Leila Lomashvili
Copyright  Leila Lomashvili 2010
A Dissertation Submitted to the Faculty of the
In Partial Fulfillment of the Requirements
For the Degree of
In the Graduate College
As members of the Dissertation Committee, we certify that we have read the dissertation
prepared by Leila Lomashvili
entitled Morphosyntax of complex predicates in South Caucasian languages
and recommend that it be accepted as fulfilling the dissertation requirement for the
Degree of Doctor of Philosophy
Date: 04/14/2010
Heidi Harley
Date: 04/14/2010
Simin Karimi
Date: 04/14/2010
Andrew Carnie
Date: 04/14/2010
Rudolph Troike
Final approval and acceptance of this dissertation is contingent upon the candidate’s
submission of the final copies of the dissertation to the Graduate College.
I hereby certify that I have read this dissertation prepared under my direction and
recommend that it be accepted as fulfilling the dissertation requirement.
________________________________________________ Date: 04/14/2010
Dissertation Director: Heidi Harley
________________________________________________ Date: 04/14/2010
Dissertation Director: Simin Karimi
This dissertation has been submitted in partial fulfillment of requirements for an
advanced degree at the University of Arizona and is deposited in the University Library
to be made available to borrowers under rules of the Library.
Brief quotations from this dissertation are allowable without special permission, provided
that accurate acknowledgment of source is made. Requests for permission for extended
quotation from or reproduction of this manuscript in whole or in part may be granted by
the copyright holder.
SIGNED: Leila Lomashvili
First of all, I would like to thank my advisors Heidi Harley and Simin Karimi in
helping me to accomplish this goal and have been a part of all I have done at the
University of Arizona Linguistics Department. They have been my mentors and to say
more precisely, my virtual guides in every step I took during my study in the program.
Heidi has been a role model, an inspiring and driving force of what I have accomplished
in this dissertation and the qualifying papers, which have dealt with various aspects of the
morphosyntax of South Caucasian languages. In these works, which naturally led to my
dissertation topic, I learned from Heidi how to approach various linguistic phenomena to
create the most comprehensive account within existing frameworks as well as to develop
my own take on them. Likewise, I am particularly indebted to Simin in learning the
structure and logic of linguistic argumentation and presenting it to a reader in the clearest
form possible. Both of them have been extremely helpful during my work on the
dissertation reading and giving comments on multiple drafts of chapters, and giving
suggestive and pithy comments whenever I needed them.
I also wish to thank my other committee members: Rudolph Troike and Andrew
Carnie. I list them in a chronological order because it was Rudy who inspired me initially
to pursue linguistics as a profession and particularly its theoretical aspect. I came to the
Program of English Language and Linguistics at the University of Arizona as a master’s
student of Russian to gain deeper understanding of linguistic theories and with the very
first classes taken from Rudy Troike I got enchanted by the depth and scope of research
in the field that Rudy presented. It was mainly he who advised me to venture into the
profession of linguist and specialize in theoretical morphosyntax, a field I fell in love
with during his lectures. Andrew has been my chief inspiring figure throughout my entire
career in the Linguistics Department and particularly, in the first year when I became
fascinated with his homeworks and linguistic tasks completed every week in his
introductory Syntax course. I remember those assignments as most challenging and at
times provoking strong emotions by solving them. Andrew was the first instructor in
syntax who supported my decision to major in syntax/morphology and probably defined
my future for many years to come. I would like to thank all these people for all their
extraordinary support in all I have been through these years.
Also, I want to thank many of my linguist friends and particularly, Elly van Gelderen
in her inspiring support during my Doctoral career. I benefited immensely with her indepth comments on my qualifying papers and wise questions during my presentations at
the conferences. Without her, many of my achievements in this program would not be
I also want to extend my gratitude to my instructors and co-instructors in the
Linguistics department: LouAnn Gerken, Cecil McKee, Amy Fountain, Ying Lin, Mike
Hammond, Massimo Piattelli-Palmarini, Diana Archangeli, Tom Bever, Adam Ussishkin
and many others in inspiring and directing me toward this goal. The discussion of various
parts of this work with my friends and classmates at different times has also been
extremely useful and without the following people this project would not be probably
possible: Tatyana Slobodchikoff, Mercedes Tubino, Alan Hogue, Brian Gordon, Collin
Gorrie, Shannon Bischoff, and friendship with Sumayya Racy, Lindsay Butler, Jaime
Parchment, Amy LaCross, Jeff Punske, Yosuke Sato, Jared Francom, Michael Anderson,
Mans Hulden, Dan Siddiqi, Jae-Hoon Choi, Jae-Hyun Sung, Sylvia Reed, Greg Key,
Deniz Tat, Dana Osborne, and many others in the Linguistics and Anthropology
I want to thank my friends outside of academia for being so supportive in my pursuits
for many years: Paulette Williams, Ken Walker, Yang Yang, Nino Buachidze, Nino
Amiridze. My family deserves a special tribute as it has been a major driving force
behind my efforts all these years: my parents, mom, Lamara and dad, Anzor for their
patience and inspiration, my twin brother Koba without whom this dissertation would be
impossible and my sister Tsisana with her constant encouragement and love. Their
contribution to my success is invaluable and cannot ever be sufficiently repaid no matter
how much effort I will put to it in the future.
