc/fem.pl you-admire<br />
‘Mary and her baby are beautiful. You see them and you admire them’<br />
(constructed)<br />
The gender resolution patterns of MG animate s<strong>in</strong>gular coord<strong>in</strong>ate nouns are summarised<br />
<strong>in</strong> table (4.2) 9 :<br />
9 The presentation of the two conjuncts <strong>in</strong> the structure does not reflect the order of the conjuncts.<br />
The same conjuncts were tested <strong>in</strong> the reverse order and the results were the same. However, we do not<br />
<strong>in</strong>clude the reverse order to avoid repetition. Also, <strong>in</strong> order to show clearly the preferred resolved gender,<br />
as was chosen by the consultants, we have used the ‘greater than’(>) symbol while the pattern <strong>in</strong>side the<br />
parenthesis is a second pattern chosen by the participants and not an unexpected one.
4.2 The <strong>Modern</strong> <strong>Greek</strong> Data 76<br />
Table 4.2: <strong>Agreement</strong> with SG Animate Coord<strong>in</strong>ate <strong>Noun</strong>s<br />
S<strong>in</strong>gular Conjuncts<br />
NP NP AP NP NP AP<br />
msg msg mpl msg fsg mpl<br />
fsg fsg fpl msg nm/f /0sg mpl<br />
nsg nsg npl fsg nm/0sg mpl<br />
fsg nf sg fpl(>mpl)<br />
Animate plural coord<strong>in</strong>ate nouns have the same expected resolution patterns as animate<br />
s<strong>in</strong>gular nouns. The table below presents the responses of the 26 consultants:<br />
Table 4.3: Animate Plural Coord<strong>in</strong>ate <strong>Noun</strong>s<br />
Consultant Number: 26<br />
NP1 NP2 Resolution NP1 NP2 Resolution<br />
Mpl Fpl Npl Mpl Fpl Npl<br />
Mpl Mpl 26 - - Mpl Fpl 26 - -<br />
Fpl Fpl - 26 - Mpl Nm/f /0pl 26 - -<br />
Npl Npl - - 26 Fpl Nm/0pl 19 - 12<br />
Nm/0pl Fpl 19 7<br />
Fpl Nf pl 5 18 5<br />
Nf pl Fpl 5 18 5<br />
When conjuncts have the same gender then that gender is used <strong>in</strong> resolution such as <strong>in</strong><br />
examples (18), (19) and (20), when conjuncts have different gender then mascul<strong>in</strong>e is the<br />
expected default gender, such as <strong>in</strong> examples (21) and (22). If all conjuncts denote females<br />
then fem<strong>in</strong><strong>in</strong>e is the resolved gender, such as <strong>in</strong> example (23). In the examples below, all<br />
consultants (26 out of 26) chose as the resolved gender the same gender as the one found<br />
<strong>in</strong> the two conjuncts.<br />
(18) I papudes<br />
ke<br />
the.fem.pl grandfathers.fem.pl and<br />
<strong>in</strong>e agapimen-i<br />
are.pl love-each-other.fem.pl<br />
‘The grandfathers and fathers love each other’<br />
(constructed)<br />
i<br />
the.fem.pl<br />
(19) I jajades<br />
ke i<br />
the.fem.pl grandmothers.fem.pl and the.fem.pl<br />
engones<br />
<strong>in</strong>e agapimen-es<br />
grandaughters.fem.pl are.pl love-each-other.fem.pl<br />
‘The grandmothers and grandaughters love each other’<br />
(constructed)<br />
(20) Ta agoria<br />
the.neut.pl<br />
xarumen-a<br />
happy.neut.pl<br />
boys.neut.pl<br />
‘The boys and girls are happy’<br />
(constructed)<br />
ke<br />
and<br />
ta<br />
the.neut.pl<br />
paterades<br />
fathers.fem.pl<br />
koritsia<br />
girls.neut.pl<br />
<strong>in</strong>e<br />
are.pl<br />
In the follow<strong>in</strong>g examples, all consultants (26 out of 26) chose a mascul<strong>in</strong>e gender when<br />
a mascul<strong>in</strong>e noun cooccurs with a fem<strong>in</strong><strong>in</strong>e or a neuter noun:
4.2 The <strong>Modern</strong> <strong>Greek</strong> Data 77<br />
(21) I<br />
antres ke<br />
the.masc.pl men.masc.pl and<br />
itan etim-i<br />
were.pl ready.masc.pl<br />
‘The men and women were ready’<br />
(constructed)<br />
(22) I<br />
babades ke<br />
the.masc.pl fathers.masc.pl and<br />
<strong>in</strong>e poli demen-i<br />
are.pl very close.masc.pl<br />
‘The fathers and boys are very close’<br />
(constructed)<br />
i<br />
the.fem.pl<br />
ta<br />
the.neut.pl<br />
j<strong>in</strong>ekes<br />
women.fem.pl<br />
agoria<br />
boys.neut.pl<br />
In the follow<strong>in</strong>g example, the majority of the consultants (18 out of 26) chose a fem<strong>in</strong><strong>in</strong>e<br />
gender as the resolved value and only 5 (out of 26) consultants chose the mascul<strong>in</strong>e gender:<br />
(23) I mamades ke<br />
the.fem.pl mothers.fem.pl and<br />
<strong>in</strong>e agapimen-es/-i<br />
are.pl love-each-other.fem/masc.pl<br />
‘The mothers and girls love each other’<br />
(constructed)<br />
ta<br />
the.neut.pl<br />
koritsia<br />
girls.neut.pl<br />
As opposed to animate s<strong>in</strong>gular coord<strong>in</strong>ate nouns, which do not present any unexpected<br />
patterns, animate plural coord<strong>in</strong>ate nouns yield two unexpected cases. The first one is when<br />
the coord<strong>in</strong>ate phrase consists of a fem<strong>in</strong><strong>in</strong>e plural and a neuter plural noun, which denotes<br />
a mascul<strong>in</strong>e referent or an unspecified referent, as shown below:<br />
(24) I jajades<br />
ke ta<br />
the.fem.pl grandmothers.fem.pl and the.neut.pl<br />
engonia<br />
<strong>in</strong>e xarumen-i/-a<br />
grandchldren.neut.pl are.pl happy.masc/neut.pl<br />
‘The grandmothers and grandchildren are happy’<br />
(constructed)<br />
The expected resolved gender is mascul<strong>in</strong>e plural but a high number of consultants<br />
(12 out of 26) chose only the neuter gender or the neuter gender alongside the mascul<strong>in</strong>e<br />
one, as the value of the predicative adjective. It could be claimed that neuter gender is a<br />
case of closest conjunct agreement s<strong>in</strong>ce the predicative adjective seems to agree with<br />
the conjunct closest to it (see Corbett (1991), Sadler (1999), Moosally (1999), Sadler<br />
(2003), Villavicencio et al. (2005) for more details). In order to test the possibility of<br />
whether the gender chosen is a case of CCA, we <strong>in</strong>cluded <strong>in</strong> the questionnaires the same<br />
patterns <strong>in</strong> reverse order, as shown <strong>in</strong> example (25). When the conjuncts are reversed as<br />
<strong>in</strong> [nm/0pl & fpl], the results showed that overall 19 (out of 26) consultants chose the<br />
mascul<strong>in</strong>e gender and 7 (out of the 26) participants chose the neuter gender. These 7 (out<br />
of 26) consultants were the same people who chose neuter resolution when the conjucts<br />
displayed the order [fpl & nm/0pl]. Therefore, when the conjuncts have the order [fpl &<br />
nm/0pl] 5 consultants (out of 26) choose the neuter gender as a result of closest conjunct<br />
agreement s<strong>in</strong>ce they chose only mascul<strong>in</strong>e when the conjucts displayed the order [nm/0pl<br />
& fpl] whereas 7 participants chose consistently the neuter gender as a result of resolution<br />
irrespective of the order of the conjuncts. This additional pattern is not a case of CCA<br />
but it is an extra agreement gender value.
4.2 The <strong>Modern</strong> <strong>Greek</strong> Data 78<br />
(25) Ta pedia<br />
ke<br />
the.neut.pl children.neut.pl and<br />
<strong>in</strong>e xarumen-i/-a<br />
are.pl happy.masc/neut.pl<br />
‘The children and mothers are happy’<br />
(constructed)<br />
i<br />
the.fem.pl<br />
mamades<br />
mothers.fem.pl<br />
Neuter gender resolution also occurs when a grammatically fem<strong>in</strong><strong>in</strong>e plural noun is coord<strong>in</strong>ated<br />
with a grammatically neuter plural noun that denotes a female. The consultants<br />
were more <strong>in</strong>cl<strong>in</strong>ed to choose the fem<strong>in</strong><strong>in</strong>e gender (18 out of 26) as the ma<strong>in</strong> resolved gender<br />
while neuter occured less often (5 out of 26) with the same frequency as the mascul<strong>in</strong>e<br />
(5 out of 26). In example (26), the grammatically fem<strong>in</strong><strong>in</strong>e noun comb<strong>in</strong>es with a grammatically<br />
neuter noun koritsia ‘girls’, which denotes a group of females, yield<strong>in</strong>g ma<strong>in</strong>ly<br />
fem<strong>in</strong><strong>in</strong>e resolution (18 out of 26 consultants). Neuter resolution occurs less often (5 out<br />
of 26) or as a second choice <strong>in</strong> these structures and it is chosen by consultants who chose<br />
neuter <strong>in</strong> all similar cases (5 out of 26). Also, some (5 out of 26) consultants chose the<br />
mascul<strong>in</strong>e gender:<br />
(26) I jajades<br />
the.fem.pl grandmothers.fem.pl<br />
<strong>in</strong>e xarumen-es/-a/-i<br />
are.pl happy.fem/neut/masc.pl<br />
‘The grandmothers and girls are happy’<br />
(constructed)<br />
ke<br />
and<br />
ta<br />
the.neut.pl<br />
koritsia<br />
girls.neut.pl<br />
The same consultants who chose neuter when the conjuncts had the word order [fpl &<br />
npl] also chose neuter as an alternative resolved gender when the conjuncts appeared <strong>in</strong><br />
the reverse order. This shows that CCA does not determ<strong>in</strong>e the gender agreement of the<br />
predicative participle. In the follow<strong>in</strong>g example, conjuncts appear <strong>in</strong> the reverse order [npl<br />
& fpl] and the agree<strong>in</strong>g gender is either fem<strong>in</strong><strong>in</strong>e (18 out of 26 consultants) or neuter plural<br />
(5 out of 26 consultants) which cooccurs with mascul<strong>in</strong>e plural (5 out of 26 consultants),<br />
favour<strong>in</strong>g the fem<strong>in</strong><strong>in</strong>e gender:<br />
(27) Ta koritsia ke i<br />
the.neut.pl girls.neut.pl and the.fem.pl<br />
<strong>in</strong>e poli demen-es/-a/-i<br />
are.pl very close.fem/neut/masc.pl<br />
‘The girls and mothers are very close’<br />
(constructed)<br />
mamades<br />
mothers.fem.pl<br />
The observed gender resolution patterns of plural animate coord<strong>in</strong>ate nouns are presented<br />
<strong>in</strong> table (4.4) 10 below:<br />
Table 4.4: <strong>Agreement</strong> with PL Animate Coord<strong>in</strong>ate <strong>Noun</strong>s<br />
Plural Conjuncts<br />
NP1 NP2 AP NP1 NP2 AP<br />
mpl mpl mpl mpl fpl mpl<br />
fpl fpl fpl mpl nm/f /0pl mpl<br />
npl npl npl fpl nm/0pl mpl>npl<br />
fpl nf pl fpl>npl/(>mpl)<br />
10 The expected patterns are shown <strong>in</strong> normal fonts. Any unexpected patterns are presented <strong>in</strong> small bold<br />
italics <strong>in</strong> the tables that follow.
4.2 The <strong>Modern</strong> <strong>Greek</strong> Data 79<br />
To summarise, the expected patterns of animate nouns are as follows: animate s<strong>in</strong>gular<br />
and plural nouns with the same gender resolve to the gender of the conjuncts and<br />
animate mixed gender conjuncts resolve to mascul<strong>in</strong>e or else to fem<strong>in</strong><strong>in</strong>e if nouns denote<br />
females. Animate s<strong>in</strong>gular nouns do not show any unexpected patterns but animate plural<br />
nouns show two unexpected patterns. When the conjucts display the comb<strong>in</strong>ations [fpl &<br />
nm/0pl] and [fpl & nf pl], the unexpected gender value <strong>in</strong> both comb<strong>in</strong>ations is neuter<br />
plural. A summary of the expected and unexpected patterns of animate s<strong>in</strong>gular and plural<br />
nouns <strong>in</strong>clud<strong>in</strong>g the percentage of the responses of the consultants is presented <strong>in</strong> a s<strong>in</strong>gle<br />
table below:<br />
Table 4.5: <strong>Agreement</strong> with SG/PL Animate Coord<strong>in</strong>ate <strong>Noun</strong>s<br />
S<strong>in</strong>gular Conjuncts<br />
NP1 NP2 AP percentage NP1 NP2 AP percentage<br />
msg msg mpl 100% msg fsg mpl 100%<br />
fsg fsg fpl 100% msg nm/f /0sg mpl 100%<br />
nsg nsg npl 100% fsg nm/0sg mpl 100%<br />
fsg nf sg fpl>mpl 80%>20%<br />
Plural Conjuncts<br />
NP1 NP2 AP percentage NP1 NP2 AP percentage<br />
mpl mpl mpl 100% mpl fpl mpl 100%<br />
fpl fpl fpl 100% mpl nm/f /0pl mpl 100%<br />
npl npl npl 100% fpl nm/0pl mpl>npl 73% > 27%<br />
fpl nf pl fpl>npl>mpl 69% >19%> 19%<br />
4.2.2 Resolution <strong>in</strong> MG Inanimate <strong>Noun</strong>s<br />
Inanimate s<strong>in</strong>gular and plural coord<strong>in</strong>ate nouns, apart from the expected patterns, admit<br />
a number of unexpected cases too. The expected patterns of s<strong>in</strong>gular and plural <strong>in</strong>animate<br />
nouns are as follows: nouns with shared gender resolve to that shared gender whereas nouns<br />
with mixed gender resolve to neuter. Both concrete and abstract <strong>in</strong>animate coord<strong>in</strong>ate<br />
nouns show the same expected patterns. Differences, however, arise only <strong>in</strong> the unexpected<br />
patterns of the two groups of nouns. Our discussion <strong>in</strong>cludes examples with both concrete<br />
and abstract nouns, which were all <strong>in</strong>cluded <strong>in</strong> the questionnaire.<br />
The discussion beg<strong>in</strong>s with <strong>in</strong>animate s<strong>in</strong>gular nouns. The table below presents the<br />
responses of the 26 consultants:<br />
Table 4.6: Inanimate S<strong>in</strong>gular Coord<strong>in</strong>ate <strong>Noun</strong>s<br />
Consultant Number: 26<br />
NP1 NP2 Resolution NP1 NP2 Resolution<br />
Mpl Fpl Npl Mpl Fpl Npl<br />
Msg Msg 25 - 10 Msg Fsg(Concrete) 12 - 14<br />
Fsg Fsg - 24 11 Msg Fsg(Abstract) 6 6 16<br />
Nsg Nsg - - 26 Msg Nsg - - 26<br />
Fsg Nsg - - 26
4.2 The <strong>Modern</strong> <strong>Greek</strong> Data 80<br />
With <strong>in</strong>animate s<strong>in</strong>gular coord<strong>in</strong>ate nouns, if both nouns are mascul<strong>in</strong>e then mascul<strong>in</strong>e<br />
is the resolved gender, as <strong>in</strong> (28), if both nouns are fem<strong>in</strong><strong>in</strong>e then resolution is fem<strong>in</strong><strong>in</strong>e, as<br />
<strong>in</strong> (29), and if both nouns are neuter then resolution is neuter, as <strong>in</strong> (30). In particular, 25<br />
(out of the 26) consultants chose the mascul<strong>in</strong>e gender when both connjuncts are mascul<strong>in</strong>e,<br />
24 (out of 26) consultants chose the fem<strong>in</strong><strong>in</strong>e gender when both conjuncts are fem<strong>in</strong><strong>in</strong>e and<br />
26 (out of 26) consultants chose the neuter when both conjuncts are neuter. Some of the<br />
examples that were given to the consultants are shown below:<br />
(28) O iroismos ke o<br />
the.masc.sg heroism.masc.sg and the.masc.sg<br />
itan kiriarx-i kata to 1821<br />
were.pl prevalent.masc.pl dur<strong>in</strong>g the 1821<br />
‘Heroism and fight<strong>in</strong>g were prevalent dur<strong>in</strong>g 1821’<br />
(G.N.:To Vima)<br />
(29) I vivliothiki ke i<br />
the.fem.sg bookcase.fem.sg and the.fem.sg<br />
<strong>in</strong>e diaforetik-es<br />
are.pl different.fem.pl<br />
‘The bookcase and d<strong>in</strong><strong>in</strong>g table are different’<br />
(constructed)<br />
(30) To aftok<strong>in</strong>ito ke<br />
the.neut.sg<br />
akriva<br />
car.neut.sg and<br />
expensive.neut.pl<br />
‘The car and house are expensive’<br />
(constructed)<br />
to<br />
the.neut.sg<br />
agonas<br />
fight.masc.sg<br />
trapezaria<br />
d<strong>in</strong><strong>in</strong>g-table.fem.sg<br />
spiti<br />
house.neut.sg<br />
<strong>in</strong>e<br />
are.pl<br />
In conjuncts with mixed gender values, the expected resolution gender is neuter. In<br />
example (31), 16 (out of 26) consultants chose the neuter gender. In example (32), (14 out<br />
of 26) chose the neuter, and <strong>in</strong> examples (33) and (34) all consultants (26 out of 26) chose<br />
the neuter gender:<br />
(31) O agonas ke i<br />
the.masc.sg fight.masc.sg and the.fem.sg<br />
itan kiriarx-a to 1821<br />
were prevalent.neut.pl <strong>in</strong> 1821<br />
‘The fight and self-sacrifice were prevalent <strong>in</strong> 1821’<br />
(G.N.:To Vima)<br />
(32) O kanapes<br />
the.masc.sg<br />
aspr-a<br />
white.neut.pl<br />
sofa.masc.sg<br />
‘The sofa and chair are white’<br />
(constructed)<br />
ke<br />
and<br />
i<br />
the.fem.sg<br />
(33) O erotas ke to<br />
the.masc.sg love.masc.sg and the.neut.sg<br />
<strong>in</strong>e aparetit-a sti zoi mas<br />
are.pl necessary.neut.pl <strong>in</strong>-the life our<br />
‘Love and passion are necessary <strong>in</strong> our life’<br />
(G.M.:VIMAgaz<strong>in</strong>o)<br />
aftothisia<br />
self-sacrifice.fem.sg<br />
karekla<br />
chair.fem.sg<br />
pathos<br />
passion.neut.sg<br />
<strong>in</strong>e<br />
are.pl
4.2 The <strong>Modern</strong> <strong>Greek</strong> Data 81<br />
(34) I prospathia ke to<br />
the.fem.sg effort.fem.sg and the.neut.sg<br />
<strong>in</strong>e apotelesmatik-a<br />
are.pl effective.neut.pl<br />
‘Effort and obst<strong>in</strong>acy have good results’<br />
(constructed)<br />
pisma<br />
obst<strong>in</strong>acy.neut.sg<br />
Three unexpected patterns arise <strong>in</strong> <strong>in</strong>animate s<strong>in</strong>gular coord<strong>in</strong>ate nouns. The first two<br />
concern coord<strong>in</strong>ate constructions whose conjuncts have the same gender, either mascul<strong>in</strong>e<br />
or fem<strong>in</strong><strong>in</strong>e. In these cases, a significant number of consultants were <strong>in</strong>cl<strong>in</strong>ed to choose<br />
the unexpected neuter resolution <strong>in</strong> addition to the expected patterns. In particular, 10<br />
(out of 26) consultants chose the neuter as an alternative form when conjuncts were both<br />
mascul<strong>in</strong>e and 11 (out of the 26) consultants chose the neuter when conjuncts were both<br />
fem<strong>in</strong><strong>in</strong>e. It is <strong>in</strong>terest<strong>in</strong>g that almost the same consultants chose the neuter gender <strong>in</strong> cases<br />
where both nouns have the same gender. Only one consultant chose neuter as a second<br />
choice when both conjuncts were fem<strong>in</strong><strong>in</strong>e and not when both conjuncts were mascul<strong>in</strong>e.<br />
Thus, <strong>in</strong>animate s<strong>in</strong>gular mascul<strong>in</strong>e nouns can resolve to mascul<strong>in</strong>e or neuter plural, and<br />
<strong>in</strong>animate s<strong>in</strong>gular fem<strong>in</strong><strong>in</strong>e nouns can resolve to fem<strong>in</strong><strong>in</strong>e or neuter plural. Some examples<br />
are shown below:<br />
(35) O p<strong>in</strong>akas ke o<br />
kanapes<br />
the.masc.sg picture.masc.sg and the.masc.sg sofa.masc.sg<br />
<strong>in</strong>e megal-i/-a<br />
ja afto to domatio ke tha<br />
are.pl big.masc.pl/neut.pl for this the room and will<br />
tus/ta<br />
valo sto alo<br />
them.masc.pl/neut.pl I-put <strong>in</strong>-the other<br />
‘The picture and the sofa are big for this room and I will put them <strong>in</strong> the other<br />
one’<br />
(constructed)<br />
(36) I kuz<strong>in</strong>a<br />
the.fem.sg<br />
kathar-es/-a<br />
kitchen.fem.sg<br />
clean.fem/neut.pl<br />
‘The kitchen and toilet are clean’<br />
(constructed)<br />
(37) I eleftheria<br />
the.fem.sg<br />
taftosim-es/-a<br />
freedom.fem.sg<br />
equal.fem/neut.pl<br />
‘Freedom and faith are equal’<br />
(G.N.:To Vima)<br />
ke<br />
and<br />
ke<br />
and<br />
i<br />
the.fem.sg<br />
i<br />
the.fem.sg<br />
tualeta<br />
toilet.fem.sg<br />
pisti<br />
faith.fem.sg<br />
<strong>in</strong>e<br />
are.sg<br />
<strong>in</strong>e<br />
are.pl<br />
The next unexpected pattern occurs <strong>in</strong> s<strong>in</strong>gular mixed gender concrete conjuncts comb<strong>in</strong><strong>in</strong>g<br />
a mascul<strong>in</strong>e and a fem<strong>in</strong><strong>in</strong>e noun. The gender output <strong>in</strong> these constructions was<br />
not only neuter but also mascul<strong>in</strong>e. In fact, 14 (out of 26) consultants chose neuter as an<br />
agree<strong>in</strong>g gender whereas 12 (out of 26) consultants chose mascul<strong>in</strong>e as an agree<strong>in</strong>g gender <strong>in</strong><br />
<strong>in</strong>animate s<strong>in</strong>gular coord<strong>in</strong>ate nouns. The follow<strong>in</strong>g examples illustrate two characteristic<br />
cases:<br />
(38) O kanapes ke<br />
the.masc.sg sofa.masc.sg and<br />
<strong>in</strong>e anet-a/-i<br />
are.pl comfortable.neut/masc.pl<br />
i<br />
the.fem.sg<br />
polithrona<br />
armchair.fem.sg
4.2 The <strong>Modern</strong> <strong>Greek</strong> Data 82<br />
‘The sofa and the armchair are comfortable’<br />
(constructed)<br />
(39) O dromos ke<br />
the.masc.sg street.masc.sg and<br />
<strong>in</strong>e kathar-a/-i<br />
are.pl clean.neut/masc.pl<br />
‘The street and the square are clean’<br />
(constructed)<br />
i<br />
the.fem.sg<br />
platia<br />
square.fem.sg<br />
Badecker (2008), who carried out research on MG <strong>in</strong>animate coord<strong>in</strong>ate nouns, argues<br />
that the number of conjuncts is a determ<strong>in</strong><strong>in</strong>g factor <strong>in</strong> the resolution result of <strong>in</strong>animate<br />
coord<strong>in</strong>ate phrases with a mascul<strong>in</strong>e and a fem<strong>in</strong><strong>in</strong>e noun. He argues that coord<strong>in</strong>ate<br />
<strong>in</strong>animate s<strong>in</strong>gular nouns with a mascul<strong>in</strong>e and a fem<strong>in</strong><strong>in</strong>e conjunct resolve to neuter but<br />
similar constructions with plural conjuncts resolve to mascul<strong>in</strong>e. Our field work certa<strong>in</strong>ly<br />
contradicts Badecker’s (2008) claim that s<strong>in</strong>gular conjuncts display only neuter resolution.<br />
The fact that almost the same number of consultants chose the mascul<strong>in</strong>e gender (12 of<br />
26) and the neuter (14 of 26) as the resolved forms is a result that cannot be ignored.<br />
An additional unexpected pattern occurs only <strong>in</strong> <strong>in</strong>animate s<strong>in</strong>gular abstract nouns<br />
when the coord<strong>in</strong>ate phrase <strong>in</strong>cludes a mascul<strong>in</strong>e and a fem<strong>in</strong><strong>in</strong>e noun. These coord<strong>in</strong>ate<br />
phrases also resolve to fem<strong>in</strong><strong>in</strong>e plural, apart from the neuter and mascul<strong>in</strong>e gender. In<br />
particular, 6 (out of 26) consultants chose the fem<strong>in</strong><strong>in</strong>e, 16 (out of 26) chose the neuter and<br />
6 (out of 26) chose the mascul<strong>in</strong>e. Therefore, although neuter is the expected resolution<br />
gender, a number of consultants realised fem<strong>in</strong><strong>in</strong>e and mascul<strong>in</strong>e on the agree<strong>in</strong>g adjective.<br />
The first th<strong>in</strong>g we need to check is whether the fem<strong>in</strong><strong>in</strong>e gender is a case of CCA. The<br />
results show that the same consultants (6 out of 26) who chose fem<strong>in</strong><strong>in</strong>e as the resolved<br />
gender <strong>in</strong> [msg & fsg] constructions, also chose fem<strong>in</strong><strong>in</strong>e when conjuncts have the reverse<br />
order [fsg & msg]. Thus, abstract s<strong>in</strong>gular mascul<strong>in</strong>e and fem<strong>in</strong><strong>in</strong>e coord<strong>in</strong>ate nouns show<br />
greater variation with regard to the agree<strong>in</strong>g gender. The examples that were used <strong>in</strong> the<br />
questionnaire are shown below:<br />
(40) O thanatos ke i<br />
the.masc.sg death.masc.sg and the.fem.sg<br />
<strong>in</strong>e simantik-a/-i/-es<br />
to 1821<br />
are.pl important.neut/masc/fem.pl <strong>in</strong> 1821<br />
‘Death and freedom are important <strong>in</strong> 1821’<br />
(G.N.:O Ependitis)<br />
(41) I eleftheria ke o<br />
the.fem.sg freedom.fem.sg and the.masc.sg<br />
itan simantik-a/-i/-es<br />
to 1821<br />
were.pl important.neut/masc/fem.pl <strong>in</strong> 1821<br />
‘Freedom and death were important <strong>in</strong> 1821’<br />
(constructed)<br />
eleftheria<br />
freedom.fem.sg<br />
thanatos<br />
death.masc.sg<br />
The gender resolution patterns of s<strong>in</strong>gular <strong>in</strong>animate nouns are presented <strong>in</strong> the follow<strong>in</strong>g<br />
table 11 :<br />
11 Table (4.7) presents <strong>in</strong> a s<strong>in</strong>gle column the concrete and abstract nouns with the same gender s<strong>in</strong>ce the<br />
results are the same <strong>in</strong> both groups of nouns. It presents separately the concrete and abstract nouns with<br />
mixed gender s<strong>in</strong>ce the results show a difference <strong>in</strong> the two groups of nouns when conjuncts are mixed <strong>in</strong><br />
gender.
4.2 The <strong>Modern</strong> <strong>Greek</strong> Data 83<br />
Table 4.7: <strong>Agreement</strong> with Inanimate SG Coord<strong>in</strong>ate <strong>Noun</strong>s<br />
S<strong>in</strong>gular Conjuncts<br />
Concrete or Abstract <strong>Noun</strong>s Concrete <strong>Noun</strong>s Abstract <strong>Noun</strong>s<br />
NP1 NP2 AP NP1 NP2 AP NP1 NP2 AP<br />
msg msg mpl>npl msg fsg npl>mpl msg fsg npl>mpl/fpl<br />
fsg fsg fpl>npl msg nsg npl msg nsg npl<br />
nsg nsg npl fsg nsg npl fsg nsg npl<br />
The expected patterns <strong>in</strong> <strong>in</strong>animate plural coord<strong>in</strong>ate nouns are the same as <strong>in</strong> <strong>in</strong>animate<br />
s<strong>in</strong>gular coord<strong>in</strong>ate nouns. The follow<strong>in</strong>g table shows the noun comb<strong>in</strong>ations and the<br />
responses of the consultants:<br />
Table 4.8: Inanimate Plural Coord<strong>in</strong>ate <strong>Noun</strong>s<br />
Consultant Number: 26<br />
NP1 NP2 Resolution NP1 NP2 Resolution<br />
Mpl Fpl Npl Mpl Fpl Npl<br />
Mpl Mpl 26 - - Mpl Fpl(Concrete) 15 - 13<br />
Fpl Fpl - 26 - Fpl Mpl(Concrete) 17 11<br />
Npl Npl - - 26 Mpl Fpl(Abstract) 15 6 13<br />
Fpl Mpl(Abstract) 16 4 11<br />
Mpl Npl - - 26<br />
Fpl Npl - - 26<br />
Coord<strong>in</strong>ate nouns with the same gender resolve to that same gender, and coord<strong>in</strong>ate<br />
nouns with mixed gender resolve to neuter. In the follow<strong>in</strong>g examples, concrete and abstract<br />
nouns with the same gender are shown where all consultants (26 out of 26) chose the same<br />
gender as the gender found <strong>in</strong> the two conjuncts:<br />
(42) I<br />
p<strong>in</strong>akes ke<br />
the.masc.pl pictures.masc.pl and<br />
<strong>in</strong>e ateriast-i<br />
are.pl unsuitable.masc.pl<br />
‘The pictures and the sofas are unsuitable’<br />
(constructed)<br />
i<br />
the.masc.pl<br />
kanapedes<br />
sofas.masc.pl<br />
(43) I<br />
kavgathes ke i<br />
the.masc.pl arguments.masc.pl and the.masc.pl<br />
xorismi<br />
<strong>in</strong>e simantiki ja t<strong>in</strong> psixologia<br />
break-ups.masc.pl are.pl important.masc.pl for the psychology<br />
ton pedion<br />
of-the children<br />
‘The arguments and break-ups are important for children’s psychology’<br />
(G.M.:To Pedi)<br />
(44) I filosofies<br />
ke i thriskies<br />
the.fem.pl philosophies.fem.pl and the.fem.pl religions.fem.pl<br />
itan kathoristikes ja tus laus<br />
were.pl determ<strong>in</strong><strong>in</strong>g.fem.pl for the people<br />
‘The philosophies and religions were determ<strong>in</strong><strong>in</strong>g for the people’<br />
(www.<strong>in</strong>.gr)<br />
(45) Ta<br />
the.neut.pl<br />
spitia<br />
houses.neut.pl<br />
ke<br />
and<br />
ta<br />
the.neut.pl<br />
aftok<strong>in</strong>ita<br />
cars.neut.pl
4.2 The <strong>Modern</strong> <strong>Greek</strong> Data 84<br />
<strong>in</strong>e<br />
are.pl<br />
pio<br />
more<br />
akriva<br />
expensive.neut.pl<br />
apo<br />
from<br />
edo<br />
now<br />
‘The houses and cars are more expensive from now on’<br />
(constructed)<br />
ke<br />
and<br />
pera<br />
on<br />
The follow<strong>in</strong>g examples illustrate <strong>in</strong>animate plural coord<strong>in</strong>ate nouns with different gender<br />
comb<strong>in</strong>ations that resolve to neuter. In the first example, which comb<strong>in</strong>es a mascul<strong>in</strong>e<br />
and a fem<strong>in</strong><strong>in</strong>e noun, 13 (out of 26) consultants choose the neuter gender (15 out of 26<br />
choose the unexpected mascul<strong>in</strong>e gender). In the next example, which comb<strong>in</strong>es a mascul<strong>in</strong>e<br />
and a neuter noun, all consultants (26 out of 26) choose the neuter gender and <strong>in</strong><br />
the last two examples, which comb<strong>in</strong>e a fem<strong>in</strong><strong>in</strong>e and a neuter, all consultants (26 out<br />
of 26) choose the neuter gender. The expected resolved gender is the neuter plural and<br />
consultants choose that gender <strong>in</strong> all cases:<br />
(46) I<br />
dromi ke i<br />
the.masc.pl roads.masc.pl and the.fem.pl<br />
<strong>in</strong>e jemat-a kosmo<br />
are.pl full-of.neut.pl people<br />
‘The roads and the squares are full of people’<br />
(Holton et al., 1997)<br />
(47) I<br />
dromi ke ta<br />
the.masc.pl roads.masc.pl and the.neut.pl<br />
<strong>in</strong>e jemat-a kosmo<br />
are.pl full-of.neut.pl people<br />
‘The roads and the alleys are full of people’<br />
(constructed)<br />
(48) Ta estiatoria ke i<br />
the.neut.pl restaurants.neut.pl and the.fem.pl<br />
itan jemat-a kosmo<br />
were.pl full-of.neut.pl people<br />
‘The restaurants and squares were full of people’<br />
(constructed)<br />
(49) I diamartiries ke ta<br />
the.fem.pl protests.fem.pl and the.neut.pl<br />
<strong>in</strong>e axrist-a s’afti t<strong>in</strong> periptosi<br />
are.pl useless.neut.pl <strong>in</strong> this case<br />
‘The protests and compla<strong>in</strong>ts are useless <strong>in</strong> this case’<br />
(G.M.:To Vima)<br />
platies<br />
squares.fem.pl<br />
sokakia<br />
alleys.neut.pl<br />
platies<br />
squares.fem.pl<br />
parapona<br />
compla<strong>in</strong>ts.neut.pl<br />
In <strong>in</strong>animate plural nouns, only two unexpected genders are confirmed by the questionnaires.<br />
These occur with mixed gender coord<strong>in</strong>ate phrases with a mascul<strong>in</strong>e and a fem<strong>in</strong><strong>in</strong>e<br />
plural noun, which resolve to the unexpected mascul<strong>in</strong>e or fem<strong>in</strong><strong>in</strong>e (less often) and not<br />
only to the expected neuter. The consultants chose both gender values, but the mascul<strong>in</strong>e<br />
gender seemed to occur more often than the neuter. In fact, <strong>in</strong> <strong>in</strong>animate concrete plural<br />
nouns, 15 (out of 26) consultants chose the unexpected mascul<strong>in</strong>e gender and 13 (out of<br />
26) consultants chose the expected neuter gender. Also, the same comb<strong>in</strong>ation of conjuncts<br />
but with abstract nouns resulted <strong>in</strong> the additional unexpected fem<strong>in</strong><strong>in</strong>e gender (6 out of 26<br />
consultants), alongside the unexpected mascul<strong>in</strong>e (15 out of 26 consultants) and expected<br />
neuter (13 out of 26 consultants) genders.
4.2 The <strong>Modern</strong> <strong>Greek</strong> Data 85<br />
(50) I<br />
dromi ke i<br />
the.masc.pl roads.masc.pl and the.fem.pl<br />
<strong>in</strong>e jemat-i/-a<br />
kosmo<br />
are.pl full-of.masc/neut.pl people<br />
‘The roads and the squares are full of people’<br />
(Holton et al., 1997)<br />
platies<br />
squares.fem.pl<br />
(51) I<br />
tifones<br />
ke i plimires<br />
the.masc.pl hurricanes.masc.pl and the.fem.pl floods.fem.pl<br />
itan katastrofik-i/-a/-es<br />
ja t<strong>in</strong> Tailandi<br />
were destructive.masc/neut/fem.pl for the Thailand<br />
‘The hurricanes and floods were destructive for Thailand’<br />
(www.<strong>in</strong>.gr)<br />
In order to test whether these genders are the outcome of resolution, we gave the<br />
same comb<strong>in</strong>ation of nouns with the reverse order [fem & masc] to the consultants. The<br />
examples that were given to the consultants are shown below:<br />
(52) I platies ke i<br />
the.fem.pl squares.fem.pl and the.masc.pl<br />
<strong>in</strong>e jemat-i/-a<br />
kosmo<br />
are.pl full-of.masc/neut.pl people<br />
‘The squares and roads are full of people’<br />
(Holton et al., 1997)<br />
dromi<br />
roads.masc.pl<br />
(53) I plimires ke i<br />
tifones<br />
the.fem.pl floods.fem.pl and the.masc.pl hurricanes.masc.pl<br />
itan katastrofik-i/-a/-es<br />
ja t<strong>in</strong> Tailandi<br />
were destructive.masc/neut/fem.pl for the Thailand<br />
‘The floods and hurricanes were destructive for Thailand’<br />
(www.<strong>in</strong>.gr)<br />
In the first example where conjuncts are concrete, 17 (out of 26) consultants chose<br />
the mascul<strong>in</strong>e gender and 11 (out of 26) consultants chose the neuter gender whereas<br />
when conjuncts are abstract 16 (out of 26) consultants chose the mascul<strong>in</strong>e gender, 11<br />
(out of 26) consultants chose the neuter gender and 4 (out of the 26) consultants showed<br />
preference to the fem<strong>in</strong><strong>in</strong>e gender. Therefore, <strong>in</strong> [fem & masc] constructions the mascul<strong>in</strong>e<br />
gender appears with higher frequency due to closest conjunct agreement. Thus, <strong>in</strong> [fem<br />
& masc] constructions the occurrence of the mascul<strong>in</strong>e gender with such high frequency<br />
is attributed both to resolution and to CCA and the occurence of the fem<strong>in</strong><strong>in</strong>e gender is<br />
attributed to resolution. Also, when abstract conjuncts appear with the order [masc &<br />
fem] the occurence of the fem<strong>in</strong><strong>in</strong>e gender <strong>in</strong> higher numbers is due to closest conjunct<br />
agreement s<strong>in</strong>ce fem<strong>in</strong><strong>in</strong>e was chosen by fewer consultants (4 out of 26) when conjuncts<br />
appear with the reverse order [fem & masc]. The table below shows both the expected<br />
and unexpected resolved genders <strong>in</strong> this group of nouns:
4.2 The <strong>Modern</strong> <strong>Greek</strong> Data 86<br />
Table 4.9: <strong>Agreement</strong> with Inanimate PL Coord<strong>in</strong>ate <strong>Noun</strong>s<br />
Plural Conjuncts<br />
Concrete or Abstract <strong>Noun</strong>s Concrete <strong>Noun</strong>s Abstract <strong>Noun</strong>s<br />
NP1 NP2 AP NP1 NP2 AP NP1 NP2 AP<br />
mpl mpl mpl mpl fpl npl>mpl mpl fpl npl>mpl/fpl<br />
fpl fpl fpl mpl npl npl mpl npl npl<br />
npl npl npl fpl npl npl fpl npl npl<br />
The data showed that plural coord<strong>in</strong>ate phrases with a mascul<strong>in</strong>e and a fem<strong>in</strong><strong>in</strong>e noun<br />
resolve to the expected neuter but more often to the unexpected mascul<strong>in</strong>e and less often<br />
to the unexpected fem<strong>in</strong><strong>in</strong>e. Thus, neuter and mascul<strong>in</strong>e resolution coexist <strong>in</strong> <strong>in</strong>animate<br />
plural nouns with a mascul<strong>in</strong>e and a fem<strong>in</strong><strong>in</strong>e gender. This contradicts Badecker (2008),<br />
who argues that a mascul<strong>in</strong>e and a fem<strong>in</strong><strong>in</strong>e plural noun resolve only to mascul<strong>in</strong>e.<br />
To summarise, <strong>in</strong>animate coord<strong>in</strong>ate nouns with s<strong>in</strong>gular or plural conjuncts show the<br />
same expected patterns. Inanimate coord<strong>in</strong>ate nouns with the same gender resolve to<br />
the gender of the conjuncts and <strong>in</strong>animate nouns with mixed gender resolve to neuter.<br />
With regard to the unexpected cases, <strong>in</strong>animate concrete coord<strong>in</strong>ate nouns differ from<br />
<strong>in</strong>animate abstract nouns s<strong>in</strong>ce abstract coord<strong>in</strong>ate nouns with a mascul<strong>in</strong>e and a fem<strong>in</strong><strong>in</strong>e<br />
noun resolve not only to the expected neuter but also to the unexpected mascul<strong>in</strong>e and<br />
fem<strong>in</strong><strong>in</strong>e genders. Concrete coord<strong>in</strong>ate nouns with the same comb<strong>in</strong>ation of nouns resolve<br />
to the expected neuter and to the unexpected mascul<strong>in</strong>e. Also, differences are found <strong>in</strong> the<br />
unexpected patterns which are related to the number of the conjuncts. Inanimate s<strong>in</strong>gular<br />
nouns yield unexpected neuter resolution when the coord<strong>in</strong>ate nouns are the same <strong>in</strong> gender,<br />
either mascul<strong>in</strong>e or fem<strong>in</strong><strong>in</strong>e. Inanimate plural nouns do not allow that. F<strong>in</strong>ally, a difference<br />
is found <strong>in</strong> the frequency of an agree<strong>in</strong>g gender. Inanimate s<strong>in</strong>gular coord<strong>in</strong>ate nouns with a<br />
mascul<strong>in</strong>e and a fem<strong>in</strong><strong>in</strong>e noun yield expected neuter resolution and unexpected mascul<strong>in</strong>e<br />
resolution (and if the conjuncts are abstract unexpected fem<strong>in</strong><strong>in</strong>e resolution), where neuter<br />
occurs with higher frequency. Inanimate plural nouns yield expected neuter resolution and<br />
unexpected mascull<strong>in</strong>e resolution (and if the conjuncts are abstract unexpected fem<strong>in</strong><strong>in</strong>e<br />
resolution) where mascul<strong>in</strong>e occurs with higher frequency than neuter. A summary of the<br />
s<strong>in</strong>gular and plural <strong>in</strong>animate nouns, their resolution values and the percentage of responses<br />
are presented below:<br />
Table 4.10: <strong>Agreement</strong> with Inanimate Coord<strong>in</strong>ate <strong>Noun</strong>s<br />
S<strong>in</strong>gular Conjuncts<br />
Concrete/Abstract <strong>Noun</strong>s Concrete <strong>Noun</strong>s Abstract <strong>Noun</strong>s<br />
NP1 NP2 AP percent NP1 NP2 AP percent NP1 NP2 AP percent<br />
msg msg mpl>npl 96%>38% msg fsg npl>mpl 53%>46% msg fsg npl>mpl/fpl 61%>23%/23%<br />
fsg fsg fpl>npl 92%>42% msg nsg npl 100% msg nsg npl 100%<br />
nsg nsg npl 100% fsg nsg npl 100% fsg nsg npl 100%<br />
Plural Conjuncts<br />
Concrete/Abstract <strong>Noun</strong>s Concrete <strong>Noun</strong>s Abstract <strong>Noun</strong>s<br />
NP1 NP2 AP percent NP1 NP2 AP percent NP1 NP2 AP percent<br />
mpl mpl mpl 100% mpl fpl mpl>npl 63%>42% mpl fpl mpl>npl>fpl 61%>42%>15%<br />
fpl fpl fpl 100% mpl npl npl 100% mpl npl npl 100%<br />
npl npl npl 100% fpl npl npl 100% fpl npl npl 100%<br />
Thus, we argue that the number of the conjuncts determ<strong>in</strong>es the resolution pattern,<br />
which agrees with Badecker (2008), but our data show that conjunct number only determ<strong>in</strong>es<br />
the preference towards a gender form over the other and does not necessarily<br />
impose neuter resolution <strong>in</strong> s<strong>in</strong>gular conjucts and mascul<strong>in</strong>e resolution <strong>in</strong> plural conjuncts.
4.2 The <strong>Modern</strong> <strong>Greek</strong> Data 87<br />
Badecker’s analysis, however, is based on a simplification which associates the number of<br />
the conjuncts with an exclusive resolved gender value.<br />
4.2.3 Mix<strong>in</strong>g Animate and Inanimate <strong>Noun</strong>s <strong>in</strong> Coord<strong>in</strong>ate Constructions<br />
In this section, we discuss the possibility of comb<strong>in</strong><strong>in</strong>g animate and <strong>in</strong>animate nouns <strong>in</strong><br />
coord<strong>in</strong>ate phrases <strong>in</strong> MG. We argue that MG allows this possibility if the nouns occur <strong>in</strong><br />
a comitative structure.<br />
Corbett and Mtenje (1987) and Corbett (1991) discussed the possibility of mix<strong>in</strong>g<br />
animate with <strong>in</strong>animate nouns and rational with nonrational nouns <strong>in</strong> subject position 12 <strong>in</strong><br />
a number of languages and argue that some languages allow animate-<strong>in</strong>animate or rationalnonrational<br />
coord<strong>in</strong>ation. English, for example, allows the coord<strong>in</strong>ation of an animate with<br />
an <strong>in</strong>animate noun and the coord<strong>in</strong>ation is acceptable both syntactically and semantically:<br />
(54) The child and his bike are stand<strong>in</strong>g there. Can you see them? (constructed)<br />
Other languages, like Chichewa, disallow the comb<strong>in</strong>ation of animate and <strong>in</strong>animate<br />
nouns. However, if this comb<strong>in</strong>ation is forced then agreement is usually with the <strong>in</strong>animate<br />
noun. This is shown <strong>in</strong> the example below by the agreement marker a or less often zi, used<br />
for non-humans:<br />
(55) Ana ndi malalanje<br />
children.1.pl and oranges.3.pl<br />
‘Children and oranges are miss<strong>in</strong>g’<br />
(Corbett and Mtenje, 1987, 33)<br />
a-ku-sowa<br />
ag-pres-miss<strong>in</strong>g<br />
Also, <strong>in</strong> languages like the Dravidian ones, <strong>in</strong> which conjuncts denote rationals and<br />
nonrationals, Corbett (1991) argues that speakers do not accept a comb<strong>in</strong>ation of a rational<br />
with a nonrational noun. Consider the follow<strong>in</strong>g example <strong>in</strong> Tamil:<br />
(56) *raaman-um naay-um<br />
Raman-and dog-and<br />
‘Raman and the dog came’<br />
(Corbett, 1991, 270)<br />
va-nt-aaÒka<br />
come.past-3rd.pl.rational<br />
When rationals and non-rationals are comb<strong>in</strong>ed an alternative form must be used which<br />
is the comitative structure (Corbett, 1991).<br />
MG resembles languages like Chichewa when it comes to the comb<strong>in</strong>ation of animate<br />
and <strong>in</strong>animate nouns s<strong>in</strong>ce it disfavours the coord<strong>in</strong>ation of animate and <strong>in</strong>animate nouns<br />
as shown below:<br />
(57) ??O kleftis ke to diamanti<br />
the thief.masc.sg and the diamond.neut.sg<br />
afant-i???/afant-os??/afant-a?<br />
disappeared.masc.pl/masc.sg/neut.pl<br />
‘The thief and the diamond are gone’<br />
(constructed)<br />
(58) ??O kleftis ke ta kosmimata<br />
the thief.masc.sg and the jewellery.neut.pl<br />
afant-i???/afant-os??/afant-a?<br />
disappeared.masc.pl/masc.sg/neut.pl<br />
12 Both Corbett and Mtenje (1987) and Corbett (1991) discuss examples <strong>in</strong> subject position. S<strong>in</strong>ce we are<br />
<strong>in</strong>terested <strong>in</strong> the gender feature agreement and this is shown when the coord<strong>in</strong>ate phrase occurs <strong>in</strong> subject<br />
position, we also focus on similar examples <strong>in</strong> MG.<br />
<strong>in</strong>e<br />
are<br />
<strong>in</strong>e<br />
are
4.3 Resolution Pr<strong>in</strong>ciples <strong>in</strong> MG 88<br />
‘The thief and the jewellery are gone’<br />
(constructed)<br />
The sentences above are strongly disprefered s<strong>in</strong>ce they comb<strong>in</strong>e an animate and an<br />
<strong>in</strong>animate noun. That structure is rarely found (<strong>in</strong> spoken language only) and if it is<br />
admitted, the least marked forms would be the verb to be <strong>in</strong> plural number and the<br />
predicative adjective/participle to show neuter plural agreement, irrespective of the number<br />
or the gender of the two nouns.<br />
Instead, when MG comb<strong>in</strong>es an animate and an <strong>in</strong>animate noun, it shows strong preference<br />
to the comitative structure where the predicative adjective/participle shows agreement<br />
with the head noun and not with the second noun phrase. This is illustrated below:<br />
(59) O kleftis me ta<br />
the thief.masc.sg with the<br />
afantos<br />
disappeared.masc.sg<br />
‘The thief with the jewellery is gone’<br />
(constructed)<br />
kosmimata<br />
jewellery.neut.pl<br />
The participle afantos ‘disappeared’ shows agreement with the head noun phrase, the<br />
animate noun kleftis ‘thief’ and not with the <strong>in</strong>animate noun kosmimata ‘jewellery’. Thus,<br />
coord<strong>in</strong>ations of animate and <strong>in</strong>animate nouns are strongly dispreferred unless they occur<br />
under the comitative structure.<br />
4.3 Resolution Pr<strong>in</strong>ciples <strong>in</strong> MG<br />
The presentation of the <strong>Modern</strong> <strong>Greek</strong> data <strong>in</strong>troduces a dist<strong>in</strong>ction between (morpho)syntactic<br />
and semantic resolution. In this section, we will argue that MG animate coord<strong>in</strong>ate<br />
nouns follow the semantic resolution pr<strong>in</strong>ciple, and MG <strong>in</strong>animate nouns follow the<br />
syntactic resolution pr<strong>in</strong>ciple. Also, we will argue that the unexpected patterns that are<br />
recorded <strong>in</strong> both groups of coord<strong>in</strong>ate constructions belong to a different pr<strong>in</strong>ciple, the<br />
referential one.<br />
4.3.1 Semantic Resolution <strong>in</strong> Animate <strong>Noun</strong>s<br />
MG animate coord<strong>in</strong>ate nouns are complicated <strong>in</strong> terms of the resolution strategies they<br />
employ. We argue that the ma<strong>in</strong> pr<strong>in</strong>ciple <strong>in</strong> the expected patterns of MG animate coord<strong>in</strong>ate<br />
nouns is the semantic pr<strong>in</strong>ciple of resolution, which is the characteristic pr<strong>in</strong>ciple<br />
found <strong>in</strong> most languages. However, there is also evidence that nouns with shared gender<br />
follow feature match<strong>in</strong>g or syntactic agreement irrespective of the referents of the nouns.<br />
F<strong>in</strong>ally, we will show that the unexpected patterns are the result of a third pr<strong>in</strong>ciple, the<br />
referential. The table below is repeated from above for reasons of clarity:<br />
<strong>in</strong>e<br />
is
4.3 Resolution Pr<strong>in</strong>ciples <strong>in</strong> MG 89<br />
Table 4.11: <strong>Agreement</strong> of SG/PL Animate Coord<strong>in</strong>ate <strong>Noun</strong>s<br />
S<strong>in</strong>gular Conjuncts<br />
NP1 NP2 AP NP1 NP2 AP<br />
msg msg mpl msg fsg mpl<br />
fsg fsg fpl msg nm/f /0sg mpl<br />
nsg nsg npl fsg nm/0sg mpl<br />
fsg nf sg fpl(>mpl)<br />
Plural Conjuncts<br />
NP1 NP2 AP NP1 NP2 AP<br />
mpl mpl mpl mpl fpl mpl<br />
fpl fpl fpl mpl nm/f /0pl mpl<br />
npl npl npl fpl nm/0pl mpl>npl<br />
fpl nf pl fpl>npl/(>mpl)<br />
Corbett (1991) (follow<strong>in</strong>g Greenberg (1966) and Schane (1970)), expla<strong>in</strong>s the semantic motivation<br />
beh<strong>in</strong>d resolution <strong>in</strong> animate coord<strong>in</strong>ate nouns by associat<strong>in</strong>g the resolved gender<br />
of a coord<strong>in</strong>ate phrase with the gender of larger groups of both sexes. Corbett (1991)(cit<strong>in</strong>g<br />
Schane (1970)), claims that French sex-differentiated plural common nouns, such as<br />
les America<strong>in</strong>s(masc.pl), denote a group of male or mixed sex Americans whereas the<br />
correspond<strong>in</strong>g fem<strong>in</strong><strong>in</strong>e one, les America<strong>in</strong>es(fem.pl), denotes only females. This type of<br />
agreement dist<strong>in</strong>ction is also found <strong>in</strong> French coord<strong>in</strong>ate phrases, which resolve to mascul<strong>in</strong>e<br />
whenever the coord<strong>in</strong>ate phrase consists of mixed gender conjuncts, and to fem<strong>in</strong><strong>in</strong>e if<br />
all conjuncts are female.<br />
Sex-differentiated plural common nouns <strong>in</strong> MG show the same k<strong>in</strong>d of dist<strong>in</strong>ction.<br />
When we refer to a group of males or a group of mixed sex, the mascul<strong>in</strong>e gender is used<br />
but when we refer to a group of females, the fem<strong>in</strong><strong>in</strong>e gender is used. For example, the<br />
noun xen-i/-es.masc/fem.pl ‘foreigners’ has a mascul<strong>in</strong>e and a fem<strong>in</strong><strong>in</strong>e gender. If we<br />
want to refer to a group of <strong>in</strong>dividuals that consists of both males and females or only<br />
males the mascul<strong>in</strong>e gender xeni.masc.pl must be used. When the same noun is used <strong>in</strong><br />
its fem<strong>in</strong><strong>in</strong>e form as <strong>in</strong> xenes.fem.pl ‘foreigners’, it denotes only a group of females 13 . This<br />
is illustrated below:<br />
(60) I Kriti exi polus ksenus<br />
the Crete has many foreigners.masc.pl<br />
‘Crete has many foreigners’ (male/mixed group of <strong>in</strong>dividuals)<br />
(constructed)<br />
(61) I Kriti exi poles ksenes<br />
the Crete has many foreigners.fem.pl<br />
‘Crete has many foreigners’ (female group of <strong>in</strong>dividuals)<br />
(constructed)<br />
Therefore, <strong>in</strong> all mixed gender comb<strong>in</strong>ations of MG animate coord<strong>in</strong>ate nouns the<br />
mascul<strong>in</strong>e agreement form is used. Thus, MG animate coord<strong>in</strong>ate nouns resolve follow<strong>in</strong>g<br />
sex-differentiated plural common nouns which is the first evidence that resolution follows<br />
the semantic pr<strong>in</strong>ciple.<br />
Wechsler and Zlatić (2003) present futher evidence that supports the existence of semantic<br />
agreement <strong>in</strong> animate coord<strong>in</strong>ate nouns of most languages. The follow<strong>in</strong>g pr<strong>in</strong>ciple,<br />
repeated from above, identifies cases where semantic agreement occurs <strong>in</strong> a language:<br />
13 Converted nouns like xen-i/-es.masc/fem.pl ‘foreigners’ <strong>in</strong>flect for gender on the basis of semantics.<br />
Once the sex of their referents can be identified the noun is assigned a gender which acts as the sole controller<br />
of any agreement type (Spencer, 2002, 297).
4.3 Resolution Pr<strong>in</strong>ciples <strong>in</strong> MG 90<br />
(62) “Gender agreement with an animate NP that lacks an <strong>in</strong>herent gender is always<br />
<strong>in</strong>terpreted semantically.”<br />
(Wechsler and Zlatić, 2003, p.181)<br />
Wechsler and Zlatić (2003) argue that this pr<strong>in</strong>ciple is triggered <strong>in</strong> two cases: when the<br />
head noun is lexically genderless or when the phrase lacks a head noun. Cases that lack an<br />
<strong>in</strong>herent grammatical gender are proper nouns, such as the French proper noun Dupont,<br />
and sex-neutral nouns, such as the French journaliste. These two groups of nouns may<br />
trigger male or female agreement depend<strong>in</strong>g on their denotation:<br />
(63) Les Dupont sont beaux/belles<br />
the Duponts are good-look<strong>in</strong>g.masc.pl/good-look<strong>in</strong>g.fem.pl<br />
‘The Duponts(male or mixed/female group) are good-look<strong>in</strong>g’<br />
(64) Les journalistes sont compétents/compétentes<br />
the journalists are competent.masc.pl/competent.fem.pl<br />
‘The journalists (male or mixed/female group) are competent’<br />
(Wechsler and Zlatić, 2003, p.181)<br />
Also, Wechsler and Zlatić (2003) claim that the same agreement forms are found <strong>in</strong><br />
French pronouns, which are mascul<strong>in</strong>e when referr<strong>in</strong>g to a male or mixed group of <strong>in</strong>dividuals<br />
or fem<strong>in</strong><strong>in</strong>e when they refer to a female group:<br />
(65) Ils/elles<br />
parlent<br />
they.masc.pl/fem.pl talk<br />
‘They are talk<strong>in</strong>g <strong>in</strong> the kitchen’<br />
dans<br />
<strong>in</strong><br />
la<br />
the<br />
cuis<strong>in</strong>e<br />
kitchen<br />
The same characteristics are also found <strong>in</strong> MG. Noncoord<strong>in</strong>ate proper names, such as<br />
Filipu or Emanuil, which are lexically genderless nouns, lack a grammatical gender and<br />
they are always <strong>in</strong>terpreted semantically. Thus, <strong>in</strong> the examples below the proper noun<br />
Filipu is <strong>in</strong>terpreted as a male or female due to the respective mascul<strong>in</strong>e/fem<strong>in</strong><strong>in</strong>e end<strong>in</strong>g<br />
<strong>in</strong> the predicative adjective and determ<strong>in</strong>er, which shows semantic agreement with the<br />
controller noun:<br />
(66) O Filipu<br />
the.masc.sg Filipu<br />
‘Filipu is good-look<strong>in</strong>g’<br />
(constructed)<br />
(67) I Filipu<br />
the.fem.sg Fillipu<br />
‘Filipu is beautiful’<br />
(constructed)<br />
<strong>in</strong>e<br />
is<br />
<strong>in</strong>e<br />
is<br />
omorfos<br />
good-look<strong>in</strong>g.masc.sg<br />
omorfi<br />
beautiful.fem.sg<br />
Also, MG pronouns display the same pattern. The pronoun afti.masc.pl ‘they’ is used<br />
to refer to a group of male or mixed-sex <strong>in</strong>dividuals whereas the pronoun aftes.fem.pl<br />
‘they’ is used to refer only to a group of female <strong>in</strong>dividuals:<br />
(68) Afti/Aftes<br />
kathonte st<strong>in</strong> kuz<strong>in</strong>a<br />
they.masc.pl/they.fem.pl are-sitt<strong>in</strong>g <strong>in</strong>-the kitchen<br />
‘They (male or mixed-sex group)/(female group) are-sitt<strong>in</strong>g <strong>in</strong> the kitchen’<br />
(constructed)
4.3 Resolution Pr<strong>in</strong>ciples <strong>in</strong> MG 91<br />
Sex-neutral nouns such as those illustrated <strong>in</strong> example (64) are not found <strong>in</strong> the <strong>Greek</strong><br />
language 14 . MG coord<strong>in</strong>ate nouns resolve accord<strong>in</strong>g to MG pronouns and lexically genderless<br />
nouns which is a second piece of evidence that they resolve semantically.<br />
The resolution patterns <strong>in</strong> mixed gender coord<strong>in</strong>ate noun phrases are themselves further<br />
evidence that animate nouns follow semantic resolution. The first case concerns the<br />
coord<strong>in</strong>ation of a fem<strong>in</strong><strong>in</strong>e noun with the neuter noun to agori ‘the boy’, which denotes a<br />
male referent. The gender resolution is mascul<strong>in</strong>e plural s<strong>in</strong>ce the second conjunct denotes<br />
a male and therefore the denotation of the conjuncts determ<strong>in</strong>es the resolved gender form:<br />
(69) I Maria<br />
the.fem.sg<br />
xarumeni<br />
Maria.fem.sg<br />
happy.masc.pl<br />
‘Mary and the boy are happy’<br />
(constructed)<br />
ke<br />
and<br />
to<br />
the.neut.sg<br />
agori<br />
boy.neut.sg<br />
<strong>in</strong>e<br />
are.pl<br />
In similar coord<strong>in</strong>ate constructions where a fem<strong>in</strong><strong>in</strong>e noun is coord<strong>in</strong>ated with the<br />
neuter noun to koritsi ‘the girl’, which denotes a female, gender resolution is fem<strong>in</strong><strong>in</strong>e.<br />
This result is motivated by the fact that the two nouns denote females and therefore the<br />
denotation of the conjuncts yields this result:<br />
(70) I Maria ke to koritsi tis<br />
the.fem.sg Maria.fem.sg and the.neut.sg girl.neut.sg her<br />
<strong>in</strong>e enthusiasmen-es ja t<strong>in</strong> kenurja tus<br />
zoi<br />
are.pl enthusiastic.fem.pl for the new their.fem/masc.pl life.<br />
‘Mary and her girl are enthusiastic about their new life’<br />
(constructed)<br />
Similarly, the grammatically fem<strong>in</strong><strong>in</strong>e proper noun Maria when it is coord<strong>in</strong>ated with<br />
the grammatically neuter noun to atomo me ti fusta ‘the person with the skirt’ yields<br />
fem<strong>in</strong><strong>in</strong>e resolution s<strong>in</strong>ce, as expected, the second conjunct refers to a female. Thus, both<br />
nouns denote female referents and the result is fem<strong>in</strong><strong>in</strong>e agreement:<br />
(71) I Maria ke to<br />
the.fem.sg Maria.fem.sg and the.neut.sg<br />
ti fusta <strong>in</strong>e paraksen-es<br />
the skirt are.pl weird.fem.pl<br />
‘Mary and the person with the skirt are weird’<br />
(constructed)<br />
atomo<br />
person.neut.sg<br />
me<br />
with<br />
Also, the coord<strong>in</strong>ation of a fem<strong>in</strong><strong>in</strong>e noun with the semantically-vacuous neuter noun<br />
to moro ‘the baby’ will result <strong>in</strong> different gender feature values, depend<strong>in</strong>g on the denotation<br />
of the second conjunct. Thus, if the noun to moro ‘the baby’ denotes a male or<br />
it is referentially unspecified then resolution is mascul<strong>in</strong>e, but if it denotes a female then<br />
resolution is fem<strong>in</strong><strong>in</strong>e. Both options are illustrated below:<br />
(72) I Maria ke<br />
the.fem.sg Mary.fem.sg and<br />
<strong>in</strong>e kukl-i/-es.<br />
are.pl beautiful.masc/fem.pl.<br />
tus/-is<br />
thavmazeis<br />
them.masc/fem.pl you-admire<br />
to moro<br />
the.neut.sg baby.neut.sg<br />
Tus/-is<br />
vlepis<br />
Them.masc/fem.pl you-see<br />
tis<br />
her<br />
ke<br />
and<br />
14 One exception is the noun star ‘star’ borrowed from English. The determ<strong>in</strong>er that precedes this noun<br />
will determ<strong>in</strong>e whether it refers to male or a female. In coord<strong>in</strong>ate structures, it behaves as the semantically-<br />
vacuous neuter nouns, which we discussed above. For more details see Holton et al. (1997).
4.3 Resolution Pr<strong>in</strong>ciples <strong>in</strong> MG 92<br />
‘Mary and her baby are beautiful. You see them and you admire them’<br />
(constructed)<br />
In all cases presented above, if the order of the conjuncts is reversed the resolved<br />
agreement form rema<strong>in</strong>s the same. We randomly take example (72) and we reverse the<br />
order <strong>in</strong> the coord<strong>in</strong>ate nouns accord<strong>in</strong>g to which the neuter noun occurs first and the<br />
fem<strong>in</strong><strong>in</strong>e noun occurs second:<br />
(73) To moro tis ke i Maria<br />
the.neut.sg baby.neut.sg her and the.fem.sg Mary.fem.sg<br />
<strong>in</strong>e kukl-i/-es.<br />
Tus/-is<br />
vlepis ke<br />
are.pl beautiful.masc/fem.pl. Them.masc/fem.pl you-see and<br />
tus/-is<br />
thavmazeis<br />
them.masc/fem.pl you-admire<br />
‘The baby and Mary are beautiful. You see them and you admire them’<br />
(constructed)<br />
Thus, the above patterns are clear cases of resolution accord<strong>in</strong>g to the semantic pr<strong>in</strong>ciple<br />
s<strong>in</strong>ce what matters <strong>in</strong> the resolution outcome is the referent of the two conjuncts and not<br />
their syntactic gender.<br />
Apart from the semantic cases above, there seems to be evidence for a preference to<br />
feature match<strong>in</strong>g or syntactic agreement <strong>in</strong> MG animate coord<strong>in</strong>ate nouns. This becomes<br />
more obvious <strong>in</strong> animate grammatically neuter conjuncts that resolve to neuter irrespective<br />
of be<strong>in</strong>g semantic or semantically-vacuous 15 neuter nouns. Semantic neuter nouns, such<br />
as to agori ‘the boy’ and to koritsi ‘the girl’, have a male and a female semantic gender,<br />
respectively. The semantic pr<strong>in</strong>ciple of resolution would consult the natural gender of the<br />
conjuncts and s<strong>in</strong>ce they have a male and a female referent, we would expect mascul<strong>in</strong>e<br />
resolution. However, none of the consultants chose the mascul<strong>in</strong>e gender but selected<br />
the neuter <strong>in</strong>stead. In fact, the mascul<strong>in</strong>e gender appears to be highly marked if not<br />
ungrammatical. The relevant example is illustrated below:<br />
(74) To agori ke to<br />
the.neut boy.neut and the.neut<br />
etima/etimi? na figun<br />
ready.neut/masc? to go<br />
‘The boy and the girl are ready to go’<br />
(constructed)<br />
koritsi<br />
girl.neut<br />
Thus, resolution here depends strictly on feature match<strong>in</strong>g s<strong>in</strong>ce it ignores the mean<strong>in</strong>g<br />
of the conjo<strong>in</strong>ed elements and depends on their grammatical gender.<br />
Similarly, when two referentially unspecified neuter nouns, such as to pedi ‘the child’<br />
and to moro ‘the baby’, are coord<strong>in</strong>ated, they resolve to neuter, irrespective of a male or<br />
female referent, of two male referents or two female referents. This is illustrated <strong>in</strong> the<br />
example below:<br />
(75) To pedi ke<br />
the.neut.sg child.neut.sg and<br />
itan klidomena sto<br />
were.pl locked.neut.pl <strong>in</strong>-the<br />
to<br />
the.neut.sg<br />
domatio<br />
room<br />
<strong>in</strong>e<br />
are<br />
moro<br />
baby.neut.sg<br />
15 The term semantic neuter noun refers to the two grammatically neuter nouns that have a semantic<br />
gender, to agori ‘the boy’ and to koritsi ‘the girl’, and the term semantically-vacuous neuter noun refers to<br />
neuter nouns that do not have a semantic gender, such as to pedi ‘the child’, to moro ‘the baby’ and to<br />
atomo ‘the person’.