Finally, I wish to thank my mom for her love and support. This dissertation is
dedicated to her.
LIST OF TABLES………………………………………………………………………..9
CHAPTER 1 INTRODUCTION………………………………………………………..13
1.0 Goals of the dissertation…………………………………………………………...13
1.1 Theoretical approaches to the study of complex predicate
1.2 Argument structure of causative and applicative constructions ………………….18
1.2.0 Introduction…………………………………………………………………..19
1.2.1 Argument structure of causatives…………………………………………….19
1.2.2 Argument structure of applicatives…………………………………………..22
1.3 Variation of causatives and applicatives in terms of complement size…………...26
1.4 Theoretical framework……………………………………………………………30
1.4.0 Distributed Morphology (DM)……………………………………………...30
1.4.1 Chomsky’s (1999, 2001) notion of phases …………………………………35
1.4.2 Contextual allomorphy as evidence for phasehood
(Embick 2009)………………………………………………………………36
1.4.3 Kratzer (1994) on the VoiceP and external arguments……………………...39
1.4.4 Cuervo (2003) on different flavors of little vs………………………………42
1.5 Status of templates and assumptions about the morphosyntax
of Georgian and related languages………………………………………………..44
1.5.0 Structure of the verbal template……………………………………………...45
1.5.1 Assumptions about language-specific templatic constraints………………...50
1.6 Case and agreement with various classes of verbs……………………………......52
1.7 Outline of the dissertation…………………………………………………………57
2.0 Introduction………………………………………………………………………..60
2.1 Organization and outline of analysis……………………………………………....60
2.2 Assumptions……………………………………………………………………….62
2.3 Inchoative-causative alternation…………………………………………………..66
2.3.0 Introduction…………………………………………………………………..66
2.3.1 Causatives of inchoative verbs……………………………………………….67
2.4 Unergative verbs of various types and their causatives…………………………...72
2.4.0 Hale & Keyser (2002) on unergatives and Noun-Incorporated
2.4.1 Adjective-Incorporated vPs and their causatives…………………………....78
2.4.2 Syntactic and iterated causatives of unergative verbs……………………….82
2.5 Transitive/causative alternation………………………………………………….. 88
2.5.0 Introduction…………………………………………………………………...88
2.5.1 X makes Y do V alternation…………………………………………………..89
2.5.2 Iterated causatives………………………………………………………….....95
2.6 Adversity causatives in Georgian………………………………………………..100
2.7 ‘Pretend’-type predicates and their causatives…………………………………...110
2.7.0 Introduction………………………………………………………………….110
2.7.1 Derivation of ‘Pretend’-type complex verbs………………………………...111
2.7.2 Causative alternation of ‘pretend’-type verbs…………………………….....115
2.8 Psych verbs and causative alternation…………………………………………....119
2.8.0 Empirical base of state, dynamic passive, and activity
Psychological predicates……………………………………………………119
2.8.1 Analysis of the morphosyntax of psych verbs and their causatives………....126
2.9 Causative predicates in related languages (Mingrelian and Svan)………………132
2.9.0 Causative alternation is Svan………………………………………………..132
2.9.1 Causative alternation in Mingrelian………………………………………....136
2.10 Conclusions……………………………………………………………………..138
3.0 Introduction……………………………………………………………………....141
3.1 Theoretical goals…………………………………………………………............141
3.1.0 Outline of the chapter………………………………………………..……..141
3.1.1 McGinnis (2004) on applicative structures and phases…………………....145
3.1.2 Transitivity restriction on high and low applicatives
(Pylkkanen 2002)………………………………………………………….149
3.1.3 Cuervo (2003) on low and high applicatives in Spanish…………………..152
3.2 Low applicatives in Georgian……………………………………………………159
3.2.0 Introduction………………………………………………………………...159
3.2.1 Low Recipient applicatives………………………………………………...159
3.2.2 Low Source applicatives…………………………………………………...170
3.2.3 Low state applicatives: Possessor Datives (AT)…………………………...172
3.3 Low applicatives of unaccusative and inchoative verbs………………………....176
3.3.0 Introduction………………………………………………………………...176
3.3.1 Empirical data and syntactic analysis………………………………...........178
3.3.2 Analysis of morphological marking ……………………………………….183
3.4 Low applicatives of Noun/Adjective-Incorporated predicates…………………..186
3.5 Reflexive applicatives……………………………………………………………190
3.5.0 Introduction…………………………………………………………………190
3.5.1 Analysis of morphosyntax of reflexive applicatives………………………..194
3.6 Possessor Datives as low applicatives…………………………………………...198