4.3 Resolution Pr<strong>in</strong>ciples <strong>in</strong> MG 93<br />
‘The child and the baby were locked <strong>in</strong> the room’<br />
(constructed)<br />
Thus, this is one more piece of evidence that <strong>in</strong> uniform gender nouns the grammatical<br />
gender determ<strong>in</strong>es resolution and not the natural one, lead<strong>in</strong>g to the conclusion that<br />
resolution <strong>in</strong> these patterns is motivated by the grammatical features of the coord<strong>in</strong>ated<br />
nouns.<br />
When the semantic neuter nouns to agori ‘the boy’ and to koritsi ‘the girl’ are coord<strong>in</strong>ated<br />
with one of the semantically-empty neuter nouns to pedi ‘the child’ and to moro<br />
‘the baby’, the resolved gender is still neuter plural, as illustrated below:<br />
(76) To agori<br />
the.neut.sg<br />
lipimena<br />
sad.neut.pl<br />
boy.neut.sg<br />
‘The boy and the baby are sad’<br />
(constructed)<br />
(77) To koritsi<br />
the.neut.sg<br />
lipimena<br />
sad.neut.pl<br />
girl.neut.sg<br />
‘The girl and the baby are sad’<br />
(constructed)<br />
ke<br />
and<br />
ke<br />
and<br />
to<br />
the.neut.sg<br />
to<br />
the.neut.sg<br />
moro<br />
baby.neut.sg<br />
moro<br />
baby.neut.sg<br />
<strong>in</strong>e<br />
are.pl<br />
<strong>in</strong>e<br />
are.pl<br />
In example (76), resolution would be semantic if the semantic neuter noun to agori ‘the<br />
boy’, which refers to a male, and the semantically-vacuous neuter noun to moro ‘the baby’<br />
resolved to mascul<strong>in</strong>e. Similarly, <strong>in</strong> (77), resolution would also follow the semantic pr<strong>in</strong>ciple<br />
if the semantic neuter noun to koritsi ‘the girl’, which denotes a female referent, and the<br />
semantically-vacuous neuter noun to moro ‘the baby’ resolved to mascul<strong>in</strong>e or fem<strong>in</strong><strong>in</strong>e.<br />
None of these cases, though, resolve accord<strong>in</strong>g to the semantic pr<strong>in</strong>ciple s<strong>in</strong>ce the coord<strong>in</strong>ate<br />
phrase copies the syntactic gender of the conjuncts. Thus, we can conclude that if animate<br />
neuter nouns or generally uniform gender nouns cooccur <strong>in</strong> a coord<strong>in</strong>ate phrase, they<br />
strictly follow feature match<strong>in</strong>g agreement and not semantic agreement, despite hav<strong>in</strong>g or<br />
not a natural gender.<br />
The analysis of the data shows that MG animate coord<strong>in</strong>ate nouns follow the semantic<br />
pr<strong>in</strong>ciple of resolution, which occurs <strong>in</strong> mixed gender comb<strong>in</strong>ations, and also feature<br />
match<strong>in</strong>g agreement, which occurs when the conjuncts share the same gender irrespective<br />
of their referent. A third pr<strong>in</strong>ciple is also present which will be discussed below.<br />
The follow<strong>in</strong>g table summarises the unexpected resolution patterns <strong>in</strong> animate coord<strong>in</strong>ate<br />
nouns which occur only <strong>in</strong> plural nouns (and not <strong>in</strong> s<strong>in</strong>gulars):<br />
Table 4.12: Animate PL <strong>Noun</strong>s: Unexpected patterns<br />
Plural Conj<br />
NP1 NP2 AP<br />
fpl nm/0pl npl<br />
fpl nf pl npl<br />
The first issue is whether these patterns are <strong>in</strong>stances of semantic resolution. We argued<br />
above that the semantic genders for the MG animate nouns are the mascul<strong>in</strong>e gender,<br />
which refers to males, and the fem<strong>in</strong><strong>in</strong>e gender, which refers to females. Thus, if resolution<br />
was accord<strong>in</strong>g to the semantic pr<strong>in</strong>ciple, we would expect mascul<strong>in</strong>e resolution <strong>in</strong> the<br />
pattern [fpl & nm/0pl] s<strong>in</strong>ce one of the conjuncts denotes a male referent, and fem<strong>in</strong><strong>in</strong>e
4.3 Resolution Pr<strong>in</strong>ciples <strong>in</strong> MG 94<br />
resolution <strong>in</strong> the pattern [fpl & nf pl] s<strong>in</strong>ce both conjuncts denote female referents. However,<br />
the resolved gender is neuter and therefore we argue that the above patterns cannot<br />
be considered cases of semantic resolution.<br />
Second, the issue is whether the above patterns follow the feature match<strong>in</strong>g or syntactic<br />
pr<strong>in</strong>ciple of agreement. Grammatical or feature match<strong>in</strong>g agreement occurs only when the<br />
conjuncts share the same gender. In the above comb<strong>in</strong>ations none of the conjuncts have<br />
the same gender and therefore neuter agreement is not a case of syntactic or a feature<br />
match<strong>in</strong>g pr<strong>in</strong>ciple. F<strong>in</strong>ally, <strong>in</strong> an <strong>in</strong>formal discussion with native speakers, they argued<br />
that the choice of the neuter gender was driven by the fact that both nouns could be<br />
referred to by the noun atoma ‘persons’, which is neuter <strong>in</strong> gender; although it does not<br />
occur <strong>in</strong> the sentence, it is implied. Thus, we assume that a different pr<strong>in</strong>ciple is present<br />
<strong>in</strong> these patterns.<br />
4.3.2 Syntactic Resolution <strong>in</strong> Inanimate <strong>Noun</strong>s<br />
Syntactic resolution means that it is the grammatical gender of the nouns that matters<br />
and not their semantic or natural gender. We will argue that the expected patterns of MG<br />
<strong>in</strong>animate nouns admit the syntactic resolution pr<strong>in</strong>ciple. Below we repeat the table of<br />
<strong>in</strong>animate nouns:<br />
Table 4.13: <strong>Agreement</strong> with Inanimate Coord<strong>in</strong>ate <strong>Noun</strong>s<br />
S<strong>in</strong>gular Conjuncts<br />
Concrete/Abstract <strong>Noun</strong>s Concrete <strong>Noun</strong>s Abstract <strong>Noun</strong>s<br />
NP1 NP2 AP NP1 NP2 AP NP1 NP2 AP<br />
msg msg mpl>npl msg fsg npl>mpl msg fsg npl>mpl>fpl<br />
fsg fsg fpl>npl msg nsg npl msg nsg npl<br />
nsg nsg npl fsg nsg npl fsg nsg npl<br />
Plural Conjuncts<br />
Concrete/Abstract <strong>Noun</strong>s Concrete <strong>Noun</strong>s Abstract <strong>Noun</strong>s<br />
NP1 NP2 AP NP1 NP2 AP NP1 NP2 AP<br />
mpl mpl mpl mpl fpl mpl>npl mpl fpl mpl>npl>fpl<br />
fpl fpl fpl mpl npl npl mpl npl npl<br />
npl npl npl fpl npl npl fpl npl npl<br />
Dalrymple and Kaplan (2000) claim that there are three recurr<strong>in</strong>g characteristics <strong>in</strong><br />
syntactic gender resolution. First, a coord<strong>in</strong>ate phrase behaves as if it has the same gender<br />
as some noncoord<strong>in</strong>ate element <strong>in</strong> the language and there are no special agreeement markers<br />
that would appear only with coord<strong>in</strong>ate phrases. Secondly, a coord<strong>in</strong>ate structure with<br />
conjuncts that share the same gender behaves as if it has the same gender found <strong>in</strong> the<br />
conjuncts, and thirdly, mixed gender conjuncts usually resolve to a gender feature value<br />
found <strong>in</strong>dependently <strong>in</strong> a language, which may or may not be the same as the gender value<br />
of the conjuncts (788).<br />
These characteristics occur <strong>in</strong> MG <strong>in</strong>animate nouns. Coord<strong>in</strong>ate structures with the<br />
same gender conjuncts result <strong>in</strong> that same gender, which is a case of feature match<strong>in</strong>g of<br />
the grammatical gender value of the <strong>in</strong>dividual conjuncts with the gender feature value of<br />
the whole coord<strong>in</strong>ate phrase:<br />
(78) O iroismos ke<br />
the.masc.sg heroism.masc.sg and<br />
itan kiriarx-i kata<br />
were.pl prevalent.masc.pl dur<strong>in</strong>g<br />
o<br />
the.masc.sg<br />
to 1821<br />
the 1821<br />
agonas<br />
fight<strong>in</strong>g.masc.sg
4.3 Resolution Pr<strong>in</strong>ciples <strong>in</strong> MG 95<br />
‘Heroism and fight<strong>in</strong>g were prevalent dur<strong>in</strong>g 1821’<br />
(G.N.: To Vima)<br />
Coord<strong>in</strong>ate structures with mixed gender conjucts also show that syntactic resolution<br />
is the ma<strong>in</strong> pr<strong>in</strong>ciple <strong>in</strong> <strong>in</strong>animate nouns s<strong>in</strong>ce mixed gender conjuncts resolve to neuter.<br />
This is a gender feature value that occurs <strong>in</strong>dependently <strong>in</strong> the language and which is not<br />
necessarily found <strong>in</strong> the <strong>in</strong>dividual conjuncts. For <strong>in</strong>stance, <strong>in</strong> example (79) the coord<strong>in</strong>ate<br />
phrase consists of a mascul<strong>in</strong>e and a fem<strong>in</strong><strong>in</strong>e noun. The predicative adjective resolves<br />
to neuter which is a different gender value that is not found <strong>in</strong> any of the two conjuncts.<br />
The same gender value occurs <strong>in</strong> example (80) where the coord<strong>in</strong>ate phrase consists of a<br />
fem<strong>in</strong><strong>in</strong>e and a neuter noun. In this case, neuter appears as the gender value only <strong>in</strong> one<br />
of the two conjuncts:<br />
(79) O kanapes ke<br />
the.masc.sg<br />
aspr-a<br />
white.neut.pl<br />
sofa.masc.sg and<br />
‘The sofa and the chair are white’<br />
(constructed)<br />
(80) I karekla ke<br />
the.fem.sg<br />
anet-a<br />
chair.fem.sg and<br />
comfortable.neut.pl<br />
‘The chair and table are comfortable’<br />
(constructed)<br />
i<br />
the.fem.sg<br />
to<br />
the.neut.sg<br />
polithrona<br />
chair.fem.sg<br />
trapezi<br />
table.neut.sg<br />
<strong>in</strong>e<br />
are.pl<br />
<strong>in</strong>e<br />
are.pl<br />
Also, the most important reason that MG <strong>in</strong>animate nouns use the syntactic pr<strong>in</strong>ciple<br />
of resolution is that <strong>in</strong>animate nouns lack a natural gender and therefore semantics<br />
cannot be consulted <strong>in</strong> order to derive resolution. Hence, the only pr<strong>in</strong>ciple they could<br />
obey is the syntactic one. For example, the MG coord<strong>in</strong>ate phrase below consists of two<br />
fem<strong>in</strong><strong>in</strong>e coord<strong>in</strong>ate nouns. The nouns that have a fem<strong>in</strong><strong>in</strong>e grammatical gender do not<br />
have any semantic gender to affect the resolution process and fem<strong>in</strong><strong>in</strong>e agreement is chosen<br />
syntactically:<br />
(81) I vivliothiki ke<br />
the.fem.sg bookcase.fem.sg and<br />
<strong>in</strong>e skur-es<br />
are.pl dark.fem.pl<br />
‘The bookcase and d<strong>in</strong><strong>in</strong>g-table are dark’<br />
i<br />
the.fem.sg<br />
trapezaria<br />
d<strong>in</strong><strong>in</strong>g-table.fem.sg<br />
Similarly, mixed gender coord<strong>in</strong>ate nouns without a natural gender admit neuter agreement<br />
s<strong>in</strong>ce the <strong>in</strong>dividual conjuncts do not have a semantic gender:<br />
(82) O thanatos ke<br />
the.masc.sg death.masc.sg and<br />
<strong>in</strong>e simantik-a<br />
are.pl important.neut.pl<br />
‘Death and freedom are important’<br />
i<br />
the.fem.sg<br />
eleftheria<br />
freedom.fem.sg<br />
Hence, the above examples make it obvious that <strong>in</strong>animate nouns do not allow cases of<br />
semantic resolution <strong>in</strong> any type of gender comb<strong>in</strong>ation.<br />
Let us consider next whether the exceptional patterns follow the syntactic resolution<br />
pr<strong>in</strong>ciple. The table below presents only the unexpected cases found <strong>in</strong> this group of nouns:
4.3 Resolution Pr<strong>in</strong>ciples <strong>in</strong> MG 96<br />
Table 4.14: Inanimate SG/PL <strong>Noun</strong>s: Unexpected patterns<br />
S<strong>in</strong>gular <strong>Noun</strong>s<br />
Concrete/Abstract <strong>Noun</strong>s Concrete <strong>Noun</strong>s Abstract <strong>Noun</strong>s<br />
NP1 NP2 AP NP1 NP2 AP NP1 NP2 AP<br />
msg msg npl msg fsg mpl msg fsg fpl/mpl<br />
fsg fsg npl<br />
Plural <strong>Noun</strong>s<br />
Concrete <strong>Noun</strong>s Abstract <strong>Noun</strong>s<br />
NP1 NP2 AP NP1 NP2 AP<br />
mpl fpl mpl mpl fpl mpl/fpl<br />
Accord<strong>in</strong>g to table (4.14), the coord<strong>in</strong>ate structures which consist of conjuncts with the<br />
same gender, either mascul<strong>in</strong>e or fem<strong>in</strong><strong>in</strong>e, resolve to the same gender of the conjuncts and<br />
to the neuter gender as well. This is shown <strong>in</strong> the examples below<br />
(83) O iroismos ke o<br />
agonas<br />
the.masc.sg heroism.masc.sg and the.masc.sg fight<strong>in</strong>g.masc.sg<br />
itan kiriarx-i/-a<br />
kata to 1821<br />
were.pl prevalent.masc/neut.pl dur<strong>in</strong>g the 1821<br />
‘Heroism and fight<strong>in</strong>g were prevalent dur<strong>in</strong>g 1821’<br />
(G.N.: To Vima)<br />
(84) I vivliothiki ke<br />
the.fem.sg bookcase.fem.sg and<br />
<strong>in</strong>e skur-es/-a<br />
are.pl dark.fem/neut.pl<br />
‘The bookcase and d<strong>in</strong><strong>in</strong>g-table are dark’<br />
(constructed)<br />
i<br />
the.fem.sg<br />
trapezaria<br />
d<strong>in</strong><strong>in</strong>g-table.fem.sg<br />
Although neuter is the syntactic gender, these cases are not syntactic s<strong>in</strong>ce we would expect<br />
<strong>in</strong> identical gender conjuncts grammatical agreement to result <strong>in</strong> the same gender found <strong>in</strong><br />
the conjuncts and not <strong>in</strong> a different gender. Thus, the unexpected neuter gender resolution<br />
<strong>in</strong> shared gender coord<strong>in</strong>ate nouns does not follow the syntactic pr<strong>in</strong>ciple.<br />
Also, mixed gender coord<strong>in</strong>ations with a mascul<strong>in</strong>e and a fem<strong>in</strong><strong>in</strong>e noun resolve to the<br />
expected neuter gender and to the unexpected mascul<strong>in</strong>e gender. The same comb<strong>in</strong>ation<br />
of nouns also resolves to the unexpected fem<strong>in</strong><strong>in</strong>e gender only if the conjuncts are abstract.<br />
These examples are shown below:<br />
(85) O kanapes ke<br />
the.masc.sg<br />
aspr-a/-i<br />
sofa.masc.sg and<br />
white.neut/masc.pl<br />
‘The sofa and the chair are white’<br />
(constructed)<br />
i<br />
the.fem.sg<br />
(86) O thanatos ke i<br />
the.masc.sg death.masc.sg and the.fem.sg<br />
<strong>in</strong>e simantik-a/-i/-es<br />
are.pl important.neut.pl/masc.pl/fem.pl<br />
‘Death and freedom are important’<br />
polithrona<br />
chair.fem.sg<br />
eleftheria<br />
freedom.fem.sg<br />
<strong>in</strong>e<br />
are.pl<br />
These gender feature values are rather unusual <strong>in</strong> <strong>in</strong>animate conjuncts s<strong>in</strong>ce the expected<br />
gender <strong>in</strong> mixed gender <strong>in</strong>animate nouns is the neuter gender, which is a syntactic
4.4 Reference Transfer <strong>in</strong> Noncoord<strong>in</strong>ate Constructions 97<br />
gender and not any of the semantic genders, such as the mascul<strong>in</strong>e and fem<strong>in</strong><strong>in</strong>e genders.<br />
The fact that the above cases resolve to a gender which is not syntactic shows that these<br />
resolution patterns do not follow the syntactic pr<strong>in</strong>ciple. In addition, s<strong>in</strong>ce <strong>in</strong>animate nouns<br />
do not resort to semantic resolution anyway, the above patterns must be evidence of the<br />
existence of a third pr<strong>in</strong>ciple. F<strong>in</strong>ally, an <strong>in</strong>formal discussion with the native speakers revealed<br />
that the choice of the unexpected genders was driven by the fact that both nouns<br />
could be referred to by an implied noun, which has agreement features similar to those<br />
of the unexpected form and it does not occur <strong>in</strong> the sentence. Thus, we assume the unexpected<br />
patterns do not belong to the syntactic or semantic pr<strong>in</strong>ciples of resolution but<br />
they are evidence of a third resolution pr<strong>in</strong>ciple which we will refer to as the referential<br />
pr<strong>in</strong>ciple. Before we analyse referential agreement <strong>in</strong> MG coord<strong>in</strong>ate nouns, we need to<br />
expla<strong>in</strong> what referential agreement is and whether English or other languages show cases<br />
of referential agreement.<br />
4.4 Reference Transfer <strong>in</strong> Noncoord<strong>in</strong>ate Constructions<br />
The aim of this section is to present the phenomenon of referential agreement or “reference<br />
transfer”(Pollard and Sag, 1994) and show how the mismatch gender agreement patterns<br />
<strong>in</strong> MG can be expla<strong>in</strong>ed <strong>in</strong> terms of this phenomenon.<br />
Mismatch agreement has been studied by a number of l<strong>in</strong>guists and a number of agreement<br />
phenomena have been noticed to be problematic for the conventional purely syntactic<br />
or semantic views. The most characteristic phenomenon of mismatch agreement, relevant<br />
to the MG data, is that of “mean<strong>in</strong>g transfer” or “reference transfer” 16 .<br />
“Mean<strong>in</strong>g transfer”, accord<strong>in</strong>g to Nunberg (1977), refers to the “productive l<strong>in</strong>guistic<br />
processes that enable us to use the same expression to refer to what are <strong>in</strong>tuitively dist<strong>in</strong>ct<br />
sorts of categories of th<strong>in</strong>gs” (Nunberg, 1995, 109). These transfers might <strong>in</strong>volve the<br />
figures that traditional rhetoric describes such as metaphors, synesthesias, metonymies<br />
and synecdotes. Nevertheless, the ma<strong>in</strong> characteristic of reference transfer is that it is a<br />
l<strong>in</strong>guistic process with specific l<strong>in</strong>guistic mechanisms. “Reference transfer” can apply to<br />
predicates of any k<strong>in</strong>d, lexical or phrasal and also to common nouns. Two characterictic<br />
examples of “transfer” are presented below:<br />
(87) a. I am parked out back<br />
b. The ham sandwich is at table 7<br />
(Nunberg, 1995, 110,115)<br />
Nunberg (1995) gives a brief analysis of how “transfer” works <strong>in</strong> the above examples.<br />
In both examples, there is a correspondence between the th<strong>in</strong>gs <strong>in</strong> one doma<strong>in</strong> and the<br />
th<strong>in</strong>gs <strong>in</strong> another doma<strong>in</strong>. Thus, <strong>in</strong> example (87a) I refers to the car parked <strong>in</strong> a specific<br />
location by referr<strong>in</strong>g to the driver of the car and not to the car itself. Similarly, <strong>in</strong> example<br />
(87b) the phrase is uttered by a restaurant waiter, who considers that customers acquire<br />
their dist<strong>in</strong>ctive properties with regard to the dishes they order. Thus, the noun phrase<br />
‘ham sandwich’ refers to the person who has ordered a ham sandwich.<br />
Nunberg (1995) attempts to expla<strong>in</strong> the l<strong>in</strong>guistic mechanisms of “transfer” <strong>in</strong> general.<br />
He dist<strong>in</strong>guishes two types of “transfer”. The first one <strong>in</strong>volves demonstratives or <strong>in</strong>dexicals<br />
and it is mentioned as “deferred ostension” or “deferred <strong>in</strong>dexical reference” (Nunberg,<br />
1995, 111). Accord<strong>in</strong>g to Nunberg (1995), this type of transfer “allows a demonstrative or<br />
<strong>in</strong>dexical to refer to an object that corresponds <strong>in</strong> a certa<strong>in</strong> way to the contextual element<br />
picked out by a demonstration or by the semantic character of the expression” (Nunberg,<br />
1995, 111). Consider the follow<strong>in</strong>g example:<br />
16 Nunberg (1977, 1995, 2005) uses the term “mean<strong>in</strong>g tranfer”. Pollard and Sag (1994) use the term<br />
“reference transfer”. Both of them refer to the same phenomenon.
4.4 Reference Transfer <strong>in</strong> Noncoord<strong>in</strong>ate Constructions 98<br />
(88) This is parked out back (hold<strong>in</strong>g up a key)<br />
(Nunberg, 1995, 110)<br />
Nunberg (1995) argues that it is more likely that the subject refers not to the key that<br />
the speaker is hold<strong>in</strong>g but to the car that the key goes with. The evidence to support<br />
this view is that the number of the demonstrative is determ<strong>in</strong>ed by the <strong>in</strong>tended referent<br />
and not the demonstratum. So even if the customer is hold<strong>in</strong>g up several keys that fit a<br />
s<strong>in</strong>gle car, he would say This is parked out back, whereas if he is hold<strong>in</strong>g up only one key<br />
that fits several cars he would say These are parked out back. Further evidence that the<br />
demonstrative refers to the car is drawn by conjo<strong>in</strong><strong>in</strong>g another predicate that describes the<br />
car, but not a predicate that describes the key:<br />
(89) a. This is parked out back and may not start<br />
b.?? This fits only the left front door and is parked out back<br />
(Nunberg, 1995, 110)<br />
Accord<strong>in</strong>g to Nunberg (1995), the demonstratives and adjectives of other languages<br />
display similar agreement patterns. Italian, for <strong>in</strong>stance, uses the fem<strong>in</strong><strong>in</strong>e noun for the<br />
word ‘key’, la chiave, and the mascul<strong>in</strong>e noun for the word ‘truck’, il camion. Thus, if a<br />
customer gives an attendant the key to a truck, the referent and not the demonstratum will<br />
determ<strong>in</strong>e the gender of the demonstrative and the adjective for ‘parked’, as seen below:<br />
(90) Hold<strong>in</strong>g up a key(la chiave,fem.sg) to refer to a truck(il camion,masc.sg)<br />
Questo.masc.sg<br />
This.masc<br />
é<br />
is<br />
‘This is parked <strong>in</strong> back’<br />
(Nunberg, 1995, 110)<br />
parcheggiato.masc.sg<br />
parked.masc<br />
<strong>in</strong><br />
<strong>in</strong><br />
dietro<br />
back<br />
A different mechanism of transfer is what is known as “predicate transfer”. In this type<br />
of transfer “the name of a property that applies to someth<strong>in</strong>g <strong>in</strong> one doma<strong>in</strong> can sometimes<br />
be used as the name of a property that applies to th<strong>in</strong>gs <strong>in</strong> another doma<strong>in</strong>, provided that<br />
the two properties correspond <strong>in</strong> a certa<strong>in</strong> way”(Nunberg, 1995, 111). This type of transfer<br />
is <strong>in</strong>troduced by the example presented below:<br />
(91) I am parked out back<br />
(Nunberg, 1995, 110)<br />
As Nunberg (1995) claims, most people would say that the transfer <strong>in</strong>volves the subject<br />
of the sentence I. A number of tests, however, <strong>in</strong>dicate that the above example <strong>in</strong>volves a<br />
transfer of the conventional mean<strong>in</strong>g of the predicate parked out back. For <strong>in</strong>stance, if the<br />
speaker had two cars he wouldn’t say:<br />
(92) a. We are parked out back<br />
(Nunberg, 1995, 111)<br />
Secondly, it is possible to conjo<strong>in</strong> any other predicate that describes the speaker, and<br />
not a predicate that describes the car:<br />
(93) a. I am parked out back and have been wait<strong>in</strong>g for 15 m<strong>in</strong>utes<br />
b. *I am parked out back and may not start<br />
(Nunberg, 1995, 111)
4.4 Reference Transfer <strong>in</strong> Noncoord<strong>in</strong>ate Constructions 99<br />
Another piece of evidence is that the verb <strong>in</strong> the sentence is am first-person s<strong>in</strong>gular<br />
and not is third-person s<strong>in</strong>gular as we would expect if the subject referred to a car. Thus,<br />
the subject of I refers to the speaker, and the transfer <strong>in</strong>volves the predicate parked out<br />
back which contributes a property of persons whose cars are parked out back (Nunberg,<br />
2005).<br />
F<strong>in</strong>ally, Nunberg (1995, 2005) claims that mean<strong>in</strong>g transfer operates on the mean<strong>in</strong>gs<br />
of common nouns as well as <strong>in</strong>dexicals, whether they are <strong>in</strong> predicate position or referr<strong>in</strong>g<br />
position. The follow<strong>in</strong>g example is similar to example (87b):<br />
(94) Who is the ham sandwich?<br />
The process of transfer <strong>in</strong> this example is that customers acquire their properties based<br />
on their relation to the dishes they order. The common noun ‘ham sandwich’ has a transferred<br />
mean<strong>in</strong>g and it contributes a property of people who have ordered ham sandwiches.<br />
Pollard and Sag (1994) also discuss the phenomenon of “reference transfer” as part<br />
of agreement mismatches, claim<strong>in</strong>g that agreement is often guided by reference transfer,<br />
rather than the grammatical (<strong>in</strong>herent) agreement properties of the phrase itself. This<br />
notion is illustrated <strong>in</strong> the two classic examples below which show how the agreement<br />
result is affected by this phenomenon:<br />
(95) a. The ham sandwich is at table 7<br />
b. The hash browns at table n<strong>in</strong>e is gett<strong>in</strong>g angry<br />
(Pollard and Sag, 1994, 85)<br />
The first example, repeated from above, displays s<strong>in</strong>gular subject-verb agreement. Even<br />
though, it seems that the verb agrees with the grammatically s<strong>in</strong>gular noun ham sandwich,<br />
this is not the case. As mentioned above, the waiter uses the specific noun to refer to a<br />
person who has ordered a ham sandwich. The person is a nonaggregate entity and the<br />
verb shows agreement with the referred implied noun. The same notion is illustrated more<br />
clearly <strong>in</strong> example (95b). The noun phrase hash browns is grammatically (<strong>in</strong>herently)<br />
plural and therefore we would expect plural verb agreement. Instead, the verb shows<br />
s<strong>in</strong>gular agreement s<strong>in</strong>ce the referent is transferred to the person who has placed the order<br />
and which is a nonaggregate entity.<br />
Pollard and Sag (1994) also <strong>in</strong>troduce another type of reference transfer which occurs<br />
<strong>in</strong> “s<strong>in</strong>gular plurals”. In this group of nouns, there is a conflict between the agreement<br />
features of the subject NP and those that the s<strong>in</strong>gular verb requires from its subject. The<br />
follow<strong>in</strong>g examples illustrate this k<strong>in</strong>d of agreement mismatch:<br />
(96) a. Eggs is my favourite breakfast<br />
b. Eggs bothers me more than okra<br />
(Pollard and Sag, 1994, 70)<br />
In both examples, even though the NP appears <strong>in</strong> plural number, the verb shows<br />
s<strong>in</strong>gular agreement. The noun eggs is referred to by a nonagreggate entity, such as ‘menu<br />
type’ <strong>in</strong> example (96a) or ‘food’ <strong>in</strong> example (96b). Accord<strong>in</strong>g to Pollard and Sag (1994),<br />
<strong>in</strong> both cases the referent of the phrase is <strong>in</strong> some contextual relation to an object usually<br />
picked out by nonmetaphorical utterances of that phrase, allow<strong>in</strong>g s<strong>in</strong>gular agreement.<br />
Similar examples are the ones below, which illustrate the same k<strong>in</strong>d of contextual<br />
relation between the NP and the referred noun:<br />
(97) a. Do<strong>in</strong>g phonology problems and dr<strong>in</strong>k<strong>in</strong>g vodka makes me sick<br />
b. Unleashed dogs on city sidewalks threatens the health and welfare of law-abid<strong>in</strong>g<br />
citizens<br />
(Pollard and Sag, 1994, 86)
4.4 Reference Transfer <strong>in</strong> Noncoord<strong>in</strong>ate Constructions 100<br />
In both cases, the two NPs refer to different nouns, such as an activity or a social<br />
problem, respectively. S<strong>in</strong>ce the referred noun is perceived as a nonagreggate entity then<br />
it yields s<strong>in</strong>gular agreement to the verb.<br />
The phenomenon of “reference transfer” described above, which affects the agreement<br />
result, seems to be closely related to the type of mismatch gender agreement found <strong>in</strong> the<br />
MG unexpected patterns described <strong>in</strong> the previous section. In the next section, we will<br />
argue that “reference transfer” is found <strong>in</strong> MG data and that it affects the gender and not<br />
the number feature.<br />
4.4.1 Reference Transfer <strong>in</strong> <strong>Modern</strong> <strong>Greek</strong><br />
In this section, we will show that all the unexpected patterns <strong>in</strong> MG are cases of agreement<br />
that fall with<strong>in</strong> the phenomenon of “reference transfer” and we adopt the term referential<br />
agreement or referential pr<strong>in</strong>ciple of resolution to refer to this type of resolution.<br />
As already mentioned above, “reference transfer” seems to affect the agreement result<br />
<strong>in</strong> noncoord<strong>in</strong>ate constructions. In the same way, resolved agreement <strong>in</strong> coord<strong>in</strong>ate<br />
constructions is not an exception and can be determ<strong>in</strong>ed by “reference transfer”. In the<br />
agreement cases discussed by Nunberg (1977, 1995) and Pollard and Sag (1994), the agreement<br />
feature that is affected by “reference transfer” is number, as <strong>in</strong> the case of “s<strong>in</strong>gular<br />
plurals”, repeated from above:<br />
(98) Eggs is my favourite breakfast<br />
(Pollard and Sag, 1994, 70)<br />
The noun eggs, which is the subject <strong>in</strong> the phrase and it is grammatically plural,<br />
is somehow referentially associated with a noun that is implied <strong>in</strong> the phrase and it is<br />
a nonagreggate entity. The fact that the implied noun is s<strong>in</strong>gular leads to a s<strong>in</strong>gular<br />
agreement <strong>in</strong> the verb.<br />
A similar k<strong>in</strong>d of mechanism is <strong>in</strong> operation <strong>in</strong> MG coord<strong>in</strong>ate nouns which result <strong>in</strong><br />
unexpected gender resolution 17 . The follow<strong>in</strong>g example discusses how “reference transfer”<br />
works <strong>in</strong> MG:<br />
(99) I<br />
dromi ke i<br />
the.masc.pl roads.masc.pl and the.fem.pl<br />
<strong>in</strong>e jemat-i/-a<br />
kosmo<br />
are.pl full-of.masc/neut.pl people<br />
‘The roads and the alleys are full of people’<br />
platies<br />
squares.fem.pl<br />
The <strong>in</strong>animate coord<strong>in</strong>ate phrase with a mascul<strong>in</strong>e plural and a fem<strong>in</strong><strong>in</strong>e plural noun<br />
yields two types of agreement, the expected neuter gender, and the unexpected mascul<strong>in</strong>e<br />
gender. The former occurs <strong>in</strong> all cases of mixed gender coord<strong>in</strong>ation <strong>in</strong> <strong>in</strong>animate nouns<br />
and it is the result of syntactic resolution, but the latter is unusual for the simple reason<br />
that neuter is the default gender <strong>in</strong> <strong>in</strong>animate nouns and not mascul<strong>in</strong>e. What allows<br />
the appearance of the mascul<strong>in</strong>e gender <strong>in</strong> mixed gender cases of <strong>in</strong>animate nouns is the<br />
phenomenon of “reference transfer”.<br />
We argue that what happens <strong>in</strong> MG is that the presence of the mascul<strong>in</strong>e gender as the<br />
agreement form <strong>in</strong> the predicative adjective is due to an implied miss<strong>in</strong>g noun, which is<br />
mascul<strong>in</strong>e <strong>in</strong> gender; it does not occur <strong>in</strong> the phrase but it is brought <strong>in</strong>to the context and<br />
denotes the coord<strong>in</strong>ate nouns. A number of consultants who chose the mascul<strong>in</strong>e gender as<br />
17 We have not come across relevant literature on reference transfer <strong>in</strong> MG non-coord<strong>in</strong>ate and coord<strong>in</strong>ate<br />
constructions. However, we assume that reference transfer <strong>in</strong> gender occurs only <strong>in</strong> coord<strong>in</strong>ate nouns <strong>in</strong> the<br />
patterns discussed here and not <strong>in</strong> non-coord<strong>in</strong>ates.
4.4 Reference Transfer <strong>in</strong> Noncoord<strong>in</strong>ate Constructions 101<br />
the agreement feature value were asked what motivated this choice for this specific gender<br />
and they confirmed that they assumed both coord<strong>in</strong>ate nouns were denoted by the common<br />
noun xori ‘spaces’, which is mascul<strong>in</strong>e <strong>in</strong> gender. The above sentence <strong>in</strong>clud<strong>in</strong>g the implied<br />
noun can be rewritten as follows:<br />
(100) I<br />
dromi ke i platies<br />
the.masc.pl roads.masc.pl and the.fem.pl squares.fem.pl<br />
<strong>in</strong>e jemat-i xori<br />
me kosmo<br />
are.pl full.masc.pl places.masc.pl with people<br />
‘The roads and the alleys are places full of people’<br />
As <strong>in</strong> the examples presented by Nunberg (1995) and Pollard and Sag (1994), the<br />
implied miss<strong>in</strong>g noun seems to have a semantic relation to the two nouns that are present<br />
<strong>in</strong> the coord<strong>in</strong>ate phrase. In the case of MG, the relation must be one of hyponymy. The<br />
implied miss<strong>in</strong>g noun functions as the superord<strong>in</strong>ate or cover term of the two nouns that<br />
are present <strong>in</strong> the coord<strong>in</strong>ate phrase and morphosyntactically determ<strong>in</strong>es the agreement<br />
result s<strong>in</strong>ce the predicative adjective is forced to show morphosyntactic agreement with the<br />
implied noun. Therefore, <strong>in</strong> example (100), the noun xori ‘spaces’ is the superord<strong>in</strong>ate or<br />
cover term of the two nouns dromi ‘roads’ and platies ‘squares’ and, although it is omitted,<br />
it determ<strong>in</strong>es the gender feature of the adjective. We need to note that the noun xori<br />
‘spaces’ is not the only possible superord<strong>in</strong>ate term. In fact, a number of consultants chose<br />
the noun perioxes’ ‘areas’ as the hypernymic term which is fem<strong>in</strong><strong>in</strong>e <strong>in</strong> gender and it can<br />
also function as a cover term for both nouns. These consultants were fewer (4 out of 26)<br />
when compared to the 15 (out of 26) consultants who chose mascul<strong>in</strong>e and the 13 (out of<br />
26) who chose neuter.<br />
The same process takes place <strong>in</strong> animate coord<strong>in</strong>ate nouns. In example (101), the<br />
two conjuncts resolve to the expected semantic gender fem<strong>in</strong><strong>in</strong>e plural and the unexpected<br />
neuter plural. The latter occurs due to the phenomenon of referential agreement. Thus, an<br />
implied miss<strong>in</strong>g noun is brought <strong>in</strong>to the context, such as atoma ‘persons’, which is neut <strong>in</strong><br />
gender and which determ<strong>in</strong>es morphosyntactically the agreement gender of the predicative<br />
adjective and functions as the superord<strong>in</strong>ate or cover term for the two conjuncts jajades<br />
‘grandmothers’ and koritsia ‘girls’:<br />
(101) I jajades<br />
ke<br />
the.fem.pl grandmothers.fem.pl and<br />
<strong>in</strong>e xarumen-es/-a<br />
are happy.fem/neut.pl<br />
‘The grandmothers and the girls are happy’<br />
ta<br />
the.neut.pl<br />
koritsia<br />
girls.neut.pl<br />
The first issue that needs to be discussed with regard to the phenomenon of referential<br />
agreement <strong>in</strong> MG is that <strong>in</strong> the specific coord<strong>in</strong>ate structures gender is the ma<strong>in</strong> feature<br />
that is affected (also found <strong>in</strong> Italian), and not number, as <strong>in</strong> the case of English. Our data<br />
showed that the number feature does not show any unexpected behaviour (i.e. s<strong>in</strong>gular<br />
agreement) s<strong>in</strong>ce number agreement is always plural with coord<strong>in</strong>ate nouns irrespective of<br />
the number of the conjuncts. In fact, s<strong>in</strong>gular verb agreement is ungrammatical when the<br />
conjuncts refer to different entities. On the other hand, the adjective has an unexpected<br />
gender form. This gender form is the same as the gender form of the implied noun that is<br />
referred to by the consultants but never occurs <strong>in</strong> the phrase. Therefore, <strong>in</strong> MG referential<br />
agreement <strong>in</strong>volves the gender feature and not the number feature, as <strong>in</strong> other languages.<br />
The second issue is that referential agreement <strong>in</strong> the MG patterns differs when compared<br />
to the correspond<strong>in</strong>g English ones with regard to the relation of the nouns that are<br />
present <strong>in</strong> the phrase and the referred implied noun. In English, it seems that the relation<br />
is one of metonymy, where the name of one entity is used to refer to another entity, as <strong>in</strong>
4.5 An Approach to Gender <strong>Agreement</strong> <strong>in</strong> MG 102<br />
example (87b), or one of hyponymy, where the implied noun functions as the superord<strong>in</strong>ate<br />
term of the two conjuncts, which are part of the coord<strong>in</strong>ate phrase, as <strong>in</strong> examples (96)<br />
and (97). This cover term, which is <strong>in</strong>directly implied, determ<strong>in</strong>es the gender value <strong>in</strong> the<br />
predicative adjective. In MG, it seems to be strictly a hyponymic relation where the two<br />
conjuncts are hyponyms of the superord<strong>in</strong>ate implied noun, at least <strong>in</strong> the cases we have<br />
presented so far.<br />
Also, <strong>in</strong> some cases the consultants do not br<strong>in</strong>g <strong>in</strong>to the discourse the same implied<br />
noun but a different one. For <strong>in</strong>stance, as already mentioned <strong>in</strong> example (100) most of the<br />
speakers brought <strong>in</strong>to the context the noun xori ‘spaces’ but some of them brought <strong>in</strong>to the<br />
context the noun perioxes’ ‘areas’. In both cases, the two nouns are considered hypernyms<br />
of the two conjuncts dromi ‘roads’ and platies ‘squares’. Thus, we argue that the consultants<br />
br<strong>in</strong>g <strong>in</strong>to the discourse SOME implied noun that it is necessarily the hypernym of<br />
the two coord<strong>in</strong>ate nouns. This noun has specific agreement features which determ<strong>in</strong>e the<br />
agreement features of the predicative adjective. The implied noun may differ from speaker<br />
to speaker depend<strong>in</strong>g on the hierarchical group<strong>in</strong>g of the lexical items <strong>in</strong> each speaker’s<br />
m<strong>in</strong>d. In most cases, however, speakers tend to br<strong>in</strong>g <strong>in</strong>to the context the same implied<br />
noun.<br />
F<strong>in</strong>ally, the data we have collected and presented to the consultants yield the unexpected<br />
patterns discussed so far. The nature of the phenomenon, which is context dependent,<br />
may allow the possibility of additional unexpected gender values result<strong>in</strong>g from other<br />
coord<strong>in</strong>ate nouns. This would require more research and a broader collection of data with<br />
a larger number of consultants to detect a comprehensive list of all the possible unexpected<br />
patterns.<br />
Thus, we have argued that the unexpected gender patterns <strong>in</strong> MG display the phenomenon<br />
of referential agreement. Referential agreement refers to the mismatch gender<br />
agreement between two coord<strong>in</strong>ate nouns and the predicative adjective/participle of the<br />
sentence. Accord<strong>in</strong>g to referential agreement, an implied noun is brought <strong>in</strong>to the context,<br />
it functions as the superord<strong>in</strong>ate term of the coord<strong>in</strong>ate nouns and its agreement features<br />
are copied by the predicative adjective/participle. Referential agreement <strong>in</strong> MG affects only<br />
the gender feature. In other languages, such as English or Italian, it affects the number<br />
feature or both the number and gender features, respectively.<br />
4.5 An Approach to Gender <strong>Agreement</strong> <strong>in</strong> MG<br />
This section presents an analysis of the expected agreement patterns <strong>in</strong> MG coord<strong>in</strong>ate<br />
nouns us<strong>in</strong>g the set-based theory. Follow<strong>in</strong>g Sadler (2006), we will assume the standard rule<br />
of gender resolution for the syntactic patterns of <strong>in</strong>animate nouns and we will <strong>in</strong>troduce<br />
a specification <strong>in</strong> the phrase structure rule for the semantic patterns of animate nouns.<br />
F<strong>in</strong>ally, we will also propose an analysis of the referential agreement patterns based on<br />
LFG.<br />
4.5.1 Problems of the Current Theories for the MG data<br />
Before we present our analysis, we will discuss the various problems of the different theorerical<br />
approaches for the MG data.<br />
Dalrymple and Kaplan’s set-based approach for gender resolution <strong>in</strong> coord<strong>in</strong>ate phrases<br />
accounts for syntactic cases of gender resolution and not for semantic cases of gender<br />
resolution found <strong>in</strong> animate coord<strong>in</strong>ate nouns <strong>in</strong> most languages. MG animate coord<strong>in</strong>ate<br />
phrases show semantic agreement, and therefore their theory of syntactic resolution cannot<br />
capture all the agreement patterns <strong>in</strong> MG.<br />
If both animate and <strong>in</strong>animate nouns are seen as com<strong>in</strong>g under the purview of the<br />
marker set approach <strong>in</strong> order to derive resolution apply<strong>in</strong>g the union operation, it is nec-
4.5 An Approach to Gender <strong>Agreement</strong> <strong>in</strong> MG 103<br />
essary to employ two different groups of marker sets for the animate and <strong>in</strong>animate nouns.<br />
In MG <strong>in</strong>animate nouns, we need to generate the resolution patterns <strong>in</strong> table (4.15) 18 :<br />
Table 4.15: <strong>Agreement</strong> of Inanimate SG/PL Coord<strong>in</strong>ate <strong>Noun</strong>s<br />
S<strong>in</strong>gular Conjuncts<br />
NP1 NP2 AP NP1 NP2 AP<br />
msg msg mpl>npl msg fsg npl>mpl/fpl<br />
fsg fsg fpl>npl msg nsg npl<br />
nsg nsg npl fsg nsg npl<br />
Plural Conjuncts<br />
NP1 NP2 AP NP1 NP2 AP<br />
mpl mpl mpl mpl fpl mpl>npl>fpl<br />
fpl fpl fpl mpl npl npl<br />
npl npl npl fpl npl npl<br />
The follow<strong>in</strong>g marker sets for masc, fem and neut <strong>in</strong>animate nouns are assigned:<br />
(102) Inanimate <strong>Noun</strong>s<br />
masc{m}<br />
fem{f}<br />
neut{m,f}<br />
Once the union operation is applied, the assignments above predict the expected resolution<br />
facts <strong>in</strong> table (4.15), but not the unexpected ones:<br />
(103) Union Operation <strong>in</strong> Inanimate <strong>Noun</strong>s<br />
{m}(MASC) ∪ {m}(MASC) = {m}(MASC)<br />
{f}(FEM) ∪ {f}(FEM) = {f}(FEM)<br />
{m,f}(NEUT) ∪ {m,f}(NEUT) = {m,f}(NEUT)<br />
{m}(MASC) ∪ {f}(FEM) = {m,f}(NEUT)<br />
{m}(MASC) ∪ {m,f}(NEUT) = {m,f}(NEUT)<br />
{f}(FEM) ∪ {m,f}(NEUT) = {m,f}(NEUT)<br />
The resolution patterns of animate nouns are different from those of <strong>in</strong>animate nouns<br />
and we need to assume different marker sets accord<strong>in</strong>g to the patterns <strong>in</strong> table (4.16),<br />
repeated from above:<br />
18 As already discussed above, the unexpected fem<strong>in</strong><strong>in</strong>e plural agreement occurs <strong>in</strong> <strong>in</strong>animate abstract<br />
s<strong>in</strong>gular and plural nouns only.
4.5 An Approach to Gender <strong>Agreement</strong> <strong>in</strong> MG 104<br />
Table 4.16: <strong>Agreement</strong> of Animate SG/PL Coord<strong>in</strong>ate <strong>Noun</strong>s<br />
S<strong>in</strong>gular Conjuncts<br />
NP1 NP2 AP NP1 NP2 AP<br />
msg msg mpl msg fsg mpl<br />
fsg fsg fpl msg nm/f /0sg mpl<br />
nsg nsg npl fsg nm/0sg mpl<br />
fsg nf sg fpl/(mpl)<br />
Plural Conjuncts<br />
NP1 NP2 AP NP1 NP2 AP<br />
mpl mpl mpl mpl fpl mpl<br />
fpl fpl fpl mpl nm/f /0pl mpl<br />
npl npl npl fpl nm/0pl mpl/npl<br />
fpl nf pl fpl/(mpl)/npl<br />
The marker sets below can produce the desired results for the animate nouns:<br />
(104) Animate <strong>Noun</strong>s<br />
masc{f,n}<br />
fem{f}<br />
neut{n}<br />
The <strong>in</strong>put is aga<strong>in</strong> the successful production of the expected resolution patterns but it<br />
is impossible to account for the semantic pattern [fpl & nf pl=fpl] and the unexpected<br />
patterns:<br />
(105) Union Operation <strong>in</strong> Animate <strong>Noun</strong>s<br />
{f,n} (MASC) ∪ {f,n}(MASC)= {f,n}(MASC)<br />
{f} (FEM) ∪ {f} (FEM) = {f} (FEM)<br />
{n} (NEUT) ∪ {n}(NEUT) = {n}(NEUT)<br />
{f,n} (MASC) ∪ {f}(FEM) = {f,n}(MASC)<br />
{f,n} (MASC) ∪ {n}(NEUT) = {f,n}(MASC)<br />
{f} (FEM) ∪ {n}(NEUT) = {f,n}(MASC)<br />
Several drawbacks result from the use of different marker sets. Firstly, the separate<br />
marker sets are not economical, as would be desirable. Also, the marker sets assumed for<br />
animate nouns cannot capture all the semantic patterns. The noun comb<strong>in</strong>ations [fsg &<br />
nf sg] and [fpl & nf pl] are not captured and therefore additional constra<strong>in</strong>ts are required<br />
to the exist<strong>in</strong>g PS rule, as <strong>in</strong> Slovene. In addition, <strong>in</strong> both animate and <strong>in</strong>animate nouns<br />
none of the unexpected patterns are accounted for by the marker sets assumed and therefore<br />
further constra<strong>in</strong>ts must be imposed so as to capture the unexpected patterns successfully.<br />
Therefore, such an approach is undesirable.<br />
Wechsler and Zlatic’s (2003) proposal also seems problematic for the gender agreement<br />
patterns found <strong>in</strong> MG. Although it accounts for resolution <strong>in</strong> animate nouns accord<strong>in</strong>g<br />
to the semantic pr<strong>in</strong>ciple and resolution <strong>in</strong> <strong>in</strong>animate nouns accord<strong>in</strong>g to the syntactic<br />
pr<strong>in</strong>ciple, their analysis seems to work only for languages with common resolution patterns<br />
<strong>in</strong> animate and <strong>in</strong>animate nouns and not for languages with different resolution patterns<br />
<strong>in</strong> the two groups of nouns like MG. Follow<strong>in</strong>g Wechsler and Zlatić (2003), we assume the<br />
logic of distributivity <strong>in</strong> MG animate nouns and we assign the semantic values ‘female’ for<br />
a female group where all members are female and ‘non-female’ for a male group or a mixed<br />
gender group. E-gender nouns, like neuter, do not have any semantic correlates and are
4.5 An Approach to Gender <strong>Agreement</strong> <strong>in</strong> MG 105<br />
exempt from the logic of distributivity. This assignment would give the desired results <strong>in</strong><br />
animate nouns.<br />
Let us turn to MG <strong>in</strong>animate nouns which resolve to neuter gender <strong>in</strong> mixed gender<br />
conjuncts. If we assign the grammatical gender values of <strong>in</strong>animates based on the correspond<strong>in</strong>g<br />
semantic values of animates, then each positively specified grammatical gender<br />
such as fem<strong>in</strong><strong>in</strong>e is represented as {f} and each negatively specified gender such as nonfem<strong>in</strong><strong>in</strong>e<br />
(i.e. mascul<strong>in</strong>e) is represented as a null set {} and neuter nouns are represented<br />
as the s<strong>in</strong>gleton set {n}. To calculate the gender of <strong>in</strong>animate nouns we need to remove the<br />
e-gender nouns by <strong>in</strong>tersect<strong>in</strong>g the calculated set with the semantic genders (s-genders).<br />
This would result <strong>in</strong> the desired patterns <strong>in</strong> coord<strong>in</strong>ate structures with conjuncts which are<br />
either both mascul<strong>in</strong>e or fem<strong>in</strong><strong>in</strong>e, such as <strong>in</strong> {f}∩{f} and {}∩{}. However, <strong>in</strong> coord<strong>in</strong>ate<br />
structures with neuter conjuncts or mixed gender conjuncts as <strong>in</strong> {}∩{f} or {}∩{n} the<br />
result is mascul<strong>in</strong>e and not neuter as required. This is illustrated as follows:<br />
(106) MG<br />
a. Set of s-genders<br />
Gs ={f}<br />
b. Set representations of the s-genders<br />
fem:{f} (
4.5 An Approach to Gender <strong>Agreement</strong> <strong>in</strong> MG 106<br />
The above rank<strong>in</strong>g would account for the different gender patterns found <strong>in</strong> animate and<br />
<strong>in</strong>animate nouns because the feature [Human] determ<strong>in</strong>es resolution only <strong>in</strong> mixed gender<br />
conjuncts and also the ident constra<strong>in</strong>t ∀c[identG(c,ph)], which accounts for identical<br />
gender conjuncts, is placed higher than the rest of the constra<strong>in</strong>ts which refer to gender.<br />
In mixed gender cases, the ident constra<strong>in</strong>t ∀c[identG(c,ph)] will be violated and then<br />
the gender choice will be determ<strong>in</strong>ed by shift<strong>in</strong>g from the least marked gender (neuter) to<br />
the next least marked (mascul<strong>in</strong>e) when the conjo<strong>in</strong>ed phrase is [+Human]. However, the<br />
above rank<strong>in</strong>g is unable to capture the pattern [fsg & nf sg=fpl] or [fpl & nf pl=fpl] <strong>in</strong><br />
animate nouns s<strong>in</strong>ce the optimal form would necessarily be mpl and not fpl because the<br />
constra<strong>in</strong>t *F is placed higher <strong>in</strong> the rank<strong>in</strong>g and not lower. In addition, Badecker (2008)<br />
argues that <strong>in</strong>animate s<strong>in</strong>gular coord<strong>in</strong>ate nouns with a mascul<strong>in</strong>e and a fem<strong>in</strong><strong>in</strong>e conjunct<br />
resolve to neuter and the correspond<strong>in</strong>g plural ones resolve to mascul<strong>in</strong>e. Our data showed<br />
that such a difference does not necessarily occur s<strong>in</strong>ce the consultants <strong>in</strong> coord<strong>in</strong>ate phrases<br />
of both numbers pick up the same resolution genders (i.e. neuter and mascul<strong>in</strong>e) but with<br />
greater preference to one or to the other possibly related to the number of conjuncts. Such<br />
a preference cannot be stated <strong>in</strong> terms of the OT ranked constra<strong>in</strong>st s<strong>in</strong>ce OT assumes<br />
only one optimal form. With regard to the unexpected patterns, Badecker’s OT analysis<br />
cannot capture these patterns too. Many different constra<strong>in</strong>ts need to be imposed and for<br />
a theory like OT, which admits only the optimal form, the several agreement choices of the<br />
exceptional referential patterns are difficult to account for.<br />
F<strong>in</strong>ally, the major drawback of Sadler’s (2006) analysis is that it cannot accommodate<br />
the MG data due to the wide range of unexpected resolution patterns which are not either<br />
syntactically or semantically motivated but referentially determ<strong>in</strong>ed. Also, it might be seen<br />
as a disadvantage the use of the additional feature semgend with values male and female<br />
s<strong>in</strong>ce it may not be required by languages that do not impose the dist<strong>in</strong>ction male/female<br />
but other dist<strong>in</strong>ctions, such as non-/rational, as <strong>in</strong> Dravidian languages. In the last case,<br />
we should postulate a different feature with different values too which might result to<br />
redundancies <strong>in</strong> the grammar. Even though her analysis could be seen as language specific,<br />
it can be easily assumed for other languages s<strong>in</strong>ce the standard PS rule is ma<strong>in</strong>ta<strong>in</strong>ed <strong>in</strong><br />
<strong>in</strong>animate nouns, which follow syntactic resolution, and we can <strong>in</strong>troduce any changes <strong>in</strong><br />
the semantic rule of animate nouns, which may differ across languages. We will attempt a<br />
similar type of analysis for the expected cases of MG.<br />
To conclude, we presented certa<strong>in</strong> problems of the theories on gender resolution with<br />
regard to the MG data, and we argue that none of them fully captures the range of<br />
resolution results that are found <strong>in</strong> MG and especially those resolution patterns that are<br />
derived referentially. Next we turn to our analysis.<br />
4.5.2 An Analysis of the MG Expected Patterns<br />
The data analysis showed that the expected patterns of MG animate and <strong>in</strong>animate coord<strong>in</strong>ates<br />
differ <strong>in</strong> the resolved gender value and the gender resolution pr<strong>in</strong>ciple. Animate<br />
coord<strong>in</strong>ate nouns with mixed gender conjuncts resolve to mascul<strong>in</strong>e unless all conjuncts<br />
denote females <strong>in</strong> which case resolution is fem<strong>in</strong><strong>in</strong>e. If conjuncts have the same gender they<br />
resolve to that shared gender, irrespective of the referents of the nouns. Thus, animate<br />
coord<strong>in</strong>ate nouns show semantic resolution. Inanimate coord<strong>in</strong>ate phrases with mixed gender<br />
conjuncts resolve to neuter. If the conjuncts share the same gender then they resolve<br />
to that shared gender. Thus, <strong>in</strong>animate coord<strong>in</strong>ate nouns follow the syntactic pr<strong>in</strong>ciple of<br />
resolution.<br />
Accord<strong>in</strong>g to Kazana (2007), the first task is to propose a unified solution for both<br />
groups of nouns <strong>in</strong> order to avoid problems of redundancy. To achieve that we need to<br />
employ the same marker sets for both animate and <strong>in</strong>animate nouns and use the set-union<br />
operation to derive the required resolution results. Assum<strong>in</strong>g the marker set decomposition,
4.5 An Approach to Gender <strong>Agreement</strong> <strong>in</strong> MG 107<br />
we represent the genders for animate and <strong>in</strong>animate nouns with the same group of sets.<br />
The sets assumed for both groups of nouns are those of the <strong>in</strong>animate nouns s<strong>in</strong>ce neuter is<br />
the resolved gender for the <strong>in</strong>animate nouns and also neuter occurs as the resolved gender<br />
<strong>in</strong> animate neuter nouns irrespective of the referents. Therefore, the marker sets used to<br />
derive the syntactic gender <strong>in</strong> <strong>in</strong>animate nouns can equally be adopted for the derivation of<br />
both the semantic patterns and the patterns that show feature match<strong>in</strong>g <strong>in</strong> animate nouns.<br />
The marker sets are the follow<strong>in</strong>g:<br />
(108) Marker Sets for Animate & Inanimate <strong>Noun</strong>s<br />
masc{m}<br />
fem {f}<br />
neut {m,f}<br />
(Kazana, 2007)<br />
Inanimate coord<strong>in</strong>ate nouns are straightforwardly derived by the above assumption<br />
account<strong>in</strong>g for all the expected non-exceptional cases. The <strong>in</strong>animate expected agreement<br />
patterns are repeated below:<br />
Table 4.17: <strong>Agreement</strong> of Inanimate Coord<strong>in</strong>ate <strong>Noun</strong>s<br />
S<strong>in</strong>gular or Plural Conjuncts<br />
NP1 NP2 AP NP1 NP2 AP<br />
msg/pl msg/pl mpl msg/pl fsg/pl npl<br />
fsg/pl fsg/pl fpl msg/pl nsg/pl npl<br />
nsg/pl nsg/pl npl fsg/pl nsg/pl npl<br />
When the specific marker sets are comb<strong>in</strong>ed through set-union the result is the neuter<br />
gender <strong>in</strong> all mixed gender coord<strong>in</strong>ate nouns and the gender of the conjuncts <strong>in</strong> all uniform<br />
gender coord<strong>in</strong>ate nouns:<br />
(109) Union Operation <strong>in</strong> Inanimate <strong>Noun</strong>s<br />
{m}(MASC) ∪ {m}(MASC) = {m}(MASC)<br />
{f}(FEM) ∪ {f}(FEM) = {f}(FEM)<br />
{m,f}(NEUT) ∪ {m,f}(NEUT) = {m,f}(NEUT)<br />
{m}(MASC) ∪ {f}(FEM) = {m,f}(NEUT)<br />
{m}(MASC) ∪ {m,f}(NEUT) = {m,f}(NEUT)<br />
{f}(FEM) ∪ {m,f}(NEUT) = {m,f}(NEUT)<br />
(Kazana, 2007)<br />
The LFG phrase structure rule that captures the necessary subset relations is the standard<br />
syntactic rule of coord<strong>in</strong>ation proposed by Dalrymple and Kaplan (2000). However,<br />
follow<strong>in</strong>g Sadler (2006, 14) we also assume that nouns have the lexical specification anim<br />
+/−. Thus, the standard syntactic rule of coord<strong>in</strong>ation is assumed:<br />
(110) Syntactic Rule - Inanimate <strong>Noun</strong>s<br />
NP → NP CONJ NP<br />
↓∈↑ ↓∈↑<br />
(↓anim)= − (↓anim)= −<br />
(↓ gend) ⊆ (↑gend) (↓gend) ⊆ (↑gend)<br />
(Kazana, 2007)<br />
This rule states that <strong>in</strong> a coord<strong>in</strong>ate phrase each NP conjunct daughter needs to be a<br />
member of the conjunct set of the mother node. In the second l<strong>in</strong>e, it is specified that each
4.5 An Approach to Gender <strong>Agreement</strong> <strong>in</strong> MG 108<br />
conjunct daughter must be <strong>in</strong>animate while the last l<strong>in</strong>e states that the set value of each<br />
conjunct daughter needs to be a subset of the set value of the mother node, def<strong>in</strong>ed as the<br />
smallest set. Thus, the expected resolution patterns <strong>in</strong> MG <strong>in</strong>animate nouns are captured<br />
us<strong>in</strong>g closed sets for the gender features and they undergo resolution us<strong>in</strong>g set union.<br />
The same marker sets are also assumed to account for the resolution patterns <strong>in</strong> MG<br />
animate nouns. The expected agreement patterns are repeated below:<br />
Table 4.18: <strong>Agreement</strong> of Animate Coord<strong>in</strong>ate <strong>Noun</strong>s<br />
S<strong>in</strong>gular or Plural Conjuncts<br />
NP1 NP2 AP NP1 NP2 AP<br />
msg/pl msg/pl mpl msg/pl fsg/pl mpl<br />
fsg/pl fsg/pl fpl msg/pl nm/f /0sg/pl mpl<br />
nsg/pl nsg/pl npl fsg/pl nm/0sg/pl mpl<br />
fsg/pl nf sg/pl fpl/(>mpl)<br />
Any attempt to apply these marker sets to capture the expected cases results partly <strong>in</strong><br />
the wrong gender feature value s<strong>in</strong>ce the desired gender feature for the mixed gender cases<br />
is the mascul<strong>in</strong>e and not the neuter one, which is generated, as shown below:<br />
(111) Union Operation <strong>in</strong> Animate <strong>Noun</strong>s<br />
{m}(MASC) ∪ {m}(MASC) = {m}(MASC)<br />
{f}(FEM) ∪ {f}(FEM) = {f}(FEM)<br />
{m,f}(NEUT) ∪ {m,f}(NEUT) = {m,f}(NEUT)<br />
{m}(MASC) ∪ {f}(FEM) = {m,f}(NEUT)*<br />
{m}(MASC) ∪ {m,f}(NEUT) = {m,f}(NEUT)*<br />
{f}(FEM) ∪ {m,f}(NEUT) = {m,f}(NEUT)*<br />
Thus, set union cannot be used for the resolution patterns of animate nouns.<br />
In order to overcome this problem, we need to employ two additional phrase structure<br />
rules for the patterns of animate nouns without us<strong>in</strong>g set union for these cases. The first<br />
rule should <strong>in</strong>corporate the necessary specifications to capture coord<strong>in</strong>ations with nouns<br />
that share the same marker sets. This rule should postulate the requirement that if all<br />
daughters share the same marker set, then use that marker set. This is formalised as<br />
follows:<br />
(112) Same Gender Rule - Animate <strong>Noun</strong>s<br />
NP → NP CONJ NP<br />
↓∈↑ ↓∈↑<br />
(↓anim)=+ (↓anim)=+<br />
(↑gend)= (↓gend) (↑gend)=(↓gend)<br />
(Kazana, 2007)<br />
In the above rule, apart from the set membership requirement, two additional specifications<br />
are <strong>in</strong>troduced. The first one requires the conjuncts to be lexically specified as<br />
anim +, and the second requires the conjunct daughters to share the same gender with<br />
the mother node.<br />
The second rule is more complicated s<strong>in</strong>ce it should capture coord<strong>in</strong>ate phrases with<br />
nouns that have different marker sets. In particular, we observed that if the genders of<br />
the conjuncts are different, then we need to use mascul<strong>in</strong>e ({m}) or optionally if conjuncts<br />
denote female <strong>in</strong>dividuals then we need to use fem<strong>in</strong><strong>in</strong>e ({f}). To achieve that we need<br />
to employ the extra feature semgend with values male and female, proposed by Sadler
4.5 An Approach to Gender <strong>Agreement</strong> <strong>in</strong> MG 109<br />
(2006). Thus, the lexical entries that refer to male <strong>in</strong>dividuals are lexically specified as<br />
(↑semgend)=male and those that refer to female <strong>in</strong>dividuals are lexically specified as<br />
(↑semgend)= female. The rule is formalised as follows:<br />
(113) Mixed Gender Rule - Animate <strong>Noun</strong>s<br />
NP → NP CONJ NP<br />
↓∈↑ ↓∈↑<br />
(↓anim)=+ (↓anim)=+<br />
((↑∈)gend)=(↓gend)<br />
{(↑gend)= {m}<br />
∨<br />
[(↑gend)= {f}<br />
¬(↑∈semgend)= female]}<br />
The first two parts of the above rule impose the set membership requirement and<br />
the specification that the nouns should be lexically specified as anim +. The third l<strong>in</strong>e<br />
postulates the restriction that the daughters have dist<strong>in</strong>ct marker sets, referr<strong>in</strong>g to mixed<br />
gender coord<strong>in</strong>ations. Then, two possibilites arise, either resolution is mascul<strong>in</strong>e ({m}),<br />
or resolution is fem<strong>in</strong><strong>in</strong>e ({f}) <strong>in</strong> which case all of the conjunct daughters should denote<br />
a female <strong>in</strong>dividual. In the last l<strong>in</strong>e, negation is expressed universally given a wide scope<br />
<strong>in</strong>terpretation:<br />
(114) ¬(↑∈semgend)= female<br />
This means that there is no member of the set whose value is non-female or else for all<br />
daughters the value is constra<strong>in</strong>ed to be female.<br />
To summarise, <strong>in</strong> this section we presented one PS rule that accounts for the expected<br />
resolution patterns <strong>in</strong> <strong>in</strong>animate nouns, us<strong>in</strong>g the crossl<strong>in</strong>guistic set-union approach, and<br />
two separate PS rules that account for the expected resolution patterns <strong>in</strong> animate nouns,<br />
<strong>in</strong>troduc<strong>in</strong>g specifications for semantic resolution.<br />
Our approach, which resembles that of Sadler (2006), has a number of advantages and<br />
disadvantages. The ma<strong>in</strong> advantage is that it captures all the expected patterns <strong>in</strong> animate<br />
and <strong>in</strong>animate nouns us<strong>in</strong>g the same marker sets for both groups of nouns. Also, it uses<br />
the standard PS rule of coord<strong>in</strong>ate structures regard<strong>in</strong>g the gender feature alongside the<br />
widely adopted set-union operation <strong>in</strong> <strong>in</strong>animate nouns. It postulates separate PS rules<br />
only for animate nouns whose resolution patterns are language specific.<br />
A problematic aspect is that we also employ the extra feature semgend <strong>in</strong> the fstructure.<br />
Although this is not advantageous for LFG as it is an extra feature, it turns<br />
out to be useful s<strong>in</strong>ce it is a placeholder for a semantic treatment of the present resolution<br />
rules s<strong>in</strong>ce it captures the dist<strong>in</strong>ction between male and female denotations of nouns.<br />
The most important problem, however, is that it does not capture the exceptional patterns<br />
<strong>in</strong> both groups of nouns. A different approach needs to be assumed for these unexpected<br />
patterns and this is where we will turn to next.<br />
4.5.3 An Approach to the MG Referential <strong>Agreement</strong> Patterns<br />
So far we presented an analysis of the expected patterns of gender resolution <strong>in</strong> coord<strong>in</strong>ate<br />
phrases. The unexpected patterns cannot be captured by the above proposal s<strong>in</strong>ce they are<br />
referentially motivated based on discourse conditions. Thus, we will propose an analysis<br />
tak<strong>in</strong>g <strong>in</strong>to account the discourse conditions imposed by the native speakers.<br />
The central idea beh<strong>in</strong>d the phenomenon of referential agreement <strong>in</strong> MG <strong>in</strong>volves SOME
4.5 An Approach to Gender <strong>Agreement</strong> <strong>in</strong> MG 110<br />
implied noun 19 brought <strong>in</strong>to the discourse by native speakers. The speakers when utter<strong>in</strong>g<br />
a coord<strong>in</strong>ate phrase modified by a predicative adjective/participle do not necessarily<br />
obey the syntactic or semantic expected patterns of resolution, but they follow referential<br />
agreement by implicitly <strong>in</strong>troduc<strong>in</strong>g <strong>in</strong>to the context a specific noun with its own morphosyntactic<br />
features. Such a noun will be referred to as the Contextually Introduced<br />
Referent 20 . The agreement target (i.e. the predicative adjective or participle) is forced<br />
to agree with the morphosyntactic features (i.e. gender and number) of the Contextually<br />
Introduced Referent, which is implicitly present <strong>in</strong> the sentence and which also functions as<br />
the superord<strong>in</strong>ate/hypernymic term of the coord<strong>in</strong>ate nouns. In MG, predicative adjectives<br />
agree grammatically with the subject of the sentence and occur <strong>in</strong> copular constructions.<br />
For the analysis of the MG copular constructions, we will follow the open function xcomp<br />
analysis 21 proposed by Dalrymple (2001) and Dalrymple et al. (2004a). This analysis assumes<br />
that the copular verb <strong>in</strong>e ‘be’ subcategorises for an xcomp only, whereas the subj<br />
occurs outside the angled brackets <strong>in</strong> the semantic form and it is not a semantic argument<br />
of the verb. This type of analysis is preferred for languages <strong>in</strong> which the postcopular complement<br />
shows agreement with the subject of the copula, as <strong>in</strong> MG.<br />
Let us first consider an expected case of gender agreement. The follow<strong>in</strong>g example<br />
consists of two s<strong>in</strong>gular mascul<strong>in</strong>e nouns that syntactically resolve to mascul<strong>in</strong>e:<br />
(115) O p<strong>in</strong>akas<br />
the.masc.sg picture.masc.sg<br />
<strong>in</strong>e megali<br />
are.pl big.masc.pl<br />
‘The picture and the sofa are big’<br />
ke<br />
and<br />
The f-structure of the sentence is presented below:<br />
o<br />
the.masc.sg<br />
kanapes<br />
sofa.masc.sg<br />
19 We use capital letters s<strong>in</strong>ce the implied noun assumed by native speakers may vary from speaker to<br />
speaker but the agreement features of different nouns may co<strong>in</strong>cide.<br />
20 We will briefly refer to the Contextually Introduced Referent by the acronym CIR.<br />
21 Dalrymple et al. (2004a) also propose the closed complement predl<strong>in</strong>k analysis which is required<br />
when the postcopular element already has a subject. Accord<strong>in</strong>g to Dalrymple et al. (2004a, 193) such<br />
constructions are that-clauses, certa<strong>in</strong> gerunds and modal uses of the copula with null pronom<strong>in</strong>al subjects.
4.5 An Approach to Gender <strong>Agreement</strong> <strong>in</strong> MG 111<br />
(116) ⎡<br />
⎤<br />
pred ‘be subj’<br />
⎢ ⎡⎧⎡<br />
⎤⎫⎤⎥<br />
⎢ ⎢<br />
pred ‘picture’<br />
⎥<br />
⎢ ⎢ ⎢<br />
⎥ ⎥⎥<br />
⎢ ⎢ ⎢def<br />
+ ⎥ ⎥⎥<br />
⎢ ⎢ ⎢<br />
⎥ ⎥⎥<br />
⎢ ⎢ ⎢<br />
⎢ ⎢ ⎢person<br />
3 ⎥ ⎥⎥<br />
⎥ ⎥⎥<br />
⎢ ⎢ ⎢<br />
⎢ ⎢ ⎣num<br />
sg<br />
⎥ ⎥⎥<br />
⎦ ⎥⎥<br />
⎢ ⎢⎪⎨<br />
⎪⎬<br />
⎥⎥<br />
⎢ ⎢ gend masc<br />
⎥⎥<br />
⎥⎥<br />
⎢ ⎢ ⎡ ⎤ ⎥⎥<br />
⎢ ⎢<br />
⎢ ⎢ pred ‘sofa’<br />
⎥⎥<br />
⎥⎥<br />
⎢ ⎢ ⎢ ⎥ ⎥⎥<br />
⎢ ⎢ ⎢def<br />
+ ⎥ ⎥⎥<br />
⎢subj<br />
⎢ ⎢ ⎥ ⎥⎥<br />
⎢ ⎢ ⎢<br />
⎢ ⎢ ⎢person<br />
3 ⎥ ⎥⎥<br />
⎥ ⎥⎥<br />
⎢ ⎢ ⎢<br />
⎢ ⎢ ⎣num<br />
sg<br />
⎥ ⎥⎥<br />
⎦ ⎥⎥<br />
⎢ ⎢<br />
⎢ ⎢<br />
⎪⎩<br />
⎪⎭<br />
⎥⎥<br />
gend masc ⎥⎥<br />
⎢ ⎢<br />
⎥⎥<br />
⎢ ⎢<br />
⎥⎥<br />
⎢ ⎢person<br />
3 ⎥⎥<br />
⎢ ⎢<br />
⎥⎥<br />
⎢ ⎢<br />
⎢ ⎢num<br />
pl ⎥⎥<br />
⎥⎥<br />
⎢ ⎢<br />
⎢ ⎣gend<br />
masc<br />
⎥⎥<br />
⎦ ⎥<br />
⎢ conj and<br />
⎥<br />
⎢ ⎡<br />
⎤⎥<br />
⎢ pred ‘big ’ ⎥<br />
⎣xcomp<br />
⎣ <br />
⎥<br />
⎦<br />
subj<br />
In the f-structure, the copular verb <strong>in</strong>e ‘be’ subcategorises for an open complement<br />
xcomp, which is the adjective megali ‘big’. S<strong>in</strong>ce the latter is predicative, it should behave<br />
as verbs do. Thus, it requires a subj, which is the same as the subj of the verb <strong>in</strong>e<br />
‘be’. This subj is the coord<strong>in</strong>ate phrase and it has a plural number and a masc gender,<br />
which is the resolved gender of the two conjuncts. The above syntactic <strong>in</strong>formation can be<br />
captured <strong>in</strong> the lexical entry of the verb <strong>in</strong>e ‘be’, illustrated below:<br />
(117) <strong>in</strong>e v (↑pred) = ‘be〈xcomp〉subj’<br />
(↑subj) = (↑xcomp subj)<br />
S<strong>in</strong>ce the subject of the verb ‘be’ is structure shared with the subject of the predicative<br />
adjective, the lexical entry of the latter should also specify the follow<strong>in</strong>g <strong>in</strong>formation:<br />
(118) megali (↑pred) = ‘big〈subj〉’<br />
(↑subj num)= pl<br />
(↑subj gend)= masc<br />
Thus, <strong>in</strong> the expected pattern of gender resolution the adjective megali ‘big’ agrees with<br />
the syntactic controller which is the subj coord<strong>in</strong>ate noun phrase. The above analysis<br />
works effectively for the expected patterns <strong>in</strong> MG.<br />
The same predicative adjective, however, also presents an alternative unexpected case<br />
of gender agreement, which has the same agreement features with those of SOME implied<br />
noun that occurs <strong>in</strong> the context. For example, <strong>in</strong> (119) the adjective shows unexpected<br />
neuter agreement, which might be the gender of the contextually implied noun pragmata<br />
‘th<strong>in</strong>gs’:<br />
(119) O p<strong>in</strong>akas<br />
the.masc.sg picture.masc.sg<br />
<strong>in</strong>e megala<br />
are.pl big.neut.pl<br />
‘The picture and the sofa are big’<br />
ke<br />
and<br />
o<br />
the.masc.sg<br />
kanapes<br />
sofa.masc.sg
4.5 An Approach to Gender <strong>Agreement</strong> <strong>in</strong> MG 112<br />
Therefore, the unexpected patterns require the predicative adjective, which is the agreement<br />
target, to display agreement features that do not agree with the syntactic controller,<br />
which is the resolved gender of the coord<strong>in</strong>ate phrase that the adjective is a predicate of,<br />
but with a contextual controller, the Contextually Introduced Referent.<br />
In order to capture the above phenomenon, we will assume that there is an extra level<br />
of representation, the Contextual level. At the Contextual level, the CIR <strong>in</strong>troduces a<br />
r(eferential)-structure with agreement feature-values that the predicative adjective shows<br />
agreement. The correspondence functions assumed with<strong>in</strong> LFG can be used to relate the<br />
Contextual level and the syntactic level as follows:<br />
Let P be the Contextually Introduced Referent and r be the f-structure of<br />
P <strong>in</strong>troduced at the Contextual level of representation C. The syntactic fstructure<br />
level of representation is mapped to the contextual level of representation<br />
through the mapp<strong>in</strong>g function d. The <strong>in</strong>verse mapp<strong>in</strong>g function d −1<br />
also holds.<br />
Thus, any f-structure f of coord<strong>in</strong>ated NPs at the syntactic level can be associated to<br />
an r-structure r <strong>in</strong>troduced at the contextual level impos<strong>in</strong>g the follow<strong>in</strong>g correspondence<br />
relation: d(f)= r. The <strong>in</strong>verse correpondence relation d −1 (r)= f will also hold denot<strong>in</strong>g<br />
the f-structure from the contextual level to the syntactic level of representation. This is<br />
schematically represented as follows:<br />
f<br />
d<br />
−→ r r<br />
d −1<br />
−−→ f<br />
Figure 4.1: Mapp<strong>in</strong>g correspondence<br />
Based on example (119), the r-structure of the Contextually Introduced Referent <strong>in</strong>troduced<br />
at the contextual level C has the follow<strong>in</strong>g agreement features:<br />
⎡ ⎤<br />
(120) pred ‘th<strong>in</strong>g’<br />
⎢ ⎥<br />
r: ⎣num<br />
pl ⎦<br />
gend neut<br />
The follow<strong>in</strong>g lexical entry will hold for the CIR:<br />
(121) pragmata (r pred) = ‘th<strong>in</strong>g’<br />
(r num)= pl<br />
(r gend)= neut<br />
In the unexpected pattern, the agreement target (i.e. predicative adjective) will agree<br />
with the contextual controller, the CIR, whose agreement features are specified <strong>in</strong> the rstructure<br />
and which occurs only at the contextual level C. In example (119), the CIR is the<br />
noun pragmata ‘th<strong>in</strong>gs’ with the r-structure <strong>in</strong> (120) and with the lexical entry <strong>in</strong> (121).<br />
Thus, the adjective megala, apart from the syntactic agreement, which results from the<br />
syntactic controller (i.e. the subject noun phrase), needs to agree with the morphosyntactic<br />
features of the CIR, which is the contextual controller and has the agreement features<br />
num=pl and gend=neut, shown <strong>in</strong> the r-structure. The lexical entry of the predicative<br />
adjective that shows referential agreement is stated as follows:<br />
(122) megala (↑pred) = ‘big’<br />
(↑d num)= pl<br />
(↑d gend)= neut
4.5 An Approach to Gender <strong>Agreement</strong> <strong>in</strong> MG 113<br />
So the general idea is that the predicative adjective will either agree with a syntactic<br />
controller or with a contextual controller. The syntactic controller is the subject noun<br />
phrase which is a coord<strong>in</strong>ate structure and it triggers agreement as a result of resolution;<br />
the contextual controller is the Contextually Introduced Referent which is some implied<br />
noun that is brought <strong>in</strong>to the discourse by the speakers and it occurs at the contextual<br />
level of representation.<br />
As proposed above, there is a mapp<strong>in</strong>g function d <strong>in</strong> the f-to-r direction and an <strong>in</strong>verse<br />
mapp<strong>in</strong>g function d −1 <strong>in</strong> the r-to-f direction. Then a relationship must hold between f,<br />
which is the f-structure of (↑subj), and r, the r-structure of CIR. This is stated as follows:<br />
(123) d (↑subj) = r<br />
Consider<strong>in</strong>g the equation <strong>in</strong> (123), we propose the follow<strong>in</strong>g lexical entry for the expected<br />
and the unexpected patterns of the predicative adjective/participle:<br />
(124) megala (↑pred) = ‘big〈subj〉’<br />
{(↑subj num)= pl<br />
(↑subj gend)= neut|<br />
((↑subj)d num)= pl<br />
((↑subj)d gend)= neut}<br />
In (124), the predicative adjective <strong>in</strong>troduces a disjunctive lexical entry. The first part<br />
specifies the expected agreement features as a result of syntactic resolution, and the second<br />
part <strong>in</strong>troduces the unexpected agreement features as a result of referential agreement<br />
with the agreement features of the CIR, which is the hypernym of the two conjuncts.<br />
The disjunctive lexical entry of the predicative adjective captures the ma<strong>in</strong> idea of the<br />
phenomenon of referential agreement <strong>in</strong> MG coord<strong>in</strong>ate noun phrases.<br />
There are some challenges that need to be discussed. The first issue concerns the status<br />
of the r-structure which occurs at the contextual level but it is represented as a syntactic<br />
structure. The r-structure, which is a contextual level structure, must be represented as an<br />
f-structure s<strong>in</strong>ce it holds all the syntactic features of the implied noun. However, it has to<br />
occur only at the contextual level and not at the syntactic level s<strong>in</strong>ce it is not syntactically<br />
present <strong>in</strong> the phrase but it is only implied by the native speakers. In addition, the pred<br />
value <strong>in</strong> the r-structure is the semantic form of some noun brought <strong>in</strong>to the context by the<br />
native speakers with a unique <strong>in</strong>stantiation. This pred value needs to be present s<strong>in</strong>ce it is<br />
a specific noun each time which functions as the superord<strong>in</strong>ate term of the two coord<strong>in</strong>ate<br />
nouns of the sentence, although it differs from speaker to speaker depend<strong>in</strong>g on speakers’<br />
perception.<br />
The second issue relates to the requirement of captur<strong>in</strong>g the semantic relation of the<br />
subj noun phrase, which occurs at the syntactic level and functions as the hyponym,<br />
with the CIR, which occurs at the contextual level and functions as the hypernym of the<br />
coord<strong>in</strong>ate subj noun phrase. This semantic relation between the Contextually Introduced<br />
Referent and the coord<strong>in</strong>ate nouns is a hypernymic-hyponymic one and could be expressed<br />
with<strong>in</strong> the lexicon <strong>in</strong> the form of a hierarchy.<br />
Constra<strong>in</strong>t-based theories like LFG and HPSG assume the template hierarchy and the<br />
type hierarchy, respectively. In LFG, the template hierarchy encodes lexical generalisations<br />
<strong>in</strong> terms of relations between descriptions of structures and not <strong>in</strong> terms of any<br />
formal <strong>in</strong>heritance relations between types. These templates are f-structure descriptions<br />
with a name which can be used <strong>in</strong> other equations to state more complicated descriptions<br />
(Dalrymple et al., 2004b, 201). In LFG, these hierarchies seem to be abbreviatory. On<br />
the other hand, HPSG builds l<strong>in</strong>guistic generalisations also us<strong>in</strong>g a type hierarchy where<br />
more specific types <strong>in</strong>herit their <strong>in</strong>formation from less specific but related types. In HPSG,
4.5 An Approach to Gender <strong>Agreement</strong> <strong>in</strong> MG 114<br />
lexical generalizations are statable as relations between elements <strong>in</strong> the type lattice, where<br />
different subtypes function as alternatives and a specific type may belong to many supertypes<br />
(Malouf, 1998; Pollard and Sag, 1994). Thus, <strong>in</strong> HPSG these hierarchies relate<br />
ontological objects of the theory and do not simply state abbreviations.<br />
In order to capture the semantic relation of the lexical items, we could assume an<br />
ontological hierarchy resembl<strong>in</strong>g those of LFG or HPSG, which will show how the various<br />
lexical items are related to each other where the more general term <strong>in</strong>cludes with<strong>in</strong> it the<br />
more specific ones. This hierarchy could be represented as follows for the relevant example<br />
<strong>in</strong> (119):<br />
THINGS<br />
DECORATION FURNITURE<br />
PAINTINGS LIGHTS TABLES<br />
Figure 4.2: A simple hierarchy represent<strong>in</strong>g ontological relations<br />
In the hierarchy above, the superord<strong>in</strong>ate term is the noun ‘th<strong>in</strong>gs’ which <strong>in</strong>cludes as<br />
hyponyms the noun ‘decoration’ and the noun ‘furniture’. With<strong>in</strong> the noun ‘furniture’,<br />
which also functions as a general term, other hyponyms are <strong>in</strong>cluded like the nouns ‘lights’<br />
and ‘tables’. Also, the noun ‘lights’ is the hyponym of the general term ‘decorations’ s<strong>in</strong>ce<br />
‘lights’ could be a piece of furniture or a piece of decoration.<br />
A second issue that arises is that the representation of this hierarchy may differ from<br />
speaker to speaker depend<strong>in</strong>g on the world knowledge and perception and on the way the<br />
various lexical items are organised <strong>in</strong> each speaker’s m<strong>in</strong>d. Thus, a person may assume<br />
a specific hierarchy with a specific superord<strong>in</strong>ate term but a different person may assume<br />
a different hierarchy and therefore a different superord<strong>in</strong>ate term for the same coord<strong>in</strong>ate<br />
nouns. As seen above, most speakers chose the noun xori ‘spaces’ as the superord<strong>in</strong>ate<br />
term for the two nouns dromi ‘streets’ and squares ‘platies’ but some speakers chose the<br />
noun perioxes ‘areas’. This clearly shows that native speakers assume a different ontological<br />
representation for the same group of nouns depend<strong>in</strong>g on their world knowledge and<br />
perception.<br />
The second issue is related to the field of ontology and could be expla<strong>in</strong>ed further from<br />
that perspective. Ontology traditionally deals with questions concern<strong>in</strong>g what entities exist<br />
or can be said to exist, and how such entities can be grouped, related with<strong>in</strong> a hierarchy,<br />
and subdivided accord<strong>in</strong>g to similarities and differences (Genesereth and Nilson, 1987).<br />
Thus, ontology focuses on the relations of different concepts with<strong>in</strong> a system. One aspect<br />
of ontological relations is also that of subord<strong>in</strong>ation and superord<strong>in</strong>ation. In Schalley and<br />
Zaefferer (2007, 4), these two relations are def<strong>in</strong>ed as “weak order<strong>in</strong>gs” as follows:<br />
Conceptual subord<strong>in</strong>ation: Concept A 22 is c-subord<strong>in</strong>ated to concept B iff every<br />
<strong>in</strong>stance of A is also an <strong>in</strong>stance of B, and<br />
Conceptual superord<strong>in</strong>ation, its converse.<br />
(Schalley and Zaefferer, 2007, 4)<br />
Accord<strong>in</strong>g to the above def<strong>in</strong>ition, <strong>in</strong> example (119) and similar ones, the native speakers<br />
carry extra-l<strong>in</strong>guistic <strong>in</strong>formation which affects the current l<strong>in</strong>guistic knowledge. This<br />
<strong>in</strong>formation refers to the CIR, which is brought forward <strong>in</strong>to the specific discourse conditions,<br />
and it is usually a superord<strong>in</strong>ate concept which <strong>in</strong>cludes with<strong>in</strong> it <strong>in</strong>stances of the two<br />
coord<strong>in</strong>ate nouns. This extral<strong>in</strong>guistic concept, the CIR, has grammatical features that determ<strong>in</strong>e<br />
the gender (and subsequently number) of the predicative adjective. In most cases,<br />
22 A, B and C stand for the related concepts.
4.6 Conclusion 115<br />
native speakers br<strong>in</strong>g <strong>in</strong>to the context the same extral<strong>in</strong>guistic concept and therefore the<br />
predicative adjective shows agreement with the features of that specific concept. There are<br />
cases, however, that native speakers br<strong>in</strong>g <strong>in</strong>to the discourse different concepts and therefore<br />
there is a divergence <strong>in</strong>to the agreement features of the predicative adjective. This is<br />
more common <strong>in</strong> abstract coord<strong>in</strong>ate nouns. In the follow<strong>in</strong>g example, the native speakers<br />
chose different gender features <strong>in</strong> the predicative adjective, mascul<strong>in</strong>e and fem<strong>in</strong><strong>in</strong>e:<br />
(125) O kopos ke i prospathia<br />
the.masc.sg effort.masc.sg and the.fem.sg trial.fem.sg<br />
aparetit-i/-es ja na proodefsei kanis<br />
necessary.masc/fem.pl for to progress someone<br />
‘The effort and trial are necessary for someone to progress’<br />
The predicative adjective shows mascul<strong>in</strong>e plural agreement s<strong>in</strong>ce native speakers may<br />
br<strong>in</strong>g <strong>in</strong>to the discourse the noun tropi.masc.pl ‘ways’ as the superord<strong>in</strong>ate term, which<br />
is mascul<strong>in</strong>e <strong>in</strong> gender. The fem<strong>in</strong><strong>in</strong>e gender is also <strong>in</strong>herited from the predicative adjective<br />
because the superord<strong>in</strong>ate term assumed by some native speakers is enies.fem.pl<br />
‘mean<strong>in</strong>gs’, which is fem<strong>in</strong><strong>in</strong>e gender.<br />
Thus, we have proposed a syntactic analysis of referential agreement and we have<br />
provided an ontological hierarchy that could be used to represent the lexical relations<br />
between the two nouns that appear <strong>in</strong> the coord<strong>in</strong>ate phrase and the CIR that is implied<br />
by native speakers and appears at the contextual level of representation. The fact that the<br />
CIR may differ from speaker to speaker due to the different way that the various native<br />
speakers perceive the world is viewed from an ontological explanation s<strong>in</strong>ce the ontological<br />
organisation of lexical items is attributed to extral<strong>in</strong>guistic conditions.<br />
4.6 Conclusion<br />
In this chapter, we presented predicate-argument agreement <strong>in</strong> MG coord<strong>in</strong>ate noun phrases,<br />
focus<strong>in</strong>g on the gender feature. The data showed that gender resolution <strong>in</strong> MG does not<br />
follow a straightforward dist<strong>in</strong>ction between syntactic resolution <strong>in</strong> <strong>in</strong>animate nouns and<br />
semantic resolution <strong>in</strong> animate ones. Instead, MG coord<strong>in</strong>ate nouns show a comb<strong>in</strong>ation of<br />
expected and unexpected resolution patterns. The expected patterns refer to cases of resolution<br />
that follow either the semantic or the syntactic pr<strong>in</strong>ciples. The unexpected patterns<br />
refer to an additional type of agreement the referential agreement which is determ<strong>in</strong>ed<br />
by contextual factors. This type of agreement depends on the assumption of a Contextually<br />
Introduced Referent on behalf of the consultants, which triggers morphosyntactic<br />
agreement <strong>in</strong> the sentential predicate.<br />
The expected patterns of <strong>in</strong>animate coord<strong>in</strong>ate nouns follow the syntactic pr<strong>in</strong>ciple of<br />
resolution and the unexpected patterns follow the referential agreement pr<strong>in</strong>ciple. Similarly,<br />
the expected patterns of animate coord<strong>in</strong>ate nouns follow the semantic pr<strong>in</strong>ciple<br />
(same gender nouns show feature match<strong>in</strong>g agreement) and the unexpected patterns follow<br />
referential agreement.<br />
To account for the variation <strong>in</strong> the expected patterns, we assumed the marker-set<br />
decomposition (Dalrymple and Kaplan, 2000) with common marker sets for both groups of<br />
nouns. In <strong>in</strong>animate nouns, syntactic resolution is derived by the standard syntactic rule<br />
for gender resolution which follows the set-union operation. In animate nouns, additional<br />
specifications are <strong>in</strong>troduced <strong>in</strong> the PS rule which capture both the grammatical and the<br />
semantic patterns. Thus, union operation is not applied.<br />
For the unexpected cases, native speakers br<strong>in</strong>g <strong>in</strong>to the discourse an extral<strong>in</strong>guistic<br />
concept the Contextually Introduced Referent. In LFG, we <strong>in</strong>troduce an extra level of<br />
<strong>in</strong>e<br />
are
4.6 Conclusion 116<br />
representation, the contextual one, where r-structure is the structure of the referred element.<br />
The agreement features of the referred element <strong>in</strong> r-structure are <strong>in</strong>troduced to the<br />
f-structure of the predicative adjective through the d mapp<strong>in</strong>g function. This prelim<strong>in</strong>ary<br />
analysis captures the basic idea of the referential agreement pr<strong>in</strong>ciple. However, we showed<br />
that the CIR, which functions as the superord<strong>in</strong>ate term of the coord<strong>in</strong>ate nouns it refers<br />
to, is extral<strong>in</strong>guistic <strong>in</strong> nature and may differ from speaker to speaker. To expla<strong>in</strong> this<br />
extral<strong>in</strong>guistic phenomenon we have proposed an ontological hierarchy similar to the template<br />
hierarchy assumed by LFG and the type hierarchy assumed by HPSG, which shows<br />
how native speakers represent the various lexical items, and we have argued that the use<br />
of a different CIR across speakers largerly depends on ontology and different perception of<br />
the world.
Chapter 5<br />
Head Modifier <strong>Agreement</strong> <strong>in</strong><br />
<strong>Modern</strong> <strong>Greek</strong><br />
5.1 Introduction<br />
In MG, head-modifier agreement requires the modifier to agree with the noun it modifies<br />
<strong>in</strong> number, gender and case. The situation is more complicated with coord<strong>in</strong>ate noun<br />
phrases. In the present chapter, we will focus on the number and gender features. In<br />
exam<strong>in</strong><strong>in</strong>g a range of modified coord<strong>in</strong>ate nouns <strong>in</strong> MG, we found that the MG def<strong>in</strong>ite<br />
determ<strong>in</strong>er differs from the rest of the modifiers. A s<strong>in</strong>gle modifier with wide scope can<br />
modify two coord<strong>in</strong>ate nouns and it agrees <strong>in</strong> number with both conjuncts but it agrees<br />
<strong>in</strong> gender with the closest conjunct (CCA), irrespective of the gender of the two nouns.<br />
The shared modifier allows the generation of jo<strong>in</strong>t and split read<strong>in</strong>gs <strong>in</strong> both s<strong>in</strong>gular and<br />
plural nouns. The def<strong>in</strong>ite determ<strong>in</strong>er, however, does not show the same behaviour s<strong>in</strong>ce<br />
each conjunct needs a separate determ<strong>in</strong>er. A shared def<strong>in</strong>ite determ<strong>in</strong>er with wide scope<br />
can modify two coord<strong>in</strong>ate nouns and it agrees <strong>in</strong> number with both nouns but it only<br />
occurs with the jo<strong>in</strong>t read<strong>in</strong>g <strong>in</strong> both s<strong>in</strong>gular and plural nouns. In terms of gender, a<br />
shared def<strong>in</strong>ite determ<strong>in</strong>er requires uniform gender conjuncts, whereas it is allowed with<br />
mixed gender conjuncts only when the two nouns are mascul<strong>in</strong>e and fem<strong>in</strong><strong>in</strong>e <strong>in</strong> gender.<br />
Some exceptional data are found when a shared determ<strong>in</strong>er with wide scope modifies two<br />
coord<strong>in</strong>ate nouns and occurs with the split read<strong>in</strong>g. We will suggest that these patterns<br />
are cases of natural coord<strong>in</strong>ation (Dalrymple and Nikolaeva, 2006).<br />
The proposal developed by K<strong>in</strong>g and Dalrymple (2004) on determ<strong>in</strong>er agreement and<br />
noun coord<strong>in</strong>ation lays out a putative crossl<strong>in</strong>guistic typology based on three different<br />
agreement systems, concord, <strong>in</strong>dex and concord/<strong>in</strong>dex that relate to the number<br />
feature. We will argue that the MG modifiers are nicely accounted for by the concord<br />
system s<strong>in</strong>ce they comply with the restrictions of that system. The MG determ<strong>in</strong>er, on the<br />
other hand, is best expla<strong>in</strong>ed by the most restrictive system, the concord/<strong>in</strong>dex. The<br />
exceptional data, which allow a shared def<strong>in</strong>ite determ<strong>in</strong>er to modify coord<strong>in</strong>ate nouns<br />
and occurs with the split read<strong>in</strong>g, could be analysed accord<strong>in</strong>g to the proposal on natural<br />
coord<strong>in</strong>ation by Dalrymple and Nikolaeva (2006). However, the syntactic structure of MG<br />
natural coord<strong>in</strong>ation patterns differs from the syntactic structure of natural coord<strong>in</strong>ation<br />
proposed by Dalrymple and Nikolaeva (2006). Thus, we follow the standard analysis of<br />
accidental coord<strong>in</strong>ation for the MG natural coord<strong>in</strong>ation cases. It is worth not<strong>in</strong>g that <strong>Modern</strong><br />
<strong>Greek</strong> grammars and traditional textbooks (Clairis et al. (2004), Holton et al. (1997),<br />
Triantaphyllidis (1994), Triantaphyllidis (2005), Joseph and Philippaki-Warburton (1987),<br />
Mackridge (1985) and Holton et al. (1997)) do not present any discussion of modifiers or<br />
the def<strong>in</strong>ite determ<strong>in</strong>er with coord<strong>in</strong>ate nouns or the phenomenon of natural coord<strong>in</strong>ation<br />
<strong>in</strong> MG.
5.2 NP Internal <strong>Agreement</strong> and Interpretation of the Coord<strong>in</strong>ate <strong>Noun</strong>s 118<br />
The first section revisits the three different agreement systems proposed by K<strong>in</strong>g and<br />
Dalrymple (2004), and discusses how the <strong>in</strong>dex num feature is associated to a semantic<br />
dist<strong>in</strong>ction between the jo<strong>in</strong>t and split read<strong>in</strong>g <strong>in</strong> coord<strong>in</strong>ate nouns modified by a shared<br />
determ<strong>in</strong>er. Next, we present MG modifiers with coord<strong>in</strong>ate nouns and we argue that they<br />
follow the concord agreement system. Then, we present the MG def<strong>in</strong>ite determ<strong>in</strong>er<br />
with coord<strong>in</strong>ate nouns and we argue that it follows the <strong>in</strong>dex/concord system. The<br />
exceptional patterns are captured assum<strong>in</strong>g the standard analysis of accidental coord<strong>in</strong>ation<br />
and not that of natural coord<strong>in</strong>ation.<br />
5.2 NP Internal <strong>Agreement</strong> and Interpretation of the Coord<strong>in</strong>ate <strong>Noun</strong>s<br />
Modification <strong>in</strong> NP coord<strong>in</strong>ate structures is rather complicated s<strong>in</strong>ce a common determ<strong>in</strong>er<br />
cannot always modify two coord<strong>in</strong>ate nouns, and if it does, it yields specific <strong>in</strong>terpretations.<br />
As discussed <strong>in</strong> Section 3.3.2, K<strong>in</strong>g and Dalrymple (2004) predict three different agreement<br />
systems based on the <strong>in</strong>dex and concord agreement features when a shared determ<strong>in</strong>er<br />
modifies coord<strong>in</strong>ate nouns <strong>in</strong> NP <strong>in</strong>ternal agreement. The concord system requires<br />
determ<strong>in</strong>ers to show agreement only with the concord features of each conjunct. This<br />
system allows a s<strong>in</strong>gular determ<strong>in</strong>er to modify coord<strong>in</strong>ated s<strong>in</strong>gular nouns and a plural<br />
determ<strong>in</strong>er to modify coord<strong>in</strong>ated plural nouns. In both cases, the conjuncts may refer to<br />
the same or different <strong>in</strong>dividuals s<strong>in</strong>ce there is no <strong>in</strong>dex feature restriction.<br />
(1) concord systems:<br />
a. S<strong>in</strong>gular determ<strong>in</strong>er<br />
(↑concord num) = sg<br />
b. Plural determ<strong>in</strong>er<br />
(↑concord num) = pl<br />
(K<strong>in</strong>g and Dalrymple, 2004, 86)<br />
The concord/<strong>in</strong>dex system is the most restrictive system and requires the determ<strong>in</strong>er<br />
to show agreement with the concord features of each conjunct and the <strong>in</strong>dex feature of<br />
the coord<strong>in</strong>ate phrase as a whole. This system imposes restrictions on both the concord<br />
and <strong>in</strong>dex agreement features and allows a s<strong>in</strong>gular determ<strong>in</strong>er to modify coord<strong>in</strong>ated<br />
s<strong>in</strong>gular nouns. A consequence of this system is that a s<strong>in</strong>gular determ<strong>in</strong>er can scope over<br />
conjuncts only when they get a jo<strong>in</strong>t or boolean read<strong>in</strong>g. The same system <strong>in</strong> plural allows<br />
a plural determ<strong>in</strong>er to modify plural coord<strong>in</strong>ate nouns, which can get either a jo<strong>in</strong>t or a<br />
split read<strong>in</strong>g.<br />
(2) concord and <strong>in</strong>dex systems:<br />
a. S<strong>in</strong>gular determ<strong>in</strong>er<br />
(↑concord num) = sg<br />
(↑<strong>in</strong>dex num) = sg<br />
b. Plural determ<strong>in</strong>er<br />
(↑concord num) = pl<br />
(↑<strong>in</strong>dex num) = pl<br />
(K<strong>in</strong>g and Dalrymple, 2004, 85)
5.2 NP Internal <strong>Agreement</strong> and Interpretation of the Coord<strong>in</strong>ate <strong>Noun</strong>s 119<br />
The <strong>in</strong>dex system requires determ<strong>in</strong>ers to show agreement only with the <strong>in</strong>dex features<br />
of the coord<strong>in</strong>ate phrase as a whole (and not with the concord features of the dist<strong>in</strong>ct<br />
conjuncts). This system allows a s<strong>in</strong>gular determ<strong>in</strong>er to modify s<strong>in</strong>gular nouns that refer<br />
to the same <strong>in</strong>dividual and a plural determ<strong>in</strong>er to modify conjo<strong>in</strong>ed s<strong>in</strong>gular nouns.<br />
(3) <strong>in</strong>dex systems:<br />
a. S<strong>in</strong>gular determ<strong>in</strong>er<br />
(↑<strong>in</strong>dex num) = sg<br />
b. Plural determ<strong>in</strong>er<br />
(↑<strong>in</strong>dex num) = pl<br />
(K<strong>in</strong>g and Dalrymple, 2004, 86)<br />
Heycock and Zamparelli (2000, 2005), whose work is discussed <strong>in</strong> K<strong>in</strong>g and Dalrymple<br />
(2004), also exam<strong>in</strong>e modification <strong>in</strong> coord<strong>in</strong>ate noun phrases and discuss the existence of<br />
two possible <strong>in</strong>terpretations when coord<strong>in</strong>ate nouns are modified by a shared determ<strong>in</strong>er.<br />
They refer to these <strong>in</strong>terpretations as the “jo<strong>in</strong>t” read<strong>in</strong>g and the “split” read<strong>in</strong>g. The jo<strong>in</strong>t<br />
read<strong>in</strong>g occurs when both nouns denote the same <strong>in</strong>dividual, and the split read<strong>in</strong>g occurs<br />
when the two nouns refer to different <strong>in</strong>dividuals. Both K<strong>in</strong>g and Dalrymple (2004) and<br />
Heycock and Zamparelli (2005) observe that there is a difference concern<strong>in</strong>g the read<strong>in</strong>gs<br />
that are generated between s<strong>in</strong>gular and plural coord<strong>in</strong>ate nouns modified by a shared<br />
def<strong>in</strong>ite determ<strong>in</strong>er <strong>in</strong> languages like Italian, Spanish, German, French and Brazilian Portuguese.<br />
In Brazilian Portuguese, for example, <strong>in</strong> s<strong>in</strong>gular conjuncts, a shared determ<strong>in</strong>er can<br />
modify two coord<strong>in</strong>ate nouns and the coord<strong>in</strong>ate phrase is <strong>in</strong>terpreted with the “jo<strong>in</strong>t”<br />
(Heycock and Zamparelli, 1999, 2000) read<strong>in</strong>g, as illustrated below:<br />
(4) O presidente e diretor da Air<br />
the.sg.masc<br />
France<br />
France<br />
president.sg.masc and director.sg.masc of Air<br />
‘The president and director of Air France’<br />
((Heycock and Zamparelli, 2000, 7) cited <strong>in</strong> (K<strong>in</strong>g and Dalrymple, 2004, 92))<br />
If a s<strong>in</strong>gle determ<strong>in</strong>er modifies coord<strong>in</strong>ate s<strong>in</strong>gular nouns and the coord<strong>in</strong>ate phrase is<br />
<strong>in</strong>terpreted with the “split” (Heycock and Zamparelli, 1999, 2000) read<strong>in</strong>g, the phrase is<br />
ungrammatical. This is shown below:<br />
(5) *A mesa e escrivan<strong>in</strong>ha estão arrumadas<br />
‘The table and the small desk are <strong>in</strong> order’<br />
(Heycock and Zamparelli, 2000, 7)<br />
The split read<strong>in</strong>g can only be admitted if a separate determ<strong>in</strong>er appears before each<br />
conjunct, as seen below:<br />
(6) A mesae a escrivan<strong>in</strong>ha estão arrumadas<br />
‘The table and the small desk are <strong>in</strong> order’<br />
(Heycock and Zamparelli, 2000, 7)<br />
The same generalisations do not hold when the conjuncts are plural. All languages<br />
discussed <strong>in</strong> Heycock and Zamparelli (2000, 2005) and K<strong>in</strong>g and Dalrymple (2004) allow<br />
both the “split” and the “jo<strong>in</strong>t” <strong>in</strong>terpretation <strong>in</strong> coord<strong>in</strong>ate plural nouns that are modified
5.2 NP Internal <strong>Agreement</strong> and Interpretation of the Coord<strong>in</strong>ate <strong>Noun</strong>s 120<br />
by a s<strong>in</strong>gle plural determ<strong>in</strong>er. The follow<strong>in</strong>g example is from Italian and it is grammatical<br />
even though a s<strong>in</strong>gle determ<strong>in</strong>er modifies both nouns and it occurs with the split read<strong>in</strong>g:<br />
(7) Gli amici e nemici di Gianni si trovano<br />
the.pl friends.pl and enemies.pl of Gianni were found<br />
d’ accordo su un solo punto<br />
<strong>in</strong> agreement on a s<strong>in</strong>gle po<strong>in</strong>t<br />
‘The friends and enemies of Gianni were <strong>in</strong>-agreement on a s<strong>in</strong>gle po<strong>in</strong>t’<br />
(Heycock and Zamparelli, 2000, 10)<br />
An important aspect of K<strong>in</strong>g and Dalrymple’s proposal, which is related to the dist<strong>in</strong>ction<br />
proposed by Heycock and Zamparelli (2000) between the jo<strong>in</strong>t and split read<strong>in</strong>g, is<br />
that the value of the <strong>in</strong>dex num feature of the coord<strong>in</strong>ate phrase represents the number of<br />
<strong>in</strong>dividuals it refers to. Thus, syntactic <strong>in</strong>dex num is associated to the semantic number<br />
of the referents <strong>in</strong> a phrase.<br />
Therefore, we can argue that as well as a syntactic agreement pattern we might f<strong>in</strong>d<br />
some additional semantic aspects. Accord<strong>in</strong>g to K<strong>in</strong>g and Dalrymple (2004, 75), <strong>in</strong> a coord<strong>in</strong>ate<br />
phrase with s<strong>in</strong>gular conjuncts, when the def<strong>in</strong>ite determ<strong>in</strong>er has a sg value <strong>in</strong><br />
the <strong>in</strong>dex num feature the coord<strong>in</strong>ate phrase necessarily refers to only one <strong>in</strong>dividual. In<br />
this case, the <strong>in</strong>terpretation of the phrase is the jo<strong>in</strong>t read<strong>in</strong>g and the def<strong>in</strong>ite determ<strong>in</strong>er<br />
imposes the syntactic feature-value requirement <strong>in</strong>dex num=sg. The most characteristic<br />
case is the English example the friend and colleague where the syntactic structure for the<br />
coord<strong>in</strong>ate phrase has the annotation <strong>in</strong>dex num=sg and the <strong>in</strong>terpretation is the jo<strong>in</strong>t<br />
read<strong>in</strong>g.<br />
On the other hand, <strong>in</strong> a coord<strong>in</strong>ate phrase with plural conjuncts, when the def<strong>in</strong>ite<br />
determ<strong>in</strong>er has a pl value <strong>in</strong> the <strong>in</strong>dex num feature, the coord<strong>in</strong>ate phrase can be <strong>in</strong>terpreted<br />
with the jo<strong>in</strong>t or the split read<strong>in</strong>g. The jo<strong>in</strong>t read<strong>in</strong>g derives from the fact that<br />
the two coord<strong>in</strong>ate nouns are both plural and they may refer to the same <strong>in</strong>dividuals. A<br />
characteristic case is the English example the friends and colleagues. In this case, the two<br />
nouns refer to the same people and the phrase occurs with the jo<strong>in</strong>t read<strong>in</strong>g, while the<br />
def<strong>in</strong>ite determ<strong>in</strong>er imposes the syntactic requirement <strong>in</strong>dex num=pl. The split read<strong>in</strong>g<br />
is also possible s<strong>in</strong>ce the two coord<strong>in</strong>ate nouns are plural and they may also refer to different<br />
<strong>in</strong>dividuals. Thus, the same example can have a split <strong>in</strong>terpretation s<strong>in</strong>ce the two<br />
nouns can refer to different groups of <strong>in</strong>dividuals and the def<strong>in</strong>ite determ<strong>in</strong>er imposes the<br />
syntactic requirement <strong>in</strong>dex num=pl. We observe that the annotation <strong>in</strong> the syntactic<br />
structure is the same irrespective of the <strong>in</strong>terpretation of the coord<strong>in</strong>ate phrase as <strong>in</strong> <strong>in</strong>dex<br />
num=pl. Thus, we conclude that <strong>in</strong> coord<strong>in</strong>ated plural nouns the syntactic structure for<br />
the coord<strong>in</strong>ate phrase has the annotation <strong>in</strong>dex num=pl and this annotation corresponds<br />
to the jo<strong>in</strong>t or the split <strong>in</strong>terpretation.<br />
The above dist<strong>in</strong>ction between split and jo<strong>in</strong>t read<strong>in</strong>gs and their association to the syntactic<br />
<strong>in</strong>dex num feature with sg or pl values are crucial for the MG data, which present<br />
a slightly different behaviour from the languages discussed by Heycock and Zamparelli<br />
(2000) and K<strong>in</strong>g and Dalrymple (2004). We will show that <strong>in</strong> MG coord<strong>in</strong>ate phrases, a<br />
shared modifier follows the concord system and <strong>in</strong>troduces the semantic requirements of<br />
that system, admit<strong>in</strong>g both jo<strong>in</strong>t and split <strong>in</strong>terpretations. The shared def<strong>in</strong>ite determ<strong>in</strong>er<br />
follows the <strong>in</strong>dex/concord system but it appears to <strong>in</strong>troduce some additional semantic<br />
specifications over and above those generated by the most restrictive syntactic agreement<br />
system of K<strong>in</strong>g and Dalrymple (2004). The def<strong>in</strong>ite determ<strong>in</strong>er allows only the jo<strong>in</strong>t read<strong>in</strong>g<br />
with both s<strong>in</strong>gular and plural conjuncts; the split <strong>in</strong>terpretation under a shared determ<strong>in</strong>er<br />
is allowed <strong>in</strong> plural number only <strong>in</strong> a specific group of data which fall with<strong>in</strong> the case of<br />
natural coord<strong>in</strong>ation (Dalrymple and Nikolaeva, 2006).<br />
The method used for the collection of data was based on a wide Internet search, on
5.3 NP Internal <strong>Agreement</strong> <strong>in</strong> <strong>Modern</strong> <strong>Greek</strong> 121<br />
newspaper articles and on personal discussions with about fifteen native speakers who<br />
confirmed the acceptability and unacceptability of the collected sentences. Consultants<br />
accepted or rejected a phrase based on whether they would use it <strong>in</strong> everyday language or<br />
they found it unaccaptable and rejected it. Next, we will discuss NP <strong>in</strong>ternal agreement<br />
<strong>in</strong> coord<strong>in</strong>ate nouns <strong>in</strong> MG.<br />
5.3 NP Internal <strong>Agreement</strong> <strong>in</strong> <strong>Modern</strong> <strong>Greek</strong><br />
The MG determ<strong>in</strong>ers and modifiers generally encode a three-way gender 1 dist<strong>in</strong>ction between<br />
mascul<strong>in</strong>e, fem<strong>in</strong><strong>in</strong>e and neuter <strong>in</strong> both s<strong>in</strong>gular and plural number 2 and agree with<br />
the nouns they modify. In example (8), the def<strong>in</strong>ite determ<strong>in</strong>er o modifies the noun andras<br />
and shows s<strong>in</strong>gular, mascul<strong>in</strong>e, nom<strong>in</strong>ative agreement with the latter. In example (9), the<br />
<strong>in</strong>def<strong>in</strong>ite determ<strong>in</strong>er merika modifies the noun pedia and shows plural, neuter, nom<strong>in</strong>ative<br />
agreement with the latter:<br />
(8) o<br />
the.sg.masc.nom<br />
‘the man’<br />
(constructed)<br />
(9) merika<br />
some.pl.neut.nom<br />
‘some children’<br />
(constructed)<br />
antras<br />
man.sg.masc.nom<br />
pedia<br />
children.pl.neut.nom<br />
Similarly, the adjectival modifier megali ‘big’ shows plural, mascul<strong>in</strong>e, nom<strong>in</strong>ative agreement<br />
with the noun andres ‘men’:<br />
(10) megali<br />
big.pl.masc.nom<br />
‘big men’<br />
(constructed)<br />
antres<br />
men.pl.masc.nom<br />
In the next section, we will discuss first shared MG modifiers with coord<strong>in</strong>ate nouns<br />
and then the def<strong>in</strong>ite determ<strong>in</strong>er o-i-to.sg.masc/fem/neut ‘the’ with coord<strong>in</strong>ate nouns.<br />
We present data with identical gender conjuncts and mixed gender conjuncts modified by<br />
a shared modifier and by a shared def<strong>in</strong>ite determ<strong>in</strong>er.<br />
5.4 The <strong>Modern</strong> <strong>Greek</strong> Modifiers<br />
Generally, <strong>in</strong> MG modifiers with coord<strong>in</strong>ate nouns, a shared modifier scopes over coord<strong>in</strong>ate<br />
nouns and it agrees <strong>in</strong> num with both nouns while it occurs either with the jo<strong>in</strong>t or the<br />
split <strong>in</strong>terpretation <strong>in</strong> both s<strong>in</strong>gular and plural conjuncts. The modifier always agrees <strong>in</strong><br />
gender with the closest conjunct. A separate modifier can also appear before each noun<br />
and agrees with the latter <strong>in</strong> both num and gender. In this case, it occurs only with<br />
a split <strong>in</strong>terpretation. We will discuss examples with various shared modifiers, such as<br />
adjectives, quantifiers and determ<strong>in</strong>ers.<br />
In coord<strong>in</strong>ated s<strong>in</strong>gular nouns with the same gender, a s<strong>in</strong>gular attributive adjective<br />
scopes over two conjuncts and requires s<strong>in</strong>gular num agreement with both nouns and<br />
gender agreement with the closest noun 3 . It can convey either a jo<strong>in</strong>t or a split read<strong>in</strong>g,<br />
1 There are a few exceptions, such as some numerals, the distributive determ<strong>in</strong>er kathe etc that do not<br />
encode this dist<strong>in</strong>ction. For more details see Mackridge (1985) and Holton et al. (1997).<br />
2 The MG determ<strong>in</strong>ers and modifiers dist<strong>in</strong>guish four cases, nom<strong>in</strong>ative, genitive, accusative and dative<br />
apart from the gender and number dist<strong>in</strong>ctions. Case is not central and we will not discuss it further.<br />
3 In uniform gender nouns, the adjective also agrees <strong>in</strong> gender with the second conjunct.
5.4 The <strong>Modern</strong> <strong>Greek</strong> Modifiers 122<br />
depend<strong>in</strong>g on whether the two nouns refer to the same or different <strong>in</strong>dividuals or entities.<br />
In example (11), the shared adjective scopes over both conjuncts and the only possible<br />
<strong>in</strong>terpretation is the jo<strong>in</strong>t one s<strong>in</strong>ce the nouns refer to a s<strong>in</strong>gle <strong>in</strong>dividual and the verb<br />
shows s<strong>in</strong>gular agreement. In example (12), a shared adjective scopes over both nouns but<br />
it conveys a split <strong>in</strong>terpretation s<strong>in</strong>ce the two conjuncts denote different <strong>in</strong>dividuals and<br />
the verb shows plural agreement 4 :<br />
(11) O Gorjias itan megalos<br />
the<br />
ritoras<br />
Gorgias was.sg great.sg.masc<br />
orator.sg.masc<br />
‘Gorgias was a great philosopher and orator’<br />
(History Book, High School 3rd year)<br />
filosofos<br />
philosopher.sg.masc<br />
ke<br />
and<br />
(12) Drastirios pateras ke ijos theorunte<br />
active.sg.masc father.sg.masc and son.sg.masc are-considered.pl<br />
o Onasis ke o Aleksantros<br />
the Onasis and the Alexander<br />
‘Active father and son are considered to be Onassis and Alexander’<br />
(Onasis, 1997)<br />
Also, a s<strong>in</strong>gular adjective cannot modify a s<strong>in</strong>gular and a plural noun simultaneously,<br />
but it only scopes the closest conjunct, which must be s<strong>in</strong>gular, and not the second conjunct:<br />
(13) Drastirios pateras ke iji<br />
theorunte<br />
active.sg.masc father.sg.masc and sons.pl.masc are-considered.pl<br />
o Latsis ke ta pedia tou<br />
the Latsis and the children his<br />
‘Active father and sons are considered Latsis and his children’<br />
In MG, a plural adjective cannot modify s<strong>in</strong>gular conjuncts, as seen is example (14):<br />
(14) *Drastirii pateras ke ijos theorunte<br />
active.pl.masc father.sg.masc and son.sg.masc are-considered.pl<br />
o Onasis ke o Aleksantros<br />
the Onasis and the Alexander<br />
‘Active father and son are considered to be Onassis and Alexander’<br />
(constructed)<br />
In coord<strong>in</strong>ated plural phrases, MG modifiers behave <strong>in</strong> a similar way. The plural<br />
modifier scopes over two conjuncts and specifies plural num agreement with both nouns<br />
and gender agreement with the closest noun. The jo<strong>in</strong>t and the split <strong>in</strong>terpretations<br />
are allowed when a shared adjective scopes over coord<strong>in</strong>ate nouns. In example (15), the<br />
conjuncts are modified by the same adjective and they refer to the same concept s<strong>in</strong>ce the<br />
conjuncts are considered synonyms. In example (16), the adjective scopes over both nouns<br />
but the <strong>in</strong>dividuals denote different groups of people and therefore the split read<strong>in</strong>g is the<br />
most plausible <strong>in</strong>terpretation:<br />
4 The follow<strong>in</strong>g examples were collected from <strong>Greek</strong> newspapers and various books. I will use the abbre-<br />
viation G.N. to stand for <strong>Greek</strong> Newspaper and G.M. to stand for <strong>Greek</strong> Magaz<strong>in</strong>e followed by the specific<br />
newspaper or magaz<strong>in</strong>e from which I have collected each example. The examples that were constructed are<br />
<strong>in</strong>dicated with the word ‘constructed’.
5.4 The <strong>Modern</strong> <strong>Greek</strong> Modifiers 123<br />
(15) Isxires antiparathesis ke antiparavoles<br />
strong.pl.fem juxtapositions.pl.fem and contrasts.pl.fem<br />
ekdilothikan st<strong>in</strong> kentriki s<strong>in</strong>edriasi tis Vulis<br />
were-expressed.pl <strong>in</strong>-the central meet<strong>in</strong>g of-the Parliament<br />
‘Strong juxtapositions and contrasts were-expressed <strong>in</strong>-the central meet<strong>in</strong>g of the<br />
Parliament’<br />
(G.N.:To Vima)<br />
(16) Megali angioplastes ke zografi<br />
great.pl.masc<br />
eona...<br />
century...<br />
sculptors.pl.masc and pa<strong>in</strong>ters.pl.masc<br />
‘Great sculptors and pa<strong>in</strong>ters of the 5th century...’<br />
(www.<strong>in</strong>.gr)<br />
tu<br />
of-the<br />
Also, a plural adjective cannot modify a plural and a s<strong>in</strong>gular noun at the same time.<br />
The plural adjective scopes over the closest conjunct, which is also plural, and not the<br />
second conjunct:<br />
(17) Megali angioplastes ke zografos<br />
great.pl.masc<br />
eona...<br />
century...<br />
sculptors.pl.masc and pa<strong>in</strong>ter.sg.masc<br />
‘Great sculptors and a pa<strong>in</strong>ter of the 5th century...’<br />
(constructed)<br />
tu<br />
of-the<br />
Apart from adjectives, the data showed that shared quantifiers and determ<strong>in</strong>ers specify<br />
the same num agreement with both nouns and gender agreement with the closet conjunct,<br />
and allow both a jo<strong>in</strong>t and a split <strong>in</strong>terpretation when they scope over coord<strong>in</strong>ate nouns. For<br />
example, the quantifier arket-os/-i/-o.sg.masc/fem/neut ‘quite a lot’ and the correlative<br />
determ<strong>in</strong>er opi-os/-a/-o.sg.masc/fem/neut ‘whoever’ have a mascul<strong>in</strong>e, a fem<strong>in</strong><strong>in</strong>e and<br />
a neuter form <strong>in</strong> both s<strong>in</strong>gular and plural numbers and can scope over coord<strong>in</strong>ated s<strong>in</strong>gular<br />
or plural nouns that admit both jo<strong>in</strong>t and split read<strong>in</strong>gs, denot<strong>in</strong>g the same or different<br />
entities, respectively. The examples below illustrate the s<strong>in</strong>gular quantifier arket-os/-i/o.sg.masc/fem/neut<br />
‘quite a lot’, which specifies s<strong>in</strong>gular agreement with both nouns<br />
and closest conjunct gender agreement:<br />
(18) Arketi<br />
anastatosi ke anisixia<br />
quite-a-lot-of.sg.fem agitation.sg.fem and worry.sg.fem<br />
proklithike eksetias tis ikonomikis krisis<br />
was-caused.sg because-of the f<strong>in</strong>ancial crisis<br />
‘Quite a lot of agitation and worry was caused because of the f<strong>in</strong>ancial crisis’<br />
(G.N.:Ethnos)<br />
(19) Arketi<br />
dulia ke organosi<br />
quite-a-lot-of.sg.fem work.sg.fem and organisation.sg.fem<br />
xriastikan sto s<strong>in</strong>gekrimeno grafio<br />
were-needed.pl <strong>in</strong>-the specific office<br />
‘Quite-a-lot of work and organisation were needed <strong>in</strong> the specific office’<br />
(G.N.:Kathimer<strong>in</strong>i)<br />
The first case denotes reference to the same entity s<strong>in</strong>ce number verb agreement is s<strong>in</strong>gular.<br />
The second case denotes two different entities and the verb shows plural number<br />
5u<br />
5th<br />
5u<br />
5th
5.4 The <strong>Modern</strong> <strong>Greek</strong> Modifiers 124<br />
agreement 5 .<br />
The plural quantifier arket-i/-es/-a.pl.masc/fem/neut ‘quite a lot’ scopes over coord<strong>in</strong>ated<br />
plural nouns, which refer either to the same or different <strong>in</strong>dividuals, and agrees <strong>in</strong><br />
gender with the closest conjunct. In example (20), the two nouns admit the jo<strong>in</strong>t read<strong>in</strong>g<br />
denot<strong>in</strong>g the same group of <strong>in</strong>dividuals, which has both properties, whereas <strong>in</strong> example<br />
(21), the nouns admit the split read<strong>in</strong>g denot<strong>in</strong>g different <strong>in</strong>dividuals:<br />
(20) Arketi<br />
ftoxi ke anergi<br />
quite-a-lot.pl.masc poor.pl.masc and unemployed.pl.masc<br />
idopiithikan prosfata apo t<strong>in</strong> eforia<br />
were-warned.pl recently from the tax-office<br />
‘Quite-a-lot of poor and unemployed were notified recently from the tax office’<br />
(G.N.:Ependitis)<br />
(21) Arketi<br />
ipurgi<br />
ke ifipurgi<br />
quite-a-lot.pl.masc m<strong>in</strong>isters.pl.masc and deputy-m<strong>in</strong>isters.pl.masc<br />
psifisan to nomosxedio<br />
voted-for.pl the bill<br />
‘Quite a lot of m<strong>in</strong>isters and deputy-m<strong>in</strong>isters voted for the bill’<br />
(G.N.:To Vima)<br />
The s<strong>in</strong>gular or plural correlative determ<strong>in</strong>er opi-os/-a/-o.sg.masc/fem/neut ‘whoever’<br />
modifies coord<strong>in</strong>ated s<strong>in</strong>gular or plural nouns, respectively, and admits both read<strong>in</strong>gs,<br />
denot<strong>in</strong>g the same or different entities. It specifies the same num agreement with both<br />
conjuncts and it always agrees <strong>in</strong> gender with the closest conjunct. Consider the follow<strong>in</strong>g<br />
examples:<br />
(22) Opios<br />
filos ke s<strong>in</strong>adelfos<br />
whoever.sg.masc friend.sg.masc and colleague.sg.masc<br />
kurasmenos dikeute adia<br />
tired is-entitled day-off<br />
‘Any tired friend and colleague is entitled a day off’<br />
(constructed)<br />
(23) Opio<br />
agori ke<br />
whoever.sg.neut boy.sg.neut and<br />
mporun na kanun etisi<br />
can to make application<br />
‘Any <strong>in</strong>terested boy and girl can apply’<br />
(constructed)<br />
koritsi<br />
girl.sg.neut<br />
<strong>in</strong>e<br />
is.sg<br />
endiaferonte<br />
are-<strong>in</strong>terested.pl<br />
The first case denotes reference to only one and the same <strong>in</strong>dividual s<strong>in</strong>ce the verb shows<br />
s<strong>in</strong>gular number agreement. The second case denotes two different <strong>in</strong>dividuals agori ke<br />
koritsi ‘boy and a girl’ and the verb shows plural number agreement.<br />
The plural correlative determ<strong>in</strong>er with coord<strong>in</strong>ated plural nouns also shows the same<br />
syntactic characteristics, while it allows reference to the same or different entities. This is<br />
illustrated <strong>in</strong> the follow<strong>in</strong>g examples:<br />
(24) Opii<br />
ftoxi<br />
whoever.pl.masc poor.pl.masc<br />
voithia dikeunte amesa<br />
help are-entitled immediately<br />
ke apori<br />
and destitutes.pl.masc<br />
epidoma anerjias<br />
jobseeker’s allowance<br />
xriazonte<br />
need.pl<br />
5 We should note that a second possibility might be available <strong>in</strong> the above examples. In particular,<br />
example (18) can also have plural verb agreement with a split <strong>in</strong>terpretation, and example (19) can also<br />
have s<strong>in</strong>gular verb agreement with a jo<strong>in</strong>t <strong>in</strong>terpretation. These were found less often.
5.4 The <strong>Modern</strong> <strong>Greek</strong> Modifiers 125<br />
‘Any poor and destitutes <strong>in</strong> need of help are immediately entitled to jobseeker’s<br />
allowance’<br />
(G.N.:Ependitis)<br />
(25) Opii<br />
ipurgi<br />
ke ifipurgi<br />
whoever.pl.masc m<strong>in</strong>isters.pl.masc and deputy-m<strong>in</strong>isters.pl.masc<br />
psifisan to nomosxedio tha m<strong>in</strong>un sti Vuli<br />
voted-for.pl the bill will stay <strong>in</strong>-the Parliament<br />
‘Any m<strong>in</strong>isters and deputy-m<strong>in</strong>isters that voted for the bill will stay <strong>in</strong> the Parliament’<br />
(G.N.:To Vima)<br />
In example (24), the two nouns refer to the same group of people, but <strong>in</strong> example (25)<br />
two separate groups are denoted.<br />
In coord<strong>in</strong>ate constructions with conjuncts of different gender the above observations<br />
also hold. A shared modifier scopes over coord<strong>in</strong>ate nouns and agrees <strong>in</strong> number with both<br />
conjuncts and <strong>in</strong> gender with the closest conjunct, while it allows both a jo<strong>in</strong>t and a split<br />
<strong>in</strong>terpretation.<br />
In example (26), the s<strong>in</strong>gular attributive adjective kalos ‘good’ agrees <strong>in</strong> number with<br />
both conjuncts but <strong>in</strong> gender with the first conjunct, the noun s<strong>in</strong>adelfo ‘colleague’. The<br />
adjective still scopes over both nouns and the two conjuncts denote different referents s<strong>in</strong>ce<br />
they <strong>in</strong>flect for a plural verb agreement:<br />
(26) Eftixos, exo kalo s<strong>in</strong>adelfo ke<br />
luckily, I-have good.sg.masc colleague.sg.masc and<br />
s<strong>in</strong>adelfisa pu <strong>in</strong>e eksipiretiki<br />
colleague.sg.fem who are.pl helpful.pl<br />
‘Luckily, I have a good man-colleague and woman-colleague who are helpful’<br />
(constructed)<br />
In example (27), the s<strong>in</strong>gular attributive adjective agrees <strong>in</strong> number with both conjuncts<br />
but <strong>in</strong> gender with the first conjunct only. Semantically, however, the adjective ipervoliki<br />
‘excessive’ scopes over both nouns and the two nouns denote the same referent s<strong>in</strong>ce they<br />
convey similar mean<strong>in</strong>gs:<br />
(27) I narkisistes xaraktirizonte apo ipervoliki<br />
the narcissists are-characterised by excessive.sg.fem<br />
aftareskia<br />
ke aftothavmasmo<br />
self-complacency.sg.fem and self-admiration sg.masc<br />
‘The narcissists are characterised by excessive self-complacency and self-admiration’<br />
(G.M.:VIMAgaz<strong>in</strong>o)<br />
Similarly, coord<strong>in</strong>ated plural nouns of the same number but different gender are modified<br />
by an adjective, which agrees <strong>in</strong> number with both nouns and <strong>in</strong> gender with the noun<br />
closest to it. Semantically, the adjective scopes over both nouns and the two conjuncts<br />
may refer to different or to identical entities, as <strong>in</strong> examples (28) and (29), respectively:<br />
(28) Katastrofiki sismi<br />
ke plimires<br />
destructive.pl.masc earthquakes.pl.masc and floods.pl.fem<br />
t<strong>in</strong> poli apo to 1965<br />
the town s<strong>in</strong>ce the 1965<br />
‘Destructive earthquakes and floods hit the town s<strong>in</strong>ce 1965’<br />
www.geocities.com/dsvolou14/ergasies/volos/volos.htm<br />
plitun<br />
hit.pl
5.4 The <strong>Modern</strong> <strong>Greek</strong> Modifiers 126<br />
(29) Kathimer<strong>in</strong>es agonies ke agxi<br />
tu<br />
daily.pl.fem agonies.pl.fem and anguishes.pl.neut of-the<br />
mesu el<strong>in</strong>a <strong>in</strong>e i dulia, i akrivia, ke ta<br />
middle<br />
dania<br />
mortgages<br />
<strong>Greek</strong> are.pl the job, the high-prices and the<br />
‘Daily agonies and anguishes of the middle class greek are jobs, high prices and<br />
mortgages’<br />
(G.M.:VIMAgaz<strong>in</strong>o)<br />
In example (28), the two nouns cannot denote the same entity s<strong>in</strong>ce they refer to<br />
different natural phenomena. In example (29), the two nouns agonies ke agxi are closely<br />
related concepts and denote the same entity.<br />
In the same way, the quantifier meriki ‘some’ modifies two plural conjuncts with mixed<br />
gender. It agrees <strong>in</strong> number with both nouns but <strong>in</strong> gender with the noun closest to it. The<br />
quantifier scopes over both nouns and allows a jo<strong>in</strong>t and a split <strong>in</strong>terpretation, respectively:<br />
(30) An ke antimetopise merikus ekfovismus<br />
even though he-faced some.pl.masc <strong>in</strong>timidations.pl.masc<br />
ke apiles s<strong>in</strong>exise to ergo tu<br />
and threats.pl.fem, he-cont<strong>in</strong>ued the work his<br />
‘Even though he faced some <strong>in</strong>timidations and threats, he cont<strong>in</strong>ued his work’<br />
(G.N.:Eleftherotipia)<br />
(31) Ja merikus daskalus<br />
ke<br />
for some.pl.masc men-teachers.pl.masc and<br />
daskales<br />
ta imerologia itan apajorevmena<br />
women-teachers.pl.fem the diaries were not-allowed<br />
‘For some men-teachers and women-teachers diaries were not allowed’<br />
(G.M.:VIMAgaz<strong>in</strong>o)<br />
In example (30), the two nouns denote the same entity, but <strong>in</strong> example (31), daskalus<br />
ke daskales ‘men-teachers and women-teachers’ denote different groups of people.<br />
The f<strong>in</strong>d<strong>in</strong>gs <strong>in</strong> MG are summarised as follows:<br />
• A s<strong>in</strong>gular modifier modifies s<strong>in</strong>gular coord<strong>in</strong>ate nouns and occurs with the jo<strong>in</strong>t and<br />
the split read<strong>in</strong>g<br />
• A plural modifier modifies plural coord<strong>in</strong>ate nouns and occurs with the jo<strong>in</strong>t and the<br />
split read<strong>in</strong>g<br />
• In both cases, the shared modifier agrees <strong>in</strong> gender with the closest conjunct<br />
• A plural modifier cannot modify s<strong>in</strong>gular coord<strong>in</strong>ate nouns and occur either with the<br />
jo<strong>in</strong>t or with the split read<strong>in</strong>g<br />
Thus, we can conclude that <strong>in</strong> MG coord<strong>in</strong>ate nouns with a shared modifier, the number<br />
feature distributes to both conjuncts s<strong>in</strong>ce both nouns need to specify the same number<br />
feature as their modifier, but the gender feature always shows closest conjunct agreement<br />
(CCA). The shared modifier scopes over coord<strong>in</strong>ate s<strong>in</strong>gular or plural nouns and occurs<br />
with the jo<strong>in</strong>t or split <strong>in</strong>terpretation. In the next section, we will present the analysis of<br />
MG shared modifiers with coord<strong>in</strong>ate nouns.
5.4 The <strong>Modern</strong> <strong>Greek</strong> Modifiers 127<br />
5.4.1 concord agreement with the MG Modifiers<br />
Next, we turn to the analysis of the MG modifiers and whether any of the agreement<br />
systems outl<strong>in</strong>ed <strong>in</strong> K<strong>in</strong>g and Dalrymple (2004), concord, concord/<strong>in</strong>dex and <strong>in</strong>dex,<br />
account for the MG modifiers.<br />
The presentation of coord<strong>in</strong>ate phrases <strong>in</strong> MG modified by a shared modifier, such as an<br />
adjective or a quantifier, showed that a shared modifier agrees <strong>in</strong> number with both nouns<br />
but <strong>in</strong> gender with the closest conjunct. The modifier scopes over coord<strong>in</strong>ate nouns and<br />
occurs with the jo<strong>in</strong>t and split read<strong>in</strong>gs, while the conjuncts refer to the same or different<br />
<strong>in</strong>dividuals, respectively. Therefore, we argue that MG modifiers follow the concord<br />
system. This system would predict the follow<strong>in</strong>g patterns <strong>in</strong> MG:<br />
Table 5.1: CONCORD SYSTEM<br />
ANIMATE INANIMATE<br />
JOINT READ <br />
MODsg<br />
Nsg & Nsg SPLIT READ <br />
MODpl<br />
JOINT READ <br />
Npl & Npl SPLIT READ <br />
MODpl<br />
JOINT READ * *<br />
Nsg & Nsg SPLIT READ * *<br />
The follow<strong>in</strong>g examples will show whether the predictions of the concord system<br />
are borne out <strong>in</strong> MG. First, we will present coord<strong>in</strong>ated s<strong>in</strong>gular nouns with the same<br />
and mixed gender, and then coord<strong>in</strong>ated plural nouns with the same and mixed gender.<br />
Accord<strong>in</strong>g to the MG data, a shared s<strong>in</strong>gular modifier modifies coord<strong>in</strong>ated s<strong>in</strong>gular nouns<br />
and agrees <strong>in</strong> gender with the closest conjunct. The shared modifier may occur with the<br />
jo<strong>in</strong>t read<strong>in</strong>g. Consider the follow<strong>in</strong>g example:<br />
(32) O Kostas <strong>in</strong>e kalos<br />
the Kostas is good.sg.masc<br />
s<strong>in</strong>adelfos<br />
colleague.sg.masc<br />
‘Kostas is a good friend and colleague’<br />
filos<br />
friend.sg.masc<br />
A s<strong>in</strong>gular attributive adjective scopes over the coord<strong>in</strong>ate nouns which denote the<br />
same <strong>in</strong>dividual. The adjective kalos ‘good’ requires a sg value for the concord num<br />
feature. The concord num features of the two nouns are also sg and match the featurevalue<br />
of the adjective. The concord gend feature of the adjective agrees with the closest<br />
conjunct, which is the noun filos ‘friend’, although <strong>in</strong> this case the second conjunct shares<br />
the same gender. The adjective does not impose any restrictions on the <strong>in</strong>dex feature and<br />
<strong>in</strong>dex agreement is not constra<strong>in</strong>ed. Therefore, it shows concord agreement only and it<br />
can occur with the jo<strong>in</strong>t <strong>in</strong>terpretation. The lexical entry and f-structure without gender<br />
agreement are as follows:<br />
(33) ‘kalos filos ke s<strong>in</strong>adelfos’<br />
kalos: (↑pred) = ‘good’<br />
((adj ∈↑) concord num) = sg<br />
ke<br />
and
5.4 The <strong>Modern</strong> <strong>Greek</strong> Modifiers 128<br />
ADJ<br />
kalos<br />
NP<br />
N<br />
filos<br />
N<br />
Conj<br />
ke<br />
N<br />
s<strong>in</strong>adelfos<br />
⎡ <br />
⎢ adj pred ‘good’<br />
⎢<br />
c: ⎢<br />
⎣<br />
⎤<br />
⎥<br />
<strong>in</strong>dex [num sg] ⎥<br />
⎧ ⎡<br />
⎤⎫<br />
⎥<br />
pred ‘friend’ ⎥<br />
⎢<br />
⎥ ⎥<br />
b: ⎣concord<br />
[num sg] ⎦ ⎥<br />
⎪⎨<br />
⎪⎬ ⎥<br />
<strong>in</strong>dex [num sg] ⎥<br />
⎡<br />
⎤ ⎥<br />
pred ‘colleague’ ⎥<br />
⎢<br />
⎥ ⎥<br />
g: ⎣concord<br />
[num sg] ⎦ ⎥<br />
⎪⎩<br />
⎪⎭<br />
⎦<br />
<strong>in</strong>dex [num sg]<br />
To account for the phenomenon of closest conjunct agreement <strong>in</strong> the gender feature,<br />
we will assume the analysis proposed by Dalrymple and Hristov (2010), which follows the<br />
basic notions of Kuhn and Sadler (2007). For these cases where the gender of the modifier<br />
is the same as the gender of one of the conjuncts only, LFG <strong>in</strong>troduces the strategy of<br />
Closest Conjunct <strong>Agreement</strong> that captures patterns where the conjuncts <strong>in</strong> a coord<strong>in</strong>ate<br />
structure differ <strong>in</strong> a feature value and the modifier agrees only with one of them usually<br />
the closest one, but scopes over both. Considerable work on CCA has been done <strong>in</strong> a<br />
number of languages by Johannessen (1996), Sadler (1999), Sadler (2003), Yatabe (2004),<br />
Villavicencio et al. (2005), Kuhn and Sadler (2007).<br />
Based on Kuhn and Sadler (2007), Dalrymple and Hristov (2010) handle the various<br />
agreement patterns by def<strong>in</strong><strong>in</strong>g metavariables <strong>in</strong> order to allow reference to peripheral<br />
conjuncts <strong>in</strong> a coord<strong>in</strong>ate phrase. They adopt the notation of Kuhn and Sadler (2007), fL<br />
and fL, where f is replaced by any expression that refers to an f-structure and they propose<br />
that these def<strong>in</strong>itions refer to the lefmost and rightmost conjunct <strong>in</strong> a coord<strong>in</strong>ate phrase,<br />
while it can also refer to a non-coord<strong>in</strong>ate phrase. They also reta<strong>in</strong> the assumption that<br />
<strong>in</strong>dex is a nondistributive feature and concord is a distributive feature, and they state<br />
specific agreement requirements based on the two features, while f represents an arbitrary<br />
f-structure. We will not cite the options for the <strong>in</strong>dex feature but only for the concord<br />
feature s<strong>in</strong>ce we are only <strong>in</strong>terested <strong>in</strong> the concord feature agreement 6 . These are:<br />
(34) • (f concord): the concord features of a noncoord<strong>in</strong>ate phrase or each conjunct<br />
of a coord<strong>in</strong>ate phrase (the standard <strong>in</strong>terpretation of this coord<strong>in</strong>ate phrase)<br />
• (fL concord): the concord features of a noncoord<strong>in</strong>ate phrase or the leftmost<br />
conjunct of a coord<strong>in</strong>ate phrase<br />
• (fR concord): the concord features of a noncoord<strong>in</strong>ate phrase or the rightmost<br />
conjunct of a coord<strong>in</strong>ate phrase<br />
In concord, Dalrymple and Hristov (2010) do not dist<strong>in</strong>guish between ‘optional’ closest<br />
conjunct agreement and exclusive closest conjunct agreement as <strong>in</strong> <strong>in</strong>dex, s<strong>in</strong>ce <strong>in</strong> cases<br />
when all conjuncts have the same concord value it is impossible to dist<strong>in</strong>guish between<br />
concord agreement with the closest conjunct and concord agreement with all conjuncts.<br />
Instead, they dist<strong>in</strong>guish only between the strongest requirement of distributive concord<br />
agreement and the weaker requirement of agreement with the closest conjunct.<br />
The expressions (f <strong>in</strong>dex) and (f concord) have their standard LFG mean<strong>in</strong>g where<br />
<strong>in</strong>dex is a nondistributive (resolv<strong>in</strong>g) feature and concord is a distributive feature. The<br />
expression f (L) has the follow<strong>in</strong>g def<strong>in</strong>ition 7 :<br />
6 The <strong>in</strong>dex feature allows for more options with regard to the concord feature s<strong>in</strong>ce it allows for<br />
optional or obligatory agreement with the closest conjunct (Dalrymple and Hristov, 2010, 197).<br />
7 ∈* picks out an arbitrarily deeply embedded member to account for nested coord<strong>in</strong>ation. Any con-<br />
stra<strong>in</strong>ts under attributes <strong>in</strong> a functional uncerta<strong>in</strong>ty path are off-path constra<strong>in</strong>ts, regulat<strong>in</strong>g the f-structures
5.4 The <strong>Modern</strong> <strong>Greek</strong> Modifiers 129<br />
(35) f (L) ≡ f ∈*<br />
¬[(←∈) f→]<br />
These def<strong>in</strong>itions refer to any leftmost or rightmost conjunct, and with embedded coord<strong>in</strong>ations<br />
they may refer to a coord<strong>in</strong>ate phrase, which is leftmost or rightmost <strong>in</strong> relation<br />
to the other conjuncts, but which conta<strong>in</strong>s conjuncts itself. Dalrymple and Hristov (2010)<br />
also add the requirement that the f-structure that is the controller of agreement must not<br />
itself be a coord<strong>in</strong>ate structure:<br />
(38) fL ≡ f ∈* : ¬(fL ∈)<br />
¬[(←∈)
5.4 The <strong>Modern</strong> <strong>Greek</strong> Modifiers 130<br />
ADJ<br />
kalos<br />
NP<br />
N<br />
filos<br />
N<br />
Conj<br />
ke<br />
N<br />
s<strong>in</strong>adelfos<br />
⎡ <br />
⎢ adj pred ‘good’<br />
⎢<br />
c: ⎢<br />
⎣<br />
⎤<br />
⎥<br />
<strong>in</strong>dex [num sg]<br />
⎥<br />
⎧ ⎡<br />
⎤⎫<br />
⎥<br />
pred ‘friend’ ⎥<br />
⎢ ⎥<br />
⎥<br />
⎢<br />
num sg ⎥ ⎥<br />
⎢concord<br />
⎥ ⎥<br />
b: ⎢ gend masc ⎥ ⎥<br />
⎥ ⎥<br />
⎢ ⎥ ⎥<br />
⎢<br />
⎣ num sg<br />
⎥ ⎥<br />
⎪⎨<br />
<strong>in</strong>dex<br />
⎦ ⎥<br />
gend masc ⎪⎬ ⎥<br />
⎡<br />
⎤ ⎥<br />
pred ‘colleague’ ⎥<br />
⎢ ⎥<br />
⎥<br />
⎢<br />
num sg ⎥ ⎥<br />
⎢concord<br />
⎥ ⎥<br />
g: ⎢ gend masc ⎥ ⎥<br />
⎥ ⎥<br />
⎢ ⎥ ⎥<br />
⎢<br />
⎣ num sg<br />
⎥ ⎥<br />
⎪⎩<br />
<strong>in</strong>dex<br />
⎦ ⎥<br />
⎦<br />
gend masc ⎪⎭<br />
Thus, the modifiers show distributed concord num agreement and closest conjunct<br />
(i.e. leftmost) agreement <strong>in</strong> gender.<br />
Coord<strong>in</strong>ated s<strong>in</strong>gular nouns are also modified by a shared modifier, which agrees <strong>in</strong><br />
number with both nouns and <strong>in</strong> gender with the closest conjunct (i.e. leftmost), while the<br />
modifier occurs with the split read<strong>in</strong>g too. Consider two s<strong>in</strong>gular concrete nouns:<br />
(40) O<br />
the<br />
Kostas<br />
Kostas<br />
exi<br />
has<br />
oreo<br />
nice.sg.neut<br />
‘Kostas has a nice car and house’<br />
aftok<strong>in</strong>ito<br />
car.sg.neut<br />
ke<br />
and<br />
spiti<br />
house.sg.neut<br />
The adjective oreo ‘nice’ scopes over coord<strong>in</strong>ate nouns that refer to different entities<br />
and specifies a sg value for the concord num feature. The nouns have a sg value <strong>in</strong> their<br />
concord num feature as specified by the adjective. Also, the adjective does not specify<br />
any restrictions on the <strong>in</strong>dex feature and therefore it shows only concord agreement.<br />
The adjective agrees <strong>in</strong> gender with the closest conjunct, which is the noun aftok<strong>in</strong>ito ‘car’,<br />
although it shows agreement with the second conjunct too due to the uniform gender of<br />
the conjuncts. The split read<strong>in</strong>g is allowed s<strong>in</strong>ce <strong>in</strong>dex agreement is not constra<strong>in</strong>ed. The<br />
relevant lexical entry and f-structure are illustrated below:<br />
(41) ‘oreo aftok<strong>in</strong>ito ke spiti’<br />
oreo: (↑pred) = ‘nice’<br />
((adj ∈↑) concord num) = sg<br />
((adj ∈↑)L concord gend) = neut
5.4 The <strong>Modern</strong> <strong>Greek</strong> Modifiers 131<br />
NP<br />
ADJ<br />
oreo N<br />
aftok<strong>in</strong>ito<br />
N<br />
Conj<br />
ke<br />
N<br />
spiti<br />
⎡ <br />
⎢adj<br />
pred ‘nice’<br />
⎢<br />
c: ⎢<br />
⎣<br />
⎤<br />
⎥<br />
<strong>in</strong>dex [num pl]<br />
⎥<br />
⎧ ⎡<br />
⎤⎫<br />
⎥<br />
pred ‘car<br />
⎥<br />
⎢ ⎥<br />
⎥<br />
⎢<br />
num sg ⎥ ⎥<br />
⎢concord<br />
⎥ ⎥<br />
b: ⎢ gend neut ⎥ ⎥<br />
⎥ ⎥<br />
⎢ ⎥ ⎥<br />
⎢<br />
⎣ num sg<br />
⎥ ⎥<br />
⎪⎨<br />
<strong>in</strong>dex<br />
⎦ ⎥<br />
gend neut ⎪⎬ ⎥<br />
⎡<br />
⎤ ⎥<br />
pred ‘house’ ⎥<br />
⎢ ⎥<br />
⎥<br />
⎢<br />
num sg ⎥ ⎥<br />
⎢concord<br />
⎥ ⎥<br />
g: ⎢ gend neut ⎥ ⎥<br />
⎥ ⎥<br />
⎢ ⎥ ⎥<br />
⎢<br />
⎣ num sg<br />
⎥ ⎥<br />
⎪⎩<br />
<strong>in</strong>dex<br />
⎦ ⎥<br />
⎦<br />
gend neut ⎪⎭<br />
Coord<strong>in</strong>ate s<strong>in</strong>gular nouns with different gender are also modified by the same modifier.<br />
The modifier shows distributive concord num agreement and occurs with a jo<strong>in</strong>t and a<br />
split read<strong>in</strong>g while it shows closest conjunct agreement <strong>in</strong> gender. In example (42), the<br />
s<strong>in</strong>gular adjective megalos agrees with both conjuncts <strong>in</strong> number and with the first conjunct<br />
only <strong>in</strong> gender. It specifies a sg value <strong>in</strong> the concord num feature and the two nouns<br />
have the same feature value. <strong>in</strong>dex agreement is not constra<strong>in</strong>ed by the adjective and<br />
the jo<strong>in</strong>t read<strong>in</strong>g is allowed. The adjective also specifies masc value <strong>in</strong> the concord<br />
gend feature agree<strong>in</strong>g only with the closest conjunct (i.e. leftmost). Thus, the adjective<br />
shows distributive concord agreement <strong>in</strong> number and closest conjunct agreement (CCA)<br />
<strong>in</strong> gender:<br />
(42) Megalos kopos ke prospathia xriazete<br />
Great.sg.masc labour.sg.masc and effort.sg.fem is-needed.sg<br />
ja na s<strong>in</strong>gedrosi to fotografiko iliko...<br />
for to gather the photographic material...<br />
‘Great labour and effort are needed to gather the photographic material...’<br />
The lexical entry and f-structure are illustrated below:<br />
(43) ‘megalos kopos ke prospathia’<br />
megalos: (↑pred) = ‘great’<br />
((adj ∈↑) concord num) = sg<br />
((adj ∈↑)L concord gend) = masc<br />
ADJ<br />
megalos<br />
NP<br />
N<br />
kopos<br />
N<br />
Conj<br />
ke<br />
N<br />
prospathia
5.4 The <strong>Modern</strong> <strong>Greek</strong> Modifiers 132<br />
⎡ <br />
⎢ adj pred ‘great’<br />
⎢<br />
c: ⎢<br />
⎣<br />
⎤<br />
⎥<br />
<strong>in</strong>dex [num sg]<br />
⎥<br />
⎧ ⎡<br />
⎤⎫<br />
⎥<br />
pred ‘labour’ ⎥<br />
⎢ ⎥<br />
⎥<br />
⎢<br />
num sg ⎥ ⎥<br />
⎢concord<br />
⎥ ⎥<br />
b: ⎢ gend masc ⎥ ⎥<br />
⎥ ⎥<br />
⎢ ⎥ ⎥<br />
⎢<br />
⎣ num sg<br />
⎥ ⎥<br />
⎪⎨<br />
<strong>in</strong>dex<br />
⎦ ⎥<br />
gend masc ⎪⎬ ⎥<br />
⎡<br />
⎤ ⎥<br />
pred ‘effort’ ⎥<br />
⎢ ⎥<br />
⎥<br />
⎢<br />
num sg ⎥ ⎥<br />
⎢concord<br />
⎥ ⎥<br />
g: ⎢ gend fem ⎥ ⎥<br />
⎥ ⎥<br />
⎢ ⎥ ⎥<br />
⎢<br />
⎣ num sg<br />
⎥ ⎥<br />
⎪⎩<br />
<strong>in</strong>dex<br />
⎦ ⎥<br />
⎦<br />
gend fem ⎪⎭<br />
Let us turn to coord<strong>in</strong>ated plural nouns with a shared modifier. Coord<strong>in</strong>ated plural<br />
nouns are modified by a plural adjective which agrees <strong>in</strong> number with both nouns and <strong>in</strong><br />
gender with the closest conjunct and occurs with the jo<strong>in</strong>t read<strong>in</strong>g. In example (44), the<br />
adjective specifies a pl value <strong>in</strong> the concord num feature. The two nouns also specify a pl<br />
value <strong>in</strong> their concord num feature. The adjective does not impose any <strong>in</strong>dex agreement<br />
requirements and the two nouns refer to the same entity, which is allowed by the concord<br />
system. Also, the adjective specifies fem value <strong>in</strong> the concord gend feature agree<strong>in</strong>g<br />
only with the closest conjunct (i.e. leftmost). The lexical entry and f-structure are shown<br />
<strong>in</strong> (45):<br />
(44) Isxires antiparathesis ke antiparavoles<br />
strong.pl.fem juxtapositions.pl.fem and contrasts.pl.fem<br />
ekdilothikan st<strong>in</strong> kentriki s<strong>in</strong>edriasi tis Vulis<br />
were-expressed.pl <strong>in</strong>-the central meet<strong>in</strong>g of-the Parliament<br />
‘Strong juxtapositions and contrasts were expressed <strong>in</strong> the central meet<strong>in</strong>g of the<br />
Parliament’<br />
(45) ‘isxires antiparathesis ke antiparavoles’<br />
isxures: (↑pred) = ‘strong’<br />
((adj ∈↑) concord num) = pl<br />
((adj ∈↑)L concord gend) = fem
5.4 The <strong>Modern</strong> <strong>Greek</strong> Modifiers 133<br />
NP<br />
ADJ<br />
isxures N<br />
antiparathesis<br />
N<br />
Conj<br />
ke<br />
N<br />
antiparavoles<br />
⎡ <br />
⎢adj<br />
pred ‘strong’<br />
⎢<br />
c: ⎢<br />
⎣<br />
⎤<br />
⎥<br />
<strong>in</strong>dex [num pl]<br />
⎥<br />
⎧ ⎡<br />
⎤⎫<br />
⎥<br />
pred ‘juxtapositions’ ⎥<br />
⎢ ⎥ ⎥<br />
⎢<br />
num pl ⎥ ⎥<br />
⎢concord<br />
⎥ ⎥<br />
b: ⎢ gend fem ⎥ ⎥<br />
⎥ ⎥<br />
⎢ ⎥ ⎥<br />
⎢<br />
⎣ num pl<br />
⎥ ⎥<br />
⎪⎨<br />
<strong>in</strong>dex<br />
⎦ ⎥<br />
gend fem ⎪⎬ ⎥<br />
⎡<br />
⎤ ⎥<br />
pred ‘contrasts’ ⎥<br />
⎢ ⎥<br />
⎥<br />
⎢<br />
num pl ⎥ ⎥<br />
⎢concord<br />
⎥ ⎥<br />
g: ⎢ gend fem ⎥ ⎥<br />
⎥ ⎥<br />
⎢ ⎥ ⎥<br />
⎢<br />
⎣ num pl<br />
⎥ ⎥<br />
⎪⎩<br />
<strong>in</strong>dex<br />
⎦ ⎥<br />
⎦<br />
gend fem ⎪⎭<br />
Coord<strong>in</strong>ated plural nouns are also modified by a plural adjective which agrees <strong>in</strong> number<br />
with both nouns and <strong>in</strong> gender with the closest one and occurs with the split read<strong>in</strong>g too.<br />
In example (46), the adjective specifies a pl value <strong>in</strong> the concord num feature. The two<br />
nouns also have a pl value <strong>in</strong> their concord num feature. The split <strong>in</strong>terpretation is<br />
generated s<strong>in</strong>ce the <strong>in</strong>dex feature is not constra<strong>in</strong>ed by the adjective. Thus, the adjective<br />
shows only concord agreement. Also, the adjective specifies neut value <strong>in</strong> the concord<br />
gend feature agree<strong>in</strong>g only with the closest conjunct (i.e. leftmost), although the second<br />
conjunct shares the same gender. The lexical entry and f-structure are shown <strong>in</strong> (47):<br />
(46) Pola mikra agoria<br />
many<br />
voithia<br />
help<br />
small.pl.neut boys.pl.neut<br />
‘Many small boys and girls need help’<br />
(47) ‘mikra agoria ke koritsia’<br />
mikra: (↑pred) = ‘young’<br />
((adj ∈↑) concord num) = pl<br />
((adj ∈↑)L concord gend) = neut<br />
ke<br />
and<br />
koritsia<br />
girls.pl.neut<br />
xriazonte<br />
need.pl
5.4 The <strong>Modern</strong> <strong>Greek</strong> Modifiers 134<br />
NP<br />
ADJ<br />
mikra N<br />
agoria<br />
N<br />
Conj<br />
ke<br />
N<br />
koritsia<br />
⎡ <br />
⎢adj<br />
pred ‘young’<br />
⎢<br />
c: ⎢<br />
⎣<br />
⎤<br />
⎥<br />
<strong>in</strong>dex [num pl]<br />
⎥<br />
⎧ ⎡<br />
⎤⎫<br />
⎥<br />
pred ‘boys ⎥<br />
⎢ ⎥<br />
⎥<br />
⎢<br />
num pl ⎥ ⎥<br />
⎢concord<br />
⎥ ⎥<br />
b: ⎢ gend neut ⎥ ⎥<br />
⎥ ⎥<br />
⎢ ⎥ ⎥<br />
⎢<br />
⎣ num pl<br />
⎥ ⎥<br />
⎪⎨<br />
<strong>in</strong>dex<br />
⎦ ⎥<br />
gend neut ⎪⎬ ⎥<br />
⎡<br />
⎤ ⎥<br />
pred ‘girls’ ⎥<br />
⎢ ⎥<br />
⎥<br />
⎢<br />
num pl ⎥ ⎥<br />
⎢concord<br />
⎥ ⎥<br />
g: ⎢ gend neut ⎥ ⎥<br />
⎥ ⎥<br />
⎢ ⎥ ⎥<br />
⎢<br />
⎣ num pl<br />
⎥ ⎥<br />
⎪⎩<br />
<strong>in</strong>dex<br />
⎦ ⎥<br />
⎦<br />
gend neut ⎪⎭<br />
Plural conjuncts with different gender are also modified by the same adjective. The<br />
adjective agrees <strong>in</strong> number with both nouns and occurs with a jo<strong>in</strong>t and a split read<strong>in</strong>g,<br />
while it agrees <strong>in</strong> gender with the closest conjunct. Below we illustrate an example with a<br />
split read<strong>in</strong>g. In example (48), the plural adjective katastrofiki specifies a plural value <strong>in</strong><br />
its concord num feature; the two coord<strong>in</strong>ate nouns have a pl value <strong>in</strong> their concord<br />
num feature, while <strong>in</strong>dex agreement is not constra<strong>in</strong>ed. The adjective also specifies a<br />
masc value <strong>in</strong> the concord gend feature, which agrees only with the first conjunct (i.e.<br />
leftmost). The modifier occurs with the split read<strong>in</strong>g s<strong>in</strong>ce the two conjuncts refer to<br />
different natural phenomena:<br />
(48) Katastrofiki sismi<br />
ke plimires<br />
destructive.pl.masc earthquakes.pl.masc and floods.pl.fem<br />
t<strong>in</strong> poli apo to 1965<br />
the town s<strong>in</strong>ce the 1965<br />
‘Destructive earthquakes and floods hit the town s<strong>in</strong>ce 1965’<br />
The lexical entry and f-structure of the coord<strong>in</strong>ate phrase are shown below:<br />
(49) ‘katastrofiki sismi ke plimires’<br />
katastrofiki: (↑pred) = ‘destructive’<br />
((adj ∈↑) concord num) = pl<br />
((adj ∈↑)L concord gend) = masc<br />
ADJ<br />
katastrofiki<br />
NP<br />
N<br />
sismi<br />
N<br />
Conj<br />
ke<br />
N<br />
plimires<br />
plitun<br />
hit.pl
5.5 The <strong>Modern</strong> <strong>Greek</strong> Def<strong>in</strong>ite Determ<strong>in</strong>er 135<br />
⎡ <br />
⎢ adj pred ‘destructive’<br />
⎢<br />
c: ⎢<br />
⎣<br />
⎤<br />
⎥<br />
<strong>in</strong>dex [num pl]<br />
⎥<br />
⎧ ⎡<br />
⎤⎫<br />
⎥<br />
pred ‘earthquakes’ ⎥<br />
⎢ ⎥<br />
⎥<br />
⎢<br />
num pl ⎥ ⎥<br />
⎢concord<br />
⎥ ⎥<br />
b: ⎢ gend masc ⎥ ⎥<br />
⎥ ⎥<br />
⎢ ⎥ ⎥<br />
⎢<br />
⎣ num pl<br />
⎥ ⎥<br />
⎪⎨<br />
<strong>in</strong>dex<br />
⎦ ⎥<br />
gend masc ⎪⎬ ⎥<br />
⎡<br />
⎤ ⎥<br />
pred ‘floods’ ⎥<br />
⎢ ⎥<br />
⎥<br />
⎢<br />
num pl ⎥ ⎥<br />
⎢concord<br />
⎥ ⎥<br />
g: ⎢ gend fem ⎥ ⎥<br />
⎥ ⎥<br />
⎢ ⎥ ⎥<br />
⎢<br />
⎣ num pl<br />
⎥ ⎥<br />
⎪⎩<br />
<strong>in</strong>dex<br />
⎦ ⎥<br />
⎦<br />
gend fem ⎪⎭<br />
To summarise, MG shared s<strong>in</strong>gular and plural modifiers show distributive concord<br />
num agreement and closest conjunct (CCA) gend agreement. The modifier agrees <strong>in</strong><br />
number with both nouns and <strong>in</strong> gender with the closest conjunct. The <strong>in</strong>terpretation of<br />
both the jo<strong>in</strong>t and split read<strong>in</strong>g is equally possible when a shared modifier scopes over<br />
coord<strong>in</strong>ate nouns s<strong>in</strong>ce the modifier does not specify any restrictions on the <strong>in</strong>dex feature<br />
of the coord<strong>in</strong>ate phrase.<br />
5.5 The <strong>Modern</strong> <strong>Greek</strong> Def<strong>in</strong>ite Determ<strong>in</strong>er<br />
5.5.1 Uniform Number and Gender <strong>in</strong> Coord<strong>in</strong>ate <strong>Noun</strong>s<br />
As opposed to MG modifiers, MG determ<strong>in</strong>ers do not have the same behaviour when they<br />
modify coord<strong>in</strong>ate nouns. In both coord<strong>in</strong>ate s<strong>in</strong>gular and plural structures, a separate<br />
determ<strong>in</strong>er needs to appear before each conjunct to allow a split read<strong>in</strong>g and each determ<strong>in</strong>er<br />
agrees with the noun it modifies <strong>in</strong> gender, number and case 8 . A shared determ<strong>in</strong>er,<br />
unlike a shared modifier, needs to agree <strong>in</strong> number and gender with both conjuncts and<br />
admits only a jo<strong>in</strong>t read<strong>in</strong>g.<br />
In coord<strong>in</strong>ate s<strong>in</strong>gular nouns, <strong>in</strong> example (50), the mascul<strong>in</strong>e def<strong>in</strong>ite determ<strong>in</strong>er o<br />
appears before each noun and modifies it. Similarly, <strong>in</strong> example (51), the fem<strong>in</strong><strong>in</strong>e def<strong>in</strong>ite<br />
determ<strong>in</strong>er i is placed before each noun and modifies it. In both cases, the determ<strong>in</strong>er agrees<br />
<strong>in</strong> number and gender with the noun it modifies. Both examples have a split <strong>in</strong>terpretation<br />
s<strong>in</strong>ce there is a separate determ<strong>in</strong>er before each conjunct and the two conjuncts refer to<br />
different <strong>in</strong>dividuals, while the verb and predicative adjective show plural agreement:<br />
(50) O sizigos<br />
the.sg.masc husband.sg.masc<br />
tis <strong>in</strong>e nei<br />
her are.pl young.pl<br />
‘Her husband and father are young’<br />
(constructed)<br />
ke<br />
and<br />
o<br />
the.sg.masc<br />
pateras<br />
father.sg.masc<br />
8 It might be thought that the occurrence of the def<strong>in</strong>ite determ<strong>in</strong>er before each conjunct is attributed<br />
to the fact that it has an affixal status. The MG def<strong>in</strong>ite determ<strong>in</strong>er, however, does not behave as an affix<br />
and therefore we cannot assume such an analysis.