3.6.0 Introduction………………………………………………………………...198
3.6.1 Low applicatives of stative eventualities…………………………………..199
3.6.2 Other applicative verbs in applicative constructions………………………203
3.7 Four-place predicates as hybrid type applicatives……………………………….205
3.7.0 Introduction………………………………………………………………...205
3.7.1 Analysis of morphosyntax of four-place applicatives……………………..210
3.8 High applicatives in Georgian…………………………………………………...215
3.8.0 Introduction………………………………………………………………….215
3.8.1 High applicatives with stative unaccusative predicates……………………..216
3.8.2 High applicatives with other unaccusative predicates………………………220
3.8.3 High applicatives with dynamic activity verbs…………………………….. 223
3.9 Applicatives in related languages (Mingrelian and Svan)………………………..227
3.9.0 Introduction…………………………………………………………………227
3.9.1 Applicatives in Mingrelian…………………………………………………227
3.9.2 Aplicatives in Svan…………………………………………………………235
3.10 Conclusions………………………………………………………………………240
CHAPTER 4. CONCLUSIONS………………………………………………………..242
REFERENCES………………………………………………………………………... 248
TABLE 1. Morphological template of the Georgian verb…………………………. …...46
TABLE 2. The structure of the series in Georgian………………………………………53
TABLE 3. Case marking of transitive verb arguments………………………………….53
TABLE 4. Georgian agreement morphology…………………………………………….55
TABLE 5. Transitive and causative predicates in Mingrelian………………………….136
TABLE 6. Morphological marking of causatives of various verb classes……………..139
TABLE 7. Low applicatives in Georgian……………………………………………....144
TABLE 8. High applicatives in Georgian………………………………………………145
1, 2, 3- Person agreement features in verbs
Abs- Absolutive case
Acc- Accusative case
AOR- Aorist series
APPLIC- Applicative/version morpheme
CAUS- Causative morpheme
CL- Clitic
Dat- Dative case
DER- Derivational affix
Erg- Ergative case
Evid- Evidentiality marker
FUT- Future tense
Gen- Genitive case
INF- Infinitive marker
Instr- Instrumental case
Loc- Locative case
Nom- Nominative case
NONACT- Nonactive Voice marker
NPN- Nominalizer affix
O (with numbers) – Object
Obl- Oblique case
Pass- Passive Voice marker
PAST- past tense mamker
PERF- Perfective series
PL- Plural number
Post- Postposition
PRES- Present series
Pretend- Pretend-type functional head
PREV- Preverbal affix denoting aspect, direction, or the intensity of an event
REECIP-Reciprocal marker
REFL- Reflexive Voice morpheme
S (with numbers) – Subject
Sg- Singular number
SCR- Screeve marker
SER- Series marker
SUBJ- Subjunctive mood marker
TAM- Tense/Aspect/Mood marker
TH- Thematic markers
Tns- Tense morpheme
VOICE- Voice marker
The argument structure of complex predicates such as causatives and applicatives is
closely associated with the functional heads that introduce core and non-core arguments:
Voice, causative and applicative. These elements merge in a sentence structure at various
cycles of derivation and take complements whose ‘size’ accounts for the meaning and
thematic interpretations of these arguments. This thesis shows that the variation in the
meaning of causatives and applicatives is not so much due to the morphological
realization of the relevant heads in these predicates (causative and applicative) ,but rather
the complement ‘size’ and the place where these heads are introduced. Specifically,
lexical causatives result from the CAUSE head selecting for a RootP complement while
syntactic causatives include two CAUSE heads taking argument-full vP as their
complement and these CAUSEs are realized by separate phonological exponents at the
Morphological Structure. In applicatives, Goal, Recipient or possessor interpretations
may be obtained not only for non-core applied arguments in double-object constructions
but for the highest external arguments of two-place predicates and also, for dative internal
arguments of internally-caused events. It is argued that external arguments take on the
meaning of non-core applied arguments in double-object constructions due to the
different type of Voice head such as reflexive. The morphological realization of these
predicates is distinct from the ones in which applied arguments represent additional, noncore elements of DOCs. The thesis shows that along with the change in morphology the
syntactic structure of various types of applicatives accounts for the resulting meaning of
complex predicates.