5.5 The <strong>Modern</strong> <strong>Greek</strong> Def<strong>in</strong>ite Determ<strong>in</strong>er 136<br />
(51) I sizigos ke<br />
the.sg.fem wife.sg.fem and<br />
<strong>in</strong>e nees<br />
are.pl young.pl<br />
‘His wife and daughter are young’<br />
(constructed)<br />
i<br />
the.sg.fem<br />
kori<br />
daughter.sg.fem<br />
A s<strong>in</strong>gle def<strong>in</strong>ite determ<strong>in</strong>er scopes over coord<strong>in</strong>ate nouns without the requirement for<br />
a second determ<strong>in</strong>er <strong>in</strong> the second conjunct, only when the coord<strong>in</strong>ate phrase has a jo<strong>in</strong>t<br />
<strong>in</strong>terpretation and the two nouns denote the same referent. The def<strong>in</strong>ite determ<strong>in</strong>er agrees<br />
<strong>in</strong> number and gender with both nouns, as shown below:<br />
(52) O filos ke s<strong>in</strong>adelfos<br />
the.sg.masc friend.sg.masc and colleague.sg.masc<br />
gnorizo xronia<br />
I-know for-years<br />
‘The friend and colleague who I know for years’<br />
(constructed)<br />
pu<br />
who<br />
tu<br />
his<br />
ton<br />
him<br />
In sentence (52), the <strong>in</strong>dividual denoted is filos ‘friend’ and s<strong>in</strong>adelfos ‘colleague’ at the<br />
same time. In MG, this pattern is rather common not only <strong>in</strong> animate nouns but also <strong>in</strong><br />
<strong>in</strong>animate nouns too 9 . For example, <strong>in</strong> (53) the s<strong>in</strong>gular def<strong>in</strong>ite determ<strong>in</strong>er nouns agrees<br />
<strong>in</strong> number and gender with both conjuncts and admits the jo<strong>in</strong>t <strong>in</strong>terpretation s<strong>in</strong>ce the<br />
two conjuncts fovos ‘fear’ and tromos ‘terror’ are two closely related notions and they refer<br />
to the same concept:<br />
(53) O fovos ke tromos emfanistike<br />
the.sg.masc fear.sg.masc and terror.sg.masc appeared.sg<br />
st<strong>in</strong> perioxi<br />
<strong>in</strong>-the area<br />
‘The fear and terror appeared aga<strong>in</strong> <strong>in</strong> the area’<br />
http://www.hri.org/E/1996/96-10-18.dir/keimena/greece/greece2.htm<br />
ksana<br />
aga<strong>in</strong><br />
Similarly, <strong>in</strong> example (54), the s<strong>in</strong>gular def<strong>in</strong>ite determ<strong>in</strong>er agrees <strong>in</strong> number and gender<br />
with the nouns grafio ‘office’ and iatrio ‘surgery’ and the determ<strong>in</strong>er admits the jo<strong>in</strong>t<br />
<strong>in</strong>terpretation s<strong>in</strong>ce the conjuncts denote the same place:<br />
(54) To grafio ke iatrio<br />
the.sg.neut office.sg.neut and surgery.sg.neut<br />
j<strong>in</strong>ekologu vriskete sto kentro<br />
gynaecologist is.sg <strong>in</strong>-the centre<br />
‘The office and surgery of the gynaecologist is <strong>in</strong> the centre’<br />
(constructed)<br />
tu<br />
of-the<br />
To summarise, when a common determ<strong>in</strong>er with wide scope modifies coord<strong>in</strong>ate s<strong>in</strong>gular<br />
nouns, it agrees with both conjuncts <strong>in</strong> number and gender and it occurs only with a jo<strong>in</strong>t<br />
<strong>in</strong>terpretation. To generate a split <strong>in</strong>terpretation, the determ<strong>in</strong>er needs to be repeated<br />
before each conjunct.<br />
In plural number, the MG def<strong>in</strong>ite determ<strong>in</strong>er behaves different from the def<strong>in</strong>ite determ<strong>in</strong>ers<br />
of languages such as Italian, German, Dutch, F<strong>in</strong>nish and Brazilian Portuguese,<br />
which fully allow a split <strong>in</strong>terpretation <strong>in</strong> addition to the jo<strong>in</strong>t one, when a shared plural<br />
9 We have exam<strong>in</strong>ed concrete and abstract nouns separately to check whether there are any differences<br />
between the two but no differences were found.
5.5 The <strong>Modern</strong> <strong>Greek</strong> Def<strong>in</strong>ite Determ<strong>in</strong>er 137<br />
determ<strong>in</strong>er modifies coord<strong>in</strong>ate plural nouns (Heycock and Zamparelli, 1999, 2000; K<strong>in</strong>g<br />
and Dalrymple, 2004). In MG, the plural determ<strong>in</strong>er imposes the same restrictions as the<br />
s<strong>in</strong>gular determ<strong>in</strong>er when it modifies coord<strong>in</strong>ate nouns. Thus, a separate plural determ<strong>in</strong>er<br />
needs to modify each conjunct to generate a split <strong>in</strong>terpretation and each determ<strong>in</strong>er agrees<br />
<strong>in</strong> number and gender with the noun it modifies, as shown below:<br />
(55) I<br />
paterades ke i<br />
the.pl.masc fathers.pl.masc and the.pl.masc<br />
papudes<br />
<strong>in</strong>e poli demeni<br />
grandfathers.pl.masc are.pl very close<br />
‘The fathers and grandfathers are very close’<br />
(constructed)<br />
A shared def<strong>in</strong>ite determ<strong>in</strong>er scopes over coord<strong>in</strong>ate plural nouns, agrees <strong>in</strong> number and<br />
gender with both nouns but it occurs only with the jo<strong>in</strong>t <strong>in</strong>terpretation, denot<strong>in</strong>g the same<br />
group of <strong>in</strong>dividuals. This is shown below:<br />
(56) I<br />
paterades ke<br />
the.pl.masc<br />
ikojenias<br />
family<br />
fathers.pl.masc and<br />
‘The fathers and protectors of the family’<br />
(constructed)<br />
prostates<br />
protectors.pl.masc<br />
tis<br />
of-the<br />
This exact same pattern occurs <strong>in</strong> <strong>in</strong>animate nouns too. In the follow<strong>in</strong>g examples, the<br />
two conjuncts are modified by a s<strong>in</strong>gle plural def<strong>in</strong>ite determ<strong>in</strong>er, which agrees with the<br />
conjuncts <strong>in</strong> number and gender, while the jo<strong>in</strong>t read<strong>in</strong>g is the only possible <strong>in</strong>terpretation:<br />
(57) Ta somata ke kalorifer thelun<br />
the.pl.neut radiators.pl.neut and heaters.pl.neut need.pl<br />
alagma me to aerio<br />
to-be-changed with the gas<br />
‘The radiators and heaters need to be changed because of the gas’<br />
(constructed)<br />
(58) I antiparathesis ke antiparavoles tis<br />
the.pl.fem contrasts.pl.fem and juxtapositions.pl.fem of<br />
DAKE itan nootropies tu parelthontos<br />
DAKE were.pl mentalities of-the past<br />
‘The contrasts and juxtapositions of DAKE were mentalities of the past’<br />
(G.N.:Eleftherotipia)<br />
In example (57), the two concrete nouns denote the same items and therefore the<br />
jo<strong>in</strong>t <strong>in</strong>terpretation is generated. In example (58), the two abstract nouns also refer to<br />
the same notion and therefore the jo<strong>in</strong>t <strong>in</strong>terpretation is only possible. Thus, a shared<br />
plural determ<strong>in</strong>er, which scopes over coord<strong>in</strong>ate plural nouns, occurs only with the jo<strong>in</strong>t<br />
<strong>in</strong>terpretation.<br />
F<strong>in</strong>ally, we need to note that a plural def<strong>in</strong>ite determ<strong>in</strong>er cannot modify coord<strong>in</strong>ate<br />
s<strong>in</strong>gular nouns. The follow<strong>in</strong>g examples are ungrammatical:<br />
(59) *I ipurgos ke ifipurgos<br />
the.pl.masc<br />
kivernisis<br />
government<br />
m<strong>in</strong>ister.sg.masc and deputy-m<strong>in</strong>ister.sg.masc<br />
‘The m<strong>in</strong>ister and deputy-m<strong>in</strong>ister of the government’<br />
(constructed)<br />
tis<br />
of-the
5.5 The <strong>Modern</strong> <strong>Greek</strong> Def<strong>in</strong>ite Determ<strong>in</strong>er 138<br />
The f<strong>in</strong>d<strong>in</strong>gs <strong>in</strong> MG are summarised as follows:<br />
• A s<strong>in</strong>gular def<strong>in</strong>ite determ<strong>in</strong>er modifies s<strong>in</strong>gular coord<strong>in</strong>ate nouns and occurs with<br />
the jo<strong>in</strong>t read<strong>in</strong>g only<br />
• A plural def<strong>in</strong>ite determ<strong>in</strong>er modifies plural coord<strong>in</strong>ate nouns and occurs with the<br />
jo<strong>in</strong>t read<strong>in</strong>g only<br />
• In both cases, the def<strong>in</strong>ite determ<strong>in</strong>er agrees <strong>in</strong> number and gender with both nouns<br />
• A plural def<strong>in</strong>ite determ<strong>in</strong>er cannot modify s<strong>in</strong>gular coord<strong>in</strong>ate nouns and occur either<br />
with the jo<strong>in</strong>t or the split read<strong>in</strong>g<br />
In the plural, there are some very limited exceptions to the generalisation stated above<br />
- namely a few cases <strong>in</strong> which a shared plural def<strong>in</strong>ite determ<strong>in</strong>er scopes over coord<strong>in</strong>ate<br />
nouns and occurs with a split <strong>in</strong>terpretation. The first exceptional pattern occurs <strong>in</strong> a<br />
specific group of animate coord<strong>in</strong>ate nouns where the two conjuncts denote closely related<br />
concepts. In the follow<strong>in</strong>g examples, the shared plural mascul<strong>in</strong>e determ<strong>in</strong>er i scopes over<br />
plural animate nouns and occurs with a split <strong>in</strong>terpretation. The def<strong>in</strong>ite determ<strong>in</strong>er agrees<br />
<strong>in</strong> number and gender with the two conjuncts:<br />
(60) I<br />
ipurgi<br />
ke ifipurgi<br />
the.pl.masc m<strong>in</strong>isters.pl.masc and deputy-m<strong>in</strong>isters.pl.masc<br />
kivernisis tis ND<br />
government of ND<br />
‘The m<strong>in</strong>isters and deputy-m<strong>in</strong>isters of the government of ND’<br />
(G.N.:Kathimer<strong>in</strong>i)<br />
(61) I<br />
ergazomeni ke<br />
the.pl.masc<br />
anaferun<br />
mention<br />
employed.pl.masc and<br />
‘The employed and unemployed mention it’<br />
(G.N.:Eleftherotipia)<br />
anerji<br />
unemployed.pl.masc<br />
(62) I<br />
iatri<br />
ke odontiatri<br />
the.pl.masc doctors.pl.masc and dentists.pl.masc<br />
simvasis me to strato<br />
contracts with the army<br />
‘The doctors and dentisists avoid any contracts with the army<br />
http://www.omhroi.gr/article2194.htm<br />
to<br />
it<br />
apofevgun<br />
avoid<br />
tis<br />
of-the<br />
In example (60), the two nouns share a s<strong>in</strong>gle def<strong>in</strong>ite determ<strong>in</strong>er but the two conjuncts<br />
refer to different groups of <strong>in</strong>dividuals, allow<strong>in</strong>g only the split <strong>in</strong>terpretation. Semantically,<br />
it is highly unlikely for the same person to be ipurgos ‘m<strong>in</strong>ister’ and ifipurgos ‘deputym<strong>in</strong>ister’<br />
at the same time, while the two nouns are closely related <strong>in</strong> mean<strong>in</strong>g s<strong>in</strong>ce one<br />
is a ‘m<strong>in</strong>ister’ and the other is a ‘deputy-m<strong>in</strong>ister’. Follow<strong>in</strong>g the same reason<strong>in</strong>g, the two<br />
conjuncts <strong>in</strong> example (61), although they are modified by a shared def<strong>in</strong>ite determ<strong>in</strong>er,<br />
they can only convey the split <strong>in</strong>terpretation s<strong>in</strong>ce the <strong>in</strong>terpretation of the sentence does<br />
not allow the people who are ergazomeni ‘employed’ to be anerji ‘unemployed’ at the same<br />
time. Also, the two nouns are semantically related s<strong>in</strong>ce they denote opposite concepts<br />
related to work. F<strong>in</strong>ally, <strong>in</strong> example (62) the two nouns share a s<strong>in</strong>gle def<strong>in</strong>ite determ<strong>in</strong>er<br />
but the split <strong>in</strong>terpretation is generated s<strong>in</strong>ce the two nouns denote two semantically related<br />
concepts iatri ‘doctors’ and odontiatri ‘dentists’.<br />
tis<br />
the
5.5 The <strong>Modern</strong> <strong>Greek</strong> Def<strong>in</strong>ite Determ<strong>in</strong>er 139<br />
The same restricted pattern also occurs <strong>in</strong> plural abstract and concrete 10 nouns which<br />
are semantically related. Thus, a s<strong>in</strong>gle def<strong>in</strong>ite determ<strong>in</strong>er scopes over coord<strong>in</strong>ated plural<br />
nouns and occurs with a split <strong>in</strong>terpretation. The def<strong>in</strong>ite determ<strong>in</strong>er agrees <strong>in</strong> gender and<br />
number with both conjuncts. Consider the follow<strong>in</strong>g examples:<br />
(63) Ta thetika ke<br />
the.pl.neut positives.pl.neut and<br />
ipothesis <strong>in</strong>e pola<br />
case are.pl many<br />
‘The pros and cons of the case are many’<br />
(G.N.:Ependitis)<br />
arnitika<br />
negatives.pl.neut<br />
(64) Ta ithi<br />
ke ethima<br />
the.pl.neut<br />
diamorfothikan<br />
were-formed.pl<br />
morals.pl.neut and customs.pl.neut<br />
‘The morals and customs as they were formed’<br />
(G.M.:VIMAgaz<strong>in</strong>o)<br />
(65) Ta maxeria ke<br />
the.pl.neut knives.pl.neut and<br />
sto trapezi<br />
the table<br />
‘The knives and forks are on the table’<br />
(G.M.:1000+1 Gefsis)<br />
pirunia<br />
forks.pl.neut<br />
opos<br />
as<br />
<strong>in</strong>e<br />
are.pl<br />
tis<br />
of-the<br />
pano<br />
on<br />
In example (63), the two <strong>in</strong>animate abstract nouns refer to different items while the<br />
def<strong>in</strong>ite determ<strong>in</strong>er used for the first conjunct thetika ‘positive’ scopes over the second<br />
conjunct too, the noun arnitika ‘negative’. The only read<strong>in</strong>g available <strong>in</strong> this case is the<br />
split one. Similarly, <strong>in</strong> examples (64) and (65), the coord<strong>in</strong>ate nouns ithi ke ethima ‘morals<br />
and customs’ and maxeria ke pirounia ‘knives and forks’ refer to different notions/entities.<br />
In both cases, a shared def<strong>in</strong>ite determ<strong>in</strong>er scopes over the two conjuncts and the phrases<br />
are acceptable. Thus, <strong>in</strong> all examples the two nouns refer to different entities 11 .<br />
To summarise, syntactically a shared def<strong>in</strong>ite determ<strong>in</strong>er agrees <strong>in</strong> number and gender<br />
with both conjuncts. Semantically, the shared def<strong>in</strong>ite determ<strong>in</strong>er modifies coord<strong>in</strong>ate<br />
nouns and occurs with the jo<strong>in</strong>t read<strong>in</strong>g <strong>in</strong> both s<strong>in</strong>gular and plural number. The split<br />
read<strong>in</strong>g is generated only when each conjunct is modified by a separate determ<strong>in</strong>er. There<br />
are some exceptional cases when a shared def<strong>in</strong>ite determ<strong>in</strong>er modifies coord<strong>in</strong>ated plural<br />
nouns. The shared def<strong>in</strong>ite determ<strong>in</strong>er agrees <strong>in</strong> gender and number with both nouns,<br />
while it scopes over both coord<strong>in</strong>ate nouns and allows a split read<strong>in</strong>g. We will show that<br />
these cases occur under specific conditions. Thus, MG differs from the other languages<br />
discussed by Heycock and Zamparelli (1999, 2000) and K<strong>in</strong>g and Dalrymple (2004) s<strong>in</strong>ce <strong>in</strong><br />
MG a shared def<strong>in</strong>ite determ<strong>in</strong>er, when modify<strong>in</strong>g coord<strong>in</strong>ated s<strong>in</strong>gular and plural nouns,<br />
generates the jo<strong>in</strong>t read<strong>in</strong>g only. In other languages (i.e. Brazilian Portuguese, French,<br />
Italian or German), however, a shared def<strong>in</strong>ite determ<strong>in</strong>er scopes over coord<strong>in</strong>ate nouns<br />
and occurs with the jo<strong>in</strong>t <strong>in</strong>terpretation <strong>in</strong> s<strong>in</strong>gular conjuncts only, but a jo<strong>in</strong>t and a split<br />
<strong>in</strong>terpretation <strong>in</strong> plural conjuncts.<br />
10 In concrete coord<strong>in</strong>ate nouns, the split read<strong>in</strong>g under a shared determ<strong>in</strong>er was found less often but it<br />
was accepted by more than half of the native speakers. S<strong>in</strong>ce the data is based on acceptability judgements,<br />
we will assume that these patterns are possible but occur rarely.<br />
11 As we will show <strong>in</strong> the 5.5.2.1, all these patterns are cases of natural coord<strong>in</strong>ation.
5.5 The <strong>Modern</strong> <strong>Greek</strong> Def<strong>in</strong>ite Determ<strong>in</strong>er 140<br />
5.5.2 Uniform Number and Different Gender <strong>in</strong> Coord<strong>in</strong>ate <strong>Noun</strong>s<br />
Next, we turn to coord<strong>in</strong>ate structures with uniform number and mixed gender nouns<br />
modified by a shared def<strong>in</strong>ite determ<strong>in</strong>er. Conjo<strong>in</strong>ed nouns of uniform number and different<br />
gender, either <strong>in</strong> s<strong>in</strong>gular or <strong>in</strong> plural, are modified by a separate determ<strong>in</strong>er each.<br />
In example (66), the s<strong>in</strong>gular animate nouns require a separate determ<strong>in</strong>er s<strong>in</strong>ce the<br />
nouns have different gender and two different <strong>in</strong>dividuals are denoted, the mascul<strong>in</strong>e noun<br />
antras ‘man’ and the neuter noun pedi ‘child’. Therefore, the sentence with a shared<br />
determ<strong>in</strong>er is ungrammatical:<br />
(66) *O antras<br />
the.sg.masc man.sg.masc<br />
‘The man and child’<br />
(constructed)<br />
ke<br />
and<br />
pedi<br />
child.sg.neut<br />
S<strong>in</strong>gular <strong>in</strong>animate conjuncts show the same behaviour. When the coord<strong>in</strong>ate nouns<br />
have a s<strong>in</strong>gular number and different gender, even if they refer to the same entity, they<br />
require a separate determ<strong>in</strong>er each, otherwise the result is ungrammatical. In example<br />
(67), each conjunct requires a separate determ<strong>in</strong>er, even though the two nouns denote the<br />
same concept. Thus, the coord<strong>in</strong>ate phrase with a shared determ<strong>in</strong>er is ungrammatical:<br />
(67) *To pisma ke isxirognomos<strong>in</strong>i<br />
the.sg.neut spite.sg.neut and stubbornness.sg.fem<br />
‘The spite and stubbornness’<br />
(constructed)<br />
In coord<strong>in</strong>ate phrases with plural conjuncts, the same restrictions hold. Animate nouns<br />
of plural number and different gender cannot occur under the same determ<strong>in</strong>er. Each<br />
conjunct requires a separate def<strong>in</strong>ite determ<strong>in</strong>er, otherwise the sentence is ungrammatical:<br />
(68) *I antres<br />
the.pl.masc men.pl.masc<br />
‘The men and children’<br />
(constructed)<br />
ke<br />
and<br />
pedia<br />
children.pl.neut<br />
Also, plural <strong>in</strong>animate nouns of uniform number and different gender cannot be modified<br />
by a shared def<strong>in</strong>ite determ<strong>in</strong>er. The follow<strong>in</strong>g example is ungrammatical:<br />
(69) *I epeni ke engomia<br />
the.pl.masc praises.pl.masc and commendations.pl.neut<br />
s<strong>in</strong>thetun to sk<strong>in</strong>iko st<strong>in</strong> metaolimpiaki Ath<strong>in</strong>a<br />
comprise.pl the scenery of-the postolympic Athens<br />
‘Praises and commendations comprise the scenery of the postolympic Athens’<br />
(constructed)<br />
Thus, coord<strong>in</strong>ations of nouns where the two conjuncts have a uniform number and<br />
different gender cannot be modified by the same def<strong>in</strong>ite article s<strong>in</strong>ce the number feature<br />
of the article will match those of both nouns but the gender feature of the article will not<br />
match the agreement features of both conjuncts. Instead, they are modified by a separate<br />
determ<strong>in</strong>er and occur with the split read<strong>in</strong>g only. The f<strong>in</strong>d<strong>in</strong>gs <strong>in</strong> MG are summarised as<br />
follows:<br />
• A s<strong>in</strong>gular def<strong>in</strong>ite determ<strong>in</strong>er cannot modify s<strong>in</strong>gular coord<strong>in</strong>ate nouns with different<br />
gender features
5.5 The <strong>Modern</strong> <strong>Greek</strong> Def<strong>in</strong>ite Determ<strong>in</strong>er 141<br />
• A plural def<strong>in</strong>ite determ<strong>in</strong>er cannot modify plural coord<strong>in</strong>ate nouns with different<br />
gender features<br />
There is aga<strong>in</strong> a strik<strong>in</strong>g pattern of exceptions to the above which allows a shared<br />
def<strong>in</strong>ite determ<strong>in</strong>er to scope over coord<strong>in</strong>ate nouns with uniform number and different<br />
gender and occurs with the split read<strong>in</strong>g. These nouns are usually plural and animate,<br />
with a mascul<strong>in</strong>e and a fem<strong>in</strong><strong>in</strong>e conjunct. These cases occur due to the <strong>in</strong>determ<strong>in</strong>ate<br />
form (identical morphological and phonological form) of the def<strong>in</strong>ite determ<strong>in</strong>er <strong>in</strong> plural<br />
number. The def<strong>in</strong>ite determ<strong>in</strong>er has the same form i ‘the’.pl.masc/fem <strong>in</strong> mascul<strong>in</strong>e and<br />
fem<strong>in</strong><strong>in</strong>e plural, which satisfies both the mascul<strong>in</strong>e plural and fem<strong>in</strong><strong>in</strong>e plural syntactic<br />
requirements 12 . Consider the follow<strong>in</strong>g:<br />
(70) I<br />
mathites<br />
ke<br />
the.pl.masc/fem boy-students.pl.masc and<br />
ton sxolion...<br />
of-the schools...<br />
‘The boy-students and girl-students of the schools...’<br />
(www.<strong>in</strong>.gr)<br />
mathitries<br />
girl-students.pl.fem<br />
(71) I<br />
athlites<br />
ke athlitries<br />
the.pl.masc/fem men-athletes.pl.masc and women-athletes.pl.fem<br />
katatasonte stis teseris katigories<br />
are-classified <strong>in</strong>to-the four categories<br />
‘The men-athletes and women-athletes are classified <strong>in</strong>to four categories’<br />
(www.<strong>in</strong>.gr)<br />
In both examples, the two nouns have different gender but plural number. The first conjunct<br />
is mascul<strong>in</strong>e and the second conjunct is fem<strong>in</strong><strong>in</strong>e while the def<strong>in</strong>ite determ<strong>in</strong>er has<br />
wide scope over both nouns without posit<strong>in</strong>g any issues of ungrammaticality.<br />
The same phenomenon occurs when the conjuncts appear <strong>in</strong> the reverse order. A shared<br />
plural def<strong>in</strong>ite determ<strong>in</strong>er scopes over coord<strong>in</strong>ated plural nouns and occurs with the split<br />
read<strong>in</strong>g when the first conjunct is fem<strong>in</strong><strong>in</strong>e and the second conjunct is mascul<strong>in</strong>e, as follows:<br />
(72) I<br />
mathitries ke<br />
the.pl.masc/fem girl-students.pl.fem and<br />
ton sxolion...<br />
of-the schools...<br />
‘The girl-students and boy-students of the schools...’<br />
(constructed)<br />
mathites<br />
boy-students.pl.masc<br />
(73) I<br />
athlitries<br />
ke athlites<br />
the.pl.masc/fem women-athletes.pl.fem and men-athletes.pl.masc<br />
katatasonte stis teseris katigories<br />
are-classified <strong>in</strong>to-the four categories<br />
‘The women-athletes and men-athletes are classified <strong>in</strong>to four categories’<br />
(constructed)<br />
The same pattern is not allowed <strong>in</strong> s<strong>in</strong>gular number when mixed gender (i.e. mascul<strong>in</strong>e<br />
and fem<strong>in</strong><strong>in</strong>e) s<strong>in</strong>gular nouns are modified by a shared s<strong>in</strong>gular def<strong>in</strong>ite determ<strong>in</strong>er.<br />
12 The MG plural def<strong>in</strong>ite determ<strong>in</strong>er is spelled and pronounced the same <strong>in</strong> the mascul<strong>in</strong>e and femi-<br />
n<strong>in</strong>e genders <strong>in</strong> plural number. The plural neuter def<strong>in</strong>ite determ<strong>in</strong>er has a different form which is ta<br />
‘the’.pl.neut.
5.5 The <strong>Modern</strong> <strong>Greek</strong> Def<strong>in</strong>ite Determ<strong>in</strong>er 142<br />
(74) *O mathitis<br />
ke mathitria<br />
the.sg.masc<br />
sxolion...<br />
schools<br />
boy-student.sg.masc and girl-student.sg.fem<br />
‘The boy-student and girl-student of the schools...’<br />
(constructed)<br />
ton<br />
of-the<br />
Thus, mixed gender but uniform number conjuncts cannot be modified by a shared<br />
def<strong>in</strong>ite determ<strong>in</strong>er and generate a split read<strong>in</strong>g. A few exceptional patterns are found <strong>in</strong><br />
plural when a shared def<strong>in</strong>ite determ<strong>in</strong>er scopes over a coord<strong>in</strong>ate phrase with a mascul<strong>in</strong>e<br />
and a fem<strong>in</strong><strong>in</strong>e plural noun and occurs with the split read<strong>in</strong>g. In the next section, we will<br />
show that the motivation for the occurrence of the exceptional patterns <strong>in</strong> the syntactically<br />
restricted NP structure <strong>in</strong> MG is that the coord<strong>in</strong>ated nouns are cases of natural<br />
coord<strong>in</strong>ation.<br />
5.5.2.1 Motivation for the Exceptional Coord<strong>in</strong>ate Constructions<br />
In this section, we will argue that the exceptional patterns, which allow a shared def<strong>in</strong>ite<br />
determ<strong>in</strong>er to scope over coord<strong>in</strong>ate nouns and occurs with a split <strong>in</strong>terpretation, are cases<br />
of natural coord<strong>in</strong>ation (Wälchli, 2005; Haspelmath, 2007; Dalrymple and Nikolaeva, 2006).<br />
We will also consider whether these patterns present any of the syntactic or semantic<br />
characteristics of natural coord<strong>in</strong>ation as proposed by Dalrymple and Nikolaeva (2006).<br />
The exceptional patterns that are cases of natural coord<strong>in</strong>ation refer to examples (60)-(65)<br />
discussed <strong>in</strong> section 5.5.1, and examples (70) and (71) discussed <strong>in</strong> section 5.5.2.<br />
Accord<strong>in</strong>g to Haspelmath (2007), natural coord<strong>in</strong>ation is a special semantic relation<br />
<strong>in</strong> which nouns form a conceptual unit and are related <strong>in</strong> mean<strong>in</strong>g. Also, Wälchli (2005)<br />
argues that natural coord<strong>in</strong>ation <strong>in</strong>volves nouns that are expected to cooccur while the<br />
parts (i.e. the conjuncts) express semantically closely related concepts, which are on the<br />
same hierarchical level and the mean<strong>in</strong>g of the whole is more general than the mean<strong>in</strong>g<br />
of the parts, such as brother-sister/‘sibl<strong>in</strong>gs’, hands-feet/‘limbs’ etc. Dalrymple and Nikolaeva<br />
(2006, 12) follow the same assumptions but add that natural coord<strong>in</strong>ation relation is<br />
also contextually specified, depend<strong>in</strong>g on <strong>in</strong>formation drawn from the specific discourse<br />
conditions, general world knowledge or utterance situation.<br />
Dalrymple and Nikolaeva (2006) propose that <strong>in</strong> addition to the above semantic characteristics,<br />
natural coord<strong>in</strong>ation is also expressed syntactically <strong>in</strong> some languages. Some<br />
languages may have different coord<strong>in</strong>ation markers <strong>in</strong> natural and accidental coord<strong>in</strong>ation<br />
(e.g. Ud<strong>in</strong>e), <strong>in</strong> other languages natural coord<strong>in</strong>ation <strong>in</strong>volves phonological reduction of<br />
the coord<strong>in</strong>ation markers (e.g. Oceanic languages). Also, it may <strong>in</strong>volve lack of determ<strong>in</strong>ers<br />
and the use of a s<strong>in</strong>gle <strong>in</strong>flectional mark<strong>in</strong>g, such as <strong>in</strong> Eastern Armenian. Another<br />
characteristic is that natural coord<strong>in</strong>ation usually allows only two conjuncts and no more<br />
than two. F<strong>in</strong>ally, accord<strong>in</strong>g to Dalrymple and Nikolaeva (2006) a natural coord<strong>in</strong>ation<br />
f-structure requires that <strong>in</strong>dex and concord feature values must be the same, while the<br />
conjuncts need to be either both s<strong>in</strong>gular or both plural. In any case, the coord<strong>in</strong>ate phrase<br />
has an f-structure that resembles the f-structure of a plural noun, as opposed to accidental<br />
coord<strong>in</strong>ation where concord is different from <strong>in</strong>dex 13 .<br />
In MG, the first exceptional structure, which concerns examples (75), (76) and (77)<br />
repeated from above, allows a shared plural def<strong>in</strong>ite determ<strong>in</strong>er to scope over coord<strong>in</strong>ate<br />
plural nouns of the same gender and occurs with a split read<strong>in</strong>g. The nouns coord<strong>in</strong>ated<br />
are not arbitrarily comb<strong>in</strong>ed conjuncts but denote entities that are characterised by, what<br />
we will call, semantic relatedness. The term semantic relatedness will be used to state that<br />
13 See section 3.3.3 for a brief summary of the characteristics of natural coord<strong>in</strong>ation and Dalrymple and<br />
Nikolaeva (2006) for a detailed analysis.
5.5 The <strong>Modern</strong> <strong>Greek</strong> Def<strong>in</strong>ite Determ<strong>in</strong>er 143<br />
the two conjuncts are <strong>in</strong> a special semantic relation with each other by be<strong>in</strong>g members<br />
of the same semantic group of nouns, such as hyponyms of the same superord<strong>in</strong>ate term.<br />
This follows from the def<strong>in</strong>ition of natural coord<strong>in</strong>ation, as proposed by Wälchli (2005),<br />
Haspelmath (2007) and Dalrymple and Nikolaeva (2006). Thus, the first semantic characteristic<br />
<strong>in</strong> MG natural coord<strong>in</strong>ation is that the nouns must be semantically related <strong>in</strong><br />
order to allow a shared def<strong>in</strong>ite determ<strong>in</strong>er to scope over coord<strong>in</strong>ate nouns and occur with<br />
a split read<strong>in</strong>g.<br />
(75) I<br />
ipurgi<br />
ke ifipurgi<br />
the.pl.masc m<strong>in</strong>isters.pl.masc and deputy-m<strong>in</strong>isters.pl.masc<br />
kivernisis tis ND<br />
government of ND<br />
‘The m<strong>in</strong>isters and deputy-m<strong>in</strong>isters of the government of ND’<br />
(G.N.:Ependitis)<br />
(76) Ta maxeria ke<br />
the.pl.neut<br />
trapezi<br />
table<br />
knives.pl.neut and<br />
‘The knives and forks are on the table’<br />
(G.M.:1000+1 Gefsis)<br />
pirunia<br />
forks.pl.neut<br />
(77) Ta ithi<br />
ke ethima<br />
the.pl.neut<br />
diamorfothikan<br />
morals.pl.neut and customs.pl.neut<br />
they-were-formed<br />
‘The morals and customs as they were formed’<br />
(G.M.:VIMAgaz<strong>in</strong>o)<br />
<strong>in</strong>e<br />
are.pl<br />
opos<br />
as<br />
sto<br />
on-the<br />
tis<br />
of-the<br />
The nouns <strong>in</strong> example (75), ipurgi ke ifipurgi ‘m<strong>in</strong>isters and deputy-m<strong>in</strong>isters’ are semantically<br />
related and they both denote entities that function as hyponyms of the same<br />
superord<strong>in</strong>ate term. The superord<strong>in</strong>ate term could be the noun ‘government’. Similarly,<br />
the nouns <strong>in</strong> examples (76), maxeria ke pirunia ‘knives and forks’ and (77), ithi ke ethima<br />
‘morals and customs’ are also semantically related and could be perceived as members of<br />
the superord<strong>in</strong>ate terms ‘cutlery’ and ‘traditions of a culture’, respectively. Thus, <strong>in</strong> all<br />
cases the two conjuncts function as hyponyms of a semantically more general term.<br />
In the follow<strong>in</strong>g example the conjuncts cannot be coord<strong>in</strong>ated under a shared def<strong>in</strong>ite<br />
determ<strong>in</strong>er with a split read<strong>in</strong>g:<br />
(78) *I jatri<br />
the.pl.masc doctors.pl.masc<br />
‘The doctors and lawyers’<br />
ke<br />
and<br />
dikigori<br />
lawyers.pl.masc<br />
The two nouns denote two professions and they could be hyponyms of the hypernymic<br />
term ‘professionals’. However, they do not express semantically closely related concepts<br />
or natural coord<strong>in</strong>ation s<strong>in</strong>ce the profession jatros ‘doctor’ does not have any common<br />
characteristics with the profession dikigoros ‘lawyer’. Therefore, the determ<strong>in</strong>er needs to<br />
be overtly specified for each conjunct <strong>in</strong> order for the phrase to be grammatical. Instead, the<br />
profession jatros ‘doctor’ is semantically related to the profession odontiatros ‘dentist’ and<br />
the profession dikigoros ‘lawyer’ is semantically related to the profession dikastis ‘judge’.<br />
Therefore, if we replace the second conjunct with a noun such as odontiatri ‘dentists’, the<br />
coord<strong>in</strong>ate phrase is acceptable even with a shared def<strong>in</strong>ite determ<strong>in</strong>er and it allows a split<br />
read<strong>in</strong>g:
5.5 The <strong>Modern</strong> <strong>Greek</strong> Def<strong>in</strong>ite Determ<strong>in</strong>er 144<br />
(79) I<br />
jatri<br />
ke odontiatri<br />
the.pl.masc doctors.pl.masc and dentists.pl.masc<br />
silogu Thesalonikis<br />
association of-Thessaloniki<br />
‘The doctors and dentists of Thessaloniki’s association’<br />
(G.N.:Eleftherotipia)<br />
tu<br />
of-the<br />
A syntactic characteristic <strong>in</strong> natural coord<strong>in</strong>ation constructions, which is also found <strong>in</strong><br />
other languages like German, is that a shared def<strong>in</strong>ite determ<strong>in</strong>er modifies two coord<strong>in</strong>ate<br />
nouns, and the nouns refer to different <strong>in</strong>dividuals (i.e. split <strong>in</strong>terpretation). In MG, this<br />
syntactic structure is admitted <strong>in</strong> accidentally coord<strong>in</strong>ate nouns and only if the nouns<br />
denote the same <strong>in</strong>dividual (i.e. jo<strong>in</strong>t <strong>in</strong>terpretation). The same structure is also admitted<br />
<strong>in</strong> the exceptional patterns discussed <strong>in</strong> which the nouns are not accidentally coord<strong>in</strong>ated<br />
but naturally coord<strong>in</strong>ated. Therefore, we could hypothesize that the presence of a s<strong>in</strong>gle<br />
def<strong>in</strong>ite determ<strong>in</strong>er scop<strong>in</strong>g over coord<strong>in</strong>ate nouns and generat<strong>in</strong>g the split read<strong>in</strong>g is one<br />
of the syntactic characteristics of natural coord<strong>in</strong>ation <strong>in</strong> MG. In other languages, natural<br />
coord<strong>in</strong>ation is expressed by the absence of a determ<strong>in</strong>er or by the use of coord<strong>in</strong>ation<br />
or phonological markers. Also, another syntactic characteristic of natural coord<strong>in</strong>ation is<br />
that the number of conjuncts allowed <strong>in</strong> these constructions is usually two (less often three<br />
conjuncts) at least <strong>in</strong> the languages discussed by Dalrymple and Nikolaeva (2006). In MG,<br />
we usually f<strong>in</strong>d two conjuncts; coord<strong>in</strong>ations of three conjuncts under a shared determ<strong>in</strong>er<br />
are not strictly ungrammatical but they are used less often and only if the nouns are<br />
semantically related:<br />
(80) ?I jatri,<br />
odontiatri ke<br />
the.pl.masc doctors.pl.masc, dentists.pl.masc and<br />
mikroviologi<br />
tu silogu Thesalonikis<br />
microbiologists.pl.masc of-the association of-Thesaloniki<br />
‘The doctors, dentists and microbiologists of Thesaloniki association’<br />
(constructed)<br />
Therefore, the presence of usually two conjuncts <strong>in</strong> the above structures <strong>in</strong> MG should<br />
be considered as one more characteristic of natural coord<strong>in</strong>ation.<br />
F<strong>in</strong>ally, a natural coord<strong>in</strong>ation f-structure requires that <strong>in</strong>dex and concord features<br />
should match and that the coord<strong>in</strong>ate phrase has an f-structure that resembles the fstructure<br />
of a plural noun. Also, the conjuncts need to be either both s<strong>in</strong>gular or both<br />
plural. In MG natural coord<strong>in</strong>ation cases, this syntactic representation does not occur. In<br />
MG, <strong>in</strong>dex and concord features are plural due to the requirement for plural conjuncts<br />
by the plural def<strong>in</strong>ite determ<strong>in</strong>er, s<strong>in</strong>ce the exceptional natural coord<strong>in</strong>ation patterns are<br />
only found <strong>in</strong> the plural. The conjuncts can never be s<strong>in</strong>gular while the determ<strong>in</strong>er specifies<br />
for a plural <strong>in</strong>dex and concord feature-value, such as <strong>in</strong> the case of F<strong>in</strong>nish. Therefore,<br />
<strong>in</strong> MG the f-structure of natural coord<strong>in</strong>ation does not resemble the f-structure of a plural<br />
noun but it appears to be the same as the f-structure of accidental coord<strong>in</strong>ation. Thus,<br />
we argue that syntactically the f-structure of MG natural coord<strong>in</strong>ation patterns seems to<br />
resemble the f-structure of accidental coord<strong>in</strong>ation.<br />
The next exceptional structure is when a plural def<strong>in</strong>ite determ<strong>in</strong>er modifies two plural<br />
conjuncts with a mascul<strong>in</strong>e and a fem<strong>in</strong><strong>in</strong>e gender. First of all, this structure is admitted<br />
due to the fact that the def<strong>in</strong>ite determ<strong>in</strong>er o, i, to ‘the’.sg.masc/fem/neut <strong>in</strong> the plural<br />
number is <strong>in</strong>determ<strong>in</strong>ate and it can show agreement with both a mascul<strong>in</strong>e and a fem<strong>in</strong><strong>in</strong>e<br />
plural noun. It has a separate form only <strong>in</strong> the neuter gender, which is ta ‘the’.pl.neut.<br />
The <strong>in</strong>determ<strong>in</strong>ate form of the determ<strong>in</strong>er allows the coord<strong>in</strong>ation of a mascul<strong>in</strong>e and a<br />
fem<strong>in</strong><strong>in</strong>e plural noun under the same plural def<strong>in</strong>ite article s<strong>in</strong>ce the latter can satisfy the<br />
syntactic requirements of both nouns. The follow<strong>in</strong>g example is repeated from above:
5.5 The <strong>Modern</strong> <strong>Greek</strong> Def<strong>in</strong>ite Determ<strong>in</strong>er 145<br />
(81) I<br />
papudes<br />
ke<br />
the.pl.masc/fem grandfathers.pl.masc and<br />
<strong>in</strong>e xarumeni<br />
are.pl happy<br />
‘The grandfathers and grandmothers are happy’<br />
(constructed)<br />
jajades<br />
grandmothers.pl.fem<br />
These patterns are considered cases of natural coord<strong>in</strong>ation s<strong>in</strong>ce a shared determ<strong>in</strong>er<br />
scopes over the two nouns and allows a split read<strong>in</strong>g. Also, these cases only occur with<br />
conjuncts that are necessarily mascul<strong>in</strong>e and fem<strong>in</strong><strong>in</strong>e, semantically related and the number<br />
of conjuncts is usually two. Any coord<strong>in</strong>ate phrase with conjuncts that are mascul<strong>in</strong>e and<br />
fem<strong>in</strong><strong>in</strong>e but not semantically related or with conjuncts that are mascul<strong>in</strong>e and neuter and<br />
semantically related are strictly ungrammatical. In these patterns, the split read<strong>in</strong>g is the<br />
only possibility. The follow<strong>in</strong>g examples are considered ungrammatical:<br />
(82) *I<br />
daskali ke mathitries<br />
the.pl.masc/fem teachers.pl.masc and girl-students.pl.fem<br />
defteras dimotiku<br />
second-year <strong>in</strong>-primary-school<br />
‘The teachers and girl-students of the second year <strong>in</strong> primary school’<br />
(constructed)<br />
(83) *I<br />
papudes<br />
ke<br />
the.pl.masc/fem grandfathers.pl.masc and<br />
to anaferun<br />
it mention.pl<br />
‘The grandfathers and grandchildren mention it’<br />
(constructed)<br />
tis<br />
of-the<br />
engonia<br />
grandchildren.pl.neut<br />
In example (82), the two conjuncts are not semantically related and therefore they<br />
cannot be coord<strong>in</strong>ated under the same determ<strong>in</strong>er. In example (83), the conjuncts are<br />
semantically related but the phrase is ungrammantical s<strong>in</strong>ce the two nouns are mascul<strong>in</strong>e<br />
and neuter <strong>in</strong> gender and they cannot occur under a shared determ<strong>in</strong>er.<br />
To summarise, the MG def<strong>in</strong>ite determ<strong>in</strong>er presents a number of exceptional patterns<br />
that occur under specific conditions. The first is when a MG shared plural def<strong>in</strong>ite determ<strong>in</strong>er<br />
modifies coord<strong>in</strong>ated plural conjuncts that are semantically related and occurs with<br />
a split <strong>in</strong>terpretation. The second exceptional pattern is when a plural def<strong>in</strong>ite determ<strong>in</strong>er<br />
modifies coord<strong>in</strong>ated plural nouns with a mascul<strong>in</strong>e and a fem<strong>in</strong><strong>in</strong>e gender that are semantically<br />
related, and also occurs with a split <strong>in</strong>terpretation. This pattern is motivated by<br />
the <strong>in</strong>determ<strong>in</strong>ate form of the def<strong>in</strong>ite determ<strong>in</strong>er <strong>in</strong> plural number, which agrees both with<br />
a mascul<strong>in</strong>e and a fem<strong>in</strong><strong>in</strong>e plural noun. These patterns should be seen under the realm<br />
of natural coord<strong>in</strong>ation s<strong>in</strong>ce they present some special characteristics - namely a common<br />
determ<strong>in</strong>er for both conjuncts and a split read<strong>in</strong>g, the presence of usually two conjuncts<br />
and semantically related conjuncts. We argued, though, that syntactically these patterns<br />
have an f-structure that seems to resemble the f-structure of accidental coord<strong>in</strong>ation and<br />
not that of natural coord<strong>in</strong>ation. In the next section, we will present the analysis of the<br />
shared def<strong>in</strong>ite determ<strong>in</strong>er with coord<strong>in</strong>ate nouns.<br />
5.5.3 <strong>in</strong>dex/concord agreement with the MG Def<strong>in</strong>ite Determ<strong>in</strong>er<br />
We now turn to the analysis of the MG def<strong>in</strong>ite determ<strong>in</strong>er with coord<strong>in</strong>ate nouns and<br />
whether any of the three systems, concord, <strong>in</strong>dex/concord and <strong>in</strong>dex, account for it.<br />
We will argue that the MG def<strong>in</strong><strong>in</strong>ite determ<strong>in</strong>er follows concord/<strong>in</strong>dex agreement. This
5.5 The <strong>Modern</strong> <strong>Greek</strong> Def<strong>in</strong>ite Determ<strong>in</strong>er 146<br />
system syntactically accounts for the MG def<strong>in</strong>ite determ<strong>in</strong>er but further semantic factors<br />
are <strong>in</strong>volved when it comes to the MG data.<br />
The MG def<strong>in</strong>ite determ<strong>in</strong>er presents specific restrictions with regards to the <strong>in</strong>dex and<br />
concord features. The only system that is restrictive is the <strong>in</strong>dex/concord, and therefore<br />
we will argue that the MG def<strong>in</strong>ite determ<strong>in</strong>er follows <strong>in</strong>dex/concord agreement.<br />
This system would predict the follow<strong>in</strong>g patterns <strong>in</strong> MG:<br />
Table 5.2: INDEX/CONCORD SYSTEM<br />
ANIMATE INANIMATE<br />
JOINT READ <br />
DETsg<br />
Nsg & Nsg SPLIT READ * *<br />
DETpl<br />
JOINT READ <br />
Npl & Npl SPLIT READ <br />
DETpl<br />
JOINT READ * *<br />
Nsg & Nsg SPLIT READ * *<br />
Next, we will exam<strong>in</strong>e whether these predications are borne out <strong>in</strong> MG. In the s<strong>in</strong>gular<br />
number, a shared MG def<strong>in</strong>ite determ<strong>in</strong>er agrees <strong>in</strong> number and gender with both nouns<br />
and scopes over coord<strong>in</strong>ated s<strong>in</strong>gular nouns that refer to a s<strong>in</strong>gle <strong>in</strong>dividual or entity.<br />
Hence, the MG def<strong>in</strong>ite determ<strong>in</strong>er requires a sg value for the <strong>in</strong>dex num feature and,<br />
redundantly, a sg value for the concord num feature. The two s<strong>in</strong>gular conjuncts also<br />
need to have a sg value <strong>in</strong> their concord num feature. F<strong>in</strong>ally, the def<strong>in</strong>ite determ<strong>in</strong>er<br />
specifies that both conjuncts have the same gender value <strong>in</strong> their concord gend feature<br />
as the determ<strong>in</strong>er. Consider the follow<strong>in</strong>g example:<br />
(84) O filos ke s<strong>in</strong>adelfos<br />
the.sg.masc friend.sg.masc and colleague.sg.masc<br />
ton gnorizo xronia<br />
him.sg.masc I-know for-years<br />
‘The friend and colleague whom I know for years’<br />
pu<br />
who<br />
S<strong>in</strong>ce the <strong>in</strong>dex feature is responsible for determ<strong>in</strong><strong>in</strong>g the distribution of the determ<strong>in</strong>er<br />
o ‘the’, the lexical entry of the determ<strong>in</strong>er specifies both <strong>in</strong>dex and concord features.<br />
Also, it specifies concord gend agreement requirements:<br />
(85) o: (↑spec) = ‘the’<br />
(↑concord num) = sg<br />
(↑concord gend) = masc<br />
(↑<strong>in</strong>dex num) = sg<br />
The f-structure would be as follows:
5.5 The <strong>Modern</strong> <strong>Greek</strong> Def<strong>in</strong>ite Determ<strong>in</strong>er 147<br />
(86)<br />
⎡<br />
⎤<br />
spec ‘the’<br />
⎢<br />
⎥<br />
⎢<strong>in</strong>dex<br />
[num sg]<br />
⎥<br />
⎢⎧<br />
⎢ ⎡<br />
⎤⎫<br />
⎥<br />
⎢ pred ‘friend ⎥<br />
⎢ ⎢ ⎥<br />
⎥<br />
⎢ ⎢<br />
⎢ ⎢<br />
num sg ⎥ ⎥<br />
⎢ ⎢concord<br />
⎥ ⎥<br />
⎢ b: ⎢ gend masc ⎥ ⎥<br />
⎥ ⎥<br />
⎢ ⎢ ⎥ ⎥<br />
⎢ ⎢<br />
⎢ ⎣ num sg<br />
⎥ ⎥<br />
c: ⎢⎪⎨<br />
<strong>in</strong>dex<br />
⎦ ⎥<br />
⎢ gend masc ⎪⎬ ⎥<br />
⎢ ⎡<br />
⎤ ⎥<br />
⎢ pred ‘colleague’<br />
⎥<br />
⎢<br />
⎢ ⎥<br />
⎥<br />
⎢<br />
⎢<br />
⎢<br />
⎢<br />
num sg ⎥ ⎥<br />
⎢<br />
⎢concord<br />
⎥ ⎥<br />
⎢<br />
g: ⎢ gend masc ⎥ ⎥<br />
⎥ ⎥<br />
⎢<br />
⎢ ⎥ ⎥<br />
⎢<br />
⎢<br />
⎣<br />
⎣ num sg<br />
⎥ ⎥<br />
⎪⎩<br />
<strong>in</strong>dex<br />
⎦ ⎥<br />
gend masc ⎪⎭<br />
⎦<br />
Thus, the MG s<strong>in</strong>gular def<strong>in</strong>ite determ<strong>in</strong>er that scopes over coord<strong>in</strong>ate s<strong>in</strong>gular nouns<br />
shows <strong>in</strong>dex/concord agreement and generates only the jo<strong>in</strong>t read<strong>in</strong>g and not the split<br />
read<strong>in</strong>g.<br />
In plural number, a shared MG def<strong>in</strong>ite determ<strong>in</strong>er agrees <strong>in</strong> number and gender with<br />
both nouns and scopes over coord<strong>in</strong>ated plural nouns, which must necessarily refer to the<br />
same <strong>in</strong>dividuals or entities. Hence, <strong>in</strong> plural number the MG def<strong>in</strong>ite determ<strong>in</strong>er requires<br />
a pl value for the <strong>in</strong>dex num feature and a pl value for the concord num feature while<br />
the two conjuncts also need to have a pl value <strong>in</strong> their concord num feature. The def<strong>in</strong>ite<br />
determ<strong>in</strong>er also requires that both conjuncts have the same gender value <strong>in</strong> their concord<br />
gend feature as the determ<strong>in</strong>er itself. Consider the follow<strong>in</strong>g example with animate nouns:<br />
(87) I<br />
fili<br />
ke<br />
the.pl.masc<br />
voithia<br />
help<br />
friends.pl.masc and<br />
‘The friends and colleagues need help’<br />
s<strong>in</strong>adelfi<br />
colleagues.pl.masc<br />
xriazonte<br />
need.pl<br />
The lexical entry of the plural def<strong>in</strong><strong>in</strong>ite determ<strong>in</strong>er specifies both the <strong>in</strong>dex and concord<br />
features s<strong>in</strong>ce the concord and redudantly the <strong>in</strong>dex feature determ<strong>in</strong>e the distribution<br />
of the determ<strong>in</strong>er:<br />
(88) i: (↑spec) = ‘the’<br />
(↑concord num) = pl<br />
(↑concord gend) = masc<br />
(↑<strong>in</strong>dex num) = pl<br />
The f-structure is as follows:
5.5 The <strong>Modern</strong> <strong>Greek</strong> Def<strong>in</strong>ite Determ<strong>in</strong>er 148<br />
(89)<br />
⎡<br />
⎤<br />
spec ‘the’<br />
⎢<br />
⎥<br />
⎢<strong>in</strong>dex<br />
[num pl]<br />
⎥<br />
⎢⎧<br />
⎢ ⎡<br />
⎤⎫<br />
⎥<br />
⎢ pred ‘friends’ ⎥<br />
⎢ ⎢ ⎥<br />
⎥<br />
⎢ ⎢<br />
⎢ ⎢<br />
num pl ⎥ ⎥<br />
⎢ ⎢concord<br />
⎥ ⎥<br />
⎢ b: ⎢ gend masc ⎥ ⎥<br />
⎥ ⎥<br />
⎢ ⎢ ⎥ ⎥<br />
⎢ ⎢<br />
⎢ ⎣ num pl<br />
⎥ ⎥<br />
c: ⎢⎪⎨<br />
<strong>in</strong>dex<br />
⎦ ⎥<br />
⎢ gend masc ⎪⎬ ⎥<br />
⎢ ⎡<br />
⎤ ⎥<br />
⎢ pred ‘colleagues’<br />
⎥<br />
⎢<br />
⎢ ⎥<br />
⎥<br />
⎢<br />
⎢<br />
⎢<br />
⎢<br />
num pl ⎥ ⎥<br />
⎢<br />
⎢concord<br />
⎥ ⎥<br />
⎢<br />
g: ⎢ gend masc ⎥ ⎥<br />
⎥ ⎥<br />
⎢<br />
⎢ ⎥ ⎥<br />
⎢<br />
⎢<br />
⎣<br />
⎣ num pl<br />
⎥ ⎥<br />
⎪⎩<br />
<strong>in</strong>dex<br />
⎦ ⎥<br />
gend masc ⎪⎭<br />
⎦<br />
Thus, the MG plural def<strong>in</strong>ite determ<strong>in</strong>er that scopes over coord<strong>in</strong>ate nouns shows <strong>in</strong>dex/concord<br />
agreement while it generates the jo<strong>in</strong>t read<strong>in</strong>g only.<br />
The analysis above showed that MG s<strong>in</strong>gular def<strong>in</strong>ite determ<strong>in</strong>er modifies coord<strong>in</strong>ate<br />
s<strong>in</strong>gular nouns and occurs only with the jo<strong>in</strong>t read<strong>in</strong>g. Also, the MG plural def<strong>in</strong>ite determ<strong>in</strong>er<br />
modifies coord<strong>in</strong>ate plural nouns and occurs only with the jo<strong>in</strong>t read<strong>in</strong>g when the<br />
coord<strong>in</strong>ate nouns represent an accidental coord<strong>in</strong>ation but it occurs with the split read<strong>in</strong>g<br />
when the coord<strong>in</strong>ate nouns are cases of natural coord<strong>in</strong>ation. Thus, we state the follow<strong>in</strong>g<br />
generalisation:<br />
Jo<strong>in</strong>t read<strong>in</strong>g <strong>in</strong> Plural Coord<strong>in</strong>ate <strong>Noun</strong>s<br />
For languages like MG, a plural def<strong>in</strong>ite determ<strong>in</strong>er imposes <strong>in</strong>dex/concord<br />
agreement and a boolean/jo<strong>in</strong>t <strong>in</strong>terpretation with an accidental coord<strong>in</strong>ation<br />
of plural nouns unless the nouns can be construed as a natural coord<strong>in</strong>ation.<br />
Therefore, we can conclude that syntactically the MG def<strong>in</strong>ite determ<strong>in</strong>er is effectivelly<br />
captured by the <strong>in</strong>dex/concord agreement system, but semantically the MG plural determ<strong>in</strong>er<br />
<strong>in</strong>troduces further semantic factors that are beyond those generated by the <strong>in</strong>dex/concord<br />
system.<br />
To summarise, the MG s<strong>in</strong>gular def<strong>in</strong>ite determ<strong>in</strong>er follows the <strong>in</strong>dex/concord system,<br />
which allows only the jo<strong>in</strong>t read<strong>in</strong>g <strong>in</strong> s<strong>in</strong>gular number. The MG plural def<strong>in</strong>ite determ<strong>in</strong>er<br />
is also syntactically captured by the <strong>in</strong>dex/concord agreement system. Semantically,<br />
the present system allows both the jo<strong>in</strong>t and split <strong>in</strong>terpretation <strong>in</strong> plural number.<br />
The MG plural def<strong>in</strong>ite determ<strong>in</strong>er, however, allows the jo<strong>in</strong>t <strong>in</strong>terpretation <strong>in</strong> accidentally<br />
coord<strong>in</strong>ated nouns and the split <strong>in</strong>terpretation <strong>in</strong> natural coord<strong>in</strong>ations. This semantic<br />
restriction is only found <strong>in</strong> MG as opposed to languages such as Brazilian Portuguese, Italian,<br />
French and German, which allow a plural def<strong>in</strong>ite determ<strong>in</strong>er with coord<strong>in</strong>ate nouns<br />
to occur with a jo<strong>in</strong>t and a split <strong>in</strong>terpretation (K<strong>in</strong>g and Dalrymple, 2004, 91). In the<br />
next section, we will discuss the patterns of natural coord<strong>in</strong>ation <strong>in</strong> MG and whether the<br />
syntactic representation of the <strong>in</strong>dex/concord system accounts for the MG data or a<br />
different syntactic structure is required.<br />
5.5.4 Natural Coord<strong>in</strong>ation <strong>in</strong> <strong>Modern</strong> <strong>Greek</strong><br />
In MG plural nouns, the split read<strong>in</strong>g emerges only when the conjuncts are candidates for<br />
natural coord<strong>in</strong>ation. Dalrymple and Nikolaeva (2006) <strong>in</strong>troduce a special f-structure <strong>in</strong>
5.5 The <strong>Modern</strong> <strong>Greek</strong> Def<strong>in</strong>ite Determ<strong>in</strong>er 149<br />
naturally coord<strong>in</strong>ated nouns, which is different from the f-structure of accidentally coord<strong>in</strong>ated<br />
nouns. As already mentioned, the natural coord<strong>in</strong>ation patterns <strong>in</strong> MG do not have<br />
the same syntactic structure as the syntactic structure of natural coord<strong>in</strong>ation proposed<br />
by Dalrymple and Nikolaeva (2006). Therefore, we will show that MG natural coord<strong>in</strong>ation<br />
cases are not captured by the f-structure proposed by Dalrymple and Nikolaeva<br />
(2006) but they are accounted for by the <strong>in</strong>dex/concord system without any additional<br />
requirements.<br />
The f-structure of natural coord<strong>in</strong>ation requires at most two conjuncts. The coord<strong>in</strong>ate<br />
nouns do not appear as members of the set, as <strong>in</strong> accidental coord<strong>in</strong>ation, but they are<br />
two <strong>in</strong>dependent conjuncts each contribut<strong>in</strong>g its own f-structure with<strong>in</strong> the f-structure of<br />
the phrase as a whole. Also, concord num and <strong>in</strong>dex num features need to have a pl<br />
value, resembl<strong>in</strong>g the f-structure for compound or plural nouns, whereas conjuncts can be<br />
s<strong>in</strong>gular or plural. F<strong>in</strong>ally, the modifier does not impose any requirements on the concord<br />
and <strong>in</strong>dex features as <strong>in</strong> accidental coord<strong>in</strong>ation.<br />
The f-structure below <strong>in</strong> F<strong>in</strong>nish, repeated from Chapter 2, admits a coord<strong>in</strong>ate phrase<br />
with a plural modifier and s<strong>in</strong>gular nouns:<br />
(90) Natural coord<strong>in</strong>ation<br />
uskolliset<br />
faithful.pl<br />
aviomies<br />
husband<br />
‘faithful [husband and wife]’<br />
AdjP<br />
Adj<br />
uskolliset<br />
NP<br />
N’<br />
N<br />
aviomies<br />
N’<br />
N<br />
Conj<br />
ja<br />
ja<br />
and<br />
N<br />
vaimo<br />
vaimo<br />
wife<br />
(Dalrymple and Nikolaeva, 2006, 842)<br />
⎡<br />
⎤<br />
pred ‘conj’<br />
⎢<br />
⎥<br />
⎢concord<br />
[num pl] ⎥<br />
⎢<br />
⎥<br />
⎢<br />
⎢<strong>in</strong>dex<br />
[num pl] ⎥<br />
⎢ ⎡<br />
⎤ ⎥<br />
⎢ pred ‘husband’ ⎥<br />
⎢ ⎢<br />
⎥⎥<br />
⎢conj1<br />
⎣concord<br />
[num sg] ⎦ ⎥<br />
⎢ <strong>in</strong>dex [num sg]<br />
⎥<br />
⎢<br />
⎡<br />
⎤⎥<br />
⎢ pred ‘wife’<br />
⎥<br />
⎢<br />
⎢<br />
⎥<br />
conj2<br />
⎢<br />
⎣concord<br />
[num sg] ⎦ ⎥<br />
⎢<br />
⎣ <strong>in</strong>dex [num sg] ⎥<br />
⎦<br />
adj [pred ‘faithful’]<br />
The same f-structure representation 14 cannot account for the MG data. First, MG does<br />
not allow patterns where a plural modifier or determ<strong>in</strong>er scopes over s<strong>in</strong>gular conjuncts.<br />
The follow<strong>in</strong>g examples, repeated from above, are ungrammatical:<br />
(91) *Drastirii pateras ke ijos theorunte<br />
active.pl.masc father.sg.masc and son.sg.masc are-considered.pl<br />
o Onasis ke o Aleksantros<br />
the Onasis and the Alexander<br />
‘Active father and son are considered to be Onassis and Alexander’<br />
(constructed)<br />
(92) *I<br />
the.pl.masc<br />
kivernisis<br />
government<br />
ipurgos<br />
m<strong>in</strong>ister.sg.masc<br />
ke<br />
and<br />
ifipurgos<br />
deputy-m<strong>in</strong>ister.sg.masc<br />
tis<br />
of-the<br />
14 Natural coord<strong>in</strong>ation <strong>in</strong> other languages like Tundra Nenets, Russian and Bahd<strong>in</strong>ani Kurdish are cap-<br />
tured by the same f-structure. See Dalrymple and Nikolaeva (2006, 843-847) for a detailed analysis.
5.5 The <strong>Modern</strong> <strong>Greek</strong> Def<strong>in</strong>ite Determ<strong>in</strong>er 150<br />
‘The m<strong>in</strong>ister and deputy-m<strong>in</strong>ister of the government’<br />
(constructed)<br />
Natural coord<strong>in</strong>ation <strong>in</strong> MG occurs only with conjuncts that are both plural and whose<br />
modifier/determ<strong>in</strong>er agrees <strong>in</strong> number 15 with the conjucts. The modifier/determ<strong>in</strong>er specifies<br />
a pl value <strong>in</strong> the concord num feature and the conjuncts need to have the same<br />
value <strong>in</strong> their concord num feature. Also, natural coord<strong>in</strong>ation patterns <strong>in</strong> MG allow a<br />
shared determ<strong>in</strong>er to scope over the two conjuncts and it usually occurs with the split <strong>in</strong>terpretation.<br />
Thus, <strong>in</strong>dex num needs to be pl s<strong>in</strong>ce the conjuncts are also plural. F<strong>in</strong>ally,<br />
an important requirement is that the coord<strong>in</strong>ate nouns must necessarily be semantically<br />
related. The follow<strong>in</strong>g examples are repeated from above:<br />
(93) I<br />
ipurgi<br />
ke ifipurgi<br />
the.pl.masc<br />
kivernisis<br />
government<br />
m<strong>in</strong>isters.pl.masc and deputy-m<strong>in</strong>isters.pl.masc<br />
‘The m<strong>in</strong>isters and deputy-m<strong>in</strong>isters of the government’<br />
(G.N.:Ependitis)<br />
(94) Ta ithi<br />
ke ethima<br />
the.pl.neut morals.pl.neut and customs.pl.neut<br />
diamorfothikan...<br />
they-were-formed<br />
‘The morals and customs as they were formed’<br />
(G.M.:VIMAgaz<strong>in</strong>o)<br />
opos<br />
as<br />
tis<br />
of-the<br />
Thus, we argue that the f-structure representation proposed by Dalrymple and Nikolaeva<br />
(2006) cannot capture the MG data s<strong>in</strong>ce it would generate patterns that are considered<br />
ungrammatical <strong>in</strong> MG, while natural coord<strong>in</strong>ation patterns <strong>in</strong> MG show different<br />
characteristics from those of the languages discussed <strong>in</strong> Dalrymple and Nikolaeva (2006,<br />
843-847).<br />
To account for the MG patterns, we will assume the standard f-structure of accidental<br />
coord<strong>in</strong>ation. We will represent the conjuncts as set elements follow<strong>in</strong>g the f-structure of<br />
accidentally coord<strong>in</strong>ate nouns <strong>in</strong> order to allow the additional number of conjuncts, which<br />
occasionally occurs <strong>in</strong> natural coord<strong>in</strong>ate structures <strong>in</strong> MG. Also, the def<strong>in</strong>ite determ<strong>in</strong>er<br />
will specify <strong>in</strong>dex and concord agreement requirements.<br />
For <strong>in</strong>stance, <strong>in</strong> example (94) the def<strong>in</strong>ite determ<strong>in</strong>er requires a pl value <strong>in</strong> the concord<br />
num feature and the nouns need to have the same feature-value concord num=pl <strong>in</strong> their<br />
f-structure. The <strong>in</strong>dex num feature must have a pl value s<strong>in</strong>ce the two nouns are plural<br />
and refer to different groups of entities. Also, the conjuncts need to be semantically related.<br />
F<strong>in</strong>ally, both conjuncts need to agree <strong>in</strong> gender with the def<strong>in</strong>ite determ<strong>in</strong>er. The lexical<br />
entry and f-structure of example (94) are as follows:<br />
(95) Natural coord<strong>in</strong>ation <strong>in</strong> MG with a def<strong>in</strong>ite determ<strong>in</strong>er<br />
(95) ta: (↑spec) = ‘the’<br />
(↑concord num) = pl<br />
(↑concord gend) = neut<br />
(↑<strong>in</strong>dex num) = pl<br />
15 Shared modifiers agree <strong>in</strong> gender with the closest conjunct but the shared def<strong>in</strong>ite determ<strong>in</strong>er agrees <strong>in</strong><br />
gender with both conjuncts.
5.5 The <strong>Modern</strong> <strong>Greek</strong> Def<strong>in</strong>ite Determ<strong>in</strong>er 151<br />
DET<br />
ta<br />
NP<br />
+ N<br />
ithi<br />
N<br />
Conj<br />
ke<br />
N<br />
ethima<br />
⎡<br />
⎤<br />
pred ‘conj’<br />
⎢<br />
⎢spec<br />
‘the’<br />
⎥<br />
⎢<br />
⎥<br />
⎢<strong>in</strong>dex<br />
[num pl]<br />
⎥<br />
⎢⎧<br />
⎢ ⎡<br />
⎤⎫<br />
⎥<br />
⎢ pred ‘morals’ ⎥<br />
⎢ ⎢ ⎥<br />
⎥<br />
⎢ ⎢<br />
⎢ ⎢<br />
num pl ⎥ ⎥<br />
⎢ ⎢concord<br />
⎥ ⎥<br />
⎢ b: ⎢ gend neut ⎥ ⎥<br />
⎥ ⎥<br />
⎢ ⎢ ⎥ ⎥<br />
⎢ ⎢<br />
⎢ ⎣ num pl<br />
⎥ ⎥<br />
⎢⎪⎨<br />
<strong>in</strong>dex<br />
⎦<br />
⎥<br />
⎢ gend neut ⎪⎬ ⎥<br />
⎢ ⎡<br />
⎤ ⎥<br />
⎢ pred ‘customs’ ⎥<br />
⎢ ⎢ ⎥<br />
⎥<br />
⎢ ⎢<br />
⎢ ⎢<br />
num pl ⎥ ⎥<br />
⎢ ⎢concord<br />
⎥ ⎥<br />
⎢ g: ⎢ gend neut ⎥ ⎥<br />
⎥ ⎥<br />
⎢ ⎢ ⎥ ⎥<br />
⎢ ⎢<br />
⎣ ⎣ num pl<br />
⎥ ⎥<br />
⎪⎩<br />
<strong>in</strong>dex<br />
⎦ ⎥<br />
⎦<br />
gend neut ⎪⎭<br />
In example (96) a shared plural def<strong>in</strong>ite determ<strong>in</strong>er scopes over a mascul<strong>in</strong>e and a<br />
fem<strong>in</strong><strong>in</strong>e animate noun and occurs with a split read<strong>in</strong>g. The determ<strong>in</strong>er requires a pl value<br />
<strong>in</strong> the concord num feature and the nouns need to have the same value <strong>in</strong> their concord<br />
num feature. <strong>in</strong>dex num should also be plural s<strong>in</strong>ce the two nouns are plural. The<br />
conjuncts need to be semantically related. Also, the def<strong>in</strong>ite determ<strong>in</strong>er is <strong>in</strong>determ<strong>in</strong>ate<br />
<strong>in</strong> gender and it agrees with both conjuncts. The follow<strong>in</strong>g example is repeated from above:<br />
(96) I<br />
mathites<br />
ke<br />
the.pl.masc/fem boy-students.pl.masc and<br />
tou sxoliu...<br />
of-the school...<br />
‘The boy-students and girl-students of the school... ’<br />
mathitries<br />
girl-students.pl.fem<br />
Follow<strong>in</strong>g Dalrymple and Kaplan (2000), we capture the fact that the def<strong>in</strong>ite determ<strong>in</strong>er<br />
is <strong>in</strong>determ<strong>in</strong>ate <strong>in</strong> gender <strong>in</strong> the plural number, by assign<strong>in</strong>g two values to the<br />
determ<strong>in</strong>er. This will allow the determ<strong>in</strong>er to agree both with a mascul<strong>in</strong>e and a fem<strong>in</strong><strong>in</strong>e<br />
conjunct 16 . The lexical entry and f-structure are presented below:<br />
(97) i: (↑spec) = ‘the’<br />
(↑concord num) = pl<br />
(↑concord gend) = {masc,fem}<br />
16 In LFG, Dalrymple and Kaplan (2000) use the set-based approach for cases of shar<strong>in</strong>g of token identical<br />
forms such as the relatives <strong>in</strong> German. For example, the relative pronoun was ‘what’ must be accusative<br />
and nom<strong>in</strong>ative <strong>in</strong> case <strong>in</strong> order to satisfy the verbs gegessen ‘eat’ and übrig ‘left’, respectively:<br />
(1) Ich habe gegessen was übrig war<br />
I have eaten what left was<br />
‘I ate what was left’ (Dalrymple and Kaplan, 2000, 759)<br />
This is captured as follows:<br />
(2) was: (↑case)= {nom,acc}<br />
gegessen: acc ∈ (↑ obj case)<br />
übrig: nom ∈ (↑ subj case) (Dalrymple and Kaplan, 2000, 766)
5.5 The <strong>Modern</strong> <strong>Greek</strong> Def<strong>in</strong>ite Determ<strong>in</strong>er 152<br />
DET<br />
i<br />
(↑<strong>in</strong>dex num) = pl<br />
N<br />
NP<br />
mathites<br />
N<br />
Conj<br />
ke<br />
N<br />
mathitries<br />
⎡<br />
⎤<br />
spec ‘the’<br />
⎢<br />
⎥<br />
⎢<strong>in</strong>dex<br />
[num pl]<br />
⎥<br />
⎢⎧<br />
⎢ ⎡<br />
⎤⎫<br />
⎥<br />
⎢ pred ‘boy-students’ ⎥<br />
⎢ ⎢ ⎥<br />
⎥<br />
⎢ ⎢<br />
⎢ ⎢<br />
num pl ⎥ ⎥<br />
⎢ ⎢concord<br />
⎥ ⎥<br />
⎢ b: ⎢ gend masc ⎥ ⎥<br />
⎥ ⎥<br />
⎢ ⎢ ⎥ ⎥<br />
⎢ ⎢<br />
⎢ ⎣ num pl<br />
⎥ ⎥<br />
c: ⎢⎪⎨<br />
<strong>in</strong>dex<br />
⎦ ⎥<br />
⎢ gend masc ⎪⎬ ⎥<br />
⎢ ⎡<br />
⎤ ⎥<br />
⎢ pred ‘girl-students’<br />
⎥<br />
⎢<br />
⎢ ⎥<br />
⎥<br />
⎢<br />
⎢<br />
⎢<br />
⎢<br />
num pl ⎥ ⎥<br />
⎢<br />
⎢concord<br />
⎥ ⎥<br />
⎢<br />
g: ⎢ gend fem ⎥ ⎥<br />
⎥ ⎥<br />
⎢<br />
⎢ ⎥ ⎥<br />
⎢<br />
⎢<br />
⎣<br />
⎣ num pl<br />
⎥ ⎥<br />
⎪⎩<br />
<strong>in</strong>dex<br />
⎦ ⎥<br />
gend fem ⎪⎭<br />
⎦<br />
Therefore, MG natural coord<strong>in</strong>ation cases with a shared def<strong>in</strong>ite determ<strong>in</strong>er follow the<br />
<strong>in</strong>dex/concord system just like MG accidental coord<strong>in</strong>ation cases.<br />
MG modifiers impose similar requirements <strong>in</strong> natural coord<strong>in</strong>ation structures with those<br />
proposed for accidental coord<strong>in</strong>ation. The follow<strong>in</strong>g example is a case of natural coord<strong>in</strong>ation<br />
with a modifier:<br />
(98) Gnosti ipurgi<br />
ke ifipurgi<br />
famous.pl.masc m<strong>in</strong>isters.pl.masc and deputy-m<strong>in</strong>isters.pl.masc<br />
parevrethikan sta engenia<br />
attended.pl the open<strong>in</strong>g<br />
‘Famous m<strong>in</strong>isters and deputy-m<strong>in</strong>isters attended the open<strong>in</strong>g’<br />
In example (98), the two nouns are coord<strong>in</strong>ated under a shared modifier and the split<br />
read<strong>in</strong>g is generated. As <strong>in</strong> accidental coord<strong>in</strong>ate structures with a modifier, the modifier<br />
requires a pl value <strong>in</strong> the concord num feature and the two conjuncts need to have the<br />
same value <strong>in</strong> their concord num feature. This is satisfied by both conjuncts s<strong>in</strong>ce both<br />
members of the set have a pl value <strong>in</strong> their concord num feature. In modifiers, the<br />
<strong>in</strong>dex num feature is not constra<strong>in</strong>ed and therefore it may denote different referents. The<br />
modifier agrees <strong>in</strong> gender with the closest conjunct (i.e. leftmost). The two nouns are also<br />
semantically related. Thus, the standard f-structure for coord<strong>in</strong>ate nouns is also assumed<br />
<strong>in</strong> natural coord<strong>in</strong>ation cases with modifiers. The f-structure of example (98) is illustrated<br />
below:<br />
(99) Natural coord<strong>in</strong>ation <strong>in</strong> MG with a modifier<br />
gnosti: (↑pred) = ‘famous’<br />
((adj∈↑) concord num) = pl<br />
((adj ∈↑)L concord gend) = masc
5.6 Conclusion 153<br />
⎡<br />
NP<br />
ADJP<br />
gnosti N<br />
ipurgi<br />
N<br />
Conj<br />
ke<br />
N<br />
ifipurgi<br />
pred ‘conj’<br />
⎢ <br />
⎢ <br />
<br />
⎢<br />
adj pred ‘famous’<br />
⎢<br />
⎣<br />
<br />
⎤<br />
⎥<br />
<strong>in</strong>dex [num pl]<br />
⎥<br />
⎧ ⎡<br />
⎤ ⎫ ⎥<br />
pred ‘m<strong>in</strong>isters’<br />
⎥<br />
⎢ <br />
⎥<br />
⎥ ⎥<br />
⎢<br />
num pl ⎥ ⎥<br />
⎢concord<br />
⎥ ⎥<br />
b: ⎢ gend masc ⎥ ⎥<br />
⎥ ⎥<br />
⎢ ⎥ ⎥<br />
⎢<br />
⎣ num pl<br />
⎥ ⎥<br />
⎪⎨<br />
<strong>in</strong>dex<br />
⎦ ⎥<br />
gend masc ⎪⎬<br />
⎥<br />
⎡<br />
⎤ ⎥<br />
pred ‘deputy-m<strong>in</strong>isters’ ⎥<br />
⎢ ⎥<br />
⎥ ⎥<br />
⎢<br />
num pl ⎥ ⎥<br />
⎢concord<br />
⎥ ⎥<br />
g: ⎢ gend masc ⎥ ⎥<br />
⎥ ⎥<br />
⎢ ⎥ ⎥<br />
⎢<br />
⎣ num pl<br />
⎥ ⎥<br />
⎪⎩<br />
<strong>in</strong>dex<br />
⎦ ⎥<br />
gend masc ⎪⎭<br />
⎦<br />
Thus, the MG natural coord<strong>in</strong>ation cases with a shared modifier follow the concord<br />
system.<br />
To summarise, the standard f-structure of accidental coord<strong>in</strong>ation is assumed for natural<br />
coord<strong>in</strong>ation patterns <strong>in</strong> MG modifiers and the MG def<strong>in</strong>ite determ<strong>in</strong>er. The plural<br />
def<strong>in</strong>ite determ<strong>in</strong>er shows <strong>in</strong>dex/concord agreement <strong>in</strong> accidental and <strong>in</strong> natural coord<strong>in</strong>ation<br />
structures, allow<strong>in</strong>g jo<strong>in</strong>t and split read<strong>in</strong>gs as this system predicts. The modifiers<br />
show concord agreement <strong>in</strong> natural and <strong>in</strong> accidental coord<strong>in</strong>ation structures. The use of<br />
the set representation allows the addition of two or more conjuncts, which is a possibility<br />
<strong>in</strong> MG. Thus, the difference between natural and accidental coord<strong>in</strong>ation is not a difference<br />
<strong>in</strong> syntactic representation for MG as <strong>in</strong> other languages. The f-structure of natural coord<strong>in</strong>ation<br />
<strong>in</strong> MG is the same as the f-structure of accidental coord<strong>in</strong>ation. A problem that<br />
still rema<strong>in</strong>s is that we do not impose any additional restrictions concern<strong>in</strong>g the relation<br />
between the two conjuncts, which must be semantically related <strong>in</strong> natural coord<strong>in</strong>ation<br />
structures. This is someth<strong>in</strong>g that must be expressed at the semantic level but it needs<br />
further <strong>in</strong>vestigation s<strong>in</strong>ce it might relate to a different level of representation with<strong>in</strong> the<br />
framework of LFG and will be left for future research.<br />
5.6 Conclusion<br />
In this chapter, we exam<strong>in</strong>ed NP <strong>in</strong>ternal agreement <strong>in</strong> MG coord<strong>in</strong>ate noun phrases,<br />
focus<strong>in</strong>g on MG modifiers and the MG def<strong>in</strong>ite determ<strong>in</strong>er. The data showed that MG<br />
modifiers behave differently from the MG def<strong>in</strong>ite determ<strong>in</strong>er. A shared MG modifier<br />
agrees <strong>in</strong> number with both conjuncts and <strong>in</strong> gender with the closest conjunct. Also, it<br />
scopes over coord<strong>in</strong>ate s<strong>in</strong>gular or plural nouns and allows two types of read<strong>in</strong>g, the jo<strong>in</strong>t<br />
and the split read<strong>in</strong>g. On the other hand, a shared MG def<strong>in</strong>ite determ<strong>in</strong>er agrees <strong>in</strong><br />
number and gender with both conjuncts and it scopes over coord<strong>in</strong>ate s<strong>in</strong>gular or plural<br />
nouns, generat<strong>in</strong>g only the jo<strong>in</strong>t read<strong>in</strong>g. A number of exceptional patterns occur when<br />
a shared MG plural determ<strong>in</strong>er scopes over coord<strong>in</strong>ate plural nouns and generates a split<br />
<strong>in</strong>terpretation. These patterns are considered cases of natural coord<strong>in</strong>ation.<br />
We showed that MG modifiers show distributive concord agreement <strong>in</strong> number and<br />
closest conjunct agreement <strong>in</strong> gender only and the best system that captures them is the<br />
concord agreement system. Also, the MG def<strong>in</strong>ite determ<strong>in</strong>er shows <strong>in</strong>dex and concord<br />
agreement <strong>in</strong> number and distributive concord agreement <strong>in</strong> gender, while it is captured
5.6 Conclusion 154<br />
by the concord/<strong>in</strong>dex agreement system. Syntactically, the MG def<strong>in</strong>ite determ<strong>in</strong>er is<br />
accounted for by the concord/<strong>in</strong>dex system but <strong>in</strong> plural number the MG determ<strong>in</strong>er<br />
<strong>in</strong>troduces further specifications that are directly related to semantics.<br />
F<strong>in</strong>ally, to account for natural coord<strong>in</strong>ation patterns we argued that we cannot follow<br />
the proposal of Dalrymple and Nikolaeva (2006). Natural coord<strong>in</strong>ation patterns <strong>in</strong> MG<br />
differ from natural coord<strong>in</strong>ation patterns <strong>in</strong> other languages and the f-structure on natural<br />
coord<strong>in</strong>ation proposed by Dalrymple and Nikolaeva (2006) cannot capture MG. Thus, we<br />
argued that the standard f-structure for accidental coord<strong>in</strong>ation is able to account for the<br />
MG data <strong>in</strong> which conjuncts are represented as members of a set, the determ<strong>in</strong>er specifies<br />
both concord and <strong>in</strong>dex agreement requirements and the modifiers specify concord<br />
agreement requirements, while the two conjuncts must be semantically related. Further<br />
work is required to exam<strong>in</strong>e whether and how it is possible to express the semantic relation<br />
between the conjuncts <strong>in</strong> natural coord<strong>in</strong>ation constructions.
Chapter 6<br />
Disjunctively Conjo<strong>in</strong>ed <strong>Noun</strong><br />
<strong>Phrases</strong><br />
6.1 Introduction<br />
Although disjunction was discussed by a number of scholars, it has not been studied as<br />
extensively as conjunction. More recently, there has been some <strong>in</strong>terest by a number of<br />
l<strong>in</strong>guists, such as Morgan (1972, 1984, 1985), Peterson (1986), Jenn<strong>in</strong>gs (1994), Eggert<br />
(2002) and Alonso-Ovalle (2006). The central issues discussed are the existence or not of a<br />
second logical operator the exclusive or, the relation of the logical operators to the natural<br />
language or and verb agreement with a disjunctive coord<strong>in</strong>ate noun phrase.<br />
In truth-conditional semantics, logicians agree on the existence of the primary logical<br />
connective or represented as ∨ and known as the <strong>in</strong>clusive or. A number of l<strong>in</strong>guists,<br />
philosophers and logicians discuss the issue of whether to add to propositional logic a<br />
second logical connective represented as and known as the exclusive or.<br />
Concern<strong>in</strong>g the relation of the logical operators ∨ (and for those who support its<br />
existence) with natural language or, the two former are b<strong>in</strong>ary operators restricted to<br />
propositional logic but the latter is not b<strong>in</strong>ary and seems to coord<strong>in</strong>ate anyth<strong>in</strong>g from<br />
sentences to bare noun phrases. Therefore, natural language or is more flexible than the<br />
two logical connectives which connect only whole sentences. F<strong>in</strong>ally, the last issue is verb<br />
agreement with a coord<strong>in</strong>ate noun phrase that consists of two disjuncts, either s<strong>in</strong>gular<br />
or mixed <strong>in</strong> number 1 . S<strong>in</strong>gular number disjuncts allow either a s<strong>in</strong>gular or a plural verb.<br />
Mixed number disjuncts follow the same pattern. All the above issues show that disjunction<br />
is a complicated but rather <strong>in</strong>trigu<strong>in</strong>g phenomenon.<br />
In this chapter, we will focus on agreement with disjunctively conjo<strong>in</strong>ed nouns. In the<br />
first section, we will discuss the dist<strong>in</strong>ction between the <strong>in</strong>clusive and exclusive or <strong>in</strong> truthconditional<br />
semantics and whether a similar dist<strong>in</strong>ction holds <strong>in</strong> natural language. Next,<br />
we will present an overview of the theories <strong>in</strong> disjunctive coord<strong>in</strong>ate nouns. F<strong>in</strong>ally, we will<br />
present disjunction phenomena <strong>in</strong> MG coord<strong>in</strong>ate noun phrases, try<strong>in</strong>g to shed some light<br />
on the factors that play a crucial role <strong>in</strong> determ<strong>in</strong><strong>in</strong>g verbal agreement with MG disjunctive<br />
nouns s<strong>in</strong>ce disjunctive coord<strong>in</strong>ate nouns and especially the issue of verb agreement and<br />
<strong>in</strong>terpretation <strong>in</strong> MG has not been discussed <strong>in</strong> any of the MG grammar books (Clairis<br />
et al. (2004), Holton et al. (1997), Triantaphyllidis (1994), Triantaphyllidis (2005), Joseph<br />
and Philippaki-Warburton (1987), Mackridge (1985) and Holton et al. (1997)).<br />
1 Uniformly plural disjuncts will not be exam<strong>in</strong>ed s<strong>in</strong>ce the resolved number is always plural and they<br />
do not present any special <strong>in</strong>terest like uniformly s<strong>in</strong>gular disjuncts or mixed number disjuncts.
6.2 Inclusive vs Exclusive or 156<br />
6.2 Inclusive vs Exclusive or<br />
The semantics of disjunction or has attracted much <strong>in</strong>terest among semanticians, logicians,<br />
philosophers and l<strong>in</strong>guists. A primary concern is whether, apart from the truth-functional<br />
<strong>in</strong>clusive or 2 , there is a purely truth-functional exclusive or as well. The second issue is<br />
if natural language adopts a similar k<strong>in</strong>d of dist<strong>in</strong>ction between an ‘<strong>in</strong>clusive-and’ and an<br />
‘exclusive-or’ <strong>in</strong>terpretation. In the current section, we present some background <strong>in</strong>formation<br />
on the logic of disjunction.<br />
In its very basic semantic <strong>in</strong>terpretation, disjunction is a logical connective which is<br />
truth-functional; the value for a complex proposition formed with a truth-functional connective<br />
is derived by the truth values of two propositions, without mak<strong>in</strong>g any reference<br />
to the content of the propositions while the best way to represent the truth-function of a<br />
complex proposition is a truth table that tabulates the graph of how disjunction works.<br />
The majority of logicians and l<strong>in</strong>guists (see Tarski (1941); Ambrose and Lazerowitz<br />
(1962); Qu<strong>in</strong>e (1972); Brody (1973); Hurford (1974); Manicas (1976); Georgakakos (1979);<br />
Barker (1985)) perceive a two way dist<strong>in</strong>ction of disjunction which is reflected <strong>in</strong> the use<br />
of natural language or. The first one is what is referred to as “<strong>in</strong>clusive disjunction” and<br />
it is commonly glossed as ‘and/or’ (Kearns, 2000) <strong>in</strong> natural language.<br />
In the follow<strong>in</strong>g table from Kearns (2000), assum<strong>in</strong>g that p and q are the two propositions,<br />
<strong>in</strong>clusive or disjunction is represented with the truth-functions below:<br />
Table 6.1: Truth Table of Inclusive or<br />
p q p∨q<br />
T T T<br />
T F T<br />
F T T<br />
F F F<br />
Inclusive or results <strong>in</strong> a truth value whenever one or more of its operands are true and results<br />
<strong>in</strong> a false value if and only if both of its operands are false. So crucially a proposition p∨q<br />
is T(rue) even if both p and q are true. The follow<strong>in</strong>g example shows how the logical<br />
disjunction corresponds to ‘<strong>in</strong>clusive disjunction’ (‘and/or’) <strong>in</strong> natural language:<br />
(1) You can get there by tra<strong>in</strong> or bus<br />
(‘You can get there by tra<strong>in</strong> or you can get there by bus’)<br />
(Kearns, 2000)<br />
In this case, the sentence is true if one of the disjuncts is true and even if both disjuncts<br />
are true.<br />
The second logical operator is what they call “exclusive disjunction” which is commonly<br />
glossed as ‘either/or’ <strong>in</strong> natural language. Exclusive or disjunction would have a different<br />
truth table, which differs from the <strong>in</strong>clusive or only when the disjuncts are both T(rue).<br />
This is illustrated below from Kearns (2000):<br />
2 We will use s<strong>in</strong>gle quotes to refer to the two <strong>in</strong>terpretations of or <strong>in</strong> natural language and italics to refer<br />
to the truh-functional or.
6.2 Inclusive vs Exclusive or 157<br />
Table 6.2: Truth Table of Exclusive or<br />
p q pq<br />
T T F<br />
T F T<br />
F T T<br />
F F F<br />
Hence, exclusive or disjunction results <strong>in</strong> a truth value if and only if exactly one of its<br />
operands is true. In other words, pq is F(alse) if both p and q have either a truth value<br />
or alternatively a false value, whereas it is true if at least or at most one of the propositions<br />
is T(rue). The sentence <strong>in</strong> example (1) illustrates how the logical exclusive or operator<br />
corresponds to natural language. The sentence conveys an exclusive <strong>in</strong>terpretation if it is<br />
paraphrased as ‘You can get there either by tra<strong>in</strong> or by bus, (but not both)’. Thus, the<br />
exclusive use of disjunction or <strong>in</strong> natural language requires the <strong>in</strong>terpretation of sentence<br />
<strong>in</strong> such a way as to allow the addition of the phrase qualification ‘but not both’ at the end<br />
of the comb<strong>in</strong>ed proposition.<br />
An oppos<strong>in</strong>g view to the general claim above is that there is only one truth-functional<br />
or and that the exclusive sense of or is not a purely truth-functional operator but it is<br />
occasionally understood by the hearer or arises <strong>in</strong> discourse conditions. Among others (see<br />
Tapscott (1976); Rub<strong>in</strong> and Young (1989) all cited <strong>in</strong> Jenn<strong>in</strong>gs (1994)), the most <strong>in</strong>fluential<br />
proponents of that view were Barret and Stanner (1971), Pelletier (1978) and more recently<br />
Jenn<strong>in</strong>gs (1994).<br />
Barret and Stanner (1971) deny completely the existence of “a purely truth-functional<br />
use of an exclusive or <strong>in</strong> English”(116), although they accept the existence of both an<br />
‘<strong>in</strong>clusive’ and an ‘exclusive’ sense of or <strong>in</strong> English. They support that the English examples<br />
found <strong>in</strong> logic textbooks that could be used to expla<strong>in</strong> a truth-functional exclusive or fail for<br />
two reasons. First, “the alternants of the alternation <strong>in</strong> question are logically <strong>in</strong>compatible<br />
and the problem of the jo<strong>in</strong>t truth cannot even arise” and second, what seems to be a<br />
truth-functional or is simply an ellipsis for a more complex expression which is not purely<br />
truth-functional (Barret and Stanner, 1971, 117) 3 . The most characteristic examples which<br />
fail for the first reason are the ones <strong>in</strong> (2), which are taken to display the exclusive or <strong>in</strong><br />
logic:<br />
(2) a. it is ra<strong>in</strong><strong>in</strong>g or it is not ra<strong>in</strong><strong>in</strong>g<br />
b. x is equal to two or x is equal to three<br />
(Barret and Stanner, 1971, 117)<br />
Barret and Stanner (1971) <strong>in</strong> order to show that the above examples are not cases of<br />
an exclusive or, they rely on the first row of the truth tables of <strong>in</strong>clusive and exclusive<br />
disjunction. As presented above, exclusive disjunction is dist<strong>in</strong>guished from <strong>in</strong>clusive only<br />
from the first row of the truth tables s<strong>in</strong>ce the former requires an F value while the latter<br />
requires a T value as the result of coord<strong>in</strong>at<strong>in</strong>g two propositions that are both T. Thus,<br />
any example that exemplifies the exclusive or needs both of its components to be T(rue)<br />
but the result to be F(alse). Both examples <strong>in</strong> (2) cannot be expla<strong>in</strong>ed by the first row of<br />
the exclusive or truth table s<strong>in</strong>ce the two alternants are <strong>in</strong>compatible and they cannot be<br />
simultaneously T(rue).<br />
An example which fails for the second reason is shown <strong>in</strong> (3). Firstly, this example<br />
cannot be a case of a truth-functional or s<strong>in</strong>ce it is not a statement and only statements<br />
are truth-functional. Barret and Stanner (1971) assume that example (4) “unpacks” the<br />
3 For more details on how they account for the nonexistence of a truth-functional or see Barret and<br />
Stanner (1971, 116).
6.2 Inclusive vs Exclusive or 158<br />
non-statement <strong>in</strong> (3). In this example, the two alternants are not <strong>in</strong>compatible and or<br />
could be seen as a case of a truth-functional exclusive or while they assume that example<br />
(5) is a paraphrase of example (4):<br />
(3) Tea or milk<br />
(4) You are permitted to have tea or milk<br />
(5) You are permitted to have tea or you are permitted to have milk<br />
(Barret and Stanner, 1971, 118)<br />
However, they show that (5) is not really a rephrasal of (4). They argue that if example<br />
(5) is comb<strong>in</strong>ed with the statement ‘you are not permitted to have tea’, it is considered<br />
a consistent statement of the form ‘p∨q: -p’, allow<strong>in</strong>g an <strong>in</strong>clusive or, or even ‘pq: -<br />
p’, allow<strong>in</strong>g an exclusive or. In this case, or is purely truth-functional and it does not<br />
represent a contradiction. For example, a waiter says ‘You are permitted to have tea or<br />
you are permitted to have milk, I’ve forgotten which’ and sometime later when he returns<br />
to the table he says ‘You are not permitted to have tea so I brought milk’ (Barret and<br />
Stanner, 1971, 119). But if example (4) is comb<strong>in</strong>ed with the same statement, the result<br />
is a self-contradictory statement and or is not purely truth-functional. For example, the<br />
waiter says ‘You are permitted to have tea or milk’ and when he returns to the table he<br />
says ‘You are not permitted to have tea unless you pay extra’ (Barret and Stanner, 1971,<br />
119). For the above reasons or <strong>in</strong> (4) cannot be a truth functional exclusive or 4 whereas<br />
or <strong>in</strong> (5) could be a truth-functional exclusive or. Thus, (5) is not a rephrasal of example<br />
(4) while the phrase tea or milk is simply an ellipsis of the phrase <strong>in</strong> (4).<br />
Similarly, Pelletier (1978) argues that the only or <strong>in</strong> English is the <strong>in</strong>clusive and that<br />
the exclusive or is derived by ellipsis or deletion. His first argument comes from examples<br />
like (6). He supports the claim that the statement is false if taxpayers file for both a s<strong>in</strong>gle<br />
and a jo<strong>in</strong>t return but this <strong>in</strong>terpretation is not due to the disjunction or but rather due<br />
to the restriction imposed by the first conjunct which states that only one return should<br />
be filed:<br />
(6) Taxpayers must file exactly one return, but it may be a s<strong>in</strong>gle or a jo<strong>in</strong>t return<br />
(Pelletier, 1978, 65)<br />
He draws further arguments from sentences which seem to convey an exclusive or as <strong>in</strong><br />
(7). First, he assumes that such a sentence is true if and only if its negation is false. A<br />
sentence with either...or negates with neither...nor <strong>in</strong> order to be false:<br />
(7) Arlene wants a marguerita or a grasshopper<br />
(8) Arlene wants neither a marguerita nor a grasshopper<br />
(Pelletier, 1978, 67)<br />
Also, sentence (8) is true <strong>in</strong> the same circumstances as the follow<strong>in</strong>g is true:<br />
(9) Arlene doesn’t want a marguerita and Arlene doesn’t want a grasshopper<br />
(Pelletier, 1978, 67)<br />
Thus, s<strong>in</strong>ce sentence (9) is <strong>in</strong>terpreted us<strong>in</strong>g the conjunction and <strong>in</strong> order to convey<br />
the equivalent mean<strong>in</strong>g of sentence (8) and <strong>in</strong> order to be the correspond<strong>in</strong>g false one of<br />
the true sentence (7), Pelletier (1978) concludes that sentence (7) is true only when both<br />
disjuncts are true and that their negative forms are false also when both disjuncts are false.<br />
4 Barret and Stanner (1971, 118-20) present more arguments aga<strong>in</strong>st the existence of truth-functional<br />
exclusive or.
6.2 Inclusive vs Exclusive or 159<br />
This shows that sentences with the disjunction or are basically <strong>in</strong>clusive and that exclusive<br />
or is drawn through the process of deletion.<br />
Jenn<strong>in</strong>gs (1994), also, extensively discusses the issues concern<strong>in</strong>g the existence of an<br />
<strong>in</strong>clusive and an exclusive or disjunction argu<strong>in</strong>g towards the view that or, like but not dual<br />
to and, does not function as a b<strong>in</strong>ary logical connective but as a k<strong>in</strong>d of a “list punctuat<strong>in</strong>g<br />
device”(293). For example, both sentences You may have tea, or cofee, or milk or You may<br />
have tea, or you may have cofee, or you may have milk <strong>in</strong>volve lists. The first <strong>in</strong>volves<br />
lists of nouns and the second lists of sentential clauses. None of these two, however, are<br />
disjunctive <strong>in</strong> an <strong>in</strong>clusive or an exlusive sense.<br />
The above argument follows from his view that the same dist<strong>in</strong>ction between an <strong>in</strong>clusive<br />
and an exclusive or disjunction cannot hold <strong>in</strong> Lat<strong>in</strong> either. This means that the lexical<br />
dist<strong>in</strong>ction <strong>in</strong> Lat<strong>in</strong> of the two words vel and aut is not one of <strong>in</strong>clusive and exclusive or.<br />
In Lat<strong>in</strong>, the orig<strong>in</strong>al view was that “the word vel expresses weak or <strong>in</strong>clusive disjunction,<br />
and the word aut corresponds to the word or <strong>in</strong> its strong or exclusive sense” (Copi, 1971,<br />
241). The follow<strong>in</strong>g example could be seen as a clear case of an exclusive or where aut<br />
is used <strong>in</strong> the formation of lists of disjo<strong>in</strong>t or contrasted or opposed items, categories or<br />
classes or states:<br />
(10) Omne enuntiatum aut verum aut falsum est<br />
Every statement is either true or false<br />
(Jenn<strong>in</strong>gs, 1994, 242)<br />
Once Jenn<strong>in</strong>gs (1994) considers a different example the behaviour of disjunction has similar<br />
problems as those <strong>in</strong> English. If the example <strong>in</strong> (11) is negated as <strong>in</strong> (12), the result<strong>in</strong>g<br />
sentence will expect a read<strong>in</strong>g along the l<strong>in</strong>es of <strong>in</strong>clusive disjunction.<br />
(11) Timebat tribunos aut plebes<br />
One feared the magistrates or the mob<br />
(12) Nemo timebat tribunos aut plebes<br />
No one feared the magistrates or the mob<br />
(Jenn<strong>in</strong>gs, 1994, 243)<br />
The example <strong>in</strong> (12) means ‘no one feared either’ and ‘not everyone either feared neither<br />
or feared both’. Thus, it is <strong>in</strong>terpreted as an <strong>in</strong>clusive disjunction and not as an exclusive<br />
disjunction. Therefore, Jenn<strong>in</strong>gs (1994) claims that Lat<strong>in</strong> does not really have two words<br />
for or but that Lat<strong>in</strong> had many words that are translated <strong>in</strong>to English as or.<br />
In search of a solution, Jenn<strong>in</strong>gs (1994) discusses nondisjunctive uses of or found <strong>in</strong> four<br />
different environments. When or occurs <strong>in</strong> the complement of a comparative adjective, as<br />
<strong>in</strong> example (13), when it occurs <strong>in</strong> expressions of preference, as <strong>in</strong> (14), when it occurs with<br />
sentences that express modality, as <strong>in</strong> (15), and when or is immediately subord<strong>in</strong>ated to<br />
if, as <strong>in</strong> (16):<br />
(13) Mary is heavier than Jack or Bob<br />
(14) I prefer shiitakes or truffles to morels or oysters<br />
(15) Jane likes Sally or Peter<br />
(16) If you drop it or if you throw it, it will break<br />
(Jenn<strong>in</strong>gs, 1994, 85-9)<br />
He actually adopts a more speech-act oriented approach plac<strong>in</strong>g or with<strong>in</strong> the realm of<br />
pragmatics and refers to the co-ord<strong>in</strong>ator or as a discourse-adverbial one. Such uses are<br />
when or is used to mark a list of alternatives, as seen below repeated from above:
6.2 Inclusive vs Exclusive or 160<br />
(17) You may have tea or coffee<br />
(Jenn<strong>in</strong>gs, 1994, 289)<br />
Associated with this use is when or is used to correct by weaken<strong>in</strong>g, seen below:<br />
(18) My brother, or my half-brother, arrived yesterday<br />
(Jenn<strong>in</strong>gs, 1994, 290)<br />
A further use is one <strong>in</strong> which possibilities are mooted. This means that or <strong>in</strong> this case<br />
goes unnoticed s<strong>in</strong>ce it <strong>in</strong>itiates whole paragraphs and therefore escapes the notice of our<br />
ma<strong>in</strong>ly sentential focus. F<strong>in</strong>ally, a last use is the exegetical one mostly found with glosses,<br />
as seen below:<br />
(19) This is a theorem schema, or meta-theorem<br />
(Jenn<strong>in</strong>gs, 1994, 290)<br />
The expression theorem schema is glossed by the expression meta-theorem. Thus,<br />
Jenn<strong>in</strong>gs (1994) argues that or is neither truth-functionally <strong>in</strong>clusive or exclusive but it is<br />
seen as discourse-adverbial.<br />
On the opposite side stand a number of scholars who support the claim that the exclusive<br />
or is the prevalent disjunction and that the <strong>in</strong>clusive or occurs less often and sometimes<br />
not at all. Lakoff (1971) claims that or is truth-functionally exclusive. One characteristic<br />
example is illustrated below:<br />
(20) Either John eats meat or Harry eats fish<br />
(Lakoff, 1971, 142)<br />
In the above example, there is only one conjunct that is true and therefore the other one<br />
is not. The <strong>in</strong>terpretation assigned to the above example is below and, as it is seen, there<br />
is only one choice:<br />
(21) a. If John doesn’t eat meat, Harry eats fish<br />
b. If John eats meat, Harry doesn’t eat fish<br />
(Lakoff, 1971, 142)<br />
Lakoff (1971) is led to assert that there is an element like choose that underlies the structure<br />
of or. She does not present any conv<strong>in</strong>c<strong>in</strong>g evidence but her argument is based on the fact<br />
that sentences conta<strong>in</strong><strong>in</strong>g or are usually ambiguous <strong>in</strong> that the choice is either assumed<br />
to be that of the speaker or that of the hearer, or that of the participant <strong>in</strong> the action<br />
or that of any other unspecified person 5 . Thus, this k<strong>in</strong>d of analysis, which assumes the<br />
existence of the verb choose <strong>in</strong> the underly<strong>in</strong>g structure of disjunction, shows that Lakoff<br />
(1971) accepts only the exclusive or and not the <strong>in</strong>clusive.<br />
The literature overview showed the different views of scholars regard<strong>in</strong>g the existence<br />
of both a truth-functional <strong>in</strong>clusive and a truth-functional exclusive or. The most common<br />
belief is that an <strong>in</strong>clusive or does exist and truth-conditional semantics accounts for that<br />
based on propositional logic. The existence of a purely truth-functional exclusive or is<br />
rather doubtful s<strong>in</strong>ce <strong>in</strong> many cases natural language expressions that could be seen as<br />
cases of a truth-functional exclusive or are <strong>in</strong>adequate to provide an explanation for its<br />
existence (Barret and Stanner, 1971; Pelletier, 1978; Jenn<strong>in</strong>gs, 1994). Next, we will discuss<br />
some issues related to the natural language or.<br />
5 For more details see Lakoff (1971).
6.2 Inclusive vs Exclusive or 161<br />
6.2.1 Natural Language or<br />
Despite the disagreement concer<strong>in</strong>g or truth-conditional semantics, none of the above scholars<br />
denies the existence of an ‘<strong>in</strong>clusive’(‘and-coord<strong>in</strong>ate’) and an ‘exclusive’(‘or’) sense of<br />
or <strong>in</strong> natural languages. A rather common l<strong>in</strong>e taken by most is that we should be careful<br />
when relat<strong>in</strong>g the truth-conditional <strong>in</strong>clusive or (∨) with or <strong>in</strong> natural languages s<strong>in</strong>ce the<br />
former connects merely propositions while the latter connects not only sentences but any<br />
grammatical type. This means that the natural language or is much more flexible than the<br />
truth-conditional <strong>in</strong>clusive or (∨).<br />
The follow<strong>in</strong>g sentence can be paraphrased accord<strong>in</strong>g to propositional logic and <strong>in</strong> this<br />
case the disjunction of the two NPs corresponds to the disjunction of two propositions. In<br />
this case or is a clear-cut case of the truth-conditional <strong>in</strong>clusive or:<br />
(22) Tom or Mary will visit me tonight<br />
(23) Tom will visit me tonight or Mary will visit me tonight<br />
(de Swart, 1998, 58)<br />
Many examples, however, do not follow that pattern, such as the follow<strong>in</strong>g taken from<br />
de Swart (1998):<br />
(24) A doctor or a dentist can write prescriptions<br />
(p ∧ q) where<br />
p = a doctor can write prescriptions<br />
q = a dentist can write prescriptions<br />
(de Swart, 1998, 58)<br />
Accord<strong>in</strong>g to de Swart (1998), the <strong>in</strong>terpretation of the above example is that both<br />
doctors and dentists can write prescriptions, mean<strong>in</strong>g that the correct translation for the<br />
above phrase is a conjunctive statement and not a truth-conditional disjunctive one. Thus,<br />
the phrase above can be paraphrased with and <strong>in</strong>stead of or.<br />
Also, Jenn<strong>in</strong>gs (1994) showed that natural language or has a wide number of uses that<br />
it can jo<strong>in</strong> almost anyth<strong>in</strong>g especially when tak<strong>in</strong>g <strong>in</strong>to consideration discourse conditions.<br />
Thus, the natural language or is much more flexible than the logical connective ∨ and, as<br />
most scholars claim, we should be careful when associat<strong>in</strong>g the logical connective with the<br />
‘<strong>in</strong>clusive-and’(conjunctive) sense of natural language or.<br />
Similarly, although most scholars try to reject the existence of an exclusive or <strong>in</strong> truthconditional<br />
semantics, there is an implicit belief that the purely disjunctive or ‘exclusive’<br />
sense of natural language or occurs across languages, especially if one considers sentences<br />
based on real world knowledge as the ones below:<br />
(25) Mary is (either) at home or at college right now<br />
In the above sentence, Mary has to be only at one place and not at two places simultaneously.<br />
Thus, the phrase conveys an ‘exclusive’ sense <strong>in</strong> natural language and it does<br />
not really need to be expla<strong>in</strong>ed based on the reason<strong>in</strong>g of truth-conditional semantics <strong>in</strong><br />
order to prove the truth of its exclusivity. Thus, we believe that an ‘exclusive’ sense of<br />
natural language or also should be seen separately from the truth-conditions approach and<br />
should be perceived <strong>in</strong> association to the discourse facts, which is what some l<strong>in</strong>guists have<br />
done. The fact that discourse conditions frequently determ<strong>in</strong>e speaker’s perception and<br />
<strong>in</strong>terpretation of various utterances shows that we cannot always rely on syntax and logic<br />
to account for the various grammatical phenomena.<br />
To conclude, we briefly presented evidence show<strong>in</strong>g that the basic controversy among<br />
l<strong>in</strong>guists, logicians and philosophers concerns whether there is an exclusive or <strong>in</strong> truthconditional<br />
semantics, based on propositional logic us<strong>in</strong>g natural language data. What is
6.3 Approaches to Disjunctive Coord<strong>in</strong>ated NPs 162<br />
commonly agreed, though, is that there is an ‘<strong>in</strong>clusive’(‘and-coord<strong>in</strong>ate) and an ‘exclusive’<br />
sense of or <strong>in</strong> natural language, proved by a number of examples that cannot be expla<strong>in</strong>ed<br />
rely<strong>in</strong>g on propositional logic but are based on discourse facts. Thus, speakers have the<br />
ability to perceive and <strong>in</strong>terpret sentences differently and this is why both <strong>in</strong>terpretations<br />
of or seem to occur freely <strong>in</strong> natural languages. This conclusion is crucial for the next<br />
sections s<strong>in</strong>ce it is a view followed by l<strong>in</strong>guists <strong>in</strong> cases where syntax seems <strong>in</strong>adequate to<br />
expla<strong>in</strong> some agreement phenomena.<br />
6.3 Approaches to Disjunctive Coord<strong>in</strong>ated NPs<br />
The present section is an overview of the different approaches that were developed regard<strong>in</strong>g<br />
verb agreement with disjunctive coord<strong>in</strong>ate NPs. The problem of verb agreement with<br />
disjunctive NPs has been approached from different views, the syntactic, semantic and<br />
pragmatic. In this section, we will present an overview of these different approaches and<br />
we will see how each of these expla<strong>in</strong>s number verb agreement <strong>in</strong> disjunctive coord<strong>in</strong>ate<br />
NPs.<br />
6.3.1 Morgan’s Analysis of Disjunctive NPs<br />
Morgan (1972) discussed the issue of number agreement <strong>in</strong> conjo<strong>in</strong>ed and disjo<strong>in</strong>ed noun<br />
phrases, focus<strong>in</strong>g on whether the variation <strong>in</strong> number verb agreement is related to mean<strong>in</strong>g<br />
or to syntactic form.<br />
Morgan (1972) argues that, although verb agreement is a syntactic phenomenon, syntax<br />
is not enough to account for the variation <strong>in</strong> verb agreement s<strong>in</strong>ce it is also determ<strong>in</strong>ed by<br />
semantic/pragmatic pr<strong>in</strong>ciples. He identifies two central problems that occur <strong>in</strong> coord<strong>in</strong>ate<br />
noun phrases, the “determ<strong>in</strong>ation” problem and the “selection” problem. We will present<br />
the “determ<strong>in</strong>ation” problem s<strong>in</strong>ce this is relevant to the current discussion. The determ<strong>in</strong>ation<br />
problem refers to the decision as to “what properties of the selected NP determ<strong>in</strong>e<br />
the appropriate form of agreement <strong>in</strong> the verb”(278).<br />
Two structures that focus on the “determ<strong>in</strong>ation” problem and show how verb agreement<br />
is affected are conjo<strong>in</strong>ed subjects and There-<strong>in</strong>sertion, and or-conjo<strong>in</strong>ed subjects 6 .<br />
In conjo<strong>in</strong>ed subjects, speakers prefer plural verb agreement. There are cases, though,<br />
where the <strong>in</strong>terpretation of the NP determ<strong>in</strong>es verb agreement. In example (26a), the<br />
coord<strong>in</strong>ate nouns are perceived as separate entities and the verb is plural, whereas <strong>in</strong><br />
example (26b) they are perceived as one entity and the verb is s<strong>in</strong>gular:<br />
(26) a. Pickles and strawberries taste good<br />
b. Pickles and strawberries tastes good<br />
(Morgan, 1972, 280)<br />
In cases of “there-<strong>in</strong>sertion”, the verb usually agrees with the noun closest to it, display<strong>in</strong>g<br />
closest conjunct agreement (i.e. leftmost). This is illustrated below:<br />
(27) There was a man and a woman <strong>in</strong> the room<br />
(28) There were two women and a man <strong>in</strong> the room<br />
(Morgan, 1972, 280-1)<br />
Thus, Morgan (1972) shows that both syntax and semantics determ<strong>in</strong>e verb agreement<br />
<strong>in</strong> coord<strong>in</strong>ate NPs.<br />
6 These are part of a group of nouns that he calls Complex NPs. Other structures that are part of<br />
Complex NPs are what he refers to as the “highest quantifiers”, “disembodied heads”, “measure phrases”.<br />
We only focus on the two structures mentioned above, which are of <strong>in</strong>terest to us.
6.3 Approaches to Disjunctive Coord<strong>in</strong>ated NPs 163<br />
In or-conjo<strong>in</strong>ed subjects, verb agreement patterns are more complicated s<strong>in</strong>ce cases of<br />
closest conjunct agreement alternate with resolved agreement, but other factors also seem<br />
to play a role <strong>in</strong> verb agreement, i.e. the presence of the preconjunction either:<br />
(29) (Either) Harry or his parents *is/are com<strong>in</strong>g<br />
(30) (Either) Harry’s parents or his wife ?is/*are com<strong>in</strong>g<br />
(31) There was (either) a bee or two flies <strong>in</strong> the room<br />
(32) There were either two flies or a bee <strong>in</strong> the room<br />
(Morgan, 1972, 281)<br />
Accord<strong>in</strong>g to Morgan (1972), <strong>in</strong> the above examples the majority of speakers choose<br />
closest conjunct agreement when either is present whereas more speakers choose resolved<br />
agreement when either is not present. Also, a number of speakers f<strong>in</strong>d both closest conjunct<br />
agreement and resolution with or without the preconjunction either ungrammatical 7 . Thus,<br />
they reject both examples below:<br />
(33) *Are/*is (either) John or his parents here?<br />
(34) *Are/*is (either) John’s parents or his wife here?<br />
(Morgan, 1972, 281)<br />
From the above data, Morgan (1972) concludes that <strong>in</strong> coord<strong>in</strong>ate noun phrases any<br />
speaker acquires a specific verb agreement pr<strong>in</strong>ciple, and if this fails then s/he uses additional<br />
ones end<strong>in</strong>g up with a set of pr<strong>in</strong>ciples. These pr<strong>in</strong>ciples may range from syntactic to<br />
semantic/pragmatic, mean<strong>in</strong>g that both syntax and semantics are <strong>in</strong>volved <strong>in</strong> verb agreement.<br />
Morgan (1985) returns to the “determ<strong>in</strong>ation problem”. His central question is whether<br />
number agreement is syntactically or semantically motivated. If it is syntactic, number<br />
agreement should be associated with the purely formal property of the NPs. If it is semantic,<br />
the question is whether it should be associated with the actual mean<strong>in</strong>g of the NP or<br />
the speaker’s beliefs concern<strong>in</strong>g the properties of the <strong>in</strong>tended referent. In fact, he draws<br />
a dist<strong>in</strong>ction between purely syntactic agreement, semantic/pragmatic agreement and a<br />
complex mixture of syntactic and semantic/pragmatic agreement. In syntactic agreement,<br />
the number properties of an NP depend on its syntactic form. In semantic agreement,<br />
number refers to the properties of expressions that are not related to particular occasions<br />
of use but rather depend on the sense or <strong>in</strong>tended referent of the NP. In the complex type<br />
of agreement, sometimes the syntactic agreement will play an active role and other times<br />
the semantic/pragmatic, while it is also possible that syntactic and semantic/pragmatic<br />
agreement may conflict (Morgan, 1985, 70-1).<br />
In or coord<strong>in</strong>ated NPs, Morgan (1985) argues that verb form is determ<strong>in</strong>ed by semantic/pragmatic<br />
agreement. He argues that two dist<strong>in</strong>ct <strong>in</strong>terpretations seem to determ<strong>in</strong>e<br />
number verb agreement; when the sentence conveys an ‘<strong>in</strong>clusive’(‘and’) sense of or, the NP<br />
forces plural verb agreement. When the sentence conveys an ‘exclusive’(‘or’) sense, the verb<br />
shows s<strong>in</strong>gular agreement. The follow<strong>in</strong>g examples display cases of an ‘<strong>in</strong>clusive’(‘and’) and<br />
an ‘exclusive’(‘or’) read<strong>in</strong>g of or:<br />
(35) John or Bill is/??are go<strong>in</strong>g to w<strong>in</strong> the race<br />
(36) I don’t th<strong>in</strong>k John or Bill are/*is go<strong>in</strong>g to w<strong>in</strong> the race<br />
(Morgan, 1985, 72)<br />
7 Morgan (1972) does not provide any details concern<strong>in</strong>g the number of participants and their responses.
6.3 Approaches to Disjunctive Coord<strong>in</strong>ated NPs 164<br />
Accord<strong>in</strong>g to Morgan (1985), <strong>in</strong> example (35) verb agreement is s<strong>in</strong>gular s<strong>in</strong>ce there<br />
must be only one w<strong>in</strong>ner <strong>in</strong> any world where it is true. He also argues that the same<br />
example can have plural verb agreement such as <strong>in</strong> (36) s<strong>in</strong>ce it may convey the mean<strong>in</strong>g<br />
that out of a group of two <strong>in</strong>dividuals, no one of them is go<strong>in</strong>g to w<strong>in</strong> the race. Thus,<br />
<strong>in</strong> both cases the <strong>in</strong>terpretation of the utterance determ<strong>in</strong>es verb agreement. Therefore,<br />
Morgan shows that <strong>in</strong> disjunctive NPs the different <strong>in</strong>terpretations of the speakers will<br />
determ<strong>in</strong>e verb agreement.<br />
To sum up, Morgan (1972, 1985) concludes that there are crucial syntactic and semantic/pragmatic<br />
pr<strong>in</strong>ciples that determ<strong>in</strong>e number verb agreement <strong>in</strong> coord<strong>in</strong>ate NPs. In<br />
conjo<strong>in</strong>ed nouns, this becomes apparent <strong>in</strong> examples where the referent is <strong>in</strong>terpreted as a<br />
s<strong>in</strong>gle entity and s<strong>in</strong>gular verb agreement is required or as two separate entities and plural<br />
verb agreement is required. In disjunctive coord<strong>in</strong>ate NPs, number verb agreement is even<br />
more undeterm<strong>in</strong>ed s<strong>in</strong>ce the choice of the agree<strong>in</strong>g verb is affected by whether disjunction<br />
conveys an ‘<strong>in</strong>clusive’(‘and’) sense or an ‘exclusive’(‘or’) sense and this <strong>in</strong>terpretation varies<br />
across speakers.<br />
Morgan’s analysis <strong>in</strong>corporates basic notions that constitute the beg<strong>in</strong>n<strong>in</strong>g for research<strong>in</strong>g<br />
verb agreement with conjunctive and disjunctive NPs from new perspectives. As it turns<br />
out, syntax is not the only factor responsible for the agreement features of the verb but a<br />
comb<strong>in</strong>ation of different pr<strong>in</strong>ciples drive verb agreement, which range from syntactic to semantic/pragmatic<br />
and relate to speaker perception and variation. One problem, however,<br />
is that, even though he discusses the existence of a pragmatic pr<strong>in</strong>ciple provid<strong>in</strong>g valid<br />
data for his claim, he avoids treat<strong>in</strong>g it as a separate pr<strong>in</strong>ciple, but <strong>in</strong>stead he assumes it is<br />
an <strong>in</strong>tegral part of semantics, <strong>in</strong>corporat<strong>in</strong>g the former with<strong>in</strong> the field of the latter. This<br />
notion ends up be<strong>in</strong>g problematic for examples where verb agreement is dependent only<br />
on contextual <strong>in</strong>terpretation. If we consider the example ‘Do you th<strong>in</strong>k John or Bill is/are<br />
happy?’, both verbs could be equally acceptable depend<strong>in</strong>g on the context and speaker<br />
<strong>in</strong>terpretation, which will largerly determ<strong>in</strong>e the choice of the verb. Thus, his analysis is<br />
very <strong>in</strong>fluential but still ignores the purely pragmatic factors that seem to play a role <strong>in</strong><br />
disjunctive coord<strong>in</strong>ate nouns.<br />
6.3.2 Peterson’s Analysis of Disjunctive NPs<br />
Peterson (1986) has more to say about number agreement with disjunctively conjo<strong>in</strong>ed<br />
nouns. He presents some <strong>in</strong>terest<strong>in</strong>g data from English and concludes that syntax itself is<br />
not enough to account for disjunctive conjo<strong>in</strong>ed NPs, but rather speakers resort to what he<br />
calls “patch -up rules” of various types. These “patch-up rules” take the form of generalised<br />
pr<strong>in</strong>ciples. Thus, he goes further than Morgan (1972) by establish<strong>in</strong>g the specific pr<strong>in</strong>ciples<br />
that speakers resort to when they decide on number verb agreement <strong>in</strong> or conjo<strong>in</strong>ed nouns.<br />
The discussion is couched <strong>in</strong> LFG.<br />
Peterson (1986) starts his analysis from the two ma<strong>in</strong> pr<strong>in</strong>ciples that have been acknowledged<br />
by a number of l<strong>in</strong>guists, who have worked on disjunctive conjo<strong>in</strong>ed nouns,<br />
“proximity” and “plural w<strong>in</strong>s”. Proximity is mentioned <strong>in</strong> Morgan (1972) as the “closest<br />
conjunct pr<strong>in</strong>ciple” 8 . Similarly, “plural w<strong>in</strong>s” is described <strong>in</strong> Partridge (1956, 221) as<br />
follows: “if either of the two subjects jo<strong>in</strong>ed by or... is plural, the ensu<strong>in</strong>g verb must be<br />
plural”.<br />
Peterson (1986) carried out a survey to confirm the existence of the two pr<strong>in</strong>ciples<br />
and check whether any extra pr<strong>in</strong>ciples are also <strong>in</strong>volved <strong>in</strong> number verb agreement with<br />
disjunctive NPs. He presented a set of data to 42 undergraduate students at the University<br />
of Newcastle (NSW), who were given a set of sentences and were asked to select one of<br />
8 A number of other l<strong>in</strong>guists discuss “proximity”, such as Langendoen (1970, 25), Quirk and Greenbaum<br />
(1973) and Fowler (1926) without giv<strong>in</strong>g it a formal name.
6.3 Approaches to Disjunctive Coord<strong>in</strong>ated NPs 165<br />
the two given verb forms that they considered to be acceptable. They could also accept<br />
or reject both forms. The questionnaires <strong>in</strong>cluded two types of sentences with or without<br />
either. Declarative sentences, with the verb follow<strong>in</strong>g the subject, such as examples (37)<br />
and (38), and <strong>in</strong>terrogative sentences, with the verb preced<strong>in</strong>g the subject, such as examples<br />
(39) and (40):<br />
(37) A rabbit or a goat has/have eaten all my lettuces<br />
(38) Either John or Bill is/are responsible for the mess<br />
(39) Is/are either Fred or Bob responsible for this mess?<br />
(40) Is/are David or Mark com<strong>in</strong>g with me?<br />
(Peterson, 1986, 234-5)<br />
Interrogative sentences without either are considered two way ambiguous. For example,<br />
(40) has two read<strong>in</strong>gs, the “alternative” and the “yes/no” read<strong>in</strong>g, paraphrased below,<br />
respectively:<br />
(41) a. Which of the two boys is com<strong>in</strong>g with me? (only 1 of the two must be chosen)<br />
b. Is either of the two boys com<strong>in</strong>g with me? (the answer is yes or no)<br />
(Peterson, 1986, 235)<br />
Read<strong>in</strong>g (41a) forces the choice of only one boy to go with the speaker; read<strong>in</strong>g (41b)<br />
requires a ‘yes’ or ‘no’ answer mean<strong>in</strong>g that either both boys or none of them will go with<br />
the speaker.<br />
The responses that were given by the consultants fell <strong>in</strong>to various patterns <strong>in</strong> terms<br />
of number verb agreement. In declarative sentences <strong>in</strong> which conjuncts have the same<br />
number, Peterson (1986) f<strong>in</strong>ds that the consultants choose the plural number whenever the<br />
two disjuncts are plural but there is greater variation <strong>in</strong> verb agreement when disjuncts<br />
are s<strong>in</strong>gular s<strong>in</strong>ce there are people who accept both a plural and a s<strong>in</strong>gular verb. Also, the<br />
presence or absence of either makes no difference to the answers of the participants. His<br />
f<strong>in</strong>d<strong>in</strong>gs, therefore, contradict Morgan’s (1972) claim that the presence of either disallows<br />
an <strong>in</strong>clusive(‘and-coord<strong>in</strong>ate’) read<strong>in</strong>g of or. Also, Peterson (1986) argues quite strongly<br />
that the <strong>in</strong>clusive(‘and’) versus exclusive(‘or’) or dist<strong>in</strong>ction plays no crucial role <strong>in</strong> the<br />
choice of number verb agreement <strong>in</strong> this group of sentences. If <strong>in</strong>terpretation played a<br />
role, he would expect a higher plural response <strong>in</strong> (42) than <strong>in</strong> (43) s<strong>in</strong>ce the first conveys<br />
more easily the <strong>in</strong>clusive(‘and-coord<strong>in</strong>ate’) read<strong>in</strong>g than the second. However, this was<br />
not confirmed, conclud<strong>in</strong>g that the <strong>in</strong>clusive/exclusive dist<strong>in</strong>ction is not obvious <strong>in</strong> positive<br />
declaratives.<br />
(42) A rabbit or a goat has/have eaten all my lettuces<br />
(43) Either John or Bill is/are responsible for this mess.<br />
(Peterson, 1986, 237)<br />
The responses of the students of the above two sentences are shown below:<br />
SENTENCE SG PL BOTH N/A<br />
(42) 36 2 4 -<br />
(43) 36 6 - -<br />
The same conclusion is drawn with regard to the <strong>in</strong>clusive/exclusive dist<strong>in</strong>ction <strong>in</strong><br />
<strong>in</strong>terrogative sentences. The two way ambiguity <strong>in</strong> <strong>in</strong>terrogatives is mapped to the <strong>in</strong>clusive/exclusive<br />
ambiguity of or. Thus, the “alternative” read<strong>in</strong>g corresponds <strong>in</strong> a way to
6.3 Approaches to Disjunctive Coord<strong>in</strong>ated NPs 166<br />
the “exclusive” read<strong>in</strong>g, and the “yes/no” one corresponds to the “<strong>in</strong>clusive”. In the follow<strong>in</strong>g<br />
sentences, the higher number of plural <strong>in</strong> example (44) than <strong>in</strong> example (45) could<br />
be expla<strong>in</strong>ed <strong>in</strong> terms of the <strong>in</strong>clusive <strong>in</strong>terpretation of the former. However, the same<br />
explanation is not applicable <strong>in</strong> example (46), which although it <strong>in</strong>cludes either and would<br />
prevent an <strong>in</strong>clusive read<strong>in</strong>g, there were participants who favoured the plu w<strong>in</strong>s strategy.<br />
(44) Is/are David or Mark com<strong>in</strong>g with me?<br />
(45) Are/Is John or Bill go<strong>in</strong>g to w<strong>in</strong> the next race?<br />
(46) Is/Are either John or Bill responsible for this mess?<br />
(Peterson, 1986, 234-5)<br />
The numbers of the responses are summarised <strong>in</strong> the follow<strong>in</strong>g table:<br />
SENTENCE SG PL BOTH N/A<br />
(44) 30 10 2 -<br />
(45) 40 1 1 -<br />
(46) 28 12 2 -<br />
Thus, Peterson (1986) argues that the <strong>in</strong>clusive/exlusive or dist<strong>in</strong>ction is not <strong>in</strong> operation<br />
even <strong>in</strong> <strong>in</strong>terrogatives. Peterson (1986) concludes that the generalisation <strong>in</strong> these<br />
three types of sentences is that “yes/no” <strong>in</strong>terrogatives favour plural and “alternative”<br />
<strong>in</strong>terrogatives favour s<strong>in</strong>gular verb agreement. So it is the contrast between “yes/no” and<br />
“alternative” <strong>in</strong>terpretation and not the contrast between an “<strong>in</strong>clusive” and “exclusive”<br />
read<strong>in</strong>g of or that is reflected <strong>in</strong> number verb agreement.<br />
Next, he discusses declarative sentences with disjunctive coord<strong>in</strong>ate nouns <strong>in</strong> which<br />
the conjuncts differ <strong>in</strong> number. The degree of uncerta<strong>in</strong>ty on the part of the consultants<br />
is much greater for these sentences. He also identifies cases where the consultants gave<br />
different answers when they were asked for a second op<strong>in</strong>ion on the same sentence. In<br />
declarative sentences, when disjuncts occur <strong>in</strong> the follow<strong>in</strong>g order [NP.sg or NP.pl] and<br />
[NP.pl or NP.sg], Peterson (1986) identifies two strategies used by the participants, plu<br />
w<strong>in</strong>s and prox. plu w<strong>in</strong>s is stronger than prox s<strong>in</strong>ce plural is a rather popular choice.<br />
In the follow<strong>in</strong>g examples, the participants would choose a plural verb more often than a<br />
s<strong>in</strong>gular:<br />
(47) a. Either this tree or those shrubs has/have to be cut down<br />
b. Either your cats or my dog has/have eaten all the daisies<br />
(Peterson, 1986, 238)<br />
The table below shows the responses of the participants:<br />
SENTENCE SG PL BOTH N/A<br />
(47a) 6 30 3 3<br />
(47b) 15 25 - 2<br />
A small number of participants used a s<strong>in</strong>gular verb form <strong>in</strong> example (47a). This choice<br />
contradicts both the plu w<strong>in</strong>s and the prox strategy. Peterson (1986) presented two<br />
possible explanations: either the consultants consistently use s<strong>in</strong>gular where at least one<br />
conjunct is s<strong>in</strong>gular (s<strong>in</strong>g w<strong>in</strong>s strategy), or they choose a verb form which agrees with<br />
the first conjunct (fc w<strong>in</strong>s strategy).<br />
S<strong>in</strong>ce there is no conv<strong>in</strong>c<strong>in</strong>g evidence for the s<strong>in</strong>g w<strong>in</strong>s strategy, Peterson (1986) assumes<br />
that the existence of an fc w<strong>in</strong>s strategy is more likely. The high choice for a<br />
s<strong>in</strong>gular verb form <strong>in</strong> example (47a) is seen as evidence that fc w<strong>in</strong>s strategy is more
6.3 Approaches to Disjunctive Coord<strong>in</strong>ated NPs 167<br />
likely to exist than the sg w<strong>in</strong>s strategy. Also, sg w<strong>in</strong>s is rejected because s<strong>in</strong>gular occurs<br />
<strong>in</strong> high numbers <strong>in</strong> (47b) as a result of the prox strategy s<strong>in</strong>ce the second conjunct is s<strong>in</strong>gular<br />
and therefore the prox strategy takes effect. Thus, Peterson (1986, 234) concludes<br />
that a third possible strategy is the one <strong>in</strong> which the verb agrees <strong>in</strong> number with the first<br />
conjunct, known as the fc w<strong>in</strong>s one. Apart from example (47a) <strong>in</strong> which s<strong>in</strong>gular was<br />
chosen by a significant number of consultants, a s<strong>in</strong>gular verb form occurs <strong>in</strong> high numbers<br />
<strong>in</strong> the <strong>in</strong>terrogative sentences below:<br />
(48) a. Have/Has the president or the senators read the documents yet?<br />
b. Was/Were the cake or the scones made by Mary?<br />
(Peterson, 1986, 240)<br />
The responses are summarised below:<br />
SENTENCE SG PL BOTH N/A<br />
(48a) 33 3 4 2<br />
(48b) 27 9 4 2<br />
Accord<strong>in</strong>g to Peterson (1986) <strong>in</strong> the above structures, which <strong>in</strong>volve subject-auxiliary<br />
<strong>in</strong>version, fc w<strong>in</strong>s and prox are equivalent strategies. But even though prox could cause<br />
the occurrence of s<strong>in</strong>gular, it cannot cause the occurrence of s<strong>in</strong>gular <strong>in</strong> example (47a).<br />
Thus, <strong>in</strong> either-or declaratives three strategies are adopted by the speakers, plu w<strong>in</strong>s,<br />
prox and fc w<strong>in</strong>s. The strongest is plu w<strong>in</strong>s, followed by the prox, while fc w<strong>in</strong>s is<br />
the weakest of the three.<br />
F<strong>in</strong>ally, <strong>in</strong> <strong>in</strong>terrogative sentences with mixed number conjuncts, the results are rather<br />
different. In <strong>in</strong>terrogatives without either and a noun order [NP.pl or NP.sg], all three<br />
strategies, plu w<strong>in</strong>s, prox and fc w<strong>in</strong>s, favour a plural verb. In <strong>in</strong>terrogatives with a<br />
noun order [NP.sg or NP.pl], prox and fc w<strong>in</strong>s are <strong>in</strong> conflict with plu w<strong>in</strong>s s<strong>in</strong>ce the<br />
first two would produce s<strong>in</strong>gular responses and plu w<strong>in</strong>s would produce plural responses.<br />
Participants, however, prefered a s<strong>in</strong>gular verb compared to the equivalent declaratives,<br />
which shows a preference for the prox strategy. In <strong>in</strong>terrogatives which conta<strong>in</strong> either,<br />
participants prefer the plu w<strong>in</strong>s strategy more than any other. Thus, for <strong>in</strong>terrogatives<br />
with either the consultants prefer the plu w<strong>in</strong>s than the prox strategy, but for <strong>in</strong>terrogatives<br />
without either the prox strategy w<strong>in</strong>s over the rest.<br />
To sum up, Peterson’s analysis showed that <strong>in</strong> sentences with subject-auxiliary <strong>in</strong>version<br />
prox strategy is stronger but <strong>in</strong> declaratives plu w<strong>in</strong>s is used more often. Thus, prox is<br />
stronger when the verb precedes the subject, and plu w<strong>in</strong>s is stronger for declarative and<br />
<strong>in</strong>terrogative phrases with either. This implies that there is a strong relation of the strength<br />
of each strategy with the type of sentence construction and these should be exam<strong>in</strong>ed<br />
together when attempt<strong>in</strong>g to analyse disjunctive coord<strong>in</strong>ate nouns 9 .<br />
Through the above analysis, Peterson (1986) concludes that syntactic rules do not<br />
function as pr<strong>in</strong>ciples for verb agreement with disjunctive conjo<strong>in</strong>ed subjects but <strong>in</strong>stead<br />
native speakers need to use a number of strategies, known as “patch-up rules”. These<br />
“patch-up rules” are not ad hoc but they can be perceived as generalised pr<strong>in</strong>ciples. These<br />
pr<strong>in</strong>ciples may have a syntactic or semantic character and they differ <strong>in</strong> strength from<br />
speaker to speaker and also across constructions.<br />
It is clear that Peterson (1986) draws similar conclusions to Morgan (1972, 1985). His<br />
analysis is <strong>in</strong>sightful s<strong>in</strong>ce it manages to shed some light <strong>in</strong>to the ma<strong>in</strong> strategies employed <strong>in</strong><br />
English, concern<strong>in</strong>g number verb agreement with disjunctive coord<strong>in</strong>ate NPs. In addition,<br />
9 Peterson (1986) also extends his <strong>in</strong>vestigation by look<strong>in</strong>g at neither-nor conjo<strong>in</strong>ed subject NPs. As he<br />
concludes, the results are rather unclear and this is why we will not present any details of his <strong>in</strong>vestigation<br />
here.
6.3 Approaches to Disjunctive Coord<strong>in</strong>ated NPs 168<br />
he identifies formally three different pr<strong>in</strong>ciples used by native speakers, prox, plu w<strong>in</strong>s<br />
and fc w<strong>in</strong>s, reject<strong>in</strong>g sg w<strong>in</strong>s as a fourth strategy with<strong>in</strong> the limitations of the English<br />
language. Another advantage of his analysis is that he stresses the importance of word<br />
order <strong>in</strong> determ<strong>in</strong><strong>in</strong>g verb agreement with disjunctive NPs, draw<strong>in</strong>g concrete conclusions<br />
from the data. F<strong>in</strong>ally, he asserts that number verb agreement with disjunctive coord<strong>in</strong>ate<br />
nouns depends on syntactic and semantic or perceptual pr<strong>in</strong>ciples.<br />
An important drawback of his analysis, however, is that he considers the patch-up<br />
rules as generalised pr<strong>in</strong>ciples and he completely ignores the purely pragmatic or discourse<br />
factors that may determ<strong>in</strong>e verb agreement. Based on his data, he argues that the type of<br />
sentence determ<strong>in</strong>es verb agreement and not the <strong>in</strong>clusive/exclusive or <strong>in</strong>terpretation. In<br />
declaratives, the <strong>in</strong>clusive/exclusive dist<strong>in</strong>ction is not salient whereas <strong>in</strong> <strong>in</strong>terrogatives the<br />
contrast between “yes/no” and “alternatives” drives verb agreement. Morgan (1972, 1985),<br />
on the other hand, appears to accept those two <strong>in</strong>terpretations cross-sententially and this<br />
is why he <strong>in</strong>corporates a semantic/pragmatic pr<strong>in</strong>ciple that determ<strong>in</strong>es verb agreement.<br />
One last th<strong>in</strong>g to note is whether these speaker strategies are crossl<strong>in</strong>guistic and how<br />
many of those are grammaticalised or not <strong>in</strong> the various languages. Concern<strong>in</strong>g the first<br />
po<strong>in</strong>t, we argue that <strong>in</strong> MG these strategies are not all activated. In fact, the flexibility<br />
of the language itself will determ<strong>in</strong>e whether it is possible for all strategies to be used or<br />
not and not all languages display the same k<strong>in</strong>d of flexibility. Secondly, as Peterson (1986)<br />
mentions, <strong>in</strong> some languages some of the strategies are grammaticalised such as the CCA<br />
strategy <strong>in</strong> Albanian 10 . This depends on the <strong>in</strong>ternal structure of the language. In MG,<br />
these strategies are not grammaticalised but they seem to be part of the speakers’ strategic<br />
choices.<br />
6.3.3 An Intermodular Theory of Disjunction<br />
Eggert (2002) exam<strong>in</strong>es disjunctive coord<strong>in</strong>ate nouns and discusses all aspects that seem to<br />
be <strong>in</strong>volved <strong>in</strong> determ<strong>in</strong><strong>in</strong>g verb agreement, such as syntax, semantics and pragmatics. He<br />
focuses on an analysis of English verb agreement with disjunctive nouns with<strong>in</strong> the framework<br />
of Autolexical Grammar (Sadock, 1991). The advantage of the specific framework is<br />
that it consists of several <strong>in</strong>dependent modules each with its own set of rules which share<br />
the same lexicon 11 . Also, the current framework is rather useful <strong>in</strong> analys<strong>in</strong>g agreement<br />
s<strong>in</strong>ce it keeps separate each aspect <strong>in</strong> agreement, e.g. syntax or semantics, allow<strong>in</strong>g an<br />
adequate representation of agreement mismatches.<br />
Autolexical Grammar <strong>in</strong>cludes five modules: L<strong>in</strong>earity, Syntax, Argument/Predicate,<br />
Logical-Scope and Morpho-syntax. In his thesis, Eggert (2002) focuses only on the first<br />
three modules, even though he asserts that the other two are still relevant for the phenomenon<br />
of coord<strong>in</strong>ation. L<strong>in</strong>earity encodes “the l<strong>in</strong>ear order of words <strong>in</strong> a sentence”;<br />
“syntax encodes the dom<strong>in</strong>ance and command relations between syntactic phrases <strong>in</strong> a<br />
10 Accord<strong>in</strong>g to Peterson (1986), <strong>in</strong> Albanian the adjectival participle must agree with the closest conjunct,<br />
which is the f<strong>in</strong>al one.<br />
11 The ma<strong>in</strong> difference between LFG and Autolexical Grammar is that LFG dist<strong>in</strong>guishes two syntactic<br />
levels: c-structure (constituent-structure) and f-structure (functional structure) but it has no syntactic<br />
level of LF and no d-structure (Horrocks, 1987); Autolexical Grammar has only one level of syntax, which<br />
corresponds to c-structure <strong>in</strong> LFG. In theoretical discussions of mismatch phenomena this difference becomes<br />
more obvious. For example, <strong>in</strong> LFG the argument-structure properties of the verb seem (as <strong>in</strong> He seems<br />
to like it) are represented <strong>in</strong> terms of f-structure relations, which are <strong>in</strong>dependent of phrase structure<br />
(reference)(Horrocks, 1987). In this case, no movement is <strong>in</strong>volved and the two levels of syntax are more<br />
<strong>in</strong>dependent of each other. In Autolexical Grammar, the <strong>in</strong>formation concern<strong>in</strong>g the argument-structure of<br />
rais<strong>in</strong>g verbs is captured <strong>in</strong> semantics. Accord<strong>in</strong>g to Sadock (1991), verbs like seem justify the dist<strong>in</strong>ction<br />
between phrase structure and semantic structure and say noth<strong>in</strong>g about additional levels of syntax (Francis<br />
and Michaelis, 2000).
6.3 Approaches to Disjunctive Coord<strong>in</strong>ated NPs 169<br />
sentence”, and Argument/Predicate “represents the dom<strong>in</strong>ance and command relations<br />
but between the semantic phrases <strong>in</strong> a sentence” (Sadock, 1991, 7). His central addition<br />
is to supplement the current framework with an extra level, the discourse, that is mapped<br />
off of semantic structure, <strong>in</strong>clud<strong>in</strong>g also a set of weighted agreement constra<strong>in</strong>ts.<br />
Eggert (2002) first presents the grammatical structure of coord<strong>in</strong>ate phrases focus<strong>in</strong>g<br />
on agreement phenomena. He argues that agreement is an <strong>in</strong>termodular phenomenon separated<br />
<strong>in</strong>to syntactic agreement, accounted for by the L<strong>in</strong>earity and Syntax modules 12 , and<br />
semantic agreement, accounted for by the Argument/Predicate module 13 with<strong>in</strong> Autolexical<br />
Grammar. A clear case of syntactic agreement concern<strong>in</strong>g number is that of closest<br />
conjunct where the verb agrees <strong>in</strong> number with the noun closest to it, shown below:<br />
(49) a. Either John or his parents were/*was here<br />
b. Either John’s parents or his brother ?was/??were here<br />
(Eggert, 2002, 17)<br />
Semantic agreement is always agreement with the entire noun phrase and depends<br />
on the mean<strong>in</strong>g of the NP. In the examples below, when the two NPs refer to different<br />
<strong>in</strong>dividuals as <strong>in</strong> (50a) (i.e. split read<strong>in</strong>g) the verb shows plural number agreement, whereas<br />
when the NPs refer to the same <strong>in</strong>dividual (i.e. jo<strong>in</strong>t read<strong>in</strong>g) the verb shows s<strong>in</strong>gular<br />
number agreement:<br />
(50) a. Her husband and her lover say that they can’t live without her<br />
b. Her longtime companion and the future editor of her papers was happy to give<br />
his approval to her request (Eggert, 2002, 16)<br />
Eggert (2002) captures the syntactic aspects of agreement <strong>in</strong> coord<strong>in</strong>ate NPs based on<br />
the L<strong>in</strong>earity and Syntax modules. He po<strong>in</strong>ts out that any difference <strong>in</strong> the hierarchical<br />
structure <strong>in</strong> l<strong>in</strong>earity will necessarily result <strong>in</strong> a difference <strong>in</strong> the l<strong>in</strong>ear order. L<strong>in</strong>earity<br />
contributes greatly s<strong>in</strong>ce it accounts for ‘closeness’ factors <strong>in</strong> agreement. The agreement<br />
constra<strong>in</strong>ts that require agreement with a particular conjunct are based on the l<strong>in</strong>ear position<br />
of the noun, which must be closest to the verb. The phrase structure rules that<br />
the L<strong>in</strong>earity module enforces are rather simple, produc<strong>in</strong>g b<strong>in</strong>ary branch<strong>in</strong>g trees (Eggert,<br />
2002, 36-7). Like the Syntax module, the structure of L<strong>in</strong>earity conforms to specific<br />
command relations. The most characteristic relation is for any two constituents α and<br />
β, α precedes β iff α c-commands β. On the other hand, the Syntax module focuses on<br />
constituent structure and command relations. After exam<strong>in</strong><strong>in</strong>g a number of syntactic theories<br />
concern<strong>in</strong>g coord<strong>in</strong>ation, such as Chomsky’s (1957) “Conjunction Reduction”(CR),<br />
Goodall’s (1987) “Three dimensional phrase structure” and Johannessen’s (1996) “Coord<strong>in</strong>ate<br />
Alpha”, he concludes that an adapted version of GPSG is more effective simply<br />
because it does not <strong>in</strong>volve transformations and coord<strong>in</strong>ate structures are base-generated.<br />
GPSG relies on two ma<strong>in</strong> pr<strong>in</strong>ciples, the Head Feature Convention and the Foot Feature<br />
Pr<strong>in</strong>ciple.<br />
The semantic aspects of agreement <strong>in</strong> coord<strong>in</strong>ate NPs are captured by the Argument/Predicate<br />
module. Eggert (2002) argues for a unified analysis for each of the two<br />
connectors and and or that will allow them to operate over conjunct sets of any type and<br />
not simply be<strong>in</strong>g propositional connectors, as has been done <strong>in</strong> the past. He favours a<br />
quantificational account of and and or based on the observation that both operators are<br />
12 Issues <strong>in</strong> morphosyntax are captured with<strong>in</strong> the syntax module. However, he does not discuss further<br />
any issues <strong>in</strong> morphology s<strong>in</strong>ce it is beyond the scope of his thesis.<br />
13 His semantic aspect relates only to the Argument/Predicate module and not the Logical-Scope one.<br />
Although, he notes that a complete analysis of coord<strong>in</strong>ation requires the Logical-Scope module too, he does<br />
not consider it relevant to agreement and he chooses not to discuss it further.