1.0 Goals of the dissertation
This dissertation focuses on the morphosyntax of complex predicate constructions in
three polysynthetic languages of the South Caucasian language family: Georgian,
Mingrelian, and Svan. In this work, I follow Baker (1996) who defines complex
predicates as verbal structures containing at least two morphemes each marking a phrasal
argument in the -grid. According to this definition, complex predicates include the
causative, applicative, and possessor raising structures. The focus of this dissertation is
limited to causative and applicative constructions. This is partly because the
morphological varieties of such constructions in these languages are sufficiently diverse
so that they could be treated in a work of this scope. Also, causatives, and applicative
structures have certain morphological similarities in that they contain morphemes
indexing both core and non-core arguments in their theta-grid, such as the external
argument, theme, and the goal/benefactee arguments. In addition to this similarity, these
constructions are viewed as syntactically and morphologically complex structures,
containing a minimum of two phrasal arguments, and at least two morphemes.
Stemming from the above definition of complex predicates, the primary goal of the
dissertation is to explore the formal properties of causative and applicative constructions
and to account for mismatches that arise between the syntax-semantics and the
morphology of these constructions. Such mismatches are common cross-linguistically.
They abound even in languages which do not morphologically mark the causative,
applicative, or any other alternation. An example of such a mismatch is when instead of
the causative or applicative morpheme itself, a Vocabulary Item (VI) realizing some
other functional head in the left periphery of the clause shows up (such as reflexive, or
voice) to mark the causative or applicative alternation. Therefore, the goal of this work is
to propose a unified theory of how the meaning is encoded in syntax and
morphophonology. The interface between these systems in causative and applicative
constructions will be subject to scrutiny in Chapters 2 and 3 of this dissertation.
1.1 Theoretical approaches to the study of complex predicate constructions
The syntax-all-the-way-down approach of Distributed Morphology is adopted here.
This approach is relatively recent and has been initiated and advocated by Baker (1985,
1988), Halle & Marantz (1993, 2002), Marantz (1997), Travis (2000), Pylkkänen (2002),
Cuervo (2003), and Harley (1995, 2008) among others for a wide range of verbal and
nominal constructions, including causative and applicative constructions. This approach
has a single engine of structure creation for both word and sentence formation.
Consequently, lexical-semantic representations are syntactic representations, and no
mapping problems arise between them. This contrasts with lexicalist approaches (for
example, Levin & Rappaport Hovav 1995), where lexical complexity has been argued to
be different from the complexity at the sentential level. Lexicalist approaches posit a need
for linking rules specifying how lexical-semantic representations map onto the syntax.
Take the following linking rule from Levin & Rappaport (1995):
(1) Linking rules
a) “Break: [[ x DO-something] Cause [y become BROKEN]
b) Immediate Cause Linking Rule
The argument of a verb that denotes the immediate cause of the eventuality
described by the verb is its external argument” (Levin & Rappaport Hovav
c) “Directed change Linking Rule
“The argument of the verb that corresponds to the entity undergoing the directed
change described by that verb is its internal argument” (Levin & Rappaport
Hovav 1995:146).
These types of rules were considered to connect the autonomous modules of the
grammar where lexical-semantic and syntactic representations are processed. The
underlying idea of the lexical-semantic approach is that certain semantic notions are
relevant to determining a verb’s argument structure. For example, in causative
constructions, the potential adicity of a verb is determined based on the distinction
between internally- and externally-caused events. The verbs denoting internally caused
changes of state (such as flower, bloom, blossom, and decay in English) in which the
cause is inherent to the natural course of development (Levin & Rappaport Hovav 1995,
henceforth, L & RH), resist the causative alternation cross-linguistically:
(2) The verbs of internal change resist causativization
a. Cactus bloomed/blossomed/flowered early.
b. * The gardener bloomed/blossomed/flowered the cactus early. (L & RH, 1995: 97)
In contrast to these verbs of internal causation, externally-caused eventualities may be
both transitive and intransitive. These regularly have transitive causative counterparts.
Nevertheless, there are exceptions:
(3) Verbs of external change resist causativization
a. The baker cut the bread.
b. *The bread cut.
c. The assassin murdered the senator.
d. *The senator murdered.
Thus, the lexical semantic properties of events, such as internal and external
causation, are of crucial importance in determining whether the verb can participate in the
argument structure alternation. L & RH argue that this distinction between externally and
internally-caused events roughly corresponds to the unaccusative/unergative distinction
too. Many unergative verbs denote internally-caused events while unaccusatives
generally are derived from externally-caused verbs. Based on this correspondence then, it
follows that unergative verbs may not participate in the causative/inchoative alternation,
as illustrated by the impossibility of causatives in the following English pairs:
(4) Causatives of unergative verbs are not possible
a. Sheila laughed.
b.* Sheila laughed John.
c. Jack danced.
d. *Jack danced Sheila.
However, L & RH argue that languages that mark the causative alternation
morphologically often allow causatives of internally-caused events, as in the Hebrew
examples in (5):
(5) Hebrew causatives
a. Hu rakad.
He danced
‘He danced.’
b. Ha-nagan
‘The musician made him dance.’
c. Hu
‘He ran.’
d. Ha-meamen
‘The coach made him run.’ (L & RH 1995: 98)
Chapter 2 of this dissertation shows that unergative verbs in Georgian and related
languages can participate in the causative alternation without any restriction, regardless
of their lexical semantics. Similarly, verbs of appearance and existence (such as exist,
appear, emerge, etc.), which have been argued to resist causativization crosslinguistically, do participate in the causative alternation in the languages under
investigation here. I argue that lexical-semantic approach does not give sufficient clearcut criteria for the classes of verbs participating in this alternation.
Hale & Keyser’s (henceforth, H & K 1993) approach removes the need for the linking
rules between the argument structure and lexical semantic interpretations, as the latter is
the same as the argument structure (a welcome consequence for the simplification of the
architecture of the grammar). This theory takes the syntax of word and sentence
derivations as the source of compositional meaning of complex expressions.
The rest of the chapter is organized as follows: in Section 1.2, the argument structure
of causative and applicative constructions is introduced; Section 1.2.1 discusses the
argument structure of causatives, and Section 1.2.2 the structure of applicatives; Section
1.3 explains the source of syntactic variation in causatives and applicatives; the
theoretical framework of the research carried out in the dissertation is discussed in
Section 1.4, which includes an overview of the framework of Distributed Morphology,
Chomsky’s (1999, 2001) notion of phases, Embick’s (2009) theory of contextual
allomorphy; Cuervo’s (2003) analysis of event-introducing heads (the varieties of little
vs) along with Kratzer’s (1994) instantiation of VoiceP is presented; Section 1.5
discusses some basic background facts of verbal morphology and the status of templates
in Georgian and in linguistics theory in general; Case and agreement morphology with
various classes of verbs are discussed in Section 1.6 and the outline of the dissertation
concludes the chapter in Section 1.7.
1.2 Argument structure of causative and applicative constructions
1.2.0 Introduction
Most work on syntax-semantics of argument structure is based on the idea that verbs
express different types of eventualities and that arguments express participants in events.
This idea naturally extends into a claim that the relations between verbs and arguments,
and between arguments themselves are built out from event structure, i.e. event types and
predicates that verbs express (Davidson, 1967). In (6a), for instance, the cause event
described by the verb melt relates the subject John to the event of ice melting, while in
(6b), the object argument of (6a) relates to the event of melting, but in this case without
any causing event:
(6) Causative/inchoative alternation of English
a. John melted ice.
b. Ice melted.
It is evident that the event structure in (6a) is more complex since it combines two
events: the causing and melting. Moreover, the two arguments (John and ice) are related
to this complex event, while in (6b) only one argument is related to the simplex event of
melting (ice). These two types of events illustrate a syntactic alternation referred to in the
literature as the inchoative/causative alternation.