6.3 Approaches to Disjunctive Coord<strong>in</strong>ated NPs 170<br />
<strong>in</strong>volved <strong>in</strong> the same types of scopal ambiguities as quantified NPs; also, <strong>in</strong> this way he can<br />
expla<strong>in</strong> patterns of non-distributive conjunction, which pose problems for other approaches<br />
(i.e. CR approach (Chomsky, 1957) and those by Gazdar (1980) and Montague (1973)).<br />
Eggert (2002) argues that conjunction and and disjunction or semantically resemble<br />
quantifiers. He proves this by treat<strong>in</strong>g and and or as analogous to all and the existentials,<br />
respectively. As is well known, all and a <strong>in</strong>teract <strong>in</strong> sentences and produce scope<br />
ambiguities, such as the ones below:<br />
(51) a. All of his friends belong to a band<br />
b. ‘all of his friends are band members’<br />
c. ‘there is a band that all of his friends belong to’<br />
(Eggert, 2002, 78)<br />
The position of the two quantifiers determ<strong>in</strong>es the difference <strong>in</strong> the two read<strong>in</strong>gs and<br />
which elements have scope over which.<br />
Similar, ambiguities arise <strong>in</strong> example (52a) with and and a, be<strong>in</strong>g ambiguous between<br />
(52b) and (52c), and <strong>in</strong> example (53a) with all and or, be<strong>in</strong>g ambiguous between (53b) and<br />
(53c):<br />
(52) a. Grant and Jacob are members of a band<br />
b. ‘Grant and Jacob are band members’<br />
c. ‘there is a band that Grant and Jacob are members of’<br />
(Eggert, 2002, 79)<br />
(53) a. All of his friends are members of Sunset Valley or Sketchy Afterdeal<br />
b. ‘All of his friends belong to one of the two bands: Sunset Valley or Sketchy<br />
Afterdeal’<br />
c. ‘All of his friends belong to Sunset Valley, or all of his friends belong to Sketchy<br />
Afterdeal’<br />
(Eggert, 2002, 79)<br />
Moreover, and and or <strong>in</strong>teract <strong>in</strong> this k<strong>in</strong>d of ambiguity:<br />
(54) a. Grant and Carm<strong>in</strong>e like Gertrude or Carmelia<br />
b. ‘Grant and Carm<strong>in</strong>e are such that they like Gertrude or Carmelia’<br />
c. ‘Gertrude or Carmelia is such that Grant and Carm<strong>in</strong>e like her’<br />
(Eggert, 2002, 79)<br />
Turn<strong>in</strong>g to a and all, it is impossible for a syntactically lower all to take scope over<br />
a syntactically higher a. For example, most people have trouble gett<strong>in</strong>g read<strong>in</strong>g (55c) for<br />
(55a):<br />
(55) a. A girl likes all the boys<br />
b. ‘A girl is such that she likes all the boys’<br />
c. ‘Every boy is such that a girl likes him’<br />
(Eggert, 2002, 80)<br />
Thus, there is a similarity between and and or and all and a.<br />
Also, and and all show similar behaviour with regard to distributivity s<strong>in</strong>ce they can<br />
be either distributive or non-distributive. In the follow<strong>in</strong>g examples, neither and nor all<br />
can be distributive:<br />
(56) a. Grant and Gertrude met <strong>in</strong> the park
6.3 Approaches to Disjunctive Coord<strong>in</strong>ated NPs 171<br />
b. *‘Grant met <strong>in</strong> the park and Gertrude met <strong>in</strong> the park’<br />
c. All the sheriffs met <strong>in</strong> the park<br />
d. *‘Sheriff1 met <strong>in</strong> the park, Sheriff2 met <strong>in</strong> the park...’<br />
(Eggert, 2002, 81)<br />
In addition, the cases where conjunction is ambiguous between a distributive and a<br />
non-distributive read<strong>in</strong>g are the same ones where sustitut<strong>in</strong>g the phrase all of the Ns is<br />
ambiguous. This is illustrated below:<br />
(57) a. Grant and Gertrude carried the box<br />
b. ‘Grant and Gertrude carried the box severally’<br />
c. ‘Grant and Gertrude carried the box together’<br />
d. All the sheriffs carried the box<br />
e. ‘All the sheriffs carried the box severally’<br />
f. ‘All the sheriffs carried the box together’<br />
(Eggert, 2002, 81)<br />
Thus, Eggert (2002) shows that and and or produce similar scopal ambiguities to<br />
quantifiers and and and all behave similarly with regard to distributivity 14 . The Argument/Predicate<br />
module, which posits a number of rules, can capture a unified treatment of<br />
both operators. The A/P rules, which correspond to arguments and predicates, state how<br />
these comb<strong>in</strong>e to form propositions, and treat conjunctions as operators that are allowed<br />
to comb<strong>in</strong>e with two or more conjuncts of any category but the conjuncts need to be of the<br />
same category. F<strong>in</strong>ally, an A/P coord<strong>in</strong>ation rule is <strong>in</strong>troduced, which treats conjunctions<br />
as n-ary and not b<strong>in</strong>ary 15 .<br />
Eggert (2002) shows that apart from the strict semantic factors, there are also discourse<br />
factors <strong>in</strong>volved when determ<strong>in</strong><strong>in</strong>g the ‘semantic number’ of an argument <strong>in</strong> coord<strong>in</strong>ate<br />
phrases. Thus, he claims that “‘semantic agreement’ should be analysed as a discourse<br />
phenomenon, which is <strong>in</strong>formed by the semantics” (Eggert, 2002, 97). Characteristic cases<br />
of discourse factors mediat<strong>in</strong>g <strong>in</strong> coord<strong>in</strong>ate NPs and determ<strong>in</strong><strong>in</strong>g verb agreement are<br />
cases of “appositive conjunction”(Hoeksema, 1988, 36) (or otherwise mentioned as jo<strong>in</strong>t<br />
<strong>in</strong>terpretation) and “deferred reference” (Nunberg, 1995, 115).<br />
A case of appositive conjunction or jo<strong>in</strong>t read<strong>in</strong>g is illustrated <strong>in</strong> example (58). The<br />
conjuncts denote the same referent and they trigger s<strong>in</strong>gular agreement. A semantic theory<br />
could capture that. However, appositive conjunction or jo<strong>in</strong>t read<strong>in</strong>g is only allowed with<br />
descriptions as <strong>in</strong> example (58) and not with proper nouns (Hoeksema, 1988, 36), as <strong>in</strong><br />
(59):<br />
(58) Grant’s former wife and his mother’s present girlfriend was on the Jerry Spr<strong>in</strong>ger<br />
show<br />
(59) ??Cassius Clay and Muhammed Ali has/have always been my father’s favorite boxer<br />
(Eggert, 2002, 93)<br />
There are also cases where coreferential proper nouns can be conjo<strong>in</strong>ed but the verb<br />
admits plural verb agreement, as below:<br />
(60) Cassius Clay and Muhammed Ali are the same person<br />
(Eggert, 2002, 93)<br />
14 For more details on how the two coord<strong>in</strong>ation markers resemble quantifiers see Eggert (2002, 78-83).<br />
15 A complete list of the rules is <strong>in</strong> Eggert (2002, 84).
6.3 Approaches to Disjunctive Coord<strong>in</strong>ated NPs 172<br />
Hoeksema (1988, 30-31) argues that these phrases <strong>in</strong>troduce two dist<strong>in</strong>ct <strong>in</strong>tensional<br />
<strong>in</strong>dividuals, which correspond to one real-world entity. In these cases, <strong>in</strong>tensionality rather<br />
than extensionality determ<strong>in</strong>e verb agreement.<br />
Also, <strong>in</strong> cases of “deferred reference” agreement is only with the <strong>in</strong>tended referent.<br />
Thus, demonstratives are s<strong>in</strong>gular or plural depend<strong>in</strong>g on the <strong>in</strong>tended referent, not the<br />
demonstratum:<br />
(61) That/*those french fries is/*are gett<strong>in</strong>g impatient<br />
(Eggert, 2002, 96)<br />
Therefore, Eggert (2002) argues that such cases cannot be handled by purely semantic<br />
theories and therefore a discourse based account should be considered.<br />
To capture the discourse and semantic factors <strong>in</strong> agreement, Eggert (2002) proposes<br />
an analysis of coord<strong>in</strong>ate phrases based on Discourse Representation Theory (Kamp and<br />
Reyle, 1993). DRT uses a discourse structure that is mapped off of grammatical structure.<br />
Such a discourse structure allows the effective <strong>in</strong>corporation of any discourse-pragmatic<br />
features <strong>in</strong>to the mean<strong>in</strong>g of sentences when the latter are uttered <strong>in</strong> a particular discourse<br />
context. DRT has the advantage of account<strong>in</strong>g for complicated phenomena, such as<br />
anaphora, negation and scope. ‘Semantic number’ is also complicated and a DRT analysis<br />
is advantageous s<strong>in</strong>ce it takes both discourse and semantic factors <strong>in</strong>to consideration when<br />
determ<strong>in</strong><strong>in</strong>g the semantic number of an argument. Thus, consider<strong>in</strong>g a given argument,<br />
its semantic number is determ<strong>in</strong>ed by whatever discourse conditions apply to the discourse<br />
referent that corresponds to the argument.<br />
Eggert (2002) develops a uniform theory for and and or <strong>in</strong> order to capture the wide<br />
range of data and also <strong>in</strong>troduce a straightforward def<strong>in</strong>ition for plurality. He achieves that<br />
by comb<strong>in</strong><strong>in</strong>g DRT with Schwarzschild’s cover theory (Schwarzschild, 1996). He treats<br />
and as a type-specific operator mean<strong>in</strong>g that it is a generic operator that is identified with<br />
“whatever operator is def<strong>in</strong>ed for the doma<strong>in</strong> of the conjo<strong>in</strong>ed terms”(Eggert, 2002, 92) and<br />
not with meet per se, as <strong>in</strong> the Boolean approach proposed by Keenan and Faltz (1985).<br />
Thus, and does not map directly <strong>in</strong>to the DRS as set-union but as a generic operator,<br />
mean<strong>in</strong>g that it operates either as set-union or as set-<strong>in</strong>tersection. He achieves that by<br />
assum<strong>in</strong>g that the doma<strong>in</strong>s of the conjo<strong>in</strong>able types are semi-lattices and not Boolean<br />
algebras. Thus, and is assigned to whichever operator is def<strong>in</strong>ed for that doma<strong>in</strong>.<br />
Next, he argues that or is not and’s dual. If they were truly duals, or should denote set<br />
<strong>in</strong>tersection just like and denotes set union. However, this is not the case s<strong>in</strong>ce the simple<br />
phrase Grant or Abigail cannot denote the <strong>in</strong>tersection of {grant} and {abigail}, which<br />
would be the empty set. Also, if and and or were duals, it would be expected to behave<br />
similarly with regard to scope. However, or resembles <strong>in</strong>def<strong>in</strong>ites s<strong>in</strong>ce it can scope above<br />
a structurally higher and, while and cannot scope above a structurally higher or. This is<br />
shown below:<br />
(62) a. Grant and Jakob danced with Rebecca or Sukie<br />
b. Grant or Jakob danced with Rebecca and Sukie<br />
(Eggert, 2002, 103)<br />
F<strong>in</strong>ally, one more argument that and and or are not duals is that there are cases they can<br />
have identical mean<strong>in</strong>gs. For example, (63) paraphrases as (64) and also as (65) accord<strong>in</strong>g<br />
to Jenn<strong>in</strong>gs (1994):<br />
(63) Grant is taller than Gertrude or Abigail<br />
(64) Grant is taller than Gertrude or Grant is taller than Abigail<br />
(65) Grant is taller than Gertrude, and Grant is taller than Abigail<br />
(Jenn<strong>in</strong>gs, 1994)
6.3 Approaches to Disjunctive Coord<strong>in</strong>ated NPs 173<br />
Initially, he adopts the DRT analysis of existentials for disjunction. Existentials trigger<br />
the <strong>in</strong>troduction of a discourse referent that is a member of its restrict<strong>in</strong>g predicate. Disjunction<br />
may also be taken to <strong>in</strong>troduce a discourse referent which is a member of the set<br />
of its disjuncts 16 . For example, we could assign the follow<strong>in</strong>g DRS for the phrase Gertrude<br />
or Abigail:<br />
(66)<br />
x,y,z<br />
gertrude(x)<br />
abigail(y)<br />
z∈{x,y}<br />
(Eggert, 2002, 106)<br />
However, Simons (1998) argues that disjunction does not <strong>in</strong>troduce any discourse referents.<br />
Consider the discouse <strong>in</strong> the example below:<br />
(67) a. Gertrude or Abigail is s<strong>in</strong>g<strong>in</strong>g tonight<br />
b. #She might dance, too<br />
The <strong>in</strong>tended read<strong>in</strong>g for (67a) is ‘whoever is s<strong>in</strong>g<strong>in</strong>g tonight (i.e. Gertrude or Abigail)<br />
might dance too’; however, this read<strong>in</strong>g does not become available.<br />
S<strong>in</strong>ce disjunction does not seem to <strong>in</strong>troduce a discourse referent (Simons, 1998), Eggert<br />
(2002) argues that disjunction must be treated as a comb<strong>in</strong>ation of the default operator,<br />
which means whatever operator is def<strong>in</strong>ed for the doma<strong>in</strong>, and a subset function, which<br />
means a function that moves from sets to subsets. By treat<strong>in</strong>g disjunction as a subset<br />
function, he avoids <strong>in</strong>troduc<strong>in</strong>g a discourse referent for disjunction. Also, he borrows from<br />
Schwarzschild (1996) the partition theory of plurals and applies that to disjunction. More<br />
precisely, Schwarzschild (1996) argues that accord<strong>in</strong>g to the partition theory of plurals<br />
the collective and distributive semantic difference <strong>in</strong> sentences comes from a contextually<br />
determ<strong>in</strong>ed variable, a partition on the universe of discourse. Schwarzschild (1996) suggests<br />
that the semantics is the same for both read<strong>in</strong>gs and <strong>in</strong>troduces a semantic theory<br />
that treats sentences as vague and not as ambiguous, which should allow for a variety of<br />
<strong>in</strong>terpretations. Therefore, all cases of semantic ambiguity are vague and are expla<strong>in</strong>ed by<br />
a variety of <strong>in</strong>terpretations. Thus, he captures problematic data and gets the distributive<br />
and collective read<strong>in</strong>gs correctly. Consider the follow<strong>in</strong>g sentence:<br />
(68) Three students carried the piano<br />
(Schwarzschild (1996) cited <strong>in</strong> Eggert (2002, 127))<br />
In some contexts, the variable partitions ‘three students’ <strong>in</strong>to three cells, one for each<br />
student, which is the distributive read<strong>in</strong>g, but <strong>in</strong> other contexts, there is only one cell<br />
<strong>in</strong>clud<strong>in</strong>g all three students, which is the collective read<strong>in</strong>g. This theory also favours the<br />
use of distributive operators and not type-rais<strong>in</strong>g ones. Thus, pragmatics will contribute<br />
<strong>in</strong>to limit<strong>in</strong>g the number of <strong>in</strong>terpretations of a sentence or even disambiguate it.<br />
Eggert (2002) after propos<strong>in</strong>g a theory of semantic number that posits a discourse level,<br />
<strong>in</strong> addition to the modular structures, he also <strong>in</strong>troduces a set of agreement constra<strong>in</strong>ts.<br />
Speakers use these constra<strong>in</strong>ts <strong>in</strong> their attempt to determ<strong>in</strong>e if a disjunctively conjo<strong>in</strong>ed<br />
noun phrase requires s<strong>in</strong>gular or plural verb agreement. He argues that agreement facts<br />
are best analysed with weighted constra<strong>in</strong>ts and not with rules or rank<strong>in</strong>g constra<strong>in</strong>ts,<br />
due to the nature of agreement data. In particular, native speakers have weak <strong>in</strong>tuitions<br />
with regard to agreement phenomena. He claims that there are cases where they are<br />
16 In DRT, disjunction is taken to <strong>in</strong>troduce a discourse referent that is a member of a set of its disjuncts<br />
(z∈{x,y}) (Kamp and Reyle, 1993)
6.3 Approaches to Disjunctive Coord<strong>in</strong>ated NPs 174<br />
<strong>in</strong>decisive as to the correct agreement choice with<strong>in</strong> a pattern. Other times they are able<br />
to state whether one pattern is preferable to another, even though they disagree about<br />
which pattern is preferable. Moreover, sometimes the role of hierarchies is crucial <strong>in</strong> order<br />
to determ<strong>in</strong>e agreement, with animacy hierarchy be<strong>in</strong>g the most important.<br />
Eggert (2002) considers that weighted constra<strong>in</strong>ts are the most effective method to<br />
account for <strong>in</strong>termediate judgements, speaker variation and the role of hierarchies <strong>in</strong> agreement<br />
facts s<strong>in</strong>ce they may <strong>in</strong>teract <strong>in</strong> different ways with different speakers and the preference<br />
of the <strong>in</strong>dividual speaker depends on the relative weights of <strong>in</strong>dividual constra<strong>in</strong>ts.<br />
Such constra<strong>in</strong>ts also <strong>in</strong>teract better with the animacy hierarchy. Rules, on the other hand,<br />
are not able to account for such variation s<strong>in</strong>ce they are restrictive <strong>in</strong> allow<strong>in</strong>g either a grammatical<br />
or ungrammatical structure. Similar problems occur with ranked constra<strong>in</strong>ts s<strong>in</strong>ce<br />
there are only three possible relations that a constra<strong>in</strong>t C1 may have with a constra<strong>in</strong>t C2<br />
when rank<strong>in</strong>g is <strong>in</strong>volved; either C1 out-ranks C2, C1 may be out-ranked by C2 or C1 may<br />
have the same rank as C2.<br />
Eggert (2002) <strong>in</strong>troduces two types of constra<strong>in</strong>ts, syntactic and semantic. The syntactic<br />
constra<strong>in</strong>ts are formnum, highnum, nearnum (Peterson’s prox) and firstnum<br />
(Peterson’s fc w<strong>in</strong>s). These are stated as follows:<br />
(69) formnum: A verb should agree <strong>in</strong> grammatical number with its syntactic subject<br />
highnum: A verb should be plural if one of the NP heads of its syntactic subject<br />
is plural<br />
nearnum: A verb should agree <strong>in</strong> grammatical number with the nearest NP head<br />
of its syntactic subject<br />
firstnum: A verb should agree <strong>in</strong> grammatical number with the first NP head of<br />
its syntactic subject<br />
(Eggert, 2002, 197)<br />
The semantic constra<strong>in</strong>ts are drnum, modelnum, disnum and multijunct while the<br />
def<strong>in</strong>ition of semnum is divided <strong>in</strong>to two parts. These are def<strong>in</strong>ed below 17 :<br />
(70) semnum1: A pred1 should be marked for semantic number<br />
semnum2: If the semantic number of a pred1 is marked it should have the same<br />
number as its semantic subject<br />
drnum: The sum of any two discourse referents should be a plural discourse referent<br />
modelnum: Where x is a discourse referent and [x] is the assignment of x <strong>in</strong> model<br />
m,|x| should be equal to |[x]|<br />
disnum: If a pred1 has a disjunctive arg for subject, it should be unspecified for<br />
number<br />
multijunct: A verb should be marked plural if its syntactic subject consists of<br />
more than two np heads<br />
(Eggert, 2002, 243)<br />
F<strong>in</strong>ally, he <strong>in</strong>troduces a constra<strong>in</strong>t which is a semantic-syntactic one:<br />
(71) semnum=formnum: If a verb is marked for number <strong>in</strong> Syntax, the correspond<strong>in</strong>g<br />
pred1 should be marked for the same number <strong>in</strong> A/P<br />
(Eggert, 2002, 243)<br />
These constra<strong>in</strong>ts capture complicated data, such as disjunctive conjo<strong>in</strong>ed nouns, which<br />
are difficult to analyse due to great variation of speakers’ judgements and <strong>in</strong>tuitions. Eggert<br />
17 In the A/P module, A/P structures show how one forms arguments and predicates and comb<strong>in</strong>es them<br />
to form propositions. Pred1 is def<strong>in</strong>ed as follows: A/P1 PRED1 → β β ∈ {run’, girl’, pretty’,...}.<br />
Therefore, PRED1 refers to any s<strong>in</strong>gle argument predicate.
6.3 Approaches to Disjunctive Coord<strong>in</strong>ated NPs 175<br />
(2002) notes that many factors determ<strong>in</strong>e the speaker’s choice from syntactic, semantic to<br />
discourse or pragmatic while most of the speakers assign weight to the constra<strong>in</strong>ts and<br />
these weights may differ considerably from speaker to speaker. Thus, some speakers use<br />
more semantically based strategies while others may use more syntactically based ones.<br />
There are also those who are between the two.<br />
On the whole, Eggert (2002) proposes an <strong>in</strong>termodular theory of disjunction that analyses<br />
disjunction from the perspective of syntax, semantics and discourse. First, he discusses<br />
the grammatical structure of disjunction <strong>in</strong> the A/P theory, the Syntax and the L<strong>in</strong>earity<br />
modules. Next, he presents a theory of semantic number, which also posits a discourse<br />
level with<strong>in</strong> DRT. F<strong>in</strong>ally, he <strong>in</strong>troduces a set of agreement constra<strong>in</strong>ts which account for<br />
the fact that speakers use these constra<strong>in</strong>ts to determ<strong>in</strong>e whether a sentence should have<br />
s<strong>in</strong>gular or plural verb agreement. Thus, his model views disjunction as a complicated<br />
phenomenon that requires the active participation of different modules <strong>in</strong> order to provide<br />
a complete account for disjunction.<br />
6.3.4 Discuss<strong>in</strong>g the three approaches<br />
In the previous sections, we decribed the three approaches to disjunction developed by<br />
Morgan (1972, 1985), Peterson (1986) and Eggert (2002) and briefly presented the ma<strong>in</strong><br />
advantages and disadvantages of these approaches. From the three approaches to disjunctive<br />
conjo<strong>in</strong>ed nouns, we believe that Eggert’s analysis is more advantageous than the other<br />
two for the simple reason that he acknowledges that discourse conditions play a crucial role<br />
<strong>in</strong> disjunction and he proposes an analysis on discourse.<br />
First of all, although Morgan (1972, 1985) discussed agreement phenomena determ<strong>in</strong>ed<br />
by discourse conditions, such as “deferred referrence”, he did not deal extensively with<br />
those but he chooses to <strong>in</strong>corporate the discourse phenomena to semantics. Therefore,<br />
cases where the <strong>in</strong>terpretation of a sentence and the choice of the verb form depend on<br />
discourse factors are attributed to semantics (and/or pragmatics) accord<strong>in</strong>g to Morgan<br />
(1972, 1985). Some of the cases that can have either s<strong>in</strong>gular or plural verb agreement<br />
depend<strong>in</strong>g on the <strong>in</strong>terpretation are the follow<strong>in</strong>g:<br />
(72) John or Bill is/are??? go<strong>in</strong>g to w<strong>in</strong> the race<br />
(73) Do you th<strong>in</strong>k John or Bill is/are go<strong>in</strong>g to come?<br />
(74) I don’t th<strong>in</strong>k John or Bill are/*is go<strong>in</strong>g to w<strong>in</strong> the race<br />
(75) Either John or Bill is/*are go<strong>in</strong>g to w<strong>in</strong> the race<br />
(Morgan, 1985, 72)<br />
Thus, purely pragmatic factors are ignored <strong>in</strong> this analysis.<br />
Similarly, Peterson (1986) avoided discuss<strong>in</strong>g cases where discourse conditions are the<br />
ma<strong>in</strong> motivation for number verb agreement. In fact, his ma<strong>in</strong> argument is that the type<br />
of sentence determ<strong>in</strong>es the choice of the verb form and not the <strong>in</strong>clusive/exclusive <strong>in</strong>terpretation<br />
of the sentence. Also, he argues that speaker strategies play a crucial role <strong>in</strong> verb<br />
agreement and he identifies three ma<strong>in</strong> strategies <strong>in</strong> English, prox, plu w<strong>in</strong>s and fc w<strong>in</strong>s.<br />
Although speaker strategies do seem to play an important role <strong>in</strong> verb agreement, there are<br />
cases that speakers are <strong>in</strong>decisive concern<strong>in</strong>g the choice of the verb form which implies that<br />
more than one strategy can be possible or even none of the strategies can clearly determ<strong>in</strong>e<br />
the agreement form. In these cases, speakers seem to consult the context and the discourse<br />
conditions, which are able to determ<strong>in</strong>e the choice of the verb form. Unfortunately, this<br />
factor is ignored by Peterson (1986). Also, the strategies <strong>in</strong>troduced by Peterson (1986)<br />
are not crossl<strong>in</strong>guistic. In fact, we will show that <strong>in</strong> MG only two strategies are activated<br />
the prox and plu w<strong>in</strong>s. Thus, the flexibility of the language itself will determ<strong>in</strong>e the<br />
strategies that will be used.
6.4 Verb <strong>Agreement</strong> <strong>in</strong> MG Disjunctive Coord<strong>in</strong>ate <strong>Noun</strong>s 176<br />
Eggert (2002), on the other hand, like Jenn<strong>in</strong>gs (1994), proposes a complete account<br />
based on a discourse based framework DRT. Eggert (2002) draws a dist<strong>in</strong>ction between<br />
purely syntactic strategies, purely semantic and a comb<strong>in</strong>ation of the two. He also takes<br />
<strong>in</strong>to consideration discourse factors that are <strong>in</strong>volved <strong>in</strong> verb agreement with disjo<strong>in</strong>t nouns,<br />
someth<strong>in</strong>g that Morgan (1972, 1985) and Peterson (1986) do not manage.<br />
Another characteristic of Eggert’s analysis is that or corresponds to a subset function,<br />
which picks up either one (i.e. result<strong>in</strong>g <strong>in</strong> sg agreement) or both (i.e. result<strong>in</strong>g <strong>in</strong> pl<br />
agreement) members of the set, be<strong>in</strong>g polysemous between two mean<strong>in</strong>gs. In fact, he<br />
does not analyse or as be<strong>in</strong>g and’s dual while he argues that there are sentences with<br />
disjunction or that do not translate as propositional disjunction, but rather as propositional<br />
conjunction. In particular, he refers to examples that have been discussed by Jenn<strong>in</strong>gs<br />
(1994), as the one below:<br />
(76) Grant is taller than Gertrude or Abigail (Eggert, 2002, 104)<br />
He claims that Boolean and clausal theories of and and or would predict only the read<strong>in</strong>g<br />
<strong>in</strong> (77) as be<strong>in</strong>g the only available one:<br />
(77) Grant is taller than Gertrude or Grant is taller than Abigail<br />
But (77) also paraphrases as <strong>in</strong> (78):<br />
(78) Grant is taller than Gertrude and Grant is taller than Abigail<br />
Thus, Eggert (2002) discussed the long stand<strong>in</strong>g issues of the different functions of or<br />
as <strong>in</strong>clusive (∨) and exclusive (∧) and also of the different <strong>in</strong>terpretations of disjunction<br />
or as ‘and-coord<strong>in</strong>ate’ and as ‘exclusive’. In fact, he places emphasis and argues <strong>in</strong> favour<br />
of the different <strong>in</strong>terpretations of or <strong>in</strong> discourse. This is taken <strong>in</strong>to consideration when<br />
posit<strong>in</strong>g the syntactic and semantic constra<strong>in</strong>ts for the account of disjunctive coord<strong>in</strong>ated<br />
nouns. Next, we turn to verb agreement <strong>in</strong> MG disjunctively conjo<strong>in</strong>ed nouns.<br />
6.4 Verb <strong>Agreement</strong> <strong>in</strong> MG Disjunctive Coord<strong>in</strong>ate <strong>Noun</strong>s<br />
In this section, we exam<strong>in</strong>e number verb agreement <strong>in</strong> disjunctive coord<strong>in</strong>ate nouns <strong>in</strong> MG<br />
and we will show that a number of factors determ<strong>in</strong>e the number of the verb. An important<br />
factor that determ<strong>in</strong>es number verb agreement is the <strong>in</strong>terpretation of disjunction. MG<br />
native speakers assign two different <strong>in</strong>terpretations to natural language disjunction or, as<br />
an ‘and-coord<strong>in</strong>ate’ so that the predicate holds of both disjuncts (this is usually the same as<br />
the <strong>in</strong>clusive <strong>in</strong>terpretation of logical disjunction ∨), and as a true ‘exclusive’ disjunction,<br />
<strong>in</strong> which the predicate is true of either one or the other of the disjuncts. Accord<strong>in</strong>g to the<br />
<strong>in</strong>terpretation, native speakers show a preference to a specific verb form, such as plural <strong>in</strong><br />
the first case and s<strong>in</strong>gular <strong>in</strong> the second.<br />
Second, verb agreement is also determ<strong>in</strong>ed by a number of syntactic factors; for example,<br />
different gender and person features found <strong>in</strong> each disjunct, the type of sentence,<br />
which may be either declarative or <strong>in</strong>terrogative, and the presence of a predicative adjective/participle,<br />
which determ<strong>in</strong>e the choice of a specific verb agreement form. Third, a<br />
number of different strategies adopted by MG native speakers are also associated to verb<br />
agreement. These strategies differ from other languages and have different weight <strong>in</strong> each<br />
language. As we will show, MG adopts two basic strategies as opposed to other languages<br />
whose native speakers use more than two.<br />
F<strong>in</strong>ally, we will briefly refer to coord<strong>in</strong>ate nouns jo<strong>in</strong>ed by the positive predisjunction<br />
and disjunction i...i/ite...ite ‘either...or’ and the negative predisjunction and disjunction<br />
oute...oute ‘neither...nor’. The positive predisjunction and disjunction i...i ‘either...or’ behaves<br />
almost the same as bare disjunction i ‘or’. However, the negative predisjunction and
6.4 Verb <strong>Agreement</strong> <strong>in</strong> MG Disjunctive Coord<strong>in</strong>ate <strong>Noun</strong>s 177<br />
disjunction oute...oute ‘neither...nor’ shows great preference to the plural verb agreement<br />
form.<br />
The data analysis was based on two questionnaires that were developed <strong>in</strong> order to<br />
test the prevalent verb agreement form <strong>in</strong> disjunctive structures. The aim was to explore<br />
the different <strong>in</strong>terpretations assigned to the coord<strong>in</strong>ate structures and the syntactic factors<br />
that determ<strong>in</strong>e verb agreement. The first questionnaire focuses on the <strong>in</strong>terpretation of the<br />
disjunctive coord<strong>in</strong>ate nouns. The second questionnaire focuses on the syntactic factors and<br />
it consists of both declarative and <strong>in</strong>terrogative sentences. The questionnaire was issued to<br />
20 native speakers who are all university graduates from different areas <strong>in</strong> Greece. Their<br />
choice depends on whether they found the phrase acceptable and they would use it <strong>in</strong><br />
everyday language or not 18 .<br />
6.4.1 Interpretation and Verb <strong>Agreement</strong> <strong>in</strong> Disjunctive Coord<strong>in</strong>ate <strong>Noun</strong>s<br />
In this section, we explore the different <strong>in</strong>terpretations assigned to disjunctive coord<strong>in</strong>ate<br />
nouns. The first group of data <strong>in</strong>volves MG disjunctive phrases with s<strong>in</strong>gular disjuncts. In<br />
MG, number verb agreement with a disjunctive phrase largerly depends on the <strong>in</strong>terpretation<br />
of disjunction as an ‘and-coord<strong>in</strong>ate’ or and as an ‘exclusive’ or. Our aim is to show<br />
that MG native speakers allow s<strong>in</strong>g verb agreement when they assign to the disjunctive<br />
phrase an ‘exclusive’ sense, and a pl verb agreement, when they assign to the phrase an<br />
‘and-coord<strong>in</strong>ate’ sense.<br />
In the follow<strong>in</strong>g example, the coord<strong>in</strong>ate phrase consists of two s<strong>in</strong>gular disjuncts, the<br />
proper noun Kostas and the proper noun Maria. The majority of the participants used a<br />
s<strong>in</strong>gular verb agreement (14/20), only 4 participants used a pl verb form while 2 participants<br />
selected both a s<strong>in</strong>gular and a plural form:<br />
(79) O Kostas i i Maria<br />
the.sg Kostas.sg or the.sg Maria.sg<br />
me to aftok<strong>in</strong>ito.<br />
with the car<br />
‘Kostas or Maria will pick me up with the car’<br />
(constructed)<br />
tha<br />
will<br />
me<br />
me<br />
pari<br />
pick-up.sg<br />
The most plausible <strong>in</strong>terpretation for the largest number of native speakers is an ‘exclusive’<br />
one. In example (79), the ‘exclusive’ <strong>in</strong>terpretation is confirmed by the fact that the<br />
sentence can accept the additional phrase as one of them is work<strong>in</strong>g tonight and it can<br />
only be <strong>in</strong>terpreted with an ‘exclusive’ sense:<br />
(80) O Kostas i i Maria tha me pari<br />
the.sg Kostas.sg or the.sg Maria.sg will me pick-up.sg<br />
me to aftok<strong>in</strong>ito, jati enas apo tus dio dulevei<br />
with<br />
apopse.<br />
tonight<br />
the car, as one of the both is-work<strong>in</strong>g<br />
‘Kostas or Maria will pick me up with the car, s<strong>in</strong>ce one of them is work<strong>in</strong>g tonight’<br />
(constructed)<br />
Therefore, example (79) is <strong>in</strong>terpreted by the majority of consultants as an action carried<br />
out by one of the two <strong>in</strong>dividuals and this is why they prefer a s<strong>in</strong>gular verb. Only 4 (out of<br />
20) consultants <strong>in</strong>terpret example (79) as if both of them ‘will pick up the speaker’, which<br />
is why we get a smaller number of consultants that prefer the plural verb.<br />
18 For more details on the methodology and rationale on construct<strong>in</strong>g the questionnaires on disjunctively<br />
conjo<strong>in</strong>ed nouns see Appendix B.
6.4 Verb <strong>Agreement</strong> <strong>in</strong> MG Disjunctive Coord<strong>in</strong>ate <strong>Noun</strong>s 178<br />
Similarly, <strong>in</strong> the follow<strong>in</strong>g example 15 (out of 20) consultants chose a s<strong>in</strong>gular verb.<br />
Therefore, the example is likely to be <strong>in</strong>terpreted with an ‘exclusively’ sense:<br />
(81) To agori i to koritsi prospathi<br />
the.sg boy.sg or the.sg girl.sg tries.sg<br />
s<strong>in</strong>estimata<br />
feel<strong>in</strong>gs<br />
‘The boy or the girl tries to suppress the feel<strong>in</strong>gs’<br />
(constructed)<br />
na<br />
to<br />
katastili<br />
suppress<br />
The exclusive <strong>in</strong>terpretation is confirmed by the presence of modifiers, such as separately,<br />
<strong>in</strong>dividually, only, that assign an ‘exclusive’ sense <strong>in</strong> the sentence. Thus, example<br />
(81) is rewritten as follows:<br />
(82) To agori i to koritsi prospathi<br />
the.sg boy.sg or the.sg girl.sg tries.sg<br />
na katastili ta s<strong>in</strong>estimata<br />
to suppress the feel<strong>in</strong>gs<br />
‘The boy or the girl alone tries to suppress the feel<strong>in</strong>gs’<br />
(constructed)<br />
mono<br />
alone/only<br />
Only 3 (out of 20) consultants chose a plural verb agreement <strong>in</strong> example (81), which<br />
is the result of an ‘and-coord<strong>in</strong>ate’ <strong>in</strong>terpretation of or. The same example with a plural<br />
verb agreement is shown below:<br />
(83) To agori i to koritsi prospathun<br />
the.sg boy.sg or the.sg girl.sg try.pl<br />
ta s<strong>in</strong>estimata<br />
the feel<strong>in</strong>gs<br />
‘The boy or the girl try to suppress the feel<strong>in</strong>gs’<br />
(constructed)<br />
na<br />
to<br />
tu<br />
its<br />
katastilun<br />
suppress<br />
Thus, the ‘and-coord<strong>in</strong>ate’ <strong>in</strong>terpretation is also acceptable and it is a natural choice<br />
for a few (3/20) native speakers but to a smaller degree. F<strong>in</strong>ally, we need to note that only<br />
2 consultants chose both the plural and the s<strong>in</strong>gular verb forms which shows that for these<br />
consultants both <strong>in</strong>terpretations are acceptable.<br />
Alternatively, <strong>in</strong> the follow<strong>in</strong>g example where the coord<strong>in</strong>ate phrase also consists of<br />
two s<strong>in</strong>gular disjuncts, the largest number of participants were <strong>in</strong>cl<strong>in</strong>ed to use the plural<br />
verb agreement form (13/20) while the s<strong>in</strong>gular verb form was chosen by fewer (7/20)<br />
consultants:<br />
(84) I j<strong>in</strong>eka i to pedi exun protereotita<br />
the woman.sg or the child.sg have.pl priority<br />
to emvolio kata tis gripis<br />
the vacc<strong>in</strong>e aga<strong>in</strong>st the flu<br />
‘The woman or child have priority for the vacc<strong>in</strong>e aga<strong>in</strong>st the flu’<br />
(G.N.: To Vima)<br />
The above example is most likely assigned an ‘and-coord<strong>in</strong>ate’ <strong>in</strong>terpretation consider<strong>in</strong>g<br />
that both groups are high-risk groups and require the vacc<strong>in</strong>e. This is confirmed by the<br />
fact that modifiers, which assign distributivity, are not used by native speakers <strong>in</strong> these<br />
cases and the sentence is paraphrased with the conjunction and, as seen below:<br />
ja<br />
for<br />
ta<br />
the
6.4 Verb <strong>Agreement</strong> <strong>in</strong> MG Disjunctive Coord<strong>in</strong>ate <strong>Noun</strong>s 179<br />
(85) I j<strong>in</strong>eka ke to pedi exun protereotita<br />
the woman.sg and the child.sg have.pl priority<br />
to emvolio kata tis gripis<br />
the vacc<strong>in</strong>e aga<strong>in</strong>st the flu<br />
‘The woman and child have priority for the vacc<strong>in</strong>e aga<strong>in</strong>st the flu’<br />
(G.N.: O Ependitis)<br />
Similarly, the example (86) is more likely <strong>in</strong>terpreted as an ‘and-coord<strong>in</strong>ate’ phrase. In<br />
particular, 15 (out of 20) consultants chose a plural verb agreement and only 5 (out of 20)<br />
consultants chose a s<strong>in</strong>gular verb agreement. However, none of the consultants chose both<br />
verb agreement forms:<br />
(86) To agori i to koritsi<br />
the.sg<br />
diavasma<br />
study<strong>in</strong>g<br />
boy.sg or the.sg girl.sg<br />
‘The boy or the girl are tired from study<strong>in</strong>g’<br />
(www.<strong>in</strong>.gr)<br />
The above example could be paraphrased as follows:<br />
(87) To agori ke to koritsi<br />
the.sg boy.sg and the.sg girl.sg<br />
to diavasma<br />
the study<strong>in</strong>g<br />
‘The boy and the girl are tired from study<strong>in</strong>g’<br />
(constructed)<br />
exun<br />
have.pl<br />
exun<br />
have.pl<br />
kurasti<br />
tired.pl<br />
kurasti<br />
tired.pl<br />
apo<br />
from<br />
ja<br />
for<br />
apo<br />
from<br />
In fact, example (86) is seen by most consultants as if the event has an effect on both<br />
‘the boy and the girl’ s<strong>in</strong>ce both are perceived to be study<strong>in</strong>g despite the fact that they<br />
may have not participated together <strong>in</strong> the same event. However, we need to note that<br />
only example (87), which can be a paraphrase of example (86), would accept modifiers<br />
which assign collectivity, such as both of them, whereas example (86) does not accept such<br />
modifiers. In fact, <strong>in</strong> example (88), if we add a modifier with a collective mean<strong>in</strong>g, the<br />
sentence is marked if not completely ungrammatical, but not <strong>in</strong> example (89):<br />
(88) ??To agori i to koritsi exun ke<br />
the.sg boy.sg or the.sg girl.sg have.pl and<br />
kurasti apo to diavasma<br />
tired.pl from the study<strong>in</strong>g<br />
‘The boy or the girl are both of them tired from study<strong>in</strong>g’<br />
(constructed)<br />
(89) To agori ke to koritsi exun ke<br />
the.sg boy.sg and the.sg girl.sg have.pl and<br />
kurasti apo to diavasma<br />
tired.pl from the study<strong>in</strong>g<br />
‘The boy and the girl are both of them tired from study<strong>in</strong>g’<br />
(constructed)<br />
ta<br />
the<br />
ta<br />
the<br />
dio<br />
both<br />
dio<br />
both<br />
The fact that some consultants chose the s<strong>in</strong>gular verb form, even though they were<br />
fewer (5/20), shows that an ‘and-coord<strong>in</strong>ate’ <strong>in</strong>terpretation is not the only possibility for<br />
the sentence. Therefore, example (86) was perceived with an ‘exclusive’ sense by a few<br />
to<br />
the
6.4 Verb <strong>Agreement</strong> <strong>in</strong> MG Disjunctive Coord<strong>in</strong>ate <strong>Noun</strong>s 180<br />
speakers. Thus, <strong>in</strong> the two examples above native speakers assigned ma<strong>in</strong>ly an ‘andcoord<strong>in</strong>ate’<br />
<strong>in</strong>terpretation followed by a plural verb with very few cases of an ‘exclusive’<br />
<strong>in</strong>terpretation followed by a s<strong>in</strong>gular verb. The responses of the consultants of the ma<strong>in</strong><br />
examples discussed are shown <strong>in</strong> the follow<strong>in</strong>g table:<br />
Table 6.3: S<strong>in</strong>gular <strong>Noun</strong>s<br />
Consultant Number: 20<br />
Examples Verb <strong>Agreement</strong> Form/Interpretation of or<br />
s<strong>in</strong>gular/‘exclusive’ plural/‘<strong>in</strong>clusive’ s<strong>in</strong>g-pl/both<br />
(79) 14 4 2<br />
(81) 15 3 2<br />
(84) 7 13 -<br />
(86) 5 15 -<br />
The table above shows that participants have a preference to a specific verb form usually<br />
related to the <strong>in</strong>terpretation they assign to the correspond<strong>in</strong>g phrase. However, this is not<br />
the only possibility but <strong>in</strong> all examples a second <strong>in</strong>terpretation is possible, which proves<br />
that the ‘exclusive/and-coord<strong>in</strong>ate’ dist<strong>in</strong>ction is not clear-cut but depends on the speaker’s<br />
perception of the specific phrase <strong>in</strong> the specific discourse conditions.<br />
There are some examples where the ‘exclusive’ and ‘and-coord<strong>in</strong>ate’ <strong>in</strong>terpretations<br />
of or do not overlap but <strong>in</strong>stead there is only one sense assigned to the phrase which is<br />
consistent across native speakers. Jenn<strong>in</strong>gs (1966) holds that a truth-functional exclusive<br />
example <strong>in</strong> English is Mary is related to either Jack or Bob (181). In the correspond<strong>in</strong>g<br />
<strong>Greek</strong> example translated from English, 20 (out of 20) consultants chose the s<strong>in</strong>gular verb<br />
agreement form, based on an ‘exclusive’ only <strong>in</strong>terpretation:<br />
(90) O Kostas i o<br />
the.sg Kostas.sg or the.sg<br />
tis Marias<br />
of-the Maria<br />
‘Kostas or Giorgos is Maria’s relative’<br />
(translation from Jenn<strong>in</strong>gs (1994))<br />
Jorgos<br />
Giorgos.sg<br />
<strong>in</strong>e<br />
is.sg<br />
s<strong>in</strong>genis<br />
relative.sg<br />
The most plausible mean<strong>in</strong>g assigned to the above sentence is, what Jenn<strong>in</strong>gs (1966) also<br />
presents for the correspond<strong>in</strong>g English example, that only one of the two must be Maria’s<br />
relative, otherwise if the sentence is <strong>in</strong>terpreted conjunctively, Maria will be related to<br />
both ‘Kostas’ and ‘Jorgos’ 19 . Logicians who support an exclusive or would claim that the<br />
possibility here is that either the first or the second disjunct is true but def<strong>in</strong>itely not both<br />
for the disjunction to be true. If both disjuncts are true then the whole disjunction is false.<br />
The same phrase, however, can be <strong>in</strong>terpreted purely pragmatically s<strong>in</strong>ce this is what one<br />
would say if they wanted to specify that one of the two people is Maria’s cous<strong>in</strong>. Thus, <strong>in</strong><br />
MG it conveys an ‘exclusive-or’ sense.<br />
Alternatively, there are examples where the ‘and-coord<strong>in</strong>ate’ <strong>in</strong>terpretation is only assigned<br />
by MG native speakers. In the correspond<strong>in</strong>g example <strong>in</strong> MG translated from<br />
English, all native speakers (20/20) used a plural verb agreement as a result of an ‘andcoord<strong>in</strong>ate’<br />
<strong>in</strong>terpretation of or assigned to the phrase:<br />
(91) O jatros<br />
the.sg<br />
farmaka.<br />
doctor.sg<br />
prescriptions<br />
i<br />
or<br />
o<br />
the.sg<br />
odontiatros<br />
dentist.sg<br />
mporun<br />
can.pl<br />
na<br />
to<br />
grapsun<br />
write<br />
19 Of course the read<strong>in</strong>g where ‘Maria’ is related to both is possible but <strong>in</strong> this case the conjunction and<br />
would be used to coord<strong>in</strong>ate the two nouns and not the disjunction or.
6.4 Verb <strong>Agreement</strong> <strong>in</strong> MG Disjunctive Coord<strong>in</strong>ate <strong>Noun</strong>s 181<br />
‘The doctor or dentist can write prescriptions’<br />
(translation from de Swart (1998))<br />
Thus, the ‘and-coord<strong>in</strong>ate’ <strong>in</strong>terpretation is the only possibility here. de Swart (1998,<br />
57) argues that <strong>in</strong> English the disjunctive coord<strong>in</strong>ate phrase above does not correspond<br />
to a disjunction of propositions <strong>in</strong> truth-conditional semantics and we should be careful<br />
when we relate the logical connective to natural language or. Even if this is the case, we<br />
cannot deny the fact that native speakers use a plural verb agreement s<strong>in</strong>ce they <strong>in</strong>terpret<br />
the sentence as ‘and-coord<strong>in</strong>ate’ consider<strong>in</strong>g the fact that both a doctor and a dentist can<br />
write prescriptions 20 .<br />
The above data suggest that there are two <strong>in</strong>terpretations for natural language or <strong>in</strong><br />
MG, the ‘and-coord<strong>in</strong>ate’ and the ‘exclusive’ one. In the majority of examples, there is<br />
not a sharp dist<strong>in</strong>ction between the two but the context and the specific conditions seem<br />
to determ<strong>in</strong>e the k<strong>in</strong>d of mean<strong>in</strong>g that will be assigned to the disjunctive phrase by the<br />
speakers. There are cases, however, that either an ‘and-coord<strong>in</strong>ate’ or an ‘exclusive’ <strong>in</strong>terpretation<br />
occurs exclusively. Some of these cases can be expla<strong>in</strong>ed us<strong>in</strong>g truth-conditional<br />
semantics s<strong>in</strong>ce the logical operators ∨ and the <strong>in</strong>terpretation for natural language or may<br />
co<strong>in</strong>cide. However, we cannot always rely on truth-conditional semantic s<strong>in</strong>ce the context<br />
is an important factor to disambiguate the mean<strong>in</strong>g of a phrase.<br />
6.4.2 Syntax and Verb <strong>Agreement</strong> <strong>in</strong> Disjunctive Coord<strong>in</strong>ate <strong>Noun</strong>s<br />
Next, it is <strong>in</strong>terest<strong>in</strong>g to exam<strong>in</strong>e whether any syntactic factors determ<strong>in</strong>e verb agreement<br />
<strong>in</strong> disjunctive coord<strong>in</strong>ate noun phrases. Parker (1983), who studies syntactic number <strong>in</strong><br />
English positive disjunctive coord<strong>in</strong>ate nouns with ‘either-or’ and negative disjunctive coord<strong>in</strong>ate<br />
nouns with ‘neither-nor’ 21 , argues that there are relevant variables that determ<strong>in</strong>e<br />
whether verb agreement is s<strong>in</strong>gular or plural. In the case of disjunctive s<strong>in</strong>gular coord<strong>in</strong>ate<br />
nouns, the variables are the gender and the person of the disjuncts. Thus, <strong>in</strong> English<br />
<strong>in</strong> mixed gender or mixed person disjuncts participants favour a syntactically plural verb<br />
form. He argues that this preference to a verb form is attributed to knowledge of the<br />
language itself and not only to knowledge of the world on the part of the speakers. This<br />
means that for Parker (1983) pure syntax also determ<strong>in</strong>es verb agreement as opposed to<br />
other l<strong>in</strong>guists who also focus on speakers’s strategies or semantic factors (Morgan (1972,<br />
1985); Peterson (1986)). Therefore, mixed gender or mixed person disjuncts should be<br />
treated as syntactically plural. Our goal is to <strong>in</strong>vestigate these claims for MG s<strong>in</strong>ce there<br />
is no similar work <strong>in</strong> other languages other than English. We will show that MG coord<strong>in</strong>ate<br />
phrases with s<strong>in</strong>gular number but mixed gender or mixed person nouns show similar<br />
variation to English and favour plural verb agreement <strong>in</strong> declarative sentences and CCA<br />
<strong>in</strong> <strong>in</strong>terrogative sentences.<br />
We will discuss first declarative sentences with mixed gender s<strong>in</strong>gular disjuncts. In<br />
Chapter 4, we saw that the expected resolved gender for mixed gender animate conjuncts<br />
is masc and for mixed gender <strong>in</strong>animate conjuncts is neut, while only adjectives and participles<br />
(not verbs) show gender agreement. In disjunctive coord<strong>in</strong>ate phrases with mixed<br />
gender disjuncts, the same requirements hold but there is a specific restriction concern<strong>in</strong>g<br />
the number feature of the verb. There are two options available <strong>in</strong> declarative sentences<br />
with s<strong>in</strong>gular mixed gender disjuncts and they relate to the presence or absence of a pred-<br />
20 Among others, Jenn<strong>in</strong>gs (1994) and de Swart (1998) argue that modality favours the <strong>in</strong>terpretation of or<br />
as and. Also, accord<strong>in</strong>g to Jenn<strong>in</strong>gs (1994) or behaves like and when or occurs <strong>in</strong> expressions of preference,<br />
when it appears <strong>in</strong> the complement of a comparative adjective and when it is immediately subord<strong>in</strong>ated to<br />
if. 21He refers to disjunctive coord<strong>in</strong>ate nouns with ‘either-or’ as “correlative constructions” and negative<br />
disjunctive coord<strong>in</strong>ate nouns with ‘neither-nor’ as “negative correlative constructions”.
6.4 Verb <strong>Agreement</strong> <strong>in</strong> MG Disjunctive Coord<strong>in</strong>ate <strong>Noun</strong>s 182<br />
icative adjective/participle, which <strong>in</strong>flects for gender. If a predicative adjective/participle is<br />
not present, s<strong>in</strong>gular and plural verb agreement varies depend<strong>in</strong>g on the <strong>in</strong>terpretation. If<br />
a predicative adjective/participle is present <strong>in</strong> the verb phrase, the number of the agree<strong>in</strong>g<br />
predicate and predicative adjective/participle is preferably the plural and therefore resolution<br />
is obeyed for both the number and gender features. The follow<strong>in</strong>g examples illustrate<br />
these patterns:<br />
(92) O fititis<br />
i i fititria<br />
the male-univ-student.masc.sg or the female-univ-student.fem.sg<br />
pu <strong>in</strong>e siguri ja kathe epilogi tha<br />
that<br />
petixun<br />
are certa<strong>in</strong>.masc.pl about every choice will<br />
succeed.masc.pl<br />
‘The male-student or female-student that is certa<strong>in</strong> about every choice will succeed’<br />
(constructed)<br />
(93) O pateras i i mitera<br />
the father.masc.sg or the mother.fem.sg<br />
enimeri ja t<strong>in</strong> katastasi<br />
<strong>in</strong>formed.masc.pl about the situation<br />
‘The father or mother are <strong>in</strong>formed about the situation’<br />
(constructed)<br />
In examples (92) and (93), both coord<strong>in</strong>ate phrases consist of two disjuncts, a mascul<strong>in</strong>e<br />
s<strong>in</strong>gular and a fem<strong>in</strong><strong>in</strong>e s<strong>in</strong>gular. In both cases, most (12/20) consultants chose plural <strong>in</strong> the<br />
agree<strong>in</strong>g verb and predicative adjective <strong>in</strong> all 4 sentences that appeared <strong>in</strong> the questionnaire.<br />
A small number of participants (4/20) opted for the s<strong>in</strong>gular. The consultants who chose<br />
a s<strong>in</strong>gular verb chose the gender of the closest conjunct. Thus, they selected the fem<strong>in</strong><strong>in</strong>e<br />
gender for the predicative adjective/participle follow<strong>in</strong>g closest conjunct agreement, as <strong>in</strong><br />
(94). When conjuncts appeared <strong>in</strong> the reverse order, the consultants who chose s<strong>in</strong>gular<br />
verb agreement opted for the mascul<strong>in</strong>e gender due to CCA, as <strong>in</strong> (95). F<strong>in</strong>ally, there<br />
were 2 (out of 20) consultants who chose the plural <strong>in</strong> 3 of the 4 questions and 2 (out of<br />
20) consultants who chose the plural <strong>in</strong> 2 of the 4 questions. The two examples discussed<br />
appear below:<br />
(94) O fititis<br />
i i fititria<br />
the male-univ-student.masc.sg or the female-univ-student.fem.sg<br />
pu <strong>in</strong>e siguri ja kathe epilogi tha petuxi<br />
who is certa<strong>in</strong>.fem.sg about every choice will succeed.sg<br />
‘The male-student or female-student who is certa<strong>in</strong> about every choice will succeed’<br />
(constructed)<br />
(95) I fititria<br />
i o fititis<br />
the female-univ-student.fem.sg or the male-univ-student.masc.sg<br />
pu <strong>in</strong>e siguros ja t<strong>in</strong> kariera tu tha<br />
that<br />
petixi<br />
is certa<strong>in</strong>.masc.sg about the career his.masc.sg will<br />
succeed.sg<br />
‘The female-student or male-student who is certa<strong>in</strong> about their career will succeed’<br />
(constructed)<br />
On the other hand, <strong>in</strong> mixed gender nouns if a predicative adjective or participle is<br />
not present, then the number of the predicate can be either s<strong>in</strong>gular or plural depend<strong>in</strong>g<br />
on the <strong>in</strong>terpretation of the phrase. For <strong>in</strong>stance, the questionnaire <strong>in</strong>cluded the follow<strong>in</strong>g<br />
<strong>in</strong>e<br />
are
6.4 Verb <strong>Agreement</strong> <strong>in</strong> MG Disjunctive Coord<strong>in</strong>ate <strong>Noun</strong>s 183<br />
two examples with mixed gender nouns and without a predicative adjective/participle, (96)<br />
and (97). In both cases, 7 (out of 20) consultants chose s<strong>in</strong>gular verb agreement, 2 (out of<br />
20) consultants chose plural verb agreement and 11 (out of 20) chose both a s<strong>in</strong>gular and<br />
a plural verb agreement. The result is still natural with either forms s<strong>in</strong>ce verbs show only<br />
number and not gender agreement <strong>in</strong> MG.<br />
(96) O mathitis<br />
i i mathitria<br />
the male-student.masc.sg or the female-student.fem.sg<br />
mpor-i/-un na enimeroth-i/-un apo tus gonis<br />
can.sg/pl to be-<strong>in</strong>formed.sg/pl by the parents<br />
‘The male-student or female-student can be <strong>in</strong>formed by the parents’<br />
(constructed)<br />
(97) O ijios i i kori<br />
the<br />
viothia<br />
help<br />
son.masc.sg or the daughter.fem.sg<br />
‘The son or the daughter need help’<br />
(constructed)<br />
xriaz-ete/-onte<br />
need.sg/pl<br />
Thus, the presence of a predicative adjective/participle, which forces agreement with the<br />
coord<strong>in</strong>ate phrase, is crucial for the verb form <strong>in</strong> MG. The table below is a summary of<br />
the above data:<br />
Table 6.4: Disjunction i<br />
S<strong>in</strong>gular Mixed Gender <strong>Noun</strong>s<br />
Consultant Number: 20<br />
Declaratives Verb <strong>Agreement</strong> Form<br />
np.sg or np.sg + verb s<strong>in</strong>gular plural s<strong>in</strong>g-pl/both<br />
(96) 7 2 11<br />
(97) 7 2 11<br />
np.sg or np.sg + verb + pred.adj/part s<strong>in</strong>gular plural s<strong>in</strong>g-pl/both<br />
(92) 4 12 4<br />
(93) 4 12 4<br />
Interrogative sentences without a predicative adjective/participle with s<strong>in</strong>gular mixed<br />
gender coord<strong>in</strong>ate NPs, occurr<strong>in</strong>g <strong>in</strong> either preverbal or <strong>in</strong> postverbal position favour s<strong>in</strong>gular<br />
verb agreement (CCA). In example (98), where the <strong>in</strong>terrogative sentence has the<br />
same word order as the declarative and the <strong>in</strong>tonation determ<strong>in</strong>es the type of the sentence,<br />
9 (out of 20) consultants chose s<strong>in</strong>gular verb agreement and only 4 (out of 20) consultants<br />
chose the plural. However, 7 (out of 20) consultants chose both a s<strong>in</strong>gular and a plural<br />
verb. In example (99), the sentence is also <strong>in</strong>terrogative but the coord<strong>in</strong>ate phrase occurs<br />
postverbally. In this case, 11 (out of 20) consultants chose a s<strong>in</strong>gular verb, 4 (out of 20)<br />
consultants chose a plural verb, and 5 (out of 20) consultants chose both a s<strong>in</strong>gular and<br />
a plural. In both sentences, the s<strong>in</strong>gular verb is preferred, with higher frequency <strong>in</strong> the<br />
second sentence.<br />
(98) I mitera i to<br />
the.sg.fem mother.sg.fem or the.sg.neut<br />
etimazet-e/-onte ja eksodo;<br />
is-gett<strong>in</strong>g-ready.sg/pl to go-out?<br />
‘Is the mother or the boy gett<strong>in</strong>g ready to go out?’<br />
(constructed)<br />
agori<br />
boy.sg.neut
6.4 Verb <strong>Agreement</strong> <strong>in</strong> MG Disjunctive Coord<strong>in</strong>ate <strong>Noun</strong>s 184<br />
(99) Etimaz-ete/-onte i mitera<br />
is-gett<strong>in</strong>g-ready.sg/pl the.sg.fem mother.sg.fem<br />
agori ja eksodo;<br />
boy.sg.neut to go-out?<br />
‘Is the mother or the boy gett<strong>in</strong>g ready to go out?’<br />
(constructed)<br />
i<br />
or<br />
to<br />
the.sg.neut<br />
In <strong>in</strong>terrogative sentences with a predicative adjective/participle, the plural verb form<br />
is favoured when the disjunctive phrase occurs preverbally, but the s<strong>in</strong>gular verb form<br />
occurs more often when the disjunctive phrase occurs postverbally. In example (100), 13<br />
(out of the 20) consultants chose the plural, 3 (out of the 20) consultants chose the s<strong>in</strong>gular<br />
and 4 (out of the 20) consultants chose both the s<strong>in</strong>gular and the plural form. From the<br />
consultants who chose both a s<strong>in</strong>gular and a plural verb form, 2 of the 4 consultants chose<br />
the plural <strong>in</strong> 2 of the 3 sentences and 2 of the 4 consultants chose the s<strong>in</strong>gular <strong>in</strong> 2 of the<br />
3 sentences. In (101), 9 (out of the 20) consultants chose the s<strong>in</strong>gular, 6 (out of the 20)<br />
consultants chose the plural and 5 (out of the 20) consultants chose both the s<strong>in</strong>gular and<br />
the plural form. From the consultants who chose both a s<strong>in</strong>gular and a plural verb form,<br />
3 of the 5 consultants chose the s<strong>in</strong>gular <strong>in</strong> 2 of the 3 sentences and 2 of the 5 consultants<br />
chose the plural <strong>in</strong> 2 of the 3 sentences.<br />
(100) I mitera i to<br />
the.sg.fem mother.sg.fem or the.sg.neut<br />
etim-i/-o<br />
ja eksodo;<br />
ready.pl.masc/sg.neut to go-out?<br />
‘Is the mother or the boy ready to go out?’<br />
(constructed)<br />
(101) Ine etimi i mitera<br />
is ready.sg.fem the.sg mother.sg.fem<br />
agori ja eksodo;<br />
boy.sg.neut to go-out?<br />
‘Is the mother or the boy ready to go out?’<br />
(constructed)<br />
In the follow<strong>in</strong>g table, we summarise the f<strong>in</strong>d<strong>in</strong>gs:<br />
i<br />
or<br />
agori<br />
boy.sg.neut<br />
to<br />
the.sg.neut<br />
<strong>in</strong>e<br />
are.pl<br />
Table 6.5: Disjunction i<br />
S<strong>in</strong>gular Mixed Gender <strong>Noun</strong>s<br />
Consultant Number: 20<br />
Interrogatives Verb <strong>Agreement</strong> Form<br />
np.sg or np.sg + verb ? s<strong>in</strong>gular plural s<strong>in</strong>g-pl/both<br />
(98) 9 4 7<br />
verb + np.sg or np.sg? s<strong>in</strong>gular plural s<strong>in</strong>g-pl/both<br />
(99) 11 4 5<br />
np.sg or np.sg + verb + pred.adj/part? s<strong>in</strong>gular plural s<strong>in</strong>g-pl/both<br />
(100) 3 13 4<br />
verb + np.sg or np.sg + pred.adj/part? s<strong>in</strong>gular plural s<strong>in</strong>g-pl/both<br />
(101) 9 6 5<br />
In the first group of <strong>in</strong>terrogative sentences, the number of native speakers who chose<br />
a s<strong>in</strong>gular verb was higher <strong>in</strong> verb-subject constructions than <strong>in</strong> subject-verb constructions.<br />
In the second group of sentences, the native speakers who chose a s<strong>in</strong>gular verb
6.4 Verb <strong>Agreement</strong> <strong>in</strong> MG Disjunctive Coord<strong>in</strong>ate <strong>Noun</strong>s 185<br />
were very few <strong>in</strong> subject-verb constructions whereas <strong>in</strong> verb-subject constructions,<br />
although the s<strong>in</strong>gular verb form was high, the plural verb form was not as low as <strong>in</strong> the<br />
correspond<strong>in</strong>g verb-subject constructions without the predicative adjective/participle.<br />
To summarise, <strong>in</strong> subject-verb <strong>in</strong>terrogatives and <strong>in</strong> verb-subject <strong>in</strong>terrogatives<br />
without a predicative adjective/participle, s<strong>in</strong>gular verb agreement occurs more frequently<br />
than plural. In <strong>in</strong>terrogative sentences with a predicative adjective/participle, the plural<br />
verb form is still preferable <strong>in</strong> subject-verb constructions but the s<strong>in</strong>gular takes over <strong>in</strong><br />
verb-subject constructions.<br />
The next group of coord<strong>in</strong>ate nouns that we will analyse is s<strong>in</strong>gular number but different<br />
person disjuncts. The person feature also determ<strong>in</strong>es the choice of verb agreement. In<br />
declarative sentences, participants show preference to a plural resolved verb form. In<br />
examples (102) and (103), 17 (out of 20) consultants opted for a plural verb and a resolved<br />
person follow<strong>in</strong>g the person hierarchy (Corbett, 1991), and 3 (out of 20) consultants chose<br />
the s<strong>in</strong>gular verb form (CCA):<br />
(102) Ego<br />
I.1.sg<br />
(103) Ego<br />
I.1.sg<br />
i<br />
or<br />
esi<br />
you.2.sg<br />
‘Me or you will need help’<br />
(constructed)<br />
i<br />
or<br />
o<br />
the<br />
tha<br />
will<br />
Kostas<br />
Kostas.3.sg<br />
‘Me or Kostas will take over the job’<br />
(constructed)<br />
xriastume<br />
need.1.pl<br />
tha<br />
will<br />
voithia<br />
help<br />
analavume<br />
take-over.1.pl<br />
t<strong>in</strong><br />
the<br />
dulia<br />
job<br />
When the sentence is <strong>in</strong>terrogative but reta<strong>in</strong>s the subject-verb order, the consultants<br />
show preference to a s<strong>in</strong>gular verb form and less often to a plural verb form. In particular,<br />
<strong>in</strong> example (104), 11 (out of 20) consultants chose a s<strong>in</strong>gular verb agreement form and 9<br />
(out of 20) consultants chose a plural verb agreement form. An example is presented below:<br />
(104) Esi<br />
you.2.sg<br />
i<br />
or<br />
i<br />
the<br />
Maria<br />
Maria.3.sg<br />
‘Will you or Maria go to the doctor?’<br />
(constructed)<br />
tha<br />
will<br />
pai<br />
go.3.sg<br />
sto<br />
to-the<br />
jatro;<br />
doctor?<br />
When the <strong>in</strong>terrogative sentence has a verb-subject order then native speakers also<br />
prefer s<strong>in</strong>gular verb agreement (and CCA for person). For <strong>in</strong>stance, <strong>in</strong> example (105), 13<br />
(out of 20) consultants chose a s<strong>in</strong>gular verb agreement form and 7 (out of 20) consultants<br />
chose a plural verb agreement form. Therefore, the s<strong>in</strong>gular verb form (and CCA for person)<br />
occurs with higher frequency <strong>in</strong> verb-subject <strong>in</strong>terrogatives than <strong>in</strong> subject-verb ones.<br />
An example is shown below:<br />
(105) Tha<br />
will<br />
pas<br />
go.2.sg<br />
esi<br />
you.2.sg<br />
i<br />
or<br />
‘Will you or Maria go to the doctor?’<br />
(constructed)<br />
i<br />
the<br />
Maria<br />
Maria.3.sg<br />
sto<br />
to-the<br />
jatro;<br />
doctor?<br />
Thus, <strong>in</strong> <strong>in</strong>terrogative sentences with s<strong>in</strong>gular and mixed person disjuncts, s<strong>in</strong>gular verb<br />
agreement is preferred when the sentence has a subject-verb order or a verb-subject<br />
order. To summarise, <strong>in</strong> declarative sentences with mixed person disjuncts speakers prefer<br />
the plural verb form than the s<strong>in</strong>gular (CCA). In <strong>in</strong>terrogative subject-verb or verbsubject<br />
sentences, the s<strong>in</strong>gular verb form occurs more often than the plural verb form<br />
with higher frequency <strong>in</strong> verb-subject sentences. The responses of the relevant examples<br />
are summarised briefly on the table below:
6.4 Verb <strong>Agreement</strong> <strong>in</strong> MG Disjunctive Coord<strong>in</strong>ate <strong>Noun</strong>s 186<br />
Table 6.6: Disjunction i<br />
S<strong>in</strong>gular Mixed Person <strong>Noun</strong>s<br />
Consultant Number: 20<br />
Affirmative Sentence Verb <strong>Agreement</strong> Form<br />
np.sg or np.sg + verb s<strong>in</strong>gular plural s<strong>in</strong>g-pl/both<br />
(102) 3 17 -<br />
(103) 3 17 -<br />
Interrogative Sentences Verb <strong>Agreement</strong> Form<br />
np.sg or np.sg + verb? s<strong>in</strong>gular plural s<strong>in</strong>g-pl/both<br />
(104) 11 9 -<br />
verb + np.sg or np.sg?<br />
(105) 13 7 -<br />
The current section presents the most important data of disjunctively conjo<strong>in</strong>ed phrases<br />
when disjuncts have a s<strong>in</strong>gular number but different gender and person features. The first<br />
observation is that a different gender and person <strong>in</strong> the disjuncts determ<strong>in</strong>es the choice of<br />
the verb form. Second, the construction itself <strong>in</strong> which the different disjuncts appear plays<br />
a critical role <strong>in</strong> verb agreement. In constructions where a predicative adjective/participle<br />
is present plural is favoured whereas when an adjective/participle is absent the choice of<br />
verb agreement is based on the <strong>in</strong>terpretation of the phrase and verb agreement can be<br />
either s<strong>in</strong>gular or plural. Also, the difference <strong>in</strong> the type of sentence seems to determ<strong>in</strong>e<br />
strongly whether the verb is plural or s<strong>in</strong>gular. Declarative sentences admit both a s<strong>in</strong>gular<br />
and a plural verb agreement whereas <strong>in</strong> <strong>in</strong>terrogative sentences the position of the subject<br />
affects number verb agreement s<strong>in</strong>ce <strong>in</strong> postverbal subjects s<strong>in</strong>gular verb agreement (CCA)<br />
<strong>in</strong>creases greatly as opposed to preverbal subjects. Thus, we can argue that verb agreement<br />
with disjunctive coord<strong>in</strong>ate nouns is also partly syntactically determ<strong>in</strong>ed and not only<br />
discourse-based 22 .<br />
6.4.3 Speakers’ Strategies and Verb <strong>Agreement</strong> <strong>in</strong> MG Disjunctive <strong>Noun</strong>s<br />
In the current part, we will focus on the agreement strategies used by MG native speakers.<br />
In uniform number disjuncts, native speakers choose either a s<strong>in</strong>gular or a plural verb,<br />
mean<strong>in</strong>g that the two possible strategies are prox and pl w<strong>in</strong>s 23 . Therefore, we will<br />
focus on coord<strong>in</strong>ate phrases consist<strong>in</strong>g of mixed number disjuncts to check whether any<br />
additional strategies occur and which strategies are stronger when compared to the rest.<br />
The data shows considerable variation with regard to the speakers’ answers. However, as <strong>in</strong><br />
other languages, there are some recurr<strong>in</strong>g stategies followed by MG native speakers when<br />
choos<strong>in</strong>g verb agreement.<br />
We will focus on subject-verb coord<strong>in</strong>ate constructions with a s<strong>in</strong>gular and a plural<br />
disjunct. Three sample sentences are shown below, one declarative and two <strong>in</strong>terrogatives:<br />
(106) To aftok<strong>in</strong>ito i ta mixanakia<br />
the.sg<br />
s<strong>in</strong>toma<br />
soon<br />
car.sg or the.pl mopeds.pl<br />
‘The car or the mopeds will be-sold soon’<br />
(constructed)<br />
tha<br />
will<br />
pulithun<br />
be-sold.pl<br />
22 We need to note that the above are only speculative remarks which need further <strong>in</strong>vestigation with<br />
corpus work and speakers’ personal op<strong>in</strong>ions (though this is more difficult when <strong>in</strong>terpretation is at issue).<br />
23 We will adopt the same terms used <strong>in</strong> Peterson (1986) for reasons of clarity.
6.4 Verb <strong>Agreement</strong> <strong>in</strong> MG Disjunctive Coord<strong>in</strong>ate <strong>Noun</strong>s 187<br />
(107) To moro i ta pedia xriazonte<br />
the.sg<br />
prosoxi;<br />
attention?<br />
baby.sg or the.pl children.pl need.pl<br />
‘Does the baby or the children need more attention?’<br />
(constructed)<br />
(108) O proedros i i ipurgi tha<br />
the.sg president.sg or the.pl m<strong>in</strong>isters.pl will.pl<br />
sto s<strong>in</strong>edrio<br />
<strong>in</strong>-the conference?<br />
‘Will the president or the m<strong>in</strong>isters talk <strong>in</strong> the conference’<br />
(constructed)<br />
perisoteri<br />
more<br />
milisun<br />
talk<br />
The data showed that all 20 (out of 20) consultants choose a plural verb irrespective<br />
of the type of the sentence 24 . The two possible strategies are pl w<strong>in</strong>s and prox but the<br />
plural verb does not show which strategy is employed by the native speakers. pl w<strong>in</strong>s is<br />
favoured for two reasons: the existence of two nouns <strong>in</strong> the noun phrase and the presence<br />
of a lexically plural disjunct. The prox strategy arises because the plural disjunct occurs<br />
close to the verb and therefore the features of that noun are selected to agree with the<br />
verb. The responses of the participants are summarised <strong>in</strong> the table below:<br />
Table 6.7: Disjunction i<br />
Mixed Number <strong>Noun</strong>s<br />
Consultant Number: 20<br />
Conjunct Order Verb <strong>Agreement</strong> Form<br />
np.sg or np.pl s<strong>in</strong>gular plural s<strong>in</strong>g-pl/both<br />
(106) - 20 -<br />
(107) - 20 -<br />
(108) - 20 -<br />
Next, <strong>in</strong> order to check which of the two strategies determ<strong>in</strong>es the choice of the consultants,<br />
we carried out two different tests. In the first test, we gave 4 different sentences<br />
with the reverse order <strong>in</strong> the disjuncts, plac<strong>in</strong>g the plural disjunct first and the s<strong>in</strong>gular<br />
disjunct second. The sample sentences, which are either declarative or <strong>in</strong>terrogative, are<br />
presented below:<br />
(109) I gates i o skilos<br />
the.pl cats.pl or the.sg dog.sg<br />
luludia mu;<br />
flowers my?<br />
‘Did your cats or your dog eat my flowers?’<br />
(constructed)<br />
(110) Ta koritsia i o ijos<br />
the.pl<br />
spiti<br />
house<br />
girls.pl or the.sg son.sg<br />
‘The girls or the son will <strong>in</strong>herit the house’<br />
(constructed)<br />
su<br />
your<br />
tha<br />
will<br />
efag-e/-an<br />
ate.sg/pl<br />
klironomis-i/-un<br />
<strong>in</strong>herit.sg/pl<br />
24 One participant chose s<strong>in</strong>gular <strong>in</strong> 1 sentence out of the 3 <strong>in</strong> the questionnaire and it is probably<br />
attributed to an <strong>in</strong>correct choice.<br />
ta<br />
the<br />
to<br />
the
6.4 Verb <strong>Agreement</strong> <strong>in</strong> MG Disjunctive Coord<strong>in</strong>ate <strong>Noun</strong>s 188<br />
(111) I<br />
the.pl<br />
karekles<br />
chairs.pl<br />
i<br />
or<br />
to<br />
the.sg<br />
trapezi<br />
table.sg<br />
‘Do the chairs or the table need to be pa<strong>in</strong>ted?’<br />
(constructed)<br />
(112) I podosferistes i o dietitis<br />
the.pl<br />
xrimata<br />
money<br />
footballers.pl or the.sg coach.sg<br />
‘The footballers or the coach earn much money’<br />
(constructed)<br />
thel-i/-un<br />
need.sg/pl<br />
kerdiz-i/-un<br />
earn.sg/pl<br />
vapsimo;<br />
pa<strong>in</strong>t<strong>in</strong>g?<br />
polla<br />
much<br />
For the participants who choose a plural verb <strong>in</strong> the four examples above pl w<strong>in</strong>s<br />
strategy is stronger than the prox. The same can be claimed for examples (106), (107)<br />
and (108) where the plural verb form is ma<strong>in</strong>ly due to the pl w<strong>in</strong>s strategy and less to the<br />
prox one. In particular, the greatest majority of consultants chose plural verb agreement<br />
<strong>in</strong> all 4 sentences whereas the s<strong>in</strong>gular occurs less often, followed by a few consultants<br />
who chose both forms <strong>in</strong> some sentences. More precisely, <strong>in</strong> sentence (109), 15 (out of 20)<br />
consultants chose a plural verb and 4 (out of 20) consultants chose a s<strong>in</strong>gular verb. Only 1<br />
(out of 20) consultant chose both a s<strong>in</strong>gular and a plural verb. In sentence (110), 13 (out<br />
of 20) consultants chose a plural verb, 5 (out of 20) consultants chose a s<strong>in</strong>gular verb and<br />
2 (out of 20) consultants chose both forms. In sentence (111), 16 (out of 20) consultants<br />
chose a plural verb, 4 (out of 20) consultants chose a s<strong>in</strong>gular verb and no consultant chose<br />
both forms. F<strong>in</strong>ally, <strong>in</strong> sentence (112), 15 (out of 20) consultants chose a plural verb, 3<br />
(out of 20) consultants chose a s<strong>in</strong>gular verb and 2 (out of 20) consultants chose both<br />
forms. Thus, <strong>in</strong> the above examples for those who chose only plural, the pl w<strong>in</strong>s strategy<br />
is stronger than the prox one. For those who chose a s<strong>in</strong>gular verb, the prox strategy is<br />
prefered, whereas for those who opted for both a s<strong>in</strong>gular and a plural verb form, both pl<br />
w<strong>in</strong>s and prox are <strong>in</strong> operation. Almost <strong>in</strong> all cases the consultants were consistent <strong>in</strong><br />
their responses choos<strong>in</strong>g the same form with a few exceptions. However, s<strong>in</strong>ce the majority<br />
chose a plural verb form pl w<strong>in</strong>s is the mostly preferred strategy even <strong>in</strong> nouns with the<br />
order [pl + sg]. The table below summarises the responses of the consultants:<br />
Table 6.8: Disjunction i<br />
Mixed Number <strong>Noun</strong>s<br />
Consultant Number: 20<br />
Conjunct Order Verb <strong>Agreement</strong> Form<br />
np.pl or np.sg s<strong>in</strong>gular plural s<strong>in</strong>g-pl/both<br />
(109) 4 15 1<br />
(110) 5 13 2<br />
(111) 4 16 -<br />
(112) 3 15 2<br />
The second test we carried out was to <strong>in</strong>clude a predicative adjective/participle that<br />
follows the coord<strong>in</strong>ate phrase. As MG encodes gender agreement <strong>in</strong> the predicative adjective<br />
and not <strong>in</strong> the predicate itself, these examples would help us test whether the plural <strong>in</strong><br />
the verb is based on the pl w<strong>in</strong>s strategy or the prox one. The follow<strong>in</strong>g example consists<br />
of a s<strong>in</strong>gular mascul<strong>in</strong>e and a plural fem<strong>in</strong><strong>in</strong>e disjunct:<br />
(113) O adelfos<br />
the.masc.sg brother.masc.sg<br />
<strong>in</strong>e taktopiimen-i/-es;<br />
are.pl settled.masc/fem.pl?<br />
i<br />
or<br />
i<br />
the.fem.pl<br />
adelfes<br />
sisters.fem.pl<br />
su<br />
your
6.4 Verb <strong>Agreement</strong> <strong>in</strong> MG Disjunctive Coord<strong>in</strong>ate <strong>Noun</strong>s 189<br />
‘Is/are your brother or your sisters settled?’<br />
(constructed)<br />
The results showed that more than half (13/20) of the native speakers chose the mascul<strong>in</strong>e<br />
plural form <strong>in</strong> the predicative adjective, which clearly <strong>in</strong>dicates that pl w<strong>in</strong>s is the ma<strong>in</strong><br />
strategy, whereas less than half (7/20) chose the fem<strong>in</strong><strong>in</strong>e plural, which shows that prox<br />
is also chosen. In fact, the results were not much different than those of the first test.<br />
Also, when revers<strong>in</strong>g the order of the conjuncts, as <strong>in</strong> (114), 16 (out of 20) consultants still<br />
choose mascul<strong>in</strong>e plural <strong>in</strong> most cases apart from 4 (out of 20) consultants who chose the<br />
s<strong>in</strong>gular verb form as a result of closest conjunct agreement.<br />
(114) I adelfes<br />
the.fem.pl sisters.fem.pl<br />
<strong>in</strong>e taktopiimeni;<br />
are.pl settled.masc.pl?<br />
i<br />
or<br />
‘Is/are your brother or your sisters settled?’<br />
(constructed)<br />
o<br />
the.masc.sg<br />
adelfos<br />
brother.masc.sg<br />
su<br />
your<br />
The above tests show that <strong>in</strong> MG coord<strong>in</strong>ate phrases with a s<strong>in</strong>gular and a plural<br />
disjunct, two strategies are <strong>in</strong> operation pl w<strong>in</strong>s and prox. The plural verb form occurs<br />
consistently <strong>in</strong> [sg + pl] constructions but with lower frequency <strong>in</strong> [pl + sg] constructions<br />
where prox seems to take over partly. Therefore, the strategies that seem to be <strong>in</strong> operation<br />
<strong>in</strong> MG preverbal coord<strong>in</strong>ate phrases are pl w<strong>in</strong>s, followed by the prox one.<br />
Next, we need to check whether “first conjunct agreement” (fc w<strong>in</strong>s) (Peterson, 1986,<br />
239) or “distant conjunct agreement” (Corbett, 2001; Sadler, 2003) are used <strong>in</strong> MG disjunctive<br />
coord<strong>in</strong>ate nouns. As seen above, coord<strong>in</strong>ate noun phrases with a noun order [sg<br />
+ pl] would never yield s<strong>in</strong>gular verb agreement. Therefore, a sentence like the one below<br />
is ungrammatical:<br />
(115) *O<br />
the.masc.sg<br />
exi dulia;<br />
has.sg job?<br />
adelfos<br />
brother.masc.sg<br />
i<br />
or<br />
i<br />
the.fem.pl<br />
‘Has/have your brother or your sisters got a job?’<br />
(constructed)<br />
adelfes<br />
sisters.fem.pl<br />
su<br />
your<br />
Further evidence that fc w<strong>in</strong>s is not a strategy adopted by MG native speakers <strong>in</strong> disjunctive<br />
conjo<strong>in</strong>ed nouns was drawn from sentences with a predicative adjective/participle<br />
where gender agreement is shown. The follow<strong>in</strong>g example <strong>in</strong>cludes a s<strong>in</strong>gular mascul<strong>in</strong>e<br />
and a plural fem<strong>in</strong><strong>in</strong>e noun:<br />
(116) O pateras i i kores<br />
the.masc.sg father.masc.sg or the.fem.pl daughters.fem.pl<br />
<strong>in</strong>e ipefth<strong>in</strong>-i/-es<br />
ja t<strong>in</strong> eteria<br />
are.pl responsible.masc/fem.pl for the company<br />
‘The father or the daughters are responsible for the company’<br />
(constructed)<br />
As a result of fc w<strong>in</strong>s, we would expect a mascul<strong>in</strong>e s<strong>in</strong>gular agreement form <strong>in</strong> the<br />
adjective above. However, <strong>in</strong> example (116), 13 (out of 20) consultants resort to a plural<br />
verb form with a mascul<strong>in</strong>e plural predicative adjective, as a result of pl w<strong>in</strong>s and gender<br />
resolution, whereas 7 (out of 20) consultants choose fem<strong>in</strong><strong>in</strong>e plural as a result of CCA.<br />
When revers<strong>in</strong>g the order of the disjuncts, an fc w<strong>in</strong>s strategy would admit a fem<strong>in</strong><strong>in</strong>e<br />
plural adjective, but <strong>in</strong>stead the plural verb form and the mascul<strong>in</strong>e plural predicative
6.4 Verb <strong>Agreement</strong> <strong>in</strong> MG Disjunctive Coord<strong>in</strong>ate <strong>Noun</strong>s 190<br />
adjective occurs ma<strong>in</strong>ly <strong>in</strong> 16 (out of 20) consultants whereas a mascul<strong>in</strong>e s<strong>in</strong>gular adjective<br />
is chosen by 4 (out of 20) consultants as a second choice due to CCA:<br />
(117) I kores<br />
i o<br />
pateras<br />
the.fem.pl daughters.fem.pl or the.masc.sg father.masc.sg<br />
<strong>in</strong>e ipefth<strong>in</strong>-i/-os<br />
ja t<strong>in</strong> eteria<br />
are.pl responsible.masc.pl/masc.sg for the company<br />
‘The daughters or the father are responsible for the company’<br />
(constructed)<br />
Thus, it appears that there is not evidence to support the existence of an fc w<strong>in</strong>s or<br />
“distant conjunct agreement” strategy <strong>in</strong> MG. Also, the above analysis has not presented<br />
any evidence for the existence of a sg w<strong>in</strong>s (Peterson, 1986, 239) strategy, which seems to<br />
be nonexistent <strong>in</strong> MG coord<strong>in</strong>ate nouns.<br />
Let us consider next disjunctive phrases with mixed number disjuncts <strong>in</strong> postverbal<br />
position. The data shows that <strong>in</strong> MG the different sentence structure does not give different<br />
results, still favour<strong>in</strong>g the pl w<strong>in</strong>s strategy more than the prox one. A relevant example<br />
is shown below:<br />
(118) Etimaz-onte ta agoria i to<br />
gett<strong>in</strong>g-prepared.pl the.pl boys.pl or the.sg<br />
ja tis eksetasis;<br />
for the exams?<br />
‘Are the boys or the girl gett<strong>in</strong>g prepared for the exams?’<br />
(constructed)<br />
koritsi<br />
girl.sg<br />
su<br />
your<br />
The results of the questionnaire showed that 20 (out of 20) consultants choose a plural<br />
verb. The plural choice is firstly motivated by the pl w<strong>in</strong>s strategy which seems to<br />
determ<strong>in</strong>e agreement patterns more widely when compared to the prox strategy. The<br />
second possibility is that the plural verb form is due to the existence of a plural disjunct<br />
close to the verb, referr<strong>in</strong>g to the prox strategy.<br />
In order to test aga<strong>in</strong> which strategies are <strong>in</strong>volved <strong>in</strong> postverbal coord<strong>in</strong>ate nouns, we<br />
carried out similar tests as above. The reverse order of the conjuncts, as <strong>in</strong> example (119),<br />
will result <strong>in</strong> the choice of a plural verb <strong>in</strong> 11 (out of the 20) consultants <strong>in</strong> the sentences<br />
<strong>in</strong>cluded <strong>in</strong> the questionnaire, show<strong>in</strong>g that pl w<strong>in</strong>s is <strong>in</strong>deed a prevalent strategy. The<br />
prox strategy also ga<strong>in</strong>s some ground <strong>in</strong> verb-subject construction as is expected. In<br />
fact, a s<strong>in</strong>gular verb was chosen <strong>in</strong> all sentences by only one consultant, which is due to<br />
the prox strategy. 8 out of the 20 consultants chose a s<strong>in</strong>gular verb <strong>in</strong> 1 of the 2 sentences<br />
and a plural verb <strong>in</strong> the other sentence. The choice of either a s<strong>in</strong>gular or a plural verb is<br />
attributed to the prox or pl w<strong>in</strong>s strategies, respectively.<br />
(119) Etimaz-ete/-onte<br />
gett<strong>in</strong>g-prepared.sg/pl<br />
ja tis eksetasis;<br />
for the exams?<br />
to<br />
the.sg<br />
koritsi<br />
girl.sg<br />
i<br />
or<br />
ta<br />
the.pl<br />
‘Are/is the girl or the boys gett<strong>in</strong>g prepared for the exams?’<br />
(constructed)<br />
agoria<br />
boys.pl<br />
su<br />
your<br />
In conclusion, <strong>in</strong> postverbal coord<strong>in</strong>ate phrases with a [pl or sg] noun order the prevalent<br />
strategy is pl w<strong>in</strong>s s<strong>in</strong>ce the data analysis showed that most native speakers chose<br />
a plural verb. The prox strategy also seems to affect the result s<strong>in</strong>ce the plural noun is<br />
closest to the plural verb, lead<strong>in</strong>g to a 100% selection of the plural verb. In coord<strong>in</strong>ate<br />
phrases with a [sg or pl] disjunct order, consultants will select either a plural only verb
6.5 Issues on the MG positive & negative predisjunction-disjunction 191<br />
form exclusively, follow<strong>in</strong>g the pl w<strong>in</strong>s strategy, a plural or a s<strong>in</strong>gular verb, balanc<strong>in</strong>g<br />
the pl w<strong>in</strong>s with the prox one, and <strong>in</strong> other cases a s<strong>in</strong>gular verb, follow<strong>in</strong>g the prox<br />
strategy.<br />
The last th<strong>in</strong>g to note is that <strong>in</strong> postverbal coord<strong>in</strong>ate phrases with a plural and a<br />
s<strong>in</strong>gular disjunct the verb cannot be s<strong>in</strong>gular, as below:<br />
(120) *Etimaz-ete ta koritsia i to<br />
gett<strong>in</strong>g-prepared.sg the.pl girl.pl or the.sg<br />
ja tis eksetasis;<br />
for the exams?<br />
‘Are the girls or the boy gett<strong>in</strong>g prepared for the exams?’<br />
(constructed)<br />
agori<br />
boy.sg<br />
su<br />
your<br />
Thus, the possibility of “distant agreement” is not part of the MG agreement system.<br />
On the whole, the above analysis showed that <strong>in</strong> both preverbal and postverbal mixed<br />
number disjunctively coord<strong>in</strong>ate nouns, MG native speakers consistently use two ma<strong>in</strong><br />
agreement strategies, pl w<strong>in</strong>s and prox. No evidence exists <strong>in</strong> MG for an fc w<strong>in</strong>s or a<br />
sg w<strong>in</strong>s strategy. In preverbal coord<strong>in</strong>ate nouns, the prevalent strategy is pl w<strong>in</strong>s and<br />
this is followed by the prox one. In postverbal coord<strong>in</strong>ate nouns, the pl w<strong>in</strong>s is still a<br />
strong strategy but prox seems to play a more active role.<br />
6.5 Issues on the MG positive & negative predisjunction-disjunction<br />
An important part of the literature <strong>in</strong> English concern<strong>in</strong>g disjunction has dealt with the<br />
claim that predisjunction and disjunction either...or behaves different from disjunction or<br />
(see Larson (1985); Morgan (1985); Peterson (1986)). This difference is reflected <strong>in</strong> verb<br />
agreement, which yields different agreement forms when the predisjunction-disjunction<br />
cooccur as opposed to bare disjunction. This section will focus on the presentation of MG<br />
data with the positive and negative predisjunction and disjunction ite...ite ‘either...or’ and<br />
oute...oute ‘neither...nor’, look<strong>in</strong>g at verb agreement.<br />
First, we will present the positive predisjunction and disjunction ite...ite. The results of<br />
the questionnaire, which <strong>in</strong>cluded sentences with the positive predisjunction and disjunction<br />
with disjuncts of the same or mixed number, gender and person, showed a few differences<br />
from the data which did not <strong>in</strong>clude the predisjunction. When the coord<strong>in</strong>ate phrase<br />
consists of two s<strong>in</strong>gular disjuncts with the same gender and person, most consultants chose<br />
a s<strong>in</strong>gular verb and fewer consultants chose a plural verb. For <strong>in</strong>stance, <strong>in</strong> example (121),<br />
14 (out of 20) consultants used a s<strong>in</strong>gular verb and 6 (out of 20) used a plural verb, and<br />
<strong>in</strong> example (122), 16 (out of 20) consultants chose a s<strong>in</strong>gular verb and only 4 consultants<br />
chose a plural verb. F<strong>in</strong>ally, <strong>in</strong> example (123), 14 (out of 20) used a s<strong>in</strong>gular verb and 6<br />
(out of 20) used a plural verb. This choice also depends on the <strong>in</strong>terpretation of the phrase,<br />
which is likely to be <strong>in</strong>terpreted as an ‘exclusive’(‘or’) or. Both verb forms are illustrated<br />
<strong>in</strong> the follow<strong>in</strong>g example:<br />
(121) I o ijos i i pateras<br />
either the.sg son.sg or the.sg father.sg<br />
par-i/-un meros sto diagonismo<br />
take.sg/pl part <strong>in</strong>-the contest<br />
‘Either my son or my father will take part <strong>in</strong> the contest’<br />
(constructed)<br />
(122) I i<br />
either the.sg<br />
analav-i/-un<br />
take-over.sg/pl<br />
mitera i<br />
mother.sg or<br />
t<strong>in</strong> epimelia<br />
the custody<br />
i<br />
the.sg<br />
thia<br />
aunt.sg<br />
mu<br />
m<strong>in</strong>e<br />
tha<br />
will<br />
tha<br />
will
6.5 Issues on the MG positive & negative predisjunction-disjunction 192<br />
‘Either the mother or the aunt will take over the custody’<br />
(constructed)<br />
(123) I to agori i to koritsi<br />
either the.sg boy.sg or the.sg girl.sg<br />
ta simer<strong>in</strong>a provlimata<br />
the today problems<br />
‘Either the boy or the girl will face today’s problems’<br />
(constructed)<br />
tha<br />
will<br />
antimetopis-i/-un<br />
face.sg/pl<br />
The number of consultants who chose consistently s<strong>in</strong>gular verb agreement <strong>in</strong> data<br />
with the predisjunction and disjunction was more or less the same as <strong>in</strong> cases without the<br />
predisjunction. The table is shown below:<br />
Table 6.9: S<strong>in</strong>gular <strong>Noun</strong>s with Identical Gender & Number<br />
Consultant Number: 20<br />
Examples Verb <strong>Agreement</strong> Form/Interpretation of or<br />
s<strong>in</strong>gular/‘exclusive’ plural/‘<strong>in</strong>clusive’ s<strong>in</strong>g-pl/both<br />
(121) 14 6 -<br />
(122) 16 4 -<br />
(123) 14 6 -<br />
Cases of mixed gender and mixed person s<strong>in</strong>gular disjuncts with the predisjunction and<br />
disjunction display also similar results as coord<strong>in</strong>ate phrases without the predisjunction.<br />
In mixed gender disjuncts, with a mascul<strong>in</strong>e and a fem<strong>in</strong><strong>in</strong>e noun, plural verb agreement is<br />
preferred if there is a predicative adjective/participle, whereas the s<strong>in</strong>gular and less often<br />
the plural agreement occur without a predicative adjective/participle. In examples (124)<br />
and (125), which have a predicative adjective, 14 (out of 20) consultants chose a plural verb,<br />
3 (out of 20) consultants chose a s<strong>in</strong>gular verb and 3 (out of 20) consultants chose both a<br />
s<strong>in</strong>gular and a plural verb. In examples (126) and (127), 9 (out of 20) consultants chose a<br />
s<strong>in</strong>gular verb, 3 (out of 20) consultants chose a plural verb and 8 (out of 20) consultants<br />
chose both a s<strong>in</strong>gular and a plural verb. The relevant examples are shown below:<br />
(124) I o<br />
pateras i i mitera<br />
either the.masc.sg father.masc.sg or the.fem.sg mother.fem.sg<br />
<strong>in</strong>e ipefth<strong>in</strong>-i/-i<br />
ja t<strong>in</strong> epimelia tu<br />
are.sg<br />
pediu<br />
child<br />
responsible.masc.pl/fem.sg for the custody of-the<br />
‘Either the father or the mother are responsible for the child’s custody’<br />
(constructed)<br />
(125) I o<br />
ijos i i<br />
either the.masc.sg son.masc.sg or the.fem.sg<br />
<strong>in</strong>e ikan-i/-i<br />
na dosun simvules<br />
are.sg able.masc.pl/fem.sg to give advise<br />
‘Either the son or the daughter are able to give any advise’<br />
(constructed)<br />
(126) I o adelfos i<br />
either the.sg brother.sg or<br />
peras-i/-un sto panepistimio<br />
enter.sg/pl the university<br />
i<br />
the.sg<br />
adelfi<br />
sister.sg<br />
kori<br />
daughter.fem.sg<br />
mu<br />
m<strong>in</strong>e<br />
tha<br />
will
6.5 Issues on the MG positive & negative predisjunction-disjunction 193<br />
‘Either my brother or my sister will enter the university’<br />
(constructed)<br />
(127) I o thios i i thia<br />
either the.sg uncle.sg or the.sg aunt.sg<br />
analav-i/-un to magazi<br />
take-over.sg/pl the shop<br />
‘Either my uncle or my aunt will take over the shop’<br />
(constructed)<br />
The table below summarises the patterns presented:<br />
mu<br />
m<strong>in</strong>e<br />
tha<br />
will<br />
Table 6.10: Predisjunction-Disjunction i...i<br />
S<strong>in</strong>gular Mixed Gender <strong>Noun</strong>s<br />
Consultant Number: 20<br />
Sentence Type Verb <strong>Agreement</strong> Form<br />
either np.sg or np.sg + verb s<strong>in</strong>gular plural s<strong>in</strong>g-pl/both<br />
(124) 3 14 3<br />
(125) 3 14 3<br />
either np.sg or np.sg + verb + pred.adj/part s<strong>in</strong>gular plural s<strong>in</strong>g-pl/both<br />
(126) 9 3 8<br />
(127) 9 3 8<br />
In mixed person disjunctive coord<strong>in</strong>ate nouns with the predisjunction and disjunction,<br />
we also f<strong>in</strong>d similar patterns to the cases without the predisjunction. Thus, the plural verb<br />
agreement is favoured <strong>in</strong> declarative sentences, follow<strong>in</strong>g person hierarchy (Corbett, 1991).<br />
In all relevant examples, 16 (out of 20) consultants chose a plural verb form and 4 (out of<br />
20) consultants chose a s<strong>in</strong>gular verb. Some examples are shown below:<br />
(128) I<br />
either<br />
ego<br />
me.1.sg<br />
i<br />
or<br />
esi<br />
you.2.sg<br />
tha<br />
will<br />
‘Either me or you will take over the bus<strong>in</strong>ess’<br />
(constructed)<br />
(129) I esi i o pateras<br />
either<br />
xartia<br />
papers<br />
you.2.sg or the father.3.sg<br />
‘Either you or the father will sign the papers’<br />
(constructed)<br />
analavume<br />
take-over.1.pl<br />
tha<br />
will<br />
t<strong>in</strong><br />
the<br />
ipograpsete<br />
sign.2.pl<br />
epixirisi<br />
bus<strong>in</strong>ess<br />
Interrogative sentences with i...i disjunctively conjo<strong>in</strong>ed nouns <strong>in</strong> preverbal or <strong>in</strong> postverbal<br />
position show s<strong>in</strong>gular and less often plural verb agreement. In particular, <strong>in</strong> example<br />
(130), 11 (out of 20) consultants chose a s<strong>in</strong>gular verb form and 9 (out of 20) consultants<br />
chose a plural verb form. In example (131), 13 (out of 20) consultants chose a s<strong>in</strong>gular<br />
verb form and 7 (out of 20) consultants chose a plural verb form. The relevant examples<br />
are shown below:<br />
(130) I esi i i Maria ex-i/-ete<br />
either you.2.sg or the Maria.3.sg has.3.sg/-ve.3.pl<br />
efth<strong>in</strong>i ja t<strong>in</strong> dulia;<br />
responsibility for the job?<br />
‘Do either you or Maria have the responsibility for the job?’<br />
(constructed)<br />
ta<br />
the<br />
t<strong>in</strong><br />
the
6.5 Issues on the MG positive & negative predisjunction-disjunction 194<br />
(131) Exis i esi i i Maria<br />
have.2.sg either you.2.sg or the Maria.3.sg<br />
efth<strong>in</strong>i ja t<strong>in</strong> dulia;<br />
responsibility for the job?<br />
‘Do either you or Maria have the responsibility for the job?’<br />
(constructed)<br />
The results of mixed person s<strong>in</strong>gular disjuncts are illustrated below:<br />
t<strong>in</strong><br />
the<br />
Table 6.11: Predisjunction-Disjunction i...i<br />
S<strong>in</strong>gular Mixed Person <strong>Noun</strong>s<br />
Consultant Number: 20<br />
Affirmative Verb <strong>Agreement</strong> Form<br />
either np.sg or np.sg + verb. s<strong>in</strong>gular plural s<strong>in</strong>g-pl/both<br />
(128) 4 16 -<br />
(129) 4 16 -<br />
Interrogatives Verb <strong>Agreement</strong> Form<br />
either np.sg or np.sg verb? s<strong>in</strong>gular plural s<strong>in</strong>g-pl/both<br />
(130) 11 9 -<br />
verb either np.sg or np.sg?<br />
(131) 13 7 -<br />
On the whole, the positive predisjunction and disjunction yields almost similar results<br />
to bare disjunction i. A s<strong>in</strong>gular verb is preferable <strong>in</strong> s<strong>in</strong>gular disjuncts and a plural verb<br />
form comes as a second option, related to the <strong>in</strong>terpretation of the phrase. In mixed<br />
gender disjuncts, s<strong>in</strong>gular verb agreement is prefered over plural when no predicative adjective/participle<br />
is present otherwise a plural verb and a resolved gender occurs more often<br />
when a predicative adjective/participle is present. F<strong>in</strong>ally, mixed person disjuncts follow<br />
ma<strong>in</strong>ly resolution <strong>in</strong> declarative sentences and s<strong>in</strong>gular or CCA <strong>in</strong> <strong>in</strong>terrogative sentences 25 .<br />
Disjunctively conjo<strong>in</strong>ed nouns with the negative disjunction oute...oute display a general<br />
preference towards a plural verb form and less often to a s<strong>in</strong>gular verb form irrespective of<br />
the sentence structure <strong>in</strong> which they occur. Parker (1983), who also discusses coord<strong>in</strong>ate<br />
nouns jo<strong>in</strong>ed with neither...nor, f<strong>in</strong>ds similar results <strong>in</strong> English and argues that there are<br />
pragmatic reasons towards this plural choice. He perceives positive or negative disjunctively<br />
conjo<strong>in</strong>ed nouns as def<strong>in</strong><strong>in</strong>g a two-member set, which is divided differently <strong>in</strong> positive and<br />
negative disjunctions. In positive disjunctive phrases, positive disjunction divides the set<br />
<strong>in</strong>to two subsets, with a s<strong>in</strong>gle member each. In negative disjunctive phrases, negative<br />
disjunction divides the two-member set <strong>in</strong>to two subsets but one of the subsets conta<strong>in</strong>s<br />
two members and the other conta<strong>in</strong>s no members at all. This dist<strong>in</strong>ction expla<strong>in</strong>s the<br />
higher percentage of plural number <strong>in</strong> negative disjunction. This is illustrated as follows:<br />
(132) either X or Y = {X} - {Y}<br />
neither X nor Y = {XY} - {∅}<br />
(Parker, 1983, 15)<br />
The <strong>Greek</strong> consultants were given sentences with the negative predisjunction and disjunction<br />
conta<strong>in</strong><strong>in</strong>g s<strong>in</strong>gular disjuncts and mixed number disjuncts either of the same or<br />
mixed gender and person. When the disjuncts are both s<strong>in</strong>gular number with the same<br />
gender, 17 (out of 20) consultants chose a plural verb whereas only 3 (out of 20) consultants<br />
chose a s<strong>in</strong>gular form <strong>in</strong> both examples. Two characteristic examples are shown below:<br />
25 More elaborate research with a wider range of data and speakers’ judgements is required to draw more<br />
concrete conclusions.
6.5 Issues on the MG positive & negative predisjunction-disjunction 195<br />
(133) Ute to agori ute to<br />
neither the.neut.sg boy.neut.sg nor the<br />
<strong>in</strong>e etim-a/-o ja afti t<strong>in</strong><br />
are/is.pl/sg ready.neut.pl/sg for this the<br />
‘Neither the boy nor the girl are ready for this change’<br />
(constructed)<br />
koritsi<br />
girl.neut.sg<br />
metavoli<br />
change<br />
(134) Ute o pateras ute o ijos<br />
neither the.neut.sg father.neut.sg nor the son.neut.sg<br />
enekr<strong>in</strong>-e/-an t<strong>in</strong> agora tis eterias<br />
approved.pl/sg the.neut.sg purchase of-the company<br />
‘Neither the father nor the son approved the company purchase’<br />
(constructed)<br />
When the disjuncts are s<strong>in</strong>gular with mixed gender and there is no predicative adjective/participle<br />
then also most (17/20) participants chose a plural verb and only 3 (out of<br />
20) chose a s<strong>in</strong>gular verb <strong>in</strong> all examples. However, when there is a predicative adjective/participle<br />
then a plural verb is used exclusively (20/20) even by those who opted for<br />
a s<strong>in</strong>gular verb agreement when the predicative adjective/participle was not present. The<br />
follow<strong>in</strong>g examples <strong>in</strong>clude sentences with and without a predicative adjective/participle,<br />
respectively:<br />
(135) Ute o<br />
antras ute i<br />
neither the.masc.sg man.masc.sg nor the.fem.sg<br />
j<strong>in</strong>eka simvivaz-onte/-ete t<strong>in</strong> simer<strong>in</strong>i epoxi<br />
woman.fem.sg compromise.pl/sg <strong>in</strong> these days<br />
‘Neither the man nor the woman compromise nowadays’<br />
(constructed)<br />
(136) Ute o<br />
ijos ute i<br />
neither the.masc.sg son.masc.sg nor the.fem.sg<br />
kori<br />
apodex-onte/-ete to diazigio<br />
daughter.fem.sg accept.pl/sg the divorce<br />
‘Neither the son nor the daughter accept the divorce’<br />
(constructed)<br />
(137) Ute o<br />
pateras ute i<br />
neither the.masc.sg father.masc.sg nor the.fem.sg<br />
mitera <strong>in</strong>e xarumen-i me afti t<strong>in</strong><br />
mother.fem.sg are.pl happy.masc.pl with this the<br />
‘Neither the father nor the mother are happy with this situation’<br />
(constructed)<br />
katastasi<br />
situation<br />
(138) Ute o<br />
pelatis<br />
ute i<br />
neither the.masc.sg male-customer.masc.sg nor the.fem.sg<br />
pelatisa<br />
<strong>in</strong>e ikanopiimen-i me t<strong>in</strong> anodo<br />
female-customer.fem.sg are.pl satisfied.masc.pl with the rise<br />
stis times<br />
<strong>in</strong> prices<br />
‘Neither the male nor the female customer are satisfied with the rise <strong>in</strong> prices’<br />
(constructed)<br />
The table below summarises the patterns presented:
6.6 Analys<strong>in</strong>g Verb <strong>Agreement</strong> <strong>in</strong> Disjunctive Coord<strong>in</strong>ate <strong>Noun</strong>s 196<br />
Table 6.12: Negative Predisjunction-Disjunction ute...ute<br />
S<strong>in</strong>gular Uniform Gender <strong>Noun</strong>s<br />
Consultant Number: 20<br />
Sentence Type Verb <strong>Agreement</strong> Form<br />
neither np.sg nor np.sg + verb s<strong>in</strong>gular plural s<strong>in</strong>g-pl/both<br />
(133) 3 17 -<br />
(134) 3 17 -<br />
neither np.sg nornp.sg + verb s<strong>in</strong>gular plural s<strong>in</strong>g-pl/both<br />
(135) 3 17 -<br />
(136) 3 17 -<br />
neither np.sg nor np.sg + verb + pred.adj/part s<strong>in</strong>gular plural s<strong>in</strong>g-pl/both<br />
(137) - 20 -<br />
(138) - 20 -<br />
Disjuncts with the same number and different person yield similar results where the majority<br />
of consultants (17 out of 20) choose a plural verb and fewer consultants (3 out of<br />
20 consultants) choose a s<strong>in</strong>gular. Therefore, the plural verb form is used more often than<br />
the s<strong>in</strong>gular and the person hierarchy is strictly obeyed <strong>in</strong> the former case. Consider the<br />
follow<strong>in</strong>g example:<br />
(139) Ute<br />
neither<br />
esi<br />
you.2.sg<br />
ute<br />
nor<br />
aftos<br />
he.3.sg<br />
‘Neither you nor he will accept the offer’<br />
(constructed)<br />
(140) Ute<br />
neither<br />
esi<br />
you.2.sg<br />
ute<br />
nor<br />
‘Neither you nor I need help’<br />
(constructed)<br />
ego<br />
I.1.sg<br />
tha<br />
will<br />
xriazomaste<br />
need<br />
dextite<br />
accept.3.pl<br />
voithia<br />
help.2.pl<br />
t<strong>in</strong><br />
the<br />
prosfora<br />
offer<br />
From the above data, we conclude that <strong>in</strong> MG the negative predisjunction and disjunction<br />
generates ma<strong>in</strong>ly plural verb agreement. These results seem to be <strong>in</strong> accordance with<br />
Parker’s view that negative disjunction applies to both members of the coord<strong>in</strong>ate phrase<br />
simultaneously and not selectively only to one of the disjuncts. This is <strong>in</strong>deed the case s<strong>in</strong>ce<br />
when attempt<strong>in</strong>g to <strong>in</strong>terpret the above examples <strong>in</strong> MG, the predisjunction will negate<br />
the first conjunct and the disjunction will simultaneously negate the second conjunct too<br />
and both conjuncts are perceived as members of the same set 26 .<br />
6.6 Analys<strong>in</strong>g Verb <strong>Agreement</strong> <strong>in</strong> Disjunctive Coord<strong>in</strong>ate <strong>Noun</strong>s<br />
We saw that there are different factors that determ<strong>in</strong>e verb agreement <strong>in</strong> disjunctive coord<strong>in</strong>ate<br />
nouns <strong>in</strong> MG, such as the <strong>in</strong>terpretation of the phrase, syntactic factors (i.e type of<br />
sentence, conjunct’s features, word order) and speaker’s specific strategies. In this section,<br />
we will propose an analysis of verb agreement with disjunctive nouns focus<strong>in</strong>g on the <strong>in</strong>terpretation<br />
of disjunction or as ‘and-coord<strong>in</strong>ate’ or as ‘exclusive-or’ (Flouraki and Kazana,<br />
2009) follow<strong>in</strong>g Eggert (2002).<br />
Eggert (2002) presents a proposal of verb agreement <strong>in</strong> disjunctive coord<strong>in</strong>ate nouns,<br />
formulated with<strong>in</strong> DRT, assum<strong>in</strong>g a similar analysis to the DRT analysis of existentials.<br />
Follow<strong>in</strong>g W<strong>in</strong>ter (1997), who uses choice function to analyse existentials <strong>in</strong> model theoretic<br />
26 The above cases also require more elaborate research with a wider range of data and speakers’ personal<br />
op<strong>in</strong>ion <strong>in</strong> order to draw more concrete conclusions.
6.6 Analys<strong>in</strong>g Verb <strong>Agreement</strong> <strong>in</strong> Disjunctive Coord<strong>in</strong>ate <strong>Noun</strong>s 197<br />
semantics, Eggert (2002) treats disjunction not as <strong>in</strong>troduc<strong>in</strong>g a discourse referent but as<br />
a choice function where “the predicate comb<strong>in</strong>es directly with the function on the set that<br />
consists of discourse referents”(107) stand<strong>in</strong>g for the two nouns. A choice function is “a<br />
function f such that for any set A=∅, f(A)∈A” (107). Thus, “a choice function is a function<br />
from a set to a member of that set”(107). This means that if disjuncts are represented as<br />
a set then or as a choice function selects one of the two disjuncts. This is presented as<br />
follows:<br />
(141) CH = {f: for all A =∅, f(A)∈A}<br />
(Eggert, 2002, 110)<br />
Consider the follow<strong>in</strong>g example which is generally problematic for most analyses of or:<br />
(142) Grant is taller than Abigail or Gertrude<br />
(Eggert, 2002, 110)<br />
If one of the disjuncts is selected, mean<strong>in</strong>g that “for some choice function f Grant is taller<br />
than f{Gertrude, Abigail}”(Eggert, 2002, 112), the DRS is as <strong>in</strong> (143):<br />
(143)<br />
x,y,z,f<br />
grant(x)<br />
abigail(y)<br />
gertrude(z)<br />
Ch(f)<br />
taller(x, f{y,z})<br />
(Eggert, 2002, 111)<br />
This represents the potential read<strong>in</strong>g e.g. ‘Grant is taller than Abigail or Gertrude-I forget<br />
which’ which is also predicted by other analyses while the choice function is existentially<br />
closed.<br />
Also, as Jenn<strong>in</strong>gs (1994) suggests or acts <strong>in</strong> the same way as the free choice any <strong>in</strong> such<br />
examples. If all/any disjuncts are selected, which is a second possible read<strong>in</strong>g, mean<strong>in</strong>g<br />
that for “all choice functions f Grant is taller than f{Gertrude, Abigail}” (Eggert, 2002,<br />
112), the DRS is as <strong>in</strong> (144):<br />
(144)<br />
x,y,z<br />
grant(x)<br />
abigail(y)<br />
gertrude(z)<br />
f<br />
Ch(f)<br />
(Eggert, 2002, 111)<br />
→ taller(x, f{y,z})<br />
This represents the read<strong>in</strong>g e.g. ‘Grant is taller than Abigail or/and Gertrude’ while choice<br />
function is universally closed.<br />
Eggert (2002) argues that the ma<strong>in</strong> disadvantage of the above analysis is that or has<br />
an amount of polysemy 27 . Thus, he proposes a modification of the analysis of or as a<br />
choice function. To account for both distributive and collective read<strong>in</strong>gs and the various<br />
<strong>in</strong>terpretations of or <strong>in</strong> a wide range of data, Eggert (2002) assumes that or is a different<br />
type of function, a subset function. First, he assumes that the universe of discourse U<br />
27 A detailed presentation of the problems of the present analysis are <strong>in</strong> Eggert (2002, 113-126).
6.6 Analys<strong>in</strong>g Verb <strong>Agreement</strong> <strong>in</strong> Disjunctive Coord<strong>in</strong>ate <strong>Noun</strong>s 198<br />
does not conta<strong>in</strong> entities, but <strong>in</strong>dividuals, which he takes to be sets of entities E. If E is<br />
a set of entities, then U is generated by E and it is the powerset of E m<strong>in</strong>us the empty<br />
set (U=P(E)-∅). He takes U to be a primitive <strong>in</strong> the model, mean<strong>in</strong>g that most one-place<br />
predicates will be functions from members of U to truth 28 . Thus, the set of one place<br />
predicates P is the doma<strong>in</strong> for terms of type 〈e,t〉 s<strong>in</strong>ce U is the doma<strong>in</strong> for terms of type<br />
〈e〉 (Eggert, 2002, 115).<br />
Follow<strong>in</strong>g Schwarzschild (1996), Eggert (2002) assumes that <strong>in</strong>dividuals (not entities)<br />
are assigned a contextually determ<strong>in</strong>ed partition on the universe of discourse <strong>in</strong> which case<br />
the relationship between arguments and predicates is not set-membership but a subset one<br />
(118). Schwarzschild (1996) suggests that the different read<strong>in</strong>gs found <strong>in</strong> a sentence (i.e.<br />
collective and distributive read<strong>in</strong>gs) are not necessarily associated to different semantic<br />
forms. Instead, he suggests that the semantics is the same for both read<strong>in</strong>gs, but the<br />
perceived difference <strong>in</strong> mean<strong>in</strong>gs comes from a contextually determ<strong>in</strong>ed variable, a partition<br />
on the universe of discourse. Follow<strong>in</strong>g Schwarzschild (1996), Eggert (2002) <strong>in</strong>troduces the<br />
follow<strong>in</strong>g relation <strong>in</strong> the DRS “for any relation r and any x,y∈U, then r(x,y) means that<br />
for some conc (i.e. union or comb<strong>in</strong>ation) of x with y,conc(x,y)⊆r” (Eggert, 2002, 132).<br />
To achieve the above assumptions, he argues that or is a subset function, mean<strong>in</strong>g that for<br />
any “non-empty set A, f(A) is a non-empty subset of A” (Eggert, 2002, 132), stated below:<br />
(145) SUB = {f: (∀ X: X = ∅) (f(X) ⊆ X ∧ f(X)= ∅)}<br />
(Eggert, 2002, 133)<br />
Eggert (2002) presents example (146), which is <strong>in</strong>terpreted <strong>in</strong> two different ways i)<br />
either Abigail notified all four, or Grant did; ii) Abigail notifie some of the four and Grant<br />
notified the rest. S<strong>in</strong>ce notify is taken to be a relation, Eggert (2002) <strong>in</strong>troduces a condition<br />
<strong>in</strong> the DRS which says ‘notify(x,y)’. Then, there must be a concatenation conc whereby<br />
conc(〚x 〛,〚y 〛)⊆〚notify〛. The possible concatenations are limited by the partitions on x<br />
and y. Tak<strong>in</strong>g y to denote the set {gertrude,jakob,beatrice,sukie}, there are three possible<br />
denotations <strong>in</strong> the present example.<br />
(146) Abigail or Grant notified Gertrude, Jacob, Beatrice, and Sukie<br />
(Eggert, 2002, 132)<br />
The DRS for the above example is the follow<strong>in</strong>g based on the present analysis:<br />
(147)<br />
t,u,v,w,x,y,f<br />
abigail(t)<br />
grant(u)<br />
gertrude(v)<br />
beatrice(w)<br />
jakob(x)<br />
sukie(y)<br />
SUB(f)<br />
notify(∪f{t,u},∪{v,w,x,y})<br />
(Eggert, 2002, 133)<br />
The possible denotations are the follow<strong>in</strong>g:<br />
(148) ∪f{t,u}: {{abigail}}, {{grant}} or {{abigail,grant}}<br />
(Eggert, 2002, 133)<br />
There are three possible denotations for example (146). If the denotation is {abigail} then<br />
the f<strong>in</strong>al condition states that for some concatenation conc:<br />
28 See Eggert (2002, 115-116, 141-147) for the two DRSs and for more details of the model he assumes.
6.6 Analys<strong>in</strong>g Verb <strong>Agreement</strong> <strong>in</strong> Disjunctive Coord<strong>in</strong>ate <strong>Noun</strong>s 199<br />
(149) conc({{abigail}},{{gertrude},{jakob},{beatrice},{sukie}}) ⊆ 〚notify〛<br />
(Eggert, 2002, 133)<br />
In fact, there is only one possible concatenation: {,,<br />
,}. For the case where the denotation is {grant},<br />
we substitute ‘abigail’ for ‘grant’ <strong>in</strong> the above formula. F<strong>in</strong>ally, if the denotation is<br />
{abigail,grant}, there are possible concatenation such as {,<br />
,,}. Accord<strong>in</strong>g to Eggert (2002),<br />
we cannot get concatenations <strong>in</strong> which {abigail,grant} is a member of an ordered pair.<br />
Also, a second example is the one <strong>in</strong> (150) repeated below, which is a problematic case<br />
for most analyses.<br />
(150) Grant is taller than Abigail or Gertrude<br />
(Eggert, 2002, 110)<br />
A s<strong>in</strong>gle DRS is proposed as <strong>in</strong> (151), which corresponds to two possible denotations.<br />
(151)<br />
x,y,z,f<br />
grant(x)<br />
abigail(y)<br />
gertrude(z)<br />
SUB(f)<br />
taller(x,∪f{y,z})<br />
(Eggert, 2002, 135)<br />
Each <strong>in</strong>terpretation depends on the assignment of f{y,z} which is as follows:<br />
(152) ∪f{y,z}: {{abigail}}, {{gertrude}} or {{abigail}, {gertrude}}<br />
(Eggert, 2002, 135)<br />
One possible <strong>in</strong>terpretation is ‘Grant is taller than Abigail or Grant is taller than Gertrude’.<br />
This read<strong>in</strong>g follows if we assume 〚f{y,z}〛={{abigail}}, which is possible s<strong>in</strong>ce {{abigail}}⊆<br />
{{abigail}, {gertrude}} or 〚f{y,z}〛={{gertrude}}, which is also possible s<strong>in</strong>ce {{gertrude}}⊆<br />
{{abigail}, {gertrude}} then two possibilities alternate:<br />
(153) conc({{grant}},{{abigail}})⊆ 〚taller〛or<br />
conc({{grant}},{{gertrude}})⊆ 〚taller〛<br />
(Eggert, 2002, 136)<br />
A second possible <strong>in</strong>terpretation, which is the preferred one <strong>in</strong> this type of example,<br />
is ‘Grant is taller than Abigail, and Grant is taller than Gertrude’ which follows<br />
if we assume 〚f({y,z})〛={{abigail},{gertrude}}. This is a possibility s<strong>in</strong>ce {{abigail},<br />
{gertrude}}⊆{{abigail},{gertrude}}. In this case the concatenation is as follows:<br />
(154) conc {, }⊆〚taller〛<br />
(Eggert, 2002, 136)<br />
The above analysis poses a s<strong>in</strong>gle DRS and determ<strong>in</strong>es semantically the different <strong>in</strong>terpretations<br />
found <strong>in</strong> disjunctively coord<strong>in</strong>ate nouns, assum<strong>in</strong>g that disjunction is a subset<br />
function of disjuncts.<br />
Our aim is to formalise the above concept <strong>in</strong> λ-DRT and Glue Semantics follow<strong>in</strong>g<br />
Dalrymple (2001) and Kokkonidis (2005)(Flouraki and Kazana, 2009). Let us consider the<br />
follow<strong>in</strong>g simple example <strong>in</strong> English which shows either s<strong>in</strong>gular or plural verb agreement<br />
depend<strong>in</strong>g on the <strong>in</strong>terpretation. The sentence is <strong>in</strong>terpreted as only one of the two is<br />
w<strong>in</strong>n<strong>in</strong>g the race (‘exclusive’ <strong>in</strong>terpretation) or out of a group of two <strong>in</strong>dividuals one of<br />
them is w<strong>in</strong>n<strong>in</strong>g (‘and-coord<strong>in</strong>ate’ <strong>in</strong>terpretation):
6.6 Analys<strong>in</strong>g Verb <strong>Agreement</strong> <strong>in</strong> Disjunctive Coord<strong>in</strong>ate <strong>Noun</strong>s 200<br />
(155) Jane or Mary is/are w<strong>in</strong>n<strong>in</strong>g.<br />
First, we will <strong>in</strong>troduce the lexical entries of the sentence above. Follow<strong>in</strong>g Kokkonidis<br />
(2005), we propose that each word conta<strong>in</strong>s <strong>in</strong>formation concern<strong>in</strong>g its syntactic specification,<br />
its compositional specification and its semantic specification. The first l<strong>in</strong>e of a lexical<br />
entry will give the syntactic category of the word and the f-structure constra<strong>in</strong>ts it comes<br />
with. The second l<strong>in</strong>e will conta<strong>in</strong> a mean<strong>in</strong>g placeholder (which is the word itself) and its<br />
compostional (Glue) type. This functions as the <strong>in</strong>terface between syntax and semantics<br />
with respect to semantic composition s<strong>in</strong>ce words have a semantic content. The third l<strong>in</strong>e<br />
of each of the lexical entries <strong>in</strong>troduces the mean<strong>in</strong>g of the word expressed <strong>in</strong> λ-DRT.<br />
Let us start with the word or. We will assume the standard mean<strong>in</strong>g constructor for or<br />
as proposed for [g-and] <strong>in</strong> Dalrymple (2001). Thus, we assume that the arguments of [or]<br />
are of type e, the type of <strong>in</strong>dividuals, and or is represented with the complex type e → (e<br />
→ e) s<strong>in</strong>ce it jo<strong>in</strong>s <strong>in</strong>dividuals. We also use the notation (↑∈)subj (Kokkonidis, 2005) for the<br />
two disjuncts s<strong>in</strong>ce both are syntactically subjects and members of the whole coord<strong>in</strong>ate<br />
phrase, which is the subject of the predicate. The lexical entry for disjunction or is the<br />
follow<strong>in</strong>g:<br />
(156) (↑conj)= ‘or’<br />
or: e (↑∈)subj ⊸ [e (↑∈)subj ⊸ e↑subj]<br />
Follow<strong>in</strong>g Eggert (2002), we treat or as a subset function, which ranges over a set of<br />
disjuncts, and the set of disjuncts is the concatenation of the members of the set while we<br />
assume that or does not <strong>in</strong>troduce a discourse referent. Thus, the mean<strong>in</strong>g assignment for<br />
or will be captured <strong>in</strong> λ-DRT as follows:<br />
(157)<br />
λx.λy<br />
f<br />
SUB(f)<br />
∪f{x,y}<br />
⊔ y ⊔ x<br />
The complete lexical entry with the syntactic representation and mean<strong>in</strong>g assignment<br />
is as follows:<br />
(158) or Conj (↑conj)= ‘or’<br />
or: e (↑∈)subj ⊸ [e (↑∈)subj ⊸ e↑subj]<br />
λx.λy<br />
f<br />
SUB(f)<br />
∪f{x,y}<br />
⊔ y ⊔ x<br />
The lexical entries for the nouns are also of type e and their lexical entries with the<br />
syntactic representation, mean<strong>in</strong>g constructors and mean<strong>in</strong>g assignment are as follows:<br />
(159) Jane PN (↑pred)= ‘Jane’<br />
Jane: e (↑∈)subj<br />
(160) Mary PN (↑pred)= ‘Mary’<br />
Mary: e (↑∈)subj<br />
λx. Jane(x)<br />
λy. Mary(y)
6.6 Analys<strong>in</strong>g Verb <strong>Agreement</strong> <strong>in</strong> Disjunctive Coord<strong>in</strong>ate <strong>Noun</strong>s 201<br />
The verbal one-place predicate is of type e → t, represented as follows:<br />
(161) w<strong>in</strong> V (↑pred)= ‘w<strong>in</strong>’<br />
w<strong>in</strong>: e (↑SUBJ) ⊸ t↑<br />
λx ′ . w<strong>in</strong>(x ′ )<br />
To derive the DRS for the whole sentence, we need to do the union of the <strong>in</strong>dividual<br />
DRSs. The process that follows shows how the DRSs are united to derive the coord<strong>in</strong>ate<br />
phrase first and then the whole sentence.<br />
First, we start with the coord<strong>in</strong>ate phrase. If we do the union of or with Jane, we get<br />
the result shown below:<br />
(162)<br />
λy.<br />
f<br />
SUB(f)<br />
∪f{x,y}<br />
⊔ y ⊔<br />
x<br />
Jane(x)<br />
The above DRS corresponds to the compositional glue type e (↑∈)subj ⊸ e↑subj s<strong>in</strong>ce only<br />
one of the two semantic resources was found and it was consumed.<br />
If we do the union of or Jane with Mary we get the follow<strong>in</strong>g result:<br />
(163)<br />
f<br />
SUB(f)<br />
∪f{x,y}<br />
⊔<br />
y<br />
Mary(y) ⊔<br />
x<br />
Jane(x)<br />
In the glue part, the second semantic resource was found and it was also consumed. Thus,<br />
the DRS above corresponds to the compositional glue type e↑subj.<br />
If we do the DRS unions, we get the follow<strong>in</strong>g DRS for the disjunctive coord<strong>in</strong>ate phrase<br />
which is of type e↑subj → t:<br />
(164)<br />
x,y,f<br />
Jane(x)<br />
Mary(y)<br />
SUB(f)<br />
∪f{x,y}<br />
Next we need to apply the verb w<strong>in</strong> to the disjunctive coord<strong>in</strong>ate phrase to get the desired<br />
result for the whole sentence:<br />
(165)<br />
x,y,f<br />
Jane(x)<br />
Mary(y)<br />
SUB(f)<br />
w<strong>in</strong>(∪f{x,y})<br />
The above DRS represents the whole coord<strong>in</strong>ate phrase which is of type t.<br />
The current approach uses λ-DRT <strong>in</strong> comb<strong>in</strong>ation with glue semantics and accounts<br />
for a simple disjunctive phrase which results <strong>in</strong> s<strong>in</strong>gular or plural verb agreement based<br />
on whether one or both disjuncts are selected. It exclusively focuses on verb agreement
6.7 Conclusion 202<br />
motivated by the <strong>in</strong>terpretation of disjunctively conjo<strong>in</strong>ed nouns. It is a rather simple<br />
analysis based on the central notions proposed by Eggert (2002), such as the assumption<br />
that arguments are treated as <strong>in</strong>dividuals assigned a contextually determ<strong>in</strong>ed partition<br />
and that or is a subset function, while it is able to account contextually for both the<br />
distributive and collective read<strong>in</strong>g of a disjunctive coord<strong>in</strong>ate phrase. Also, it borrows<br />
from Kokkonidis (2005), who comb<strong>in</strong>es λ-DRT with glue semantics, lead<strong>in</strong>g to a simple<br />
modular design without chang<strong>in</strong>g the basic concepts <strong>in</strong> the glue part and allow<strong>in</strong>g the<br />
syntactic <strong>in</strong>formation to be imported <strong>in</strong>to the mean<strong>in</strong>g representation language. Further<br />
work is required to account for disjunctive phrases with more than two disjuncts while<br />
more research is needed to <strong>in</strong>clude more complicated types than s<strong>in</strong>gle <strong>in</strong>dividuals and tie<br />
it up with syntax <strong>in</strong> due course (i.e. <strong>in</strong>dex agreement etc.).<br />
6.7 Conclusion<br />
The above prelim<strong>in</strong>ary proposal captures effectively verb agreement variation which depends<br />
on <strong>in</strong>terpretational factors. However, the data showed that verb agreement with<br />
MG disjunctive conjo<strong>in</strong>ed nouns is rather complicated simply because there are many factors<br />
<strong>in</strong>volved when speakers choose the agree<strong>in</strong>g number of the verb.<br />
Apart from the <strong>in</strong>terpretational factor, verb agreement with MG disjunctive conjo<strong>in</strong>ed<br />
nouns is determ<strong>in</strong>ed by syntactic factors and speaker’s <strong>in</strong>dividual strategies. Thus, these<br />
factors should be considered <strong>in</strong> the analysis of verb agreement with disjunctive nouns.<br />
So far even the most <strong>in</strong>termodular and flexible theoretical framework would encounter<br />
difficulties <strong>in</strong> try<strong>in</strong>g to capture the MG data. Eggert’s analysis can cover issues of syntax,<br />
semantics and pragmatics. Despite its flexibility, though, it would still face problems<br />
<strong>in</strong> account<strong>in</strong>g for number verb agreement <strong>in</strong> MG. First, it cannot predict patterns when<br />
disjuncts have different agreement features or when the structure of the sentence is different<br />
(i.e. presence or absence of a predicative adjective/participle). Also, it cannot expla<strong>in</strong><br />
when verb agreement variation depends on the type of the sentence (i.e. declarative or<br />
<strong>in</strong>terrogative) or on the subject position (i.e. preverbal or postverbal subjects).<br />
Similarly, Morgan’s analysis acknowledges that syntax, semantics and pragmatics are<br />
<strong>in</strong>volved <strong>in</strong> verb agreement with disjunctively conjo<strong>in</strong>ed nouns. Still, though, his analysis<br />
cannot expla<strong>in</strong> any of the agreement patterns <strong>in</strong> MG where verb agreement is related to<br />
differences <strong>in</strong> the agreement features of the two conjuncts, <strong>in</strong> the structure of the sentences<br />
such as the presence or absence of a predicative adjective, <strong>in</strong> the type of sentence, (i.e.<br />
declarative or <strong>in</strong>terrogative) and <strong>in</strong> the subject position. Also, Peterson’s “patch-up rules”<br />
cover a wide range of data <strong>in</strong> English and account for various patterns <strong>in</strong> MG such as CCA<br />
and when verb agreement partly depends on the type of the sentences (i.e. declarative or<br />
<strong>in</strong>terrogative). They still encounter problems, though, with cases <strong>in</strong>volv<strong>in</strong>g the <strong>in</strong>ternal<br />
structure of the sentence or cases related to the <strong>in</strong>terpretation. Hence, verb agreement<br />
with disjunctive conjo<strong>in</strong>ed nouns is complicated and it may <strong>in</strong>volve various factors rang<strong>in</strong>g<br />
from syntactic to semantic and pragmatic.<br />
Our proposal, on the other hand, does not consider any other factors apart from the<br />
pragmatic one and would face difficulties when verb agreement is determ<strong>in</strong>ed by purely<br />
syntactic factors or the different <strong>in</strong>dividual strategies chosen by native speakers. However,<br />
we believe that <strong>in</strong> MG the <strong>in</strong>terpretation of the disjunctive noun phrase is significant <strong>in</strong> determ<strong>in</strong><strong>in</strong>g<br />
verb agreement and our analysis captures effectively the variation <strong>in</strong> the agree<strong>in</strong>g<br />
verb which arises <strong>in</strong> a wide range of data. We acknowledge, though, that a more complete<br />
analysis is requred for the complicated phenomenon of verb agreement <strong>in</strong> disjunctive coord<strong>in</strong>ate<br />
nouns. In particular, we showed that number verb agreement with MG disjunctive<br />
conjo<strong>in</strong>ed nouns is determ<strong>in</strong>ed by multiple different factors. Firstly, the preference of the<br />
MG native speakers towards a resolved verb form <strong>in</strong> mixed gender conjuncts of declarative
6.7 Conclusion 203<br />
sentences and the choice of the closest conjunct agreement form <strong>in</strong> mixed gender conjuncts<br />
of preverbal or postverbal subject <strong>in</strong>terrogative sentences is certanly directly related to notions<br />
of l<strong>in</strong>earity <strong>in</strong> word order and prosodic factors. Also, the different strategies adopted<br />
by native speakers <strong>in</strong> verb agreement determ<strong>in</strong>ed by the <strong>in</strong>ternal agreement features of<br />
the conjuncts and their position <strong>in</strong> the coord<strong>in</strong>ate phrase are also cases that should be<br />
syntactically treated. F<strong>in</strong>ally, the specific type of each sentence which appears with or<br />
without a predicative adjective/participle is directly related to syntactic issues s<strong>in</strong>ce it is<br />
the presence or absence of the target that becomes apparent to the native speakers and<br />
leads to the choice of either a s<strong>in</strong>gular or a plural verb agreement form. Therefore, these<br />
issues should be seriously considered when attempt<strong>in</strong>g to analyse disjunctively conjo<strong>in</strong>ed<br />
noun phrases <strong>in</strong> MG and <strong>in</strong> other languages.<br />
To conclude, MG verb agreement <strong>in</strong> disjunctive coord<strong>in</strong>ate nouns turns out to be even<br />
more complicated than verb agreement <strong>in</strong> conjunctive coord<strong>in</strong>ate nouns. The factors that<br />
seem to lead to a specific verb agreement form are related to syntax, semantics, pragmatics<br />
or discourse factors and even prosodic factors <strong>in</strong> some cases. Our analysis focuses only on<br />
the discourse conditions, even though it would be favourable to seek a theory that would<br />
<strong>in</strong>volve all these issues. A lot of work needs to be done towards that direction <strong>in</strong> order to<br />
provide an account consider<strong>in</strong>g all the factors that are <strong>in</strong>volved <strong>in</strong> verb agreement.
Chapter 7<br />
Conclud<strong>in</strong>g Remarks<br />
The aim of the thesis was to make a contribution to the l<strong>in</strong>guistic area of agreement <strong>in</strong> noun<br />
phrase coord<strong>in</strong>ate constructions <strong>in</strong> MG. The discussion was couched with<strong>in</strong> the framework<br />
of Lexical-Functional Grammar.<br />
In agreement <strong>in</strong> MG predicate-argument relations, our ma<strong>in</strong> focus was the gender<br />
feature. We argued that gender resolution <strong>in</strong> MG is not only determ<strong>in</strong>ed by syntactic<br />
and semantic resolution but also by a third pr<strong>in</strong>ciple, which we refered to as the referential<br />
pr<strong>in</strong>ciple. MG <strong>in</strong>animate nouns follow syntactic resolution also found <strong>in</strong> <strong>in</strong>animates<br />
of other languages. MG animate nouns follow the semantic resolution pr<strong>in</strong>ciple. A number<br />
of exceptional patterns are also found <strong>in</strong> both animate and <strong>in</strong>animate nouns which<br />
follow the referential pr<strong>in</strong>ciple. This pr<strong>in</strong>ciple is determ<strong>in</strong>ed by a Contextually Introduced<br />
Referent (CIR), which functions as a superord<strong>in</strong>ate term of the two conjuncts and whose<br />
agreement features are <strong>in</strong>herited by the predicative adjective/participle. Thus, the gender<br />
(and number) of the predicative adjective/participle is derived contextually. To account<br />
for the syntactic and semantic patterns, we proposed an analysis us<strong>in</strong>g the set-based approach,<br />
which comb<strong>in</strong>es the syntactic and semantic resolution. Concern<strong>in</strong>g the referential<br />
agreement, we proposed an analysis based on the formalism of LFG. Also, we proposed an<br />
ontological hierarchy which captures the semantic relation of the coord<strong>in</strong>ate nouns with<br />
the hypernymic term. This seems to expla<strong>in</strong> the different gender of the predicative adjective/participle<br />
from the <strong>in</strong>dividual conjuncts <strong>in</strong> referential agreement.<br />
In agreement <strong>in</strong> MG head-modifier relations, we focused on the number and gender<br />
features. We exam<strong>in</strong>ed MG modifiers and the MG def<strong>in</strong>ite determ<strong>in</strong>er and we showed<br />
that they behave different. Follow<strong>in</strong>g the three agreement systems proposed by K<strong>in</strong>g and<br />
Dalrymple (2004), we argued that MG modifiers agree <strong>in</strong> number with both nouns and<br />
<strong>in</strong> gender with the closest conjunct. Thus, modifiers follow the concord system which<br />
allows a shared modifier to scope over coord<strong>in</strong>ate nouns and occurs with a jo<strong>in</strong>t and a<br />
split <strong>in</strong>terpretation. On the other hand, the MG def<strong>in</strong>ite determ<strong>in</strong>er agrees <strong>in</strong> number<br />
and gender with both conjuncts. The MG shared determ<strong>in</strong>er follows the <strong>in</strong>dex/concord<br />
system and occurs with the jo<strong>in</strong>t read<strong>in</strong>g <strong>in</strong> both s<strong>in</strong>gular and plural number when it scopes<br />
over coord<strong>in</strong>ate nouns. A second important f<strong>in</strong>d<strong>in</strong>g is that there are some exceptional<br />
patterns when a shared def<strong>in</strong>ite determ<strong>in</strong>er scopes over coord<strong>in</strong>ate nouns and admits a split<br />
<strong>in</strong>terpretation. These patterns occur only when the conjuncts are a natural coord<strong>in</strong>ation.<br />
For the MG natural coord<strong>in</strong>ation patterns, the special f-structure for naturally coord<strong>in</strong>ate<br />
nouns proposed <strong>in</strong> Dalrymple and Nikolaeva (2006) cannot provide an adequate account<br />
so we assumed the standard f-structure of accidentally coord<strong>in</strong>ate nouns where conjuncts<br />
are members of a set and the concord and <strong>in</strong>dex features do not match. Thus, the<br />
concord/<strong>in</strong>dex system effectively accounts for these patterns too.<br />
F<strong>in</strong>ally, <strong>in</strong> verb agreement with disjunctive coord<strong>in</strong>ate nouns our ma<strong>in</strong> focus was number.<br />
We argued that number verb agreement variation <strong>in</strong> MG depends on the <strong>in</strong>terpre-
205<br />
tation of disjunction as an ‘and-coord<strong>in</strong>ate’ or as an ‘exclusive’ or. Hence, we proposed an<br />
analysis of verb agreement <strong>in</strong> λ-DRT, a discourse model comb<strong>in</strong>ed with λ-calculus, follow<strong>in</strong>g<br />
Eggert (2002) who treats or as a subset function. This analysis captures the patterns<br />
<strong>in</strong> question but it needs further improvements to extend to more complicated disjunctive<br />
coord<strong>in</strong>ate phrases. However, we argued that verb agreement <strong>in</strong> MG disjunctive coord<strong>in</strong>ate<br />
nouns is the result of the effect of other factors too, such as the <strong>in</strong>ternal structure of<br />
a sentence or the type of sentence, the different features of the conjuncts and f<strong>in</strong>ally the<br />
subject position. Thus, more research is required to provide a solution that will <strong>in</strong>volve all<br />
these factors.<br />
There are, however, some rema<strong>in</strong><strong>in</strong>g issues for further research that we need to outl<strong>in</strong>e.<br />
First, it is the status of the r-structure <strong>in</strong> referential agreement <strong>in</strong> predicate-argument<br />
relations. We showed that LFG accounts for these patterns by <strong>in</strong>troduc<strong>in</strong>g a specific lexical<br />
entry where the predicative adjective agrees syntactically with the noun that is present <strong>in</strong><br />
the phrase or referentially with the Contextually Introduced Referent, which occurs only at<br />
the contextual level and not at the syntactic level. The r-structure of the CIR, which occurs<br />
at the contextual level, has a syntactic representation s<strong>in</strong>ce the syntactic features of the<br />
CIR are required to be copied by the predicative adjective/participle and this might be a<br />
shortcom<strong>in</strong>g <strong>in</strong> our analysis. Thus, further work needs to be done towards the structure of<br />
this extra level that relates to context. A second important issue is the ontological relation<br />
between the CIR and the two coord<strong>in</strong>ate nouns. The CIR is the superord<strong>in</strong>ate term and<br />
the coord<strong>in</strong>ate nouns are the hyponyms. In order to capture this relation, we used an<br />
ontological hierarchy, based on an ontological approach. An important question is whether<br />
the <strong>in</strong>formation of the CIR and its relation to the conjuncts can be captured at a different<br />
level with<strong>in</strong> LFG, which is a more abstract level. It could be possible, however, that LFG<br />
does not have the adequate mechanisms to provide such k<strong>in</strong>d of abstract <strong>in</strong>formation like<br />
other theoretical frameworks. These issues need more work and are directly related to the<br />
<strong>in</strong>ternal structure of the framework of LFG.<br />
Second, <strong>in</strong> head-modifier agreement <strong>in</strong> MG coord<strong>in</strong>ate nouns, we showed that the MG<br />
def<strong>in</strong>ite determ<strong>in</strong>er is syntactically captured by the <strong>in</strong>dex/concord system. The MG<br />
def<strong>in</strong>ite determ<strong>in</strong>er imposes a further semantic restriction concern<strong>in</strong>g the <strong>in</strong>terpretation<br />
of a shared def<strong>in</strong>ite determ<strong>in</strong>er when it scopes over coord<strong>in</strong>ate nouns. In particular, we<br />
showed that the jo<strong>in</strong>t <strong>in</strong>terpretation occurs <strong>in</strong> s<strong>in</strong>gular and plural accidentally coord<strong>in</strong>ate<br />
nouns but the split <strong>in</strong>terpretation occurs only <strong>in</strong> plural naturally coord<strong>in</strong>ate nouns. Thus,<br />
despite the fact that the <strong>in</strong>dex/concord system accounts for the MG def<strong>in</strong>ite determ<strong>in</strong>er,<br />
the latter imposes further specifications that are related to the field of semantics. S<strong>in</strong>ce<br />
semantics is not the area of exam<strong>in</strong>ations these may be <strong>in</strong>vestigated <strong>in</strong> the future. A<br />
second important issue is that <strong>in</strong> MG natural coord<strong>in</strong>ation cases the two nouns cooccur<br />
under specific semantic conditions s<strong>in</strong>ce they need to be semantically related. In LFG, this<br />
type of relation is not expressed <strong>in</strong> the analysis of natural coord<strong>in</strong>ation, but s<strong>in</strong>ce it plays a<br />
significant role <strong>in</strong> natural coord<strong>in</strong>ation patterns of some languages it should be given some<br />
attention. Nevertheless, it requires further work s<strong>in</strong>ce it <strong>in</strong>volves the <strong>in</strong>ternal structure of<br />
the framework itself.<br />
F<strong>in</strong>ally, the proposal <strong>in</strong> verb agreement with disjunctive coord<strong>in</strong>ate nouns captures the<br />
<strong>in</strong>terpretation of or as ‘and-coord<strong>in</strong>ate’ and as ‘exclusive-or’ and allows both a s<strong>in</strong>gular or<br />
a plural verb agreement form. However, it is a prelim<strong>in</strong>ary analysis focus<strong>in</strong>g on <strong>in</strong>dividuals<br />
of type e only and we need to work further on its modification so as to extend to other<br />
types of disjunctive phrases, such as quantified coord<strong>in</strong>ate nouns or verb phrases. Also, we<br />
saw that other factors are <strong>in</strong>volved <strong>in</strong> disjunctive coord<strong>in</strong>ate nouns which probably need<br />
to be <strong>in</strong>corporated <strong>in</strong>to the analysis of disjunctively coord<strong>in</strong>ate nouns. Hence, the analysis<br />
provided so far needs to be extended <strong>in</strong> order to overcome the above problems and to cover<br />
a wider range of data <strong>in</strong> other languages which may have similar behaviour. These issues
will be left for future research.<br />
206
Appendix A<br />
Questionnaires on Predicate<br />
Argument <strong>Agreement</strong><br />
A.1 Discussion of questionnnaires<br />
The data analysis was based on four different questionnaires that were developed <strong>in</strong> order<br />
to test the gender resolution <strong>in</strong> the same gender or mixed gender animate and <strong>in</strong>animate<br />
coord<strong>in</strong>ate nouns. The method used to draw the results of the gender values <strong>in</strong> both<br />
groups of nouns is based on the responses of 26 <strong>Greek</strong> native speakers who answered the<br />
four different questionnaires. The participants are all university graduates from North and<br />
South Greece. The different geographical orig<strong>in</strong>s of the participants did not play any active<br />
role. The questionnaires appeared <strong>in</strong> different forms such as ‘multiple choice’ and ‘fill <strong>in</strong><br />
the gap’. The consultants filled <strong>in</strong> the questionnaires based on the choice of the form they<br />
found most appropriate. They were also asked to choose a second possibility if they found<br />
it acceptable.<br />
Some issues need to be discussed regard<strong>in</strong>g questions that were raised about the questionnaires<br />
and how we approached these. The first issue is the type of the questionnaire<br />
that would be given to the participants. We decided to issue four different questionnaires;<br />
the first three appeared <strong>in</strong> the form of a ‘multiple choice’ where each sentence conta<strong>in</strong>ed<br />
three different choices with regard to the gender feature. The participants were asked to<br />
give one or two possible answers whenever they found it appropriate. The fourth questionnaire<br />
appeared <strong>in</strong> the form of ‘fill <strong>in</strong> the gaps’. In this questionnaire the participants were<br />
asked to respond choos<strong>in</strong>g the best possible answer. Most participants were will<strong>in</strong>g to give<br />
more than one answer. The reason the fourth questionnaire was different was to confirm<br />
some of the responses of the consultants.<br />
The second issue concerned the fact that we needed to check the resolution genders<br />
<strong>in</strong> both animate and <strong>in</strong>animate nouns and <strong>in</strong> s<strong>in</strong>gular and plural conjuncts. We decided<br />
to <strong>in</strong>clude only sentences with s<strong>in</strong>gular animate and <strong>in</strong>animate coord<strong>in</strong>ate nouns <strong>in</strong> the<br />
first questionnaire and sentences with plural animate and <strong>in</strong>animate coord<strong>in</strong>ate nouns <strong>in</strong><br />
the second questionnaire. The other two questionnaires <strong>in</strong>cluded mixed sentences either<br />
with s<strong>in</strong>gular animate or <strong>in</strong>animate coord<strong>in</strong>ate nouns or with plural animate or <strong>in</strong>animate<br />
coord<strong>in</strong>ate nouns.<br />
Another issue concerned the time that the questionnaires should be given to the consultants<br />
s<strong>in</strong>ce we wanted to leave a short period of time <strong>in</strong> between the first two questionnaires<br />
and the next two questionnaires. Our aim was to test the consistency of the responses of<br />
the participants. Thus, we issued only the first two questionnaires to the consultants and<br />
then we issued the other two questionnaires after some time to the same consultants <strong>in</strong><br />
order to check whether their responses were consistent. Also, some of the sentences were<br />
repeated <strong>in</strong> the last two questionnaires <strong>in</strong> order to see whether the same form had been
A.2 Predicate Argument <strong>Agreement</strong> sets of examples 208<br />
chosen by the consultants previously. F<strong>in</strong>ally, the ma<strong>in</strong> issue was to avoid controll<strong>in</strong>g the<br />
responses of the participants. The best way to adm<strong>in</strong>ister that was to provide the consultants<br />
with different types of questionnaires and without necessarily provid<strong>in</strong>g a choice of<br />
forms. Therefore, the first three questionnaires <strong>in</strong>cluded sentences where the consultants<br />
had to choose one or two possible forms out of the three forms provided, and <strong>in</strong> the fourth<br />
questionnaire the consultants had to th<strong>in</strong>k and decide themselves the form they would<br />
choose to fill <strong>in</strong> the gap.<br />
The problems that could be found <strong>in</strong> the above questionnaires are summarized as follows.<br />
Some of the examples were repeated especially <strong>in</strong> the third and fourth questionnaire.<br />
Possibly a wider range of examples should be <strong>in</strong>cluded to avoid the problem of repetition.<br />
Second, we presented sentences with animate coord<strong>in</strong>ate nouns separately from sentences<br />
with <strong>in</strong>animate coord<strong>in</strong>ate nouns <strong>in</strong> the first two questionnaires and this separation may<br />
have the effect of controll<strong>in</strong>g the responses of the consultants to some extent. In order<br />
to m<strong>in</strong>imise the effect of controll<strong>in</strong>g their responses, we issued the last two questionnaires<br />
where we presented a mix of sentences with animate and <strong>in</strong>animate coord<strong>in</strong>ate nouns.<br />
F<strong>in</strong>ally, the sentences did not appear <strong>in</strong> a specific context and the responses of the<br />
participants were based only on the sentence they were presented with. However, <strong>in</strong> the<br />
sentences where we exam<strong>in</strong>ed the phenomenon of gender resolution this is not an important<br />
problem s<strong>in</strong>ce we do not focus on the <strong>in</strong>terpretation of the phrase but rather on the resolved<br />
gender form which is derived based on the gender of the two coord<strong>in</strong>ate nouns that are<br />
present <strong>in</strong> the phrase.<br />
A.2 Predicate Argument <strong>Agreement</strong> sets of examples<br />
1. Father.m and mother.f are tired. Their life is difficult and demand<strong>in</strong>g<br />
2. Kostas.m and the baby.n are ready to depart<br />
3. Kostas.m and the girl.n are ready to depart<br />
4. Mary.f and her girl.n are enthusiastic about their new life<br />
5. Mary.f and the boy/baby.n are happy<br />
6. Mary.f and her baby.nf are beautiful. You see them and you admire them<br />
7. Mary.f and the person.n with the skirt are weird<br />
8. The child.n and the baby.n were locked <strong>in</strong> the room<br />
9. The boy.n and the girl.n were seated at the back<br />
10. The men.m and the women.f are ready<br />
11. The boys.n and girls.n are happy<br />
12. The fathers.f and boys.n are very close<br />
13. The grandmothers.f and grandchildren.n are happy<br />
14. The children.n and mothers.f are happy<br />
15. The grandmothers.f and girls.n are happy<br />
16. The girls.n and the mothers.f are very close<br />
17. Heroism.m and fight<strong>in</strong>g.m were prevalent dur<strong>in</strong>g 1821<br />
28. The picture.m and the sofa.m are big for this room and I will put them <strong>in</strong> the other<br />
19. The fight.m and self-sacrifice.f were prevalent <strong>in</strong> 1821<br />
20. Death.m and freedom.f are important<br />
21. Freedom.f and death.m were important <strong>in</strong> 1821<br />
22. The sofa.m and the table.n are white<br />
23. The bookcase.f and d<strong>in</strong><strong>in</strong>g-table.f are different<br />
24. Freedom.f and faith.f are equal<br />
25. Effort.f and obst<strong>in</strong>ancy.n have good results<br />
26. Love.f and passion.n are necessary <strong>in</strong> our life<br />
27. The sofa.m and the armchair.f are comfortable
A.3 Tables 209<br />
28. The street.m and the square.f are clean<br />
29. The pictures.m and sofas.m are unsuitable<br />
30. The philosophies.f and religions.f were determ<strong>in</strong><strong>in</strong>g for the people<br />
31. The houses.n and cars.n are more expensive from now on<br />
32. The roads.m and the squares.f are full of people<br />
33. The squares.f and roads.m are full of people<br />
34. The roads.m and the alleys.n are full of people<br />
35. The hurricanes.m and floods.f were destructive for Tailand<br />
36. The protests.f and compla<strong>in</strong>ts.n are useless <strong>in</strong> this case<br />
A.3 Tables<br />
Table A.1: List of consultants of questionnaires 1,2,3,4<br />
Consultant Number: 26<br />
Consultant Number age gender education orig<strong>in</strong><br />
c1 37 F BA,MA,PhD Athens<br />
c2 35 F BA Thessaloniki<br />
c3 34 F BA,MA,PhD Crete<br />
c4 35 M BA,MA,PhD Volos<br />
c5 31 F BA Thessaloniki<br />
c6 33 F BA Thessaloniki<br />
c7 25 F BA Thessaloniki<br />
c8 32 F BA,MA,PhD Athens<br />
c9 31 M BA,MA Thessaloniki<br />
c10 55 F BA Thessaloniki<br />
c11 60 M BA Thessaloniki<br />
c12 24 F BA Thessaloniki<br />
c13 36 M BA,MA Thessaloniki<br />
c14 30 M BA,MA Thessaloniki<br />
c15 32 M BA,MA Thessaloniki<br />
c16 36 F BA,MA Orestiada<br />
c17 30 F BA,MA,PhD Orestiada<br />
c18 31 M BA Thessaloniki<br />
c19 27 M BA,MA Thessaloniki<br />
c20 62 F BA Thessaloniki<br />
c21 36 F BA,MA Thessaloniki<br />
c22 32 F BA,MA,PhD Crete<br />
c23 31 M BA,MA Peloponese<br />
c24 24 F BA Thessaloniki<br />
c25 24 F BA Thessaloniki<br />
c26 24 F BA Thessaloniki
A.3 Tables 210<br />
Table A.2: Results of crucial sets of examples<br />
Consultant Number: 26<br />
examples expected unexpected<br />
No. m f n m f n<br />
1 26 - - - - -<br />
c1-c26<br />
2 26 - - - - -<br />
c1-c26 - - - - -<br />
3 26 - - - - -<br />
c1-c26 - - - - -<br />
4 5 21 - - - -<br />
c4,c14,c18 c1,c2,c3,c5,c6,c7,<br />
c20,c24 c8,c9,c10,c11,c12<br />
c13,c15,c16,c17,c19<br />
c21,c22,c23,c25,c26<br />
5 26 - - - - -<br />
c1-c26 - - - - -<br />
6 5 21 - - - -<br />
c4,c15,c18 c1,c2,c3,c5,c6,c7,<br />
c20,c26 c8,c9,c10,c11,c12<br />
c13,c14,c16,c17,c19<br />
c21,c22,c23,c24,c25<br />
7 - 26 - - - -<br />
c1-c26 - - - -<br />
8 - - 26 - - -<br />
c1-c26<br />
9 - - 26 - - -<br />
c1-c26<br />
10 26 - - - - -<br />
c1-c26 - - - - -<br />
11 - - 26 - - -<br />
- - c1-c26 -<br />
12 26 - - - - -<br />
c1-c26 - - -<br />
13 19 - - - - 12<br />
c2,c4,c5,c6,c7, c1,c3,c6<br />
c9,c11,c12,c13, c8,c9,c10<br />
c14,c16,c18,c19 c15,c16,<br />
c20,c21,c22,c23, c17,c20<br />
c25,c26 c24,c26<br />
14 19 - - - - 7<br />
c2,c4,c5,c6,c7, c1,c3,c8,<br />
c9,c11,c12,c13, c10,c15,<br />
c14,c16,c18,c19 c17,c24<br />
c20,c21,c22,c23,<br />
c25,c26<br />
15 5 18 - - - 5<br />
c4,c14,c18 c2,c5,c6,c7 c1,c3,c8<br />
c20,c24 c8,c9,c10 c17,c24<br />
c11,c12,c13<br />
c15,c16,c17<br />
c19,c21,c22<br />
c23,c25,c26<br />
16 5 18 - - - 5<br />
c4,c15,c18 c2,c5,c6,c7, c1,c3,c8<br />
Cont<strong>in</strong>ued on next page
A.3 Tables 211<br />
Table A.2: Results of crucial sets of examples<br />
Consultant Number: 26<br />
examples expected unexpected<br />
No. m f n m f n<br />
c20,c26 c8,c9,c10, c10,c24<br />
c11,c12,c13,<br />
c14,c16,c17,<br />
c19,c21,c22,<br />
c23,c24,c25<br />
17 25 - - - - 10<br />
c1,c2,c3,c6 c3,c4,c5<br />
c7,c8,c9 c7,c8,c9<br />
,c10,c11,c12 c10,c11,<br />
c13,c14,<br />
c15,c16,c17, c12,c13<br />
c18,c19,c20,<br />
c21,c22,c23,<br />
c24,c25,c26<br />
18 25 - - - - 10<br />
c1,c3,c4 c2,c4,c5,<br />
c6,c7,c8 c7,c8,c9<br />
c9,c10,c11<br />
c12,c13,c14 c10,c11<br />
c15,c16,c17 c20,c24<br />
c18,c19,c20<br />
,c21,c22,c23<br />
c24,c25,c26<br />
19 - - 16 6 6 -<br />
c2,c3,c5,c6 c7,c8 c1,c4,<br />
c7,c8,c9,c11 c12,c14 c10,c16<br />
c13,c17,c19 c15,c18 c21,c24<br />
c20,c22,c23<br />
c25,c26<br />
20 - - 16 6 6 -<br />
c2,c3,c5,c6 c7,c8 c1,c4<br />
c7,c8,c9,c12 c11,c14, c10,c19<br />
c13,c16,c17 c15,c18 c21,c24<br />
c20,c22,c23<br />
c25,c26<br />
21 - - 18 5 5 -<br />
c2,c3,c5,c6,c7 c7,c8 c1,c4<br />
c8,c9,c11,c13 c12,c14 c10,<br />
c15,c16,c17, c18 c21<br />
c19,c20,c22, c24<br />
c23,c25,c26<br />
22 - - 26 - - -<br />
- - c1-c26 -<br />
23 - 24 - - - 11<br />
c1,c2,c3,c5,c6,c7 c3,c4,c7<br />
c8,c9,c10,c11,c13 c8,c9<br />
c14,c15,c16,c17 c10,c11<br />
c18,c19,c20,c21 c12,c13<br />
c22,c23,c24,c25 c20,c24<br />
c26<br />
24 - 24 - - - 11<br />
c1,c2,c3,c5,c6,c7 c3,c4,c5<br />
Cont<strong>in</strong>ued on next page
A.3 Tables 212<br />
Table A.2: Results of crucial sets of examples<br />
Consultant Number: 26<br />
examples expected unexpected<br />
No. m f n m f n<br />
c8,c9,c10,c11,c12 c8,c9<br />
c14,c15,c16,c17 c10,c11<br />
c18,c19,c20,c21 c12,c13<br />
c22,c23,c24,c25 c20,c24<br />
c26<br />
25 - - 26 - - -<br />
- - c1-c26 -<br />
26 - - 26 - - -<br />
- - c1-c26 -<br />
27 - - 14 12 - -<br />
c1,c2,c6, c3,c4,c5<br />
c7,c12,c14, c8,c9,c10,<br />
c15,c16,c17, c11,c13,c21<br />
c18,c19,c22 c20,c24,c26<br />
c23,c25<br />
28 - - 14 12 -<br />
c1,c2,c6,c7 c3,c4,c5<br />
c12,c14, c8,c9,c10,<br />
c15,c16,c17, c11,c13,c21,<br />
c18,c19,c22, c20,c24,c26<br />
c23,c25<br />
29 26 - - - - -<br />
c1-c26 - - -<br />
30 - 26 - - - -<br />
c1-c26 - -<br />
31 - - 26 - - -<br />
- - c1-c26 -<br />
32 - - 13 15 6 -<br />
c1,c2,c6,c7, c3,c4,c5,c8, c1,c2<br />
c14,c15,c16 c9,c10,c11, c7,c18,<br />
c17,c18,c19, c12,c13,c14, c19,c22<br />
c22,c23,c25 c17,c20,c21,<br />
c24,c26<br />
33 - - 11 17 4 -<br />
c1,c2,c6, c3,c4,c5,c7,c8 c1,c2,<br />
c14,c15,c16 c9,c10,c11,c12 c18,c19,<br />
c17,c18,c19, c13,c14,c17<br />
c22,c23 c20,c21,c22,<br />
c24,c25,c26<br />
34 - - 26 - - -<br />
- - c1-c26 -<br />
35 - - 13 15 6 -<br />
c1,c2,c6,c7, c3,c4,c5,c8, c1,c2,<br />
c14,c15,c16 c9,c10,c11 c7,c18<br />
c17,c18,c19, c12,c13,c14, c19,c22<br />
c22,c23,c25 c17,c20,c21<br />
c24,c26<br />
36 - - 26 - - -<br />
- - c1-c26 -
Appendix B<br />
Questionnaires on agreement with<br />
Disjunctively Conjo<strong>in</strong>ed <strong>Noun</strong>s<br />
B.1 Discussion of questionnnaires<br />
The data analysis is based on questionnaires that were developed <strong>in</strong> order to test the<br />
prevalent verb agreement form <strong>in</strong> disjunctively coord<strong>in</strong>ate nouns. The aim is to explore<br />
the different <strong>in</strong>terpretations assigned to the coord<strong>in</strong>ate structures and the syntactic factors<br />
that determ<strong>in</strong>e verb agreement. Two different questionnaires were produced for that. The<br />
first questionnaire focused on the choice of verb agreement related to the <strong>in</strong>terpretation of<br />
the disjunctively conjo<strong>in</strong>ed nouns and on the choice of verb agreement related to syntactic<br />
factors, it consists of both declarative and <strong>in</strong>terrogative sentences and it <strong>in</strong>cludes sentences<br />
with s<strong>in</strong>gular disjuncts of the same or different gender and of the same or different person<br />
while the second part consists of mixed number, gender and person disjuncts. The<br />
second questionnaire <strong>in</strong>cludes sentences with the positive predisjunction and disjunction<br />
i...i/ite...ite ‘either...or’ and the negative predisjunction and disjunction oute...oute ‘neither...nor’<br />
with s<strong>in</strong>gular disjuncts of the same or different gender and person and mixed<br />
number disjuncts also of the same or different gender and person. The questionnaires were<br />
issued to 20 <strong>Greek</strong> native speakers who are all university graduates from different areas <strong>in</strong><br />
Greece. The participants were asked to make a choice from three possible verb forms and<br />
were asked to consider two verb forms whenever they found it appropriate.<br />
Some issues need to be discussed regard<strong>in</strong>g questions that were raised about the questionnaires<br />
on disjunctively conjo<strong>in</strong>ed nouns. The first issue concerns the type of the questionnaire<br />
that would be given to the participants. We decided to issue two different questionnaires<br />
which appeared <strong>in</strong> the form of a ‘multiple choice’. Consultants were given three<br />
different choices and were asked to choose one or two possible answers. The second issue<br />
concerns how we would <strong>in</strong>vestigate any different semantic or syntactic phenomena. The<br />
first questionnaire <strong>in</strong>cludes sentences which were related to the <strong>in</strong>terpretation of the disjunctively<br />
conjo<strong>in</strong>ed phrase; these sentences consisted of s<strong>in</strong>gular disjuncts of the same or<br />
different gender and of the same or different person. Also, it <strong>in</strong>cludes sentences with a<br />
s<strong>in</strong>gular and a plural disjunct that <strong>in</strong>vestigates the speakers strategies, and sentences with<br />
s<strong>in</strong>gular disjuncts but with different person and different gender that <strong>in</strong>vestigates other syntactic<br />
factors (i.e. mixed person or gender and verb agreement). The second questionnaire<br />
<strong>in</strong>cludes sentences with the positive predisjunction and disjunction i...i/ite...ite ‘either...or’<br />
and the negative predisjunction and disjunction oute...oute ‘neither...nor’ with s<strong>in</strong>gular disjuncts<br />
of the same or different gender and person, and mixed number disjuncts also of the<br />
same or different gender and person. The aim was to test any difference between the bare<br />
disjunction and the predisjunction and disjunction, and the negative predisjunction and<br />
disjunction <strong>in</strong> relation to verb agreement. A third issue was raised <strong>in</strong> the first questionnaire
B.2 Disjunctive Coord<strong>in</strong>ate <strong>Noun</strong>s sets of examples 214<br />
which focused on the <strong>in</strong>terpretation of the disjunctively conjo<strong>in</strong>ed nouns. In order to make<br />
sure how the consultants <strong>in</strong>terpreted the disjunctively conjo<strong>in</strong>ed nouns they were asked to<br />
give further <strong>in</strong>formation on how they would <strong>in</strong>terpret the particular disjunctive coord<strong>in</strong>ate<br />
phrase.<br />
Some problems arise <strong>in</strong> these questionnaires. Sentences did not appear with<strong>in</strong> a specific<br />
context as would be desirable and the responses of the participants were based only on the<br />
sentence they were presented with. However, s<strong>in</strong>ce the issue of <strong>in</strong>terpretation is crucial<br />
we asked the consultants to give their reason for the choice of the specific verb form by<br />
provid<strong>in</strong>g their personal <strong>in</strong>terpretation of the sentence, such as whether it <strong>in</strong>volves both<br />
coord<strong>in</strong>ate nouns or a s<strong>in</strong>gle noun. A second problem is that <strong>in</strong> disjunctively conjo<strong>in</strong>ed<br />
nouns we discussed both <strong>in</strong>terpretational factors and syntactic factors and therefore a<br />
wider range of examples should be <strong>in</strong>cluded to have more accurate results. We need to<br />
note, though, that this was only a prelim<strong>in</strong>ary research which can be expanded further<br />
<strong>in</strong> the future <strong>in</strong> order to test accurately both the pragmatic factors and the syntactic<br />
factors. Also, some of the data were repeated <strong>in</strong> both questionnaires. Probably a wider<br />
range of sentences would give more precise and exact results concern<strong>in</strong>g the outcome of the<br />
research. F<strong>in</strong>ally, a major disadvantage is the fact that the participants who responded to<br />
the follow<strong>in</strong>g questionnaires were around 20, although the questionnaires were orig<strong>in</strong>ally<br />
sent to around 30 people. This was a serious problem s<strong>in</strong>ce a larger number of participants<br />
would help us make out possible differences on the way the language is used and draw more<br />
concrete conclusions concern<strong>in</strong>g our results.<br />
B.2 Disjunctive Coord<strong>in</strong>ate <strong>Noun</strong>s sets of examples<br />
1. Kostas or Maria will pick me up with the car<br />
2. The boy or the girl tries to suppress the feel<strong>in</strong>gs<br />
3. The woman or child have priority for the vacc<strong>in</strong>e aga<strong>in</strong>st flu<br />
4. The boy or the girl are tired from study<strong>in</strong>g<br />
5. Kostas or Giorgos is Maria’s relative<br />
6. The doctor or dentist can write prescriptions<br />
7. The male-student or female-student that is certa<strong>in</strong> about every choice will succeed<br />
8. The male-student or female-student can be <strong>in</strong>formed by the parents<br />
9. The son or the daughter need help<br />
10. Is the mother or the boy gett<strong>in</strong>g ready to go out<br />
11. The mother or the boy ready to go out<br />
12. Me or you will need help<br />
13. Me or Kostas will over the job<br />
14. Will you or Maria go to the doctor?<br />
15. The car or the mopeds will be-sold soon<br />
16. Your cats or your dog ate my owers<br />
17. The footballers or the coach earn much money<br />
18. Are your brother or your sisters settled?<br />
19. Is/are your sisters or your brother settled?<br />
20. The father or the daughters are responsible for the company<br />
21. The daughters or the father are responsible for the company?<br />
22. Are the boys or the girl gett<strong>in</strong>g prepared for the exams<br />
23. Are the girl or the boys gett<strong>in</strong>g prepared for the exams<br />
24. Either my son or my father will take part <strong>in</strong> the contest<br />
25. Either the mother or the aunt will take over the custody<br />
26. Either the father or the mother are responsible for the child’s custody<br />
27. Either my brother or my sister will enter the university
B.3 Tables 215<br />
28. Either me or you will take over the bus<strong>in</strong>ess<br />
29. Either you or the father will sign the papers<br />
30. Do either you or Maria have the responsibility for the job<br />
31. Neither the boy nor the girl are ready for this change<br />
32. Neither the father nor the mother are happy with this situation<br />
33. Neither the man nor the woman compromise nowadays<br />
34. Neither you nor he will accept the offer<br />
B.3 Tables<br />
Table B.1: List of consultants for questionnaires 1,2<br />
Consultant Number: 20<br />
Consultant Number age gender education orig<strong>in</strong><br />
c1 60 F - Thessaloniki<br />
c2 32 M BA,MA Thessaloniki<br />
c3 54 F - Thessaloniki<br />
c4 32 F BA Thessaloniki<br />
c5 33 F BA,MA,PhD Crete<br />
c6 32 F BA Thessaloniki<br />
c7 31 M BA,MA Patras<br />
c8 60 M BA Thessaloniki<br />
c9 35 F BA,MA Thessaloniki<br />
c10 35 F - Thessaloniki<br />
c11 40 M BA,MA Thessaloniki<br />
c12 32 F BA,MA Messolongi<br />
c13 26 F BA,MA Thessaloniki<br />
c14 36 F BA,MA,PhD Athens<br />
c15 30 F BA,MA,PhD Orestiada<br />
c16 35 M BA,MA,PhD Volos<br />
c17 27 M BA,MA Thessaloniki<br />
c18 32 F BA Athens<br />
c19 32 F BA,MA Thessaloniki<br />
c20 32 F BA,MA Thessaloniki<br />
Table B.2: Results of crucial sets of examples<br />
Consultant Number: 20<br />
examples sg pl both<br />
1 14 4 2<br />
c1,c2,c3,c4 c8,c11, c17,c19<br />
c5,c6,c7,c9 c16,c20<br />
c10,c12,c13<br />
c14,c15,c18<br />
Cont<strong>in</strong>ued on next page
B.3 Tables 216<br />
Table B.2: Results of crucial sets of examples<br />
Consultant Number: 20<br />
examples sg pl both<br />
2 15 3 2<br />
c1,c2,c3,c5 c8,c11,c16 c4,c12<br />
c6,c7,c9,c10<br />
c13,c14,c15,c17<br />
c18,c19,c20<br />
3 7 13 -<br />
c5,c6,c11,c12 c1,c2,c3,c4<br />
c13,c14,c15 c7,c8,c9,c10<br />
c16,c17,c18<br />
c19,c20<br />
4 5 15 -<br />
c5,c6,c11 c1,c2,c3,c4<br />
c12,c13 c7,c8,c9,c10<br />
c14,c15,c16,<br />
c17,c18,c19,c20<br />
5 20 - -<br />
c1-c20<br />
6 - 20 -<br />
c1-c20<br />
7 4 12 4<br />
c6,c7, c1,c2,c3,c4 c17,c18<br />
c14,c15 c5,c8,c9,c10 c19,c20<br />
c11,c12,c13,<br />
c16<br />
8 7 2 11<br />
c9,c12,c13,c14 c3,c6 c1,c2,c4,c5<br />
c15,c18,c20 c7,c8,c9,c10<br />
c11,c16,c17,c19<br />
9 7 2 11<br />
c9,c12,c13,c14 c3,c6 c1,c2,c4,c5<br />
c15,c18,c20 c7,c8,c9,c10<br />
c11,c16,c17,c19<br />
10 9 4 7<br />
c2,c4,c5,c6 c1,c3,c8,c10 c11,c12,c13,c17<br />
c7,c9,c14,c15 c18,c19,c20<br />
c16<br />
11 3 13 4<br />
c9,c10,c14 c1,c2,c3,c4 c11,c12,c13,c17<br />
c5,c6,c7,c8<br />
c15,c16,c18<br />
c19,c20<br />
12 3 17 -<br />
c6,c7,c9 c1,c2,c3,c4<br />
c5,c8,c10,c11<br />
c12,c13,c14,c15<br />
c16,c17,c18,c19<br />
Cont<strong>in</strong>ued on next page
B.3 Tables 217<br />
Table B.2: Results of crucial sets of examples<br />
Consultant Number: 20<br />
examples sg pl both<br />
c20<br />
13 3 17 -<br />
c6,c7,c9 c1,c2,c3,c4<br />
c5,c8,c10,c11<br />
c12,c13,c14,c15<br />
c16,c17,c18,c19<br />
c20<br />
14 11 9 -<br />
c2,c3,c4,c5 c1,c8,c10,c11<br />
c6,c7,c9,c13 c12,c17,c18,c19<br />
c14,c15,c16 c20<br />
15 - 20 -<br />
c1-c20<br />
16 4 15 1<br />
c6,c8,c12,c17 c1,c2,c3,c4 c19<br />
c5,c7,c9,c10<br />
c11,c13,c14,c15<br />
c16,c18,c20<br />
17 3 15 2<br />
c2,c4,c6 c1,c3,c5,c7 c17,c19<br />
c8,c9,c10,c11<br />
c12,c13,c14,c15<br />
c16,c18,c20<br />
18 7 13 -<br />
c12,c13,c14 c1,c2,c3,c4<br />
c16,c17,c18,c19 c5,c6,c7,c8<br />
c9,c10,c11<br />
c15,c20<br />
19 4 16 -<br />
c6,c9,c10,c11 c1,c2,c3,c4<br />
c5,c7,c8,c12<br />
c13,c14,c15,c16<br />
c17,c18,c19,c20<br />
20 7 13 -<br />
c8,c9,c10,c11 c1,c2,c3,c4<br />
c14,c16,c17 c5,c6,c7,c12<br />
c13,c15,c18<br />
c19,c20<br />
21 4 16 -<br />
c9,c10,c11,c16 c1,c2,c3,c4<br />
c5,c6,c7,c8<br />
c12,c13,c14<br />
c15,c17,c18<br />
c19,c20<br />
22 - 20 -<br />
c1-c20<br />
Cont<strong>in</strong>ued on next page
B.3 Tables 218<br />
Table B.2: Results of crucial sets of examples<br />
Consultant Number: 20<br />
examples sg pl both<br />
23 1 11 8<br />
c1 c2,c3,c7,c12 c4,c5,c6,c8<br />
c13,c14,c16 c9,10,c11,c15<br />
c17,c18,c19,c20<br />
24 14 6 -<br />
c1,c2,c3,c4 c6,c7,c8,c12<br />
c5,c9,c10,c11 c13,c14<br />
c15,c16,c17,c18<br />
c19,c20<br />
25 16 4 -<br />
c1,c2,c3,c4 c6,c7,c8,c12<br />
c5,c9,c10,c11<br />
c13,c14,c15,c16<br />
c17,c18,c19,c20<br />
26 5 12 3<br />
c2,c3,c5,c9 c1,c4,c11,c12 c6,c7,c8<br />
c10 c13,c14,c15,c16<br />
c17,c18,c19,c20<br />
27 9 3 8<br />
c1,c3,c4,c9 c2,c12,c13 c5,c6,c7,c8<br />
c10,c14,c15 c11,c17,c19,c20<br />
c16,c18<br />
28 4 16 -<br />
c6,c7,c9,c10 c1,c2,c3,c4<br />
c5,c8,c11,c12<br />
c13,c14,c15,c16<br />
c17,c18,c19,c20<br />
29 4 16 -<br />
c6,c7,c9,c10 c1,c2,c3,c4<br />
c5,c8,c11,c12<br />
c13,c14,c15,c16<br />
c17,c18,c19,c20<br />
30 11 9 -<br />
c1,c3,c4,c5,c6 c2,c7,c12,c13,c14<br />
c8,c9,c10,c11 c15,c18,c19,c20<br />
c16,c17<br />
31 3 17 -<br />
c7,c10,c11 c1,c2,c3,c4<br />
c5,c6,c8,c9<br />
c12,c13,c14,c15<br />
c16,c17,c18,c19<br />
c20<br />
32 - 20 -<br />
c1-c20<br />
33 3 17 -<br />
c7,c10,c11 c1,c2,c3,c4<br />
Cont<strong>in</strong>ued on next page
B.3 Tables 219<br />
Table B.2: Results of crucial sets of examples<br />
Consultant Number: 20<br />
examples sg pl both<br />
c5,c6,c8,c9<br />
c12,c13,c14,c15<br />
c16,c17,c18,c19<br />
c20<br />
34 3 17 -<br />
c7,c10,c11 c1,c2,c3,c4<br />
c5,c6,c8,c9<br />
c12,c13,c14,c15<br />
c16,c17,c18,c19<br />
c20
Appendix C<br />
Orig<strong>in</strong>al Questionnaires <strong>in</strong> <strong>Greek</strong>
Predicate Argument <strong>Agreement</strong> Questionnaires<br />
QUESTIONNAIRE 1<br />
SINGULAR NOUNS<br />
Choose any one or two of the follow<strong>in</strong>g items that might fill <strong>in</strong> the gaps!<br />
1. Το τραπέζι και το γραφείο είναι πολύ .................... . Θα ....... µεταφέρω.<br />
a. µεγάλο b. µεγάλοι c. µεγάλα<br />
a. τον b. τα c. τους<br />
2. Ο ηρωισµός και ο θάνατος είναι ....................... στο χώµα της Ελλάδος.<br />
a. χαραγµένος b. χαραγµένα c. χαραγµένοι<br />
3. Ο πίνακας και ο καναπές είναι .................. γι’αυτό το δωµάτιο.<br />
a. µεγάλες b. µεγάλοι c. µεγάλα<br />
4. Η βιβλιοθήκη και η τραπεζαρία είναι ........................<br />
a. διαφορετικοί b. διαφορετικές c. διαφορετικά<br />
5. Η ελευθερία και η πίστη είναι .....................<br />
a. ταυτόσηµες b. ταυτόσηµα c. Tαυτόσηµοι<br />
6. Ο αγώνας και η αυτοθυσία ήταν ................... κατα το 1821.<br />
a. κυρίαρχες b. κυρίαρχοι c. κυρίαρχα<br />
7. Ο θάνατος και η ελευθερία είναι ..............................<br />
a. σηµαντικές b. σηµαντικά c. σηµαντικοί<br />
8. Ο δρόµος και η πλατεία είναι .....................<br />
a. καθαρές b. καθαρά c. καθαροί<br />
9. Ο καναπές και η πολυθρόνα είναι .................<br />
a. άνετες b. άνετοι c.άνετα<br />
10. Η δικαιοσύνη και ο φιλελευθερισµός ήτανε .......................<br />
a. πρωτόγονες b. πρωτόγονοι c. πρωτόγονα<br />
11. Η καρέκλα και ο καναπές είναι ………………..<br />
α. άσπρες b. άσπροι c. άσπρα<br />
12. Ο έρωτας και το πάθος είναι .................. στη ζωή µας. .......... ζεί κανείς λίγες φορές.<br />
a. απαραίτητος b. απαραίτητα c. απαραίτητοι<br />
a. τα b. τις c. το<br />
13. Η προσπάθεια και το πείσµα είναι ...........................<br />
a. αποτελεσµατικές b. αποτελεσµατικοί c. αποτελεσµατικά
14. Ο Πέτρος και ο Γιάννης είναι ...................... Η δουλειά τους έχει εκσαντλήσει.<br />
a. κουρασµένος b. κουρασµένοι c. Κουρασµένα<br />
15. Το αγόρι και το κορίτσι βρίσκονται .......................... πίσω.<br />
a. καθισµένοι b. καθισµένα c. καθισµένες<br />
16. Το παιδί και το µωρό ήταν ………………….. στο δωµάτιο<br />
a. κλειδωµένες b. κλειδωµένοι c. κλειδωµένα<br />
17. Ο πατέρας και η µητέρα είναι ................... Η ζωή τους είναι δύσκολη και απαιτητική.<br />
a. κουρασµένες b. κουρασµένοι c. κουρασµένα<br />
18. Ο Κώστας και το κορίτσι είναι ................. για αναχώρηση.<br />
a. έτοιµοι b. έτοιµες c. έτοιµα<br />
19. Ο πατέρας και το παιδί βρέθηκαν .......................<br />
a. δολοφονηµένες b. δολοφονηµένοι c. δολοφονηµένα<br />
20. Ο Κώστας και το µωρό είναι ................. για αναχώρηση.<br />
a. έτοιµοι b. έτοιµες c. έτοιµα<br />
21. Η Μαρία και το κορίτσι έιναι ................ για την καινούργια τους ζωή.<br />
a. ενθουσιασµένοι b. ενθουσιασµένες c. ενθουσιασµένα<br />
22. Η µητέρα και το µωρό της είναι ...................<br />
a. χαρούµενοι b. χαρούµενες c. χαρούµενα<br />
23. Η Μαρία και το µωρό (αγόρι) της είναι ..................βλέπεις και ........ θαυµάζεις.<br />
a. όµορφοι b. όµορφες c. όµορφα<br />
a. τούς b. τίς<br />
a. τούς b. τίς<br />
24. Η Μαρία και το µώρο (κορίτσι) της είναι ................... ....... βλέπεις και ........<br />
θαυµάζεις.<br />
a. όµορφοι b. όµορφες c. όµορφα<br />
a. τούς b. τίς<br />
a. τούς b. τίς<br />
25. Η Μαρία και το άτοµο µε τη φούστα είναι ..........................<br />
a. περίεργες b. περίεργα c. περίεργοι<br />
Please write your name: …………………………
QUESTIONNAIRE 2<br />
PLURAL NOUNS<br />
Choose any one or two of the follow<strong>in</strong>g items that might fill <strong>in</strong> the gaps!<br />
1. Οι γιαγιάδες και τα εγγόνια ήταν ............................<br />
α. χαρούµενα β. χαρούµενοι γ. χαρούµενες<br />
2. Οι άνδρες και οι γυναίκες ήταν .............................<br />
α. έτοιµες β. έτοιµοι γ. έτοιµα<br />
3. Οι παππούδες και τα εγγόνια είναι ...........................<br />
α. χαρούµενα β. χαρούµενοι γ. χαρούµενες<br />
4. Οι γιαγιάδες και τα κορίτσια είναι πολύ............................<br />
α. γελαστοί β. γελαστά γ. γελαστές<br />
5. Τα κορίτσια και οι γιαγιάδες είναι ............................<br />
α. χαρούµενοι β. χαρούµενες γ. χαρούµενα<br />
6. Τα παδιά και οι µαµάδες είναι πολύ ...............................<br />
α. χαρούµενα β. χαρούµενοι γ. χαρούµενες<br />
7. Τα παιδιά και οι µπαµπάδες είναι πολύ .............................<br />
α. δεµένοι β. δεµένα γ. δεµένες<br />
8. Οι µπαµπάδες και τα αγόρια είναι πολύ .................................<br />
α. δεµένοι β. δεµένα γ. δεµένες<br />
9. Τα αγόρια και τα κορίτσια είναι ………………….<br />
α. χαρούµενοι β. χαρούµενες γ. χαρούµενα<br />
10. Οι παππούδες και οι πατεράδες είναι ………………<br />
α. αγαπηµένοι β. αγαπηµένες γ. αγαπηµένα<br />
11. Οι πίνακες και οι καναπέδες είναι .......................<br />
α. αταίριαστοι β. αταίριαστες γ. αταίριαστα<br />
12. Οι καυγάδες και οι χωρισµοί είναι …………για την ψυχολογία των παιδιών.<br />
α. σηµαντικές β. σηµαντικά γ. σηµαντικοί<br />
13. Οι κουζίνες και οι τουαλέτες είναι ..............................<br />
α. καθαρές β. καθαρά γ. καθαροί<br />
14. Οι φιλοσοφίες και οι θρησκείες ήταν ........................ για τους λαούς<br />
α. καθοριστικοί β. καθοριστικά γ. καθοριστικές
15. Οι δρόµοι και οι πλατείες ήταν ............................ κόσµο<br />
α. γεµάτοι β. γεµάτα γ. γεµάτες<br />
16. Οι τυφώνες και οι πληµµύρες ήταν .............................. για την Ταϊλάνδη<br />
α. καταστροφικές β. καταστροφικά γ. καταστροφικοί<br />
17. Οι δρόµοι και τα σοκάκια ήταν .......................... κόσµο<br />
α. γεµάτοι β. γεµάτα γ. γεµάτες<br />
18. Οι διαµαρτυρίες και τα παράπονα είναι ................... σ’ αυτήν τη περίπτωση.<br />
α. άχρηστες β. άχρηστοι γ. Άχρηστα<br />
19. Οι πλατείες και οι δρόµοι ήταν ............................ κόσµο.<br />
α. γεµάτοι β. γεµάτα γ. γεµάτες<br />
20. Τα εστιατόρια και οι πλατείες ήταν ........................... κόσµο<br />
α. γεµάτες β. γεµάτα γ. γεµάτοι<br />
21. Τα σοκάκια και οι πεζόδροµοι ήταν ............................<br />
α. άδειοι β. άδεια γ. άδειες<br />
22. Τα σπίτια και τα αυτοκίνητα είναι πιο ...................... απο εδώ και πέρα.<br />
α. ακριβά β. ακριβοί γ. ακριβές<br />
Please write your name: .................................................
QUESTIONNAIRE 3<br />
SG & PL IN/ANIMATE NOUNS<br />
Choose any one or two of the follow<strong>in</strong>g items that might fill <strong>in</strong> the gaps!<br />
1. Το αγόρι και η µητέρα ήταν ................................<br />
α. χαρούµενοι β. χαρούµενες γ. χαρούµενα<br />
2. Η µητέρα και το κορίτσι είναι πολύ ........................µεταξύ τους.<br />
α. δεµένοι β. δεµένες γ. δεµένα<br />
3. Ο κόπος και η προσπάθεια είναι ..................... για να προοδεύσει κανείς.<br />
α. απαραίτητοι β. απαραίτητες γ. απαραίτητα<br />
4. Η κουζίνα και ο πάγκος είναι πολύ ............................<br />
α. βρώµικοι β. βρώµικες γ. βρώµικα<br />
5. Οι καυγάδες και τα µπλεξίµατα είναι ....................στις µέρες µας.<br />
α. επικίνδυνοι β. επικίνδυνες γ. επικίνδυνα<br />
6. Τα µυστήρια και οι θάνατοι παραµένουν .............................<br />
α. ανεξιχνίαστοι β. ανεξιχνίαστες γ. ανεξιχνίαστα<br />
7. Οι απερισκεψίες και τα λάθη είναι ............................για την επιτυχία.<br />
α. καταστροφικοί β. καταστροφικά γ. καταστροφικές<br />
8. Τα παράπονα και οι διαµαρτυρίες είναι .................. σ’ αυτή την περίπτωση<br />
α. άχρηστοι β. άχρηστα γ. άχρηστες<br />
9. Τα τραπέζια και τα θρανία θα βαφτούν ........................<br />
α. άσπροι β. άσπρες γ.άσπρα<br />
10. Οι καθρέφτες και οι νιπτήρες είναι .........................<br />
α. βρώµικοι β. βρώµικες γ. βρώµικα<br />
11. Το παιδί και η γιαγιά είναι .......................<br />
α. αγαπηµένοι β. αγαπηµένες γ.αγαπηµένα<br />
12. Τα αιτήµατα και οι απαιτήσεις είναι ..................... αυτή την ώρα.<br />
α. ανούσιες β. ανούσια γ.ανούσιοι<br />
13. Τα κορίτσια και οι µαµάδες είναι ............................<br />
α. χαρούµενοι β. χαρούµενες γ. χαρούµενα<br />
14. Οι άνδρες και τα κορίτσια είναι ...................................<br />
α. αγαπηµένα β. αγαπηµένες γ. αγαπηµένα
15. Η ελευθερία και ο θάνατος ήταν ……………. το 1821<br />
α. σηµαντικά β. σηµαντικοί γ. σηµαντικές<br />
16. Η αγάπη και το πάθος είναι ………………. στη ζωή µας.<br />
α. απαραίτητα β. απαραίτητες γ.απαραίτητοι<br />
17. Ο καναπές και το τραπέζι είναι .......................<br />
α. άσπρα β. άσπρες γ. άσπροι<br />
18. Οι µαµάδες και τα κορίτσια είναι …………………….<br />
α. αγαπηµένοι β. αγαπηµένες γ.αγαπηµένα<br />
Please write your name: ………………
QUESTIONNAIRE 4<br />
SG & PL IN/ANIMATE NOUNS<br />
Fill <strong>in</strong> the correct form of the adjective<br />
1. Η γιαγιά και το εγγόνι ήταν γελαστ…………………<br />
2. Η γυναίκα και ο άνδρας είναι ερωτευµέν………………<br />
3. Η καρέκλα και η πολυθρόνα είναι άνετ……………...<br />
4. Η πίστη και η θρησκεία ήταν απαραίτητ…............για την Ελλάδα.<br />
5. Ο καθρέφτης και ο νιπτήρας είναι καθαρ....................<br />
6. Ο τρόµος και ο φόβος ήταν µεγάλ....................... για τους κατοίκους.<br />
7. Η ελευθερία και ο θάνατος ήταν σηµαντικ..............το 1821.<br />
8. Οι γυναίκες και οι άνδρες είναι σκληρά εργαζόµεν..................<br />
9. Τα αγόρια και τα κορίτσια είναι έτοιµ................ για τον αγώνα<br />
10. Τα αγόρια και οι µαµάδες είναι πολύ δεµέν.....................<br />
11. Τα κορίτσια και οι µαµάδες δεν είναι πολύ ...................... αντιθέτως.<br />
12. Οι γυναίκες και τα παιδιά ήταν χαρούµεν……………..<br />
13. Οι δρόµοι και οι πλατείες ήταν γεµάτ...................... κόσµο.<br />
14. Οι πλατείες και τα σοκάκια της Πάρου ήταν άδει............χθές.<br />
15. Οι πλατείες και οι πεζόδροµοι ήταν πολύ βρώµικ........... µετά τη διαδήλωση.<br />
16. Οι φιλοσοφίες και οι θρησκείες ήταν καθοριστικ......... για τους λαούς.<br />
17. Οι τυφώνες και οι πληµµύρες ήταν καταστροφικ....... για την Ταιλάνδη.<br />
18. Οι καυγάδες και οι χωρισµοί είναι καθοριστικ............... γιατις ανθρώπινες σχέσεις.<br />
19. Οι απιστίες και οι χωρισµοί είναι συνηθισµέν................... σήµερα.<br />
20. Οι αποφάσεις και οι νόµοι αυτής της κυβέρνησης είναι ασύµβατ...............<br />
Please write your name: ………………
Disjunctive Coord<strong>in</strong>ate <strong>Noun</strong>s Questionnaires<br />
QUESTIONNAIRE 1<br />
DISJUNCTION<br />
Choose any one or two of the follow<strong>in</strong>g items that might fill <strong>in</strong> the gaps!<br />
1. Ο Κώστας ή η Μαρία θα µε πάρει/πάρουν µε το αυτοκίνητο.<br />
2. Το αγόρι ή το κορίτσι προσπαθεί /προσπαθούν να καταστείλει /καταστείλουν τα<br />
συναισθήµατα<br />
3. Η γυναίκα ή το παιδί έχει/έχουν προτεραιότητα για το εµβόλιο κατά της γρίπης<br />
4. Το αγόρι ή το κορίτσι έχει/έχουν κούραση από το διάβασµα.<br />
5. Ο Κώστας ή ο Γιώργος έχει/έχουν συγγενή την Μαρία<br />
6. Ο γιατρός ή ο οδοντίατρος µπορεί/µπορούν να γράψει/γράψουν φάρµακα<br />
7. Η Μαρία ή η Κατερίνα έχει/έχουν σχέση µε τον Κώστα<br />
8. Ο δικηγόρος ή ο συµβολαιογράφος παρευρίσκεται/παραβρίσκονται στην υπογραφή<br />
συµβολαίων<br />
9. Ο φοιτητής ή η φοιτήτρια που είναι σίγουρος/σίγουρη/σίγουροι για κάθε επιλογή θα<br />
πετύχει/πετύχουν<br />
10. Ο πατέρας ή η µητέρα είναι ενήµερος/ενήµερη/ενήµεροι για την κατάσταση<br />
11. Ο άνδρας ή η γυναίκα που είναι ανήσυχος/ανήσυχη/ ανήσυχοι να<br />
µείνει/µείνουν εδώ.<br />
12. Η φοιτήτρια ή ο φοιτητής που είναι σίγουρος/ σίγουρη/σίγουροι για κάθε επιλογή θα<br />
πετύχει/πετύχουν<br />
13. Ο µαθητής ή η µαθήτρια µπορεί/µπορούν να ενηµερωθεί/ενηµερωθούν από τους γονείς<br />
για την αιτία.<br />
14. Ο γιος ή η κόρη χρειάζεται/χρειάζονται βοήθεια.<br />
15. Η µητέρα ή το αγόρι ετοιµάζεται/ετοιµάζονται για έξοδο;<br />
16. Eτοιµάζεται/ετοιµάζονται η µητέρα ή το αγόρι για έξοδο;<br />
17. Η µητέρα ή το αγόρι είναι έτοιµη/έτοιµο/έτοιµοι για έξοδο;<br />
18. Είναι έτοιµη/έτοιµο/έτοιµοι η µητέρα ή το αγόρι για έξοδο;<br />
19. Ο άνδρας ή η γυναίκα είναι έτοιµος/έτοιµη/έτοιµοι να ξεκινήσει/ξεκινήσουν;<br />
20. Ο µαθητής ή η µαθήτρια είναι πρόθυµος/πρόθυµη/πρόθυµοι να συνεχίσει/συνεχίσουν τη<br />
δουλειά;<br />
21. Εγώ ή εσύ θα χρειαστώ/χρειαστείς/χρειαστούµε βοήθεια.
22. Εγώ ή ο Κώστας θα αναλάβω/αναλάβει/αναλάβουµε τη δουλειά.<br />
23. Εσύ ή ο Γιώργος θα αποφασίσεις/αποφασίσει/αποφασίσετε για το σπίτι.<br />
24. Εσύ ή η Μαρία θα πας/πάει/πάµε στο γιατρό;<br />
25. Εγώ ή ο Κώστας θα πάρω/πάρει/πάρουµε την επιχείρηση;<br />
26. Εσύ ή εγώ θέλεις/θέλω/θέλουµε βοήθεια;<br />
27. Θα πας/πάει/πάµε εσύ ή η Μαρία στο γιατρό;<br />
28. Θα αναλάβω/αναλάβεις/αναλάβουµε εγώ ή εσύ την εταιρεία;<br />
29. Το αυτοκίνητο ή τα µηχανάκια θα πουληθεί/πουληθούν σύντοµα.<br />
30. Το µωρό ή τα παιδιά χρειάζεται/χρειάζονται πιο πολύ προσοχή;<br />
31. Ο πρόεδρος ή οι υπουργοί θα µιλήσει/µιλήσουν στο συνέδριο;<br />
32. Οι γάτες σου ή ο σκύλος µου έφαγε/έφαγαν τα λουλούδια µου;<br />
33. Τα κορίτσια ή ο γιος θα κληρονοµήσει/κληρονοµήσουν το σπίτι.<br />
34. Οι καρέκλες ή το τραπέζι θέλει/θέλουν βάψιµο;<br />
35. Οι ποδοσφαιριστές ή ο διαιτητής κερδίζει/κερδίζουν πολλά χρήµατα<br />
36. Οι αδελφές ή ο αδελφός σου είναι ταχτοποιηµένοι/ταχτοποιηµένες/ταχτοποιηµένος<br />
37. Ο αδελφός ή οι αδελφές σου έχει/έχουν δουλειά;<br />
38. Ο πατέρας ή οι κόρες είναι υπεύθυνος/υπεύθυνοι/υπεύθυνες για την εταιρεία.<br />
39. Οι κόρες ή ο πατέρας είναι υπεύθυνος/υπεύθυνοι/υπεύθυνες για την εταιρεία.<br />
40. Ετοιµάζεται/ετοιµάζονται το κορίτσι ή τα αγόρια σου για τις εξετάσεις;<br />
41. Ετοιµάζεται/ετοιµάζονται τα κορίτσια ή το αγόρι σου για τις εξετάσεις;<br />
NAME:………………………………………………………………………<br />
THANKS A LOT FOR YOUR HELP!!!
QUESTIONNAIRΕ 2<br />
DISJUNCTION<br />
Choose any one or two of the follow<strong>in</strong>g items!<br />
1. Ή ο γιος ή η κόρη µου θα πάρει/πάρουν µέρος στο διαγωνισµό.<br />
2. Ή η µητέρα ή η γιαγιά θα αναλάβει/αναλάβουν την επιµέλεια<br />
3. Ή το αγόρι ή το κορίτσι θα αντιµετωπίσει/αντιµετωπίσουν τα σηµερινά προβλήµατα.<br />
4. Ή ο πατέρας ή η µητέρα είναι υπεύθυνοι/υπεύθυνος/υπεύθυνη για την επιµέλεια του παιδιού.<br />
5. Ή ο γιος ή η κόρη είναι ικανός/ικανοί/ικανή να δώσουν συµβουλές.<br />
6. Ή ο αδελφός ή η αδελφή µου θα περάσει/περάσουν στο πανεπιστήµιο.<br />
7. Ή ο θείος ή η θεία µου θα αναλάβει/αναλάβουν το µαγαζί.<br />
8. Ή εγώ ή εσύ έχω/έχεις/έχουµε δικαίωµα συµµετοχής στον αγώνα.<br />
9. Ή εγώ ή εσύ θα αναλάβω/αναλάβεις/αναλάβουµε την επιχείρηση.<br />
10. Ή εσύ ή ο πατέρας θα υπογράψεις/υπογράψετε/υπογράψει τα χαρτιά.<br />
11. Ή εσύ ή η Μαρία έχεις/έχετε/έχει την ευθύνη για τη δουλειά.<br />
12. Έχεις/έχουν/έχει ή εσύ ή η Μαρία την ευθύνη για τη δουλειά;<br />
13. Θέλεις/θέλω/θέλουµε ή εγώ ή εσύ να πάρεις/πάρω/πάρουµε τη δουλειά;<br />
14. Ή εσύ ή αυτός θα πάρει/πάρεις/πάρετε µετεγγραφή.<br />
15. Ή οι κόρες ή η µητέρα θα πάει/πάνε εκδροµή.<br />
16.. Ή εµείς ή αυτοί θα συνεργαστούµε/συνεργαστούν µαζί τους.<br />
17. Ή τα µωρά ή το παιδί θα πάρει/πάρουν δώρο φέτος.<br />
18. Ή ο πατέρας ή τα παιδιά θα δουλέψει/δουλέψουν στην επιχείρηση.<br />
19. Ή εγώ ή αυτός αγαπάω/αγαπάει/αγαπάµε την δουλειά του/µάς.<br />
20. Ή εσείς ή τα παιδιά θα αποφασίσετε/αποφασίσουν για την επόµενη χρονιά.<br />
21. Ούτε το αγόρι ούτε το κορίτσι είναι έτοιµο/έτοιµα γι αυτήν την αλλαγή.<br />
22. Ούτε ο πατέρας ούτε ο γιος ενέκρινε/ενέκριναν την αγορά της εταιρείας.<br />
23. Ούτε ο άνδρας ούτε η γυναίκα συµβιβάζεται/συµβιβάζονται την σηµερινή εποχή.<br />
24. Ούτε ο πατέρας ούτε η µητέρα είναι χαρούµενη/χαρούµενος/χαρούµενοι µε αυτή τη<br />
κατάσταση.<br />
25. Ούτε ο πελάτης ούτε η πελάτισσα είναι ικανοποιηµένος/ικανοποιηµένη/ικανοποιηµένοι µε
την άνοδο στις τιµές.<br />
26. Ούτε εσύ ούτε αυτός θα δεχτείς/δεχτεί/δεχτείτε την προσφορά<br />
27. Ούτε εσύ ούτε εγώ χρειάζοµαι/χρειάζεσαι/χρειαζόµαστε βοήθεια.<br />
28. Ούτε τα κορίτσια ούτε η µητέρα είναι σύµφωνη/σύµφωνες για τη µετακόµιση.<br />
NAME:………………………………………………………………………<br />
THANKS A LOT FOR YOUR HELP!!!
Bibliography<br />
Alonso-Ovalle, Luis. 2006. Disjunction <strong>in</strong> Alternative Semantics. Ph.D. thesis, University<br />
of Massachusetts Amherst.<br />
Ambrose, Alice and Morris Lazerowitz. 1962. Fundamentals of Symbolic Logic. New York:<br />
Holt, R<strong>in</strong>ehart and W<strong>in</strong>ston.<br />
Aust<strong>in</strong>, Peter K. 2001. Lexical Functional Grammar. In N. Smelser and P. Baltes, eds.,<br />
International Encyclopedia of the Social and Behavioural Sciences, pages 8748–8754.<br />
http://www.l<strong>in</strong>guistics.unimelb.edu.au/contact/staff/petter/Elsevier.pdf: Elsevier.<br />
Bach, Emmon. 1989. Informal Lectures on Formal Semantics. Albany, NY: State University<br />
of New York Press.<br />
Badecker, William. 2007. A feature pr<strong>in</strong>ciple for partial agreement. L<strong>in</strong>gua 117:1541–1565.<br />
Badecker, William. 2008. Gender resolution: The role of markedness and feature uniformity<br />
<strong>in</strong> agreement with conjo<strong>in</strong>ed noun phrases. John Hopk<strong>in</strong>s University, Draft.<br />
Barker, Stephen. 1985. The Elements of Logic. New York: McGraw-Hill.<br />
Barret, Robert B. and Alfred J. Stanner. 1971. The myth of the exclusive ‘or’. M<strong>in</strong>d<br />
80(317):116–121.<br />
Bateman, Nicoleta and Maria Pol<strong>in</strong>sky. 2005. Rumanian as a two-gender language. Unpublished<br />
LSA handout.<br />
Benor, Sarah B. and Roger Levy. 2006. The chicken or the egg? A probabilistic analysis<br />
of English b<strong>in</strong>om<strong>in</strong>als. Language 82(2):233–278.<br />
Bloomfield, Leonard. 1962. Language. New York: Holt, Re<strong>in</strong>hart and W<strong>in</strong>ston.<br />
Bresnan, Joan. 1997. Mixed categories as head shar<strong>in</strong>g constructions. In M. Butt and<br />
T. H. K<strong>in</strong>g, eds., On-l<strong>in</strong>e Proceed<strong>in</strong>gs of the LFG97 Conference. Stanford, CA: CSLI<br />
Publications.<br />
Bresnan, Joan. 2001. Lexical-Functional Syntax. Oxford: Blackwell Publish<strong>in</strong>g Ltd.<br />
Bresnan, Joan. 2001c. Optimal syntax. In J. Dekkers, F. van der Leeuw, and J. van de<br />
Weijer, eds., Optimality Theory: Phonology, Syntax and Acquisition, chap. 10. Oxford:<br />
Oxford University Press.<br />
Bresnan, Joan, Ronald Kaplan, and Peter Peterson. 1985b. Coord<strong>in</strong>ation and the flow of<br />
<strong>in</strong>formation through phrase structure. Palo Alto, CA: Xerox Park.<br />
Brody, Baruch A. 1973. Logic: Theoretical and Applied. Englewood Cliffs, N.J.: Prentice-<br />
Hall.
BIBLIOGRAPHY 233<br />
Chomsky, Noam. 1957. Syntactic Structures. The Hague: Mouton.<br />
Clairis, Christos, George D. Bab<strong>in</strong>iotis, (<strong>in</strong> cooperation with Amalia Mozer, Aikater<strong>in</strong>i<br />
Bakakou-Orfanou, and Stavro Skopetea). 2004. A Grammar of <strong>Modern</strong> <strong>Greek</strong>: a<br />
Structural-Functional-Communicative Approach. Athens: Ell<strong>in</strong>ika Grammata.<br />
Copi, Irv<strong>in</strong>g M. 1971. Introduction to Logic. New York: Macmillan.<br />
Corbett, Greville G. 1979. The <strong>Agreement</strong> Hierarchy. Journal of L<strong>in</strong>guistics 15:203–224.<br />
Corbett, Greville G. 1983a. Hierarchies, Targets and Controllers: <strong>Agreement</strong> Patterns <strong>in</strong><br />
Slavic. London: Croom Helm Ltd.<br />
Corbett, Greville G. 1983b. Resolution rules: <strong>Agreement</strong> <strong>in</strong> person, number and gender. In<br />
G. Gazdar, E. Kle<strong>in</strong>, and G. K. Pullum, eds., Order, Concord and Constituency, pages<br />
175–206. Dordrecht: Foris.<br />
Corbett, Greville G. 1991. Gender. Cambridge, UK: Cambridge University Press.<br />
Corbett, Greville G. 2000. Number. Cambridge, UK: Cambridge University Press.<br />
Corbett, Greville G. 2001. <strong>Agreement</strong>: Terms and boundaries. In William Griff<strong>in</strong>, ed.,<br />
The Role of <strong>Agreement</strong> <strong>in</strong> Natural Language: Proceed<strong>in</strong>gs of the 2001 Texas L<strong>in</strong>guistic<br />
Society Conference, pages 109–122. Aust<strong>in</strong>, Texas.<br />
Corbett, Greville G. and Norman M. Fraser. 1993. Network morphology: a DATR account<br />
of Russian nom<strong>in</strong>al <strong>in</strong>flection. Journal of L<strong>in</strong>guistics 29:113–142.<br />
Corbett, Greville G. and A. D. Mtenje. 1987. Gender agreement <strong>in</strong> Chichewa. Studies <strong>in</strong><br />
African L<strong>in</strong>guistics 18:1–38.<br />
Dalrymple, Mary. 2001. Lexical Functional Grammar, vol. 42 of Syntax and Semantics<br />
Series. Stanford, CA: Academic Press.<br />
Dalrymple, Mary, Helge Dyvik, and Tracy H. K<strong>in</strong>g. 2004a. Copular complements: closed or<br />
open? In M. Butt and T. H. K<strong>in</strong>g, eds., On-l<strong>in</strong>e Proceed<strong>in</strong>gs of the LFG04 Conference,<br />
pages 188–198. Stanford, CA: CSLI Publications.<br />
Dalrymple, Mary and Bozhil Hristov. 2010. <strong>Agreement</strong> patterns and coord<strong>in</strong>ation <strong>in</strong> Lexical<br />
Functional Grammar. In M. Butt and T. H. K<strong>in</strong>g, eds., Proceed<strong>in</strong>gs of the LFG10<br />
Conference, pages 186–206. Stanford, CA: CSLI Publications.<br />
Dalrymple, Mary and Ronald Kaplan. 2000. Feature <strong>in</strong>determ<strong>in</strong>acy and feature resolution.<br />
Journal of Language 76(4):759–798.<br />
Dalrymple, Mary, Ronald Kaplan, and Tracy H. K<strong>in</strong>g. 2004b. L<strong>in</strong>guistic generalisations<br />
over descriptions. In M. Butt and T. H. K<strong>in</strong>g, eds., On-l<strong>in</strong>e Proceed<strong>in</strong>gs of the LFG04<br />
Conference, pages 199–208. Stanford, CA: CSLI Publications.<br />
Dalrymple, Mary, Ronald Kaplan, John T. Maxwell, and Annie Zaenen, eds. 1995. Formal<br />
Issues <strong>in</strong> Lexical-Functional Grammar. Stanford, CA: CSLI Publications.<br />
Dalrymple, Mary, John Lamp<strong>in</strong>g, Fernando Pereira, and Vijay Saraswat. 2002. L<strong>in</strong>ear logic<br />
for mean<strong>in</strong>g assembly. In S. Manandhar, G. P. Lopes, and W. Nutt, eds., Proceed<strong>in</strong>gs of<br />
Computational Logic for Natural Language Process<strong>in</strong>g. Ed<strong>in</strong>burgh.<br />
Dalrymple, Mary and Irena Nikolaeva. 2006. Syntax of natural and accidental coord<strong>in</strong>ation:<br />
evidence from agreement. Language 82(4):824–849.
BIBLIOGRAPHY 234<br />
de Swart, Henriette. 1998. Introduction to Natural Language Semantics. Stanford, CA:<br />
CSLI Publications.<br />
Dougherty, Ray C. 1970. A grammar of coord<strong>in</strong>ate conjo<strong>in</strong>ed structures, Part I. Language<br />
46(4):850–898.<br />
Dougherty, Ray C. 1971. A grammar of coord<strong>in</strong>ate conjo<strong>in</strong>ed structures, Part II. Language<br />
47(2):298–339.<br />
Eggert, Randall. 2002. Disconcordance: The Syntax, Semantics and Pragmatics of or-<br />
<strong>Agreement</strong>. Ph.D. thesis, University of Chicago, Chicago: Ill<strong>in</strong>ois.<br />
Falk, Yehuda N. 1984. The English auxiliary system: A Lexical-Functional analysis. Language<br />
60(3):483–509.<br />
Falk, Yehuda N. 2001. Lexical-Functional Grammar: An Introduction to Parallel<br />
Constra<strong>in</strong>t-based Syntax. Stanford, CA: CSLI Publications.<br />
Farkas, Donka F. 1990. Two cases of underspecification <strong>in</strong> morphology. L<strong>in</strong>guistic Inquiry<br />
4(21):539–550.<br />
Farkas, Donka F. and D. Zec. 1995. <strong>Agreement</strong> and pronom<strong>in</strong>al reference. In G. C<strong>in</strong>que and<br />
G. Giusti, eds., Advances <strong>in</strong> Rumanian L<strong>in</strong>guistics, pages 83–101. Philadelphia: John<br />
Benjam<strong>in</strong>s.<br />
Fassi-Fehri, Abdelkader. 1981. Complémentation et Anaphore en Arabe <strong>Modern</strong>e: Une<br />
Approche Lexicale Fonctionelle. Ph.D. thesis, Univeristé de Paris III, Paris, France.<br />
Flouraki, Maria and Desp<strong>in</strong>a Kazana. 2009. Constra<strong>in</strong><strong>in</strong>g disjunctive constructions <strong>in</strong><br />
<strong>Modern</strong> <strong>Greek</strong>. In M. Butt and T. H. K<strong>in</strong>g, eds., On-l<strong>in</strong>e Proceed<strong>in</strong>gs of the LFG09<br />
Conference, pages 282–296. Stanford, CA: CSLI Publications.<br />
Fowler, Henry W. 1926. A Dictionary of <strong>Modern</strong> English Usage. London: Oxford University<br />
Press.<br />
Francis, Ela<strong>in</strong>e J. and Laura A. Michaelis. 2000. Approaches to mismatch: <strong>in</strong>troduction.<br />
In M. Butt and T. H. K<strong>in</strong>g, eds., Proceed<strong>in</strong>gs of the BFG00 Conference Workshops. The<br />
University of California, Berkeley: CSLI Publications.<br />
Gazdar, Gerald. 1980. A cross-categorial semantics for coord<strong>in</strong>ation. L<strong>in</strong>guistics and Philosophy<br />
3(3):407–409.<br />
Genesereth, Michael R. and Nils J. Nilson. 1987. Logical Foundations of Artificial Intelligence.<br />
San Mateo, CA: Morgan Kaufmann Publishers.<br />
Georgakakos, George. 1979. Elementary Formal Logic. New York.<br />
Givon, Talmy. 1970. The resolution of gender conflicts <strong>in</strong> Bantu conjunction: when syntax<br />
and semantics clash. In Papers from the Sixth Regional Meet<strong>in</strong>g Chicago L<strong>in</strong>guistic<br />
Society, pages 250–261. Chicago: Chicago L<strong>in</strong>guistic Society.<br />
Greenberg, Joseph. 1966. Language Universals: With Special Reference to Feature Hierarchies.<br />
The Hague: Mouton.<br />
Haspelmath, Mart<strong>in</strong>. 2004a. Coord<strong>in</strong>at<strong>in</strong>g Constructions. Amsterdam: John Benjam<strong>in</strong>s.<br />
Haspelmath, Mart<strong>in</strong>. 2007. Coord<strong>in</strong>ation. In T. Shopen, ed., Language Typology and<br />
Syntactic Description, vol. 2, pages 1–51. Cambridge: Cambridge University Press, 2nd<br />
edn.
BIBLIOGRAPHY 235<br />
Heycock, Carol<strong>in</strong>e and Roberto Zamparelli. 1999. Toward a unified analysis of DP conjunction.<br />
In P. Dekker, ed., Proceed<strong>in</strong>gs of the Twelfth Amsterdam Colloquium, pages<br />
127–132. University of Amsterdam: ILLC.<br />
Heycock, Carol<strong>in</strong>e and Roberto Zamparelli. 2000. Friends and colleagues: plurality and<br />
NP coord<strong>in</strong>ation. In M. Hirotani, A. Coetzee, N. Hall, and J.-Y. Kim, eds., Proceed<strong>in</strong>gs<br />
of the 30th Annual Meet<strong>in</strong>g of the North Eastern L<strong>in</strong>guistic Society, vol. 30. Rutgers<br />
University.<br />
Heycock, Carol<strong>in</strong>e and Roberto Zamparelli. 2003. Coord<strong>in</strong>ated bare def<strong>in</strong>ites. L<strong>in</strong>guistic<br />
Inquiry 34(3):443–469.<br />
Heycock, Carol<strong>in</strong>e and Roberto Zamparelli. 2005. Friends and colleagues: plurality, coord<strong>in</strong>ation<br />
and the structure of DP. Natural Language Semantics 13(3):201–270.<br />
Hoeksema, Jack. 1988. The semantics of non-Boolean ‘AND’. Journal of Semantics<br />
6(1):19–40.<br />
Holton, David, Peter Mackridge, and Irene Philippaki-Warburton. 1997. <strong>Greek</strong>: A Comprehensive<br />
Grammar of the <strong>Modern</strong> Language. London: Routledge.<br />
Horrocks, Geoffrey. 1987. Generative Grammar. London: Longman.<br />
Hurford, James R. 1974. Exclusive or <strong>in</strong>clusive disjunction. Foundations of Language<br />
11(3):409–411.<br />
Jenn<strong>in</strong>gs, Ray E. 1966. Or. Analysis 26:181–184.<br />
Jenn<strong>in</strong>gs, Ray E. 1994. The Genealogy of Disjunction. USA: Oxford University Press.<br />
Johannessen, Janne Bondi. 1996. Partial agreement and coord<strong>in</strong>ation. L<strong>in</strong>guistic Inquiry<br />
27(4):661–676.<br />
Joseph, Brian D. and Irene Philippaki-Warburton. 1987. <strong>Modern</strong> <strong>Greek</strong>. Croom Helm<br />
Descriptive Grammar Series. London: Routledge Kegan and Paul.<br />
Kameyama, M. 1985. Zero Anaphora: The Case of Japanese. Ph.D. thesis, Stanford<br />
University.<br />
Kamp, Hans and Uwe Reyle. 1993. From Discourse to Logic: An Introduction to Model<br />
Theoretic Semantics of Natural Language, Formal Logic and Discourse Representation<br />
Theory, vol. 1,2 of Studies <strong>in</strong> L<strong>in</strong>guistics and Philosophy. Dordrecht: Kluwer Academic<br />
Publishers.<br />
Kaplan, Ronald. 1995. The formal architecture of Lexical-Functional Grammar. In M. Dalrymple,<br />
R. Kaplan, J. T. M. III, and A. Zaenen, eds., Formal Issues <strong>in</strong> Lexical-Functional<br />
Grammar, pages 7–27. Stanford, CA: CSLI Publications.<br />
Kaplan, Ronald and Joan Bresnan. 1982. Lexical Functional Grammar: A formal system<br />
for grammatical representation. In J. Bresnan, ed., The Mental Representation of<br />
Grammatical Relations, chap. 4, pages 173–281. Cambridge, MA: The MIT Press.<br />
Kaplan, Ronald and John T. Maxwell. 1988. Constituent coord<strong>in</strong>ation <strong>in</strong> Lexical-<br />
Functional Grammar. In Proceed<strong>in</strong>gs of the 12th International Conference on Computational<br />
L<strong>in</strong>guistics (COLING88), vol. 1, pages 303–305. Budapest.
BIBLIOGRAPHY 236<br />
Kathol, Andreas. 1999. <strong>Agreement</strong> and the syntax-morphology <strong>in</strong>terface <strong>in</strong> HPSG. In<br />
R. Lev<strong>in</strong>e and G. Green, eds., Studies <strong>in</strong> Contemporary Phrase Structure Grammar.<br />
New York: Cambridge University Press.<br />
Kazana, Desp<strong>in</strong>a. 2007. Gender resolution <strong>in</strong> <strong>Modern</strong> <strong>Greek</strong>. In Essex Graduate Student<br />
Papers <strong>in</strong> Language and L<strong>in</strong>guistics, vol. 9, pages 39–56. University of Essex: Department<br />
of Language and L<strong>in</strong>guistics.<br />
Kearns, Kate. 2000. Semantics. <strong>Modern</strong> L<strong>in</strong>guistics Series. England: Macmillan.<br />
Keenan, Edward L. and Leonard M. Faltz. 1985. Boolean Semantics for Natural Language.<br />
Studies <strong>in</strong> L<strong>in</strong>guistics and Philosophy. Boston: D. Reidel.<br />
K<strong>in</strong>g, Tracy H. and Mary Dalrymple. 2004. Determ<strong>in</strong>er agreement and noun conjunction.<br />
Journal of L<strong>in</strong>guistics 40(1):69–104.<br />
Kokkonidis, Miltiadis. 2005. Why glue a donkey to an f-structure when you can constra<strong>in</strong><br />
it and b<strong>in</strong>d it <strong>in</strong>stead? In M. Butt and T. H. K<strong>in</strong>g, eds., On-l<strong>in</strong>e Proceed<strong>in</strong>gs of the<br />
LFG05 Conference, pages 238–252. Stanford, CA: CSLI Publications.<br />
Kuhn, Jonas and Louisa Sadler. 2007. S<strong>in</strong>gle conjunct agreement and the formal treatment<br />
of coord<strong>in</strong>ation <strong>in</strong> LFG. In M. Butt and T. H. K<strong>in</strong>g, eds., On-l<strong>in</strong>e Proceed<strong>in</strong>gs of the<br />
LFG07 Conference, pages 302–322. Stanford, CA: CSLI Publications.<br />
Lakoff, Rob<strong>in</strong>. 1971. If’s, and’s and but’s about conjunction. In Fillmore and Langendeon,<br />
eds., Studies <strong>in</strong> L<strong>in</strong>guistic Semantics, pages 115–149. New York: Holt, Re<strong>in</strong>hart and<br />
W<strong>in</strong>ston.<br />
Lambrecht, Knud. 1984. Formulaicity, frame semantics and pragmatics <strong>in</strong> German b<strong>in</strong>omial<br />
expressions. Language 60(4):753–796.<br />
Langendoen, D. Terence. 1970. Essentials of English Grammar. New York: Holt, R<strong>in</strong>ehart,<br />
W<strong>in</strong>ston.<br />
Larson, Richard. 1985. On the syntax of disjunction scope. Natural Language and L<strong>in</strong>guistic<br />
Theory 3(2):217–264.<br />
Lumsden, John. 1992. Undespecification <strong>in</strong> grammatical and natural gender. L<strong>in</strong>guistic<br />
Inquiry 3(23):469–86.<br />
Mackridge, Peter. 1985. The <strong>Modern</strong> <strong>Greek</strong> Language. Oxford: Oxford University Press.<br />
Malouf, Robert. 1998. Categories, prototypes and default <strong>in</strong>heritance. In G. Bouma, G.-J.<br />
Kruijff, and R. Oehrle, eds., Proceed<strong>in</strong>gs of the Jo<strong>in</strong>t Conference on Formal Grammar,<br />
Head-Driven Phrase Structure Grammar and Categorial Grammar, pages 207–216. Saarbrucken.<br />
Manicas, Peter T. 1976. Logic: The Essentials. New York: McGraw-Hill.<br />
McCloskey, James. 1986. Inflection and conjunction <strong>in</strong> <strong>Modern</strong> Irish. Natural Language<br />
and L<strong>in</strong>guistic Theory 4(2):245–282.<br />
M<strong>in</strong>assian, Martiros. 1980. Grammaire d’ Armenien Oriental. Delmar, NY: Caravan.<br />
Montague, Richard. 1973. The proper treatment of quantification <strong>in</strong> English. In P. Suppes,<br />
J. Moravcsik, and J. H<strong>in</strong>tikka, eds., Approaches to Natural Language. Dordrecht: Reidel.
BIBLIOGRAPHY 237<br />
Moosally, Michelle. 1998. <strong>Noun</strong> Phrase Coord<strong>in</strong>ation: Ndebele <strong>Agreement</strong> Patterns and<br />
Cross-L<strong>in</strong>guistic Variation. Ph.D. thesis, University of Texas at Aust<strong>in</strong>, Aust<strong>in</strong>.<br />
Moosally, Michelle. 1999. Subject and object coord<strong>in</strong>ation <strong>in</strong> Ndbele: An HPSG analysis.<br />
In S. Bird, A. Carnie, J. D. Haugen, and P. Norquest, eds., Proceed<strong>in</strong>gs of the WCCFL<br />
18 Conference. Somerville, MA: Cascadilla Press.<br />
Morgan, Jerry. 1972. Verb agreement as a rule of English. In P. Peranteau, J. Levi,<br />
and G. Phares, eds., Papers from the Eighth Regional Meet<strong>in</strong>g of the Chicago L<strong>in</strong>guistic<br />
Society, pages 278–286. Chicago: Chicago L<strong>in</strong>guistic Society.<br />
Morgan, Jerry. 1984. Some problems of agreement <strong>in</strong> English and Albanian. In C. Brugman<br />
and M. Macaulay, eds., Proceed<strong>in</strong>g of the Berkeley L<strong>in</strong>guistic Society, vol. 10, pages 233–<br />
247. Berkeley: Berkeley L<strong>in</strong>guistics Society.<br />
Morgan, Jerry. 1985. Some problems of determ<strong>in</strong>ation <strong>in</strong> English number agreement. In<br />
G. Alvarez, B. Brondie, and T. McCoy, eds., Proceed<strong>in</strong>g of the First Eastern States<br />
Conference on L<strong>in</strong>guistics, pages 69–78. Columbus, Ohio: Ohio State University.<br />
Moyse-Faurie, Claire and John Lynch. 2004. Coord<strong>in</strong>ation <strong>in</strong> Oceanic languages and Proto<br />
Oceanic. In M. Haspelmath, ed., Coord<strong>in</strong>at<strong>in</strong>g Constructions, pages 445–498. Amsterdam:<br />
John Benjam<strong>in</strong>s.<br />
Nordl<strong>in</strong>ger, Rachel and Joan Bresnan. To appear. Lexical-Functional Grammar: Interactions<br />
between morphology and syntax. In R. D. Borsley and K. Börjars, eds., Non-<br />
Transformational Syntax: A Guide to Current Models. Oxford: Blackwells.<br />
Nunberg, Geoffrey. 1977. The Pragmatics of Reference. Ph.D. thesis, City University of<br />
New York, New York.<br />
Nunberg, Geoffrey. 1995. Transfers of mean<strong>in</strong>g. Journal of Semantics 12(2):109–132.<br />
Nunberg, Geoffrey. 2005. The pragmatics of deferred <strong>in</strong>terpretation. In L. R. Horn and<br />
G. Ward, eds., The Handbook of Pragmatics, chap. 15, pages 344–364. UK: Blackwell.<br />
Parker, Frank. 1983. Number <strong>in</strong> <strong>in</strong>def<strong>in</strong>ite pronouns and correlatives. University of South<br />
Florida Language Quarterly 22(1-2):13–16.<br />
Partee, Barbara H., Alice ter Meulen, and Robert E. Wall. 1993. Mathematical Methods<br />
<strong>in</strong> L<strong>in</strong>guistics. Studies <strong>in</strong> L<strong>in</strong>guistics and Philosophy. Dordrecht: Kluwer Academic<br />
Publishers, 2nd edn.<br />
Partridge, Eric. 1956. Usage and Abusage: A Guide to Good English. London: Hamish<br />
Hamilton.<br />
Pelletier, Francis J. 1978. Or. Theoretical L<strong>in</strong>guistics 4(1).<br />
Peterson, Peter. 1986. Establish<strong>in</strong>g verb agreement with disjunctively conjo<strong>in</strong>ed subjects:<br />
strategies vs. pr<strong>in</strong>ciples. Australian Journal of L<strong>in</strong>guistics 6(2):231–249.<br />
Pollard, Carl and Ivan A. Sag. 1994. Head-Driven Phrase Structure Grammar. Chicago:<br />
The University of Chicago Press.<br />
Pr<strong>in</strong>ce, Alan and Paul Smolensky. 2002. Optimality Theory: Constra<strong>in</strong>t Interaction <strong>in</strong><br />
Generative Grammar. Maden, MA: Blackwell.<br />
Qu<strong>in</strong>e, William V. 1972. Methods of Logic. New York: Holt, R<strong>in</strong>ehart and W<strong>in</strong>ston.
BIBLIOGRAPHY 238<br />
Quirk, Randolph and Sidney Greenbaum. 1973. A University Grammar of English. London:<br />
Longman.<br />
Rub<strong>in</strong>, Ronald and Charles M. Young. 1989. Formal Logic: a Model of English. California:<br />
Mayfield Publish<strong>in</strong>g Co.<br />
Sadler, Louisa. 1997. Clitics and the structure-function mapp<strong>in</strong>g. In M. Butt and T. H.<br />
K<strong>in</strong>g, eds., On-l<strong>in</strong>e proceed<strong>in</strong>gs of the LFG97 Conference. Stanford, CA: CSLI Publications.<br />
Sadler, Louisa. 1999. Non-distributive features and coord<strong>in</strong>ation <strong>in</strong> Welsh. In M. Butt<br />
and T. H. K<strong>in</strong>g, eds., On-l<strong>in</strong>e proceed<strong>in</strong>gs of the LFG99 Conference. Stanford, CA: CSLI<br />
Publications.<br />
Sadler, Louisa. 2003. Coord<strong>in</strong>ation and asymmetric agreement <strong>in</strong> Welsh. In M. Butt<br />
and T. H. K<strong>in</strong>g, eds., Nom<strong>in</strong>als: Inside and Out, pages 85–118. Stanford, CA: CSLI<br />
Publications.<br />
Sadler, Louisa. 2006. Gender resolution <strong>in</strong> Rumanian. In M. Butt, M. Dalrymple, and T. H.<br />
K<strong>in</strong>g, eds., Intelligent L<strong>in</strong>guistic Architectures: Variations on Themes by Ron Kaplan,<br />
pages 437–454. Stanford, CA: CSLI Publications.<br />
Sadock, Jerrold M. 1991. Autolexical Syntax. Chicago: Chicago University Press.<br />
Schalley, Andrea C. and Dietmar Zaefferer, eds. 2007. Ontol<strong>in</strong>guistics: How Ontological<br />
Status Shapes the L<strong>in</strong>guistics Cod<strong>in</strong>g Concepts. Trends <strong>in</strong> L<strong>in</strong>guistics: Studies and<br />
Monographs. Berl<strong>in</strong>: Mouton de Gruyer.<br />
Schane, Sanford A. 1970. Phonological and morphological markedness. In M. Bierwisch<br />
and K. E. Heidolph, eds., Progress <strong>in</strong> L<strong>in</strong>guistics, pages 286–294. The Hague: Mouton.<br />
Schwarzschild, Roger. 1996. Pluralities. Dordrecht: Kluwer Academic Publishers.<br />
Siewierska, Anna. 2004. Person. Cambridge Textbooks <strong>in</strong> L<strong>in</strong>guistics. Cambridge, UK:<br />
Cambridge University Press.<br />
Simons, Mandy. 1998. Or: Issues <strong>in</strong> the Semantics and Pragmatics of Disjunction. Ph.D.<br />
thesis, Cornell University.<br />
Spencer, Andrew. 2002. Gender as an <strong>in</strong>flectional category. Journal of L<strong>in</strong>guistics 38:279–<br />
312.<br />
Stassen, Leon. 2000. AND-languages and WITH-languages. L<strong>in</strong>guistic Typology 4(1):1–54.<br />
Tapscott, Bangs L. 1976. Elementary Applied Symbolic Logic. New Jersey: Prentice-Hall,<br />
Englewoods Cliffs.<br />
Tarski, Alfred. 1941. Introduction to Logic and to the Methodology of Deductive Sciences.<br />
New York: Oxford University Press, Revised 1946.<br />
Triantaphyllidis, Manolis. 1994. Neo-elleniki Grammatiki. Athens: Organismos Ekdoseon<br />
Didaktikon Biblion.<br />
Triantaphyllidis, Manolis. 2005. Neoelleniki Grammatiki tis Dimotikis. Thessaloniki: Aristotle<br />
University of Thessaloniki, Institute of Manolis Triantafillidis, 2nd edn.
BIBLIOGRAPHY 239<br />
Villavicencio, Al<strong>in</strong>e, Louisa Sadler, and Doug Arnold. 2005. An HPSG account of closest<br />
conjunct agreement <strong>in</strong> NP coord<strong>in</strong>ation <strong>in</strong> Portuguese. In S. Muller, ed., The Proceed<strong>in</strong>gs<br />
of the 12th International Conference on Head-Driven Phrase Structure Grammar, pages<br />
427–447. Stanford, CA: CSLI Publications.<br />
V<strong>in</strong>cent, Nigel and Kersti Börjars. 2000. Feature resolution and the content of features. In<br />
M. Butt and T. H. K<strong>in</strong>g, eds., On-l<strong>in</strong>e proceed<strong>in</strong>gs of the LFG00 Conference. Stanford,<br />
CA: CSLI Publications.<br />
Wälchli, Bernhard. 2005. Co-compounds and Natural Coord<strong>in</strong>ation. Oxford: Oxford University<br />
Press.<br />
Wechsler, Stephen. 2009. ‘Elsewhere’ <strong>in</strong> gender resolution. In K. Hanson and S. Inkelas,<br />
eds., The Nature of the Word-Essays <strong>in</strong> Honor of Paul Kiparsky, pages 567–586.<br />
Cambridge, MA: MIT Press.<br />
Wechsler, Stephen and Larisa Zlatić. 2000. A theory of agreement and its application to<br />
Serbo-Croatian. Language 76(4):799–832.<br />
Wechsler, Stephen and Larisa Zlatić. 2003. The Many Faces of <strong>Agreement</strong>. Stanford, CA:<br />
CSLI Publications.<br />
W<strong>in</strong>ter, Yoad. 1997. Choice functions and the scopal semantics of <strong>in</strong>def<strong>in</strong>ites. L<strong>in</strong>guistics<br />
and Philosophy 20(4):399–467.<br />
Yatabe, Shûichi. 2004. A comprehensive theory of coord<strong>in</strong>ation of unlikes. In S. Muller, ed.,<br />
The Proceed<strong>in</strong>gs of the 11th International Conference on Head-Driven Phrase Structure<br />
Grammar. Stanford, CA: CSLI Publications.<br />
Zwicky, Arnold M. 1977. Hierarchies of person. In W. A. Beach, S. E. Fox, and S. Philosoph,<br />
eds., Papers from the Thirteenth Regional Meet<strong>in</strong>g, Chicago L<strong>in</strong>guistic Society, pages<br />
714–733. Chicago.