Consider next how the dative argument expressing the possession or the
Benefactive/Recipient relation is licensed in the following:
(7) Applicative alternation
a. John bought Anne a pastry.
b. John bought a pastry. (Adapted from Levin & Rappaport 1995)
In these sentences, the event structure is not composed of two separate events, as in
(6a), where the verb combines the causing and the melting events. So we need to
determine what is licensing the arguments in this construction.
The insights incorporated in the analysis of the argument structure of complex
predicates in this work are mainly drawn from H & K’s (1993) work and developed by,
among others, H & K (2002), Borer (1994), Harley (1995), Marantz (1997), Travis
(2000), Nash (2002), Pylkkänen (2002), and Cuervo (2003). Within this framework,
syntactic elements such as functional heads and DPs are the building blocks of event
structure, whose meaning is interpreted compositionally.
In what follows, I present some basic assumptions about the argument structure of
these constructions.
1.2.1 Argument structure of causatives
Hale & Keyser (1993, 2002), among others, posit that the argument structure of
causatives involves an additional argument that is interpreted as a causer of the event
described by the verb. The following sentences from English and Georgian illustrate this:
(8) English
a. Inchoative
b. Causative
The cake baked.
Lisa baked the cake.
(9) Georgian
a. Inchoative
da- lp’a.
‘Potatoes rotted.’
b. Causative
‘Dato rotted potatoes.’
prev- CAUS- rotted
As seen in (8)-(9), the inchoative/causative alternation involves the addition of the
causer argument to the argument structure of a one-place inchoative verb. Syntactically
the causative structures in these examples are non-distinct in English and Georgian.
However, morphologically we see a difference. In Georgian, the causative head in
inchoative/causative alternation is marked morphologically (affix a-), while in English it
is not. Chapter 2 explores the conditions on the realization of this morpheme in Georgian
and whether there is a correlation between the syntax of causatives and their
morphological shape.
I follow Pylkkanen (2002) in claiming that the causative head need not introduce the
causer argument, and that causativization does not always increase the number of a verb’s
syntactic arguments (Pylkkänen 2002:75). Rather the causative head (CAUSE),
introduces a syntactically implicit event argument that contributes to the semantics of the
causative verb. Pylkkänen considers two sources of variation in terms of argumentprojecting properties of the causative head: in languages like Finnish and Japanese,
where the CAUSE is independent of the Voice 1, the CAUSE never introduces a causer.
Instead, in these languages, she argues that the causer is projected by the Voice head
(Marantz (1986), Kratzer (1994) among others)). In English, by contrast, since the
CAUSE bundles with the Voice0 as a single syntactic head, both introduce causative
event semantics and the causer argument. Observe these differences in (10):
(10) Variation in Voice-bundling in causatives (Pylkkänen 2002:76)
a. Non-Voice-bundling causative
b. Voice-bundling causative
(Japanese, Finnish)
[Voice, CAUSE]
Georgian and related languages with their non-zero CAUSE morphemes fit the
typology of languages with the non-Voice bundling causative head. The relevant question
is what evidence supports the separation of CAUSE from Voice in languages like
Japanese, Finnish, Georgian, etc. Pylkkänen identifies Japanese adversity and Finnish
desiderative causatives as structures where the causer argument is not projected at all
even by the Voice head. In these constructions, the nominative subject is interpreted as an
affected argument of the event described by the verb, but not as a causer:
(11) Japanese adversity causative
See Section 1.6.4 for the motivation why the little v head is distinct from Voice.
X in (7) is an external causer argument projected by the Voice0.
(i) ‘Taro caused his son to die.’
(ii) ‘Taro’s son died on him.’ 3 (Pylkkänen 2002: 81)
Taro is introduced neither by the CAUSE head, nor by Voice0, as the adversity
causative lacks external arguments. The causing event is clearly present and
morphologically the CAUSE is realized as –(s)ase, but this head does not relate any
participant to the causing event. Pylkkänen demonstrates that the separation of the
CAUSE from the Voice is a natural consequence of the inability of CAUSE to introduce
a causer argument in languages which have similar structures. In chapter 2, I pursue a
similar analysis of the CAUSE in adversity causatives (in which CAUSE does not
introduce an external argument), but the analysis is still different from Pylkkanen’s in
that, in Georgian, adversity causatives are productive and their meaning is predictable.
This will be argued to be a result of their syntax, specifically, the kind of complement
that the CAUSE takes in these structures.
To summarize my assumptions about causative constructions:
a. causative structures universally contain the causative functional head (CAUSE)
and this head introduces a syntactically implicit event argument;
b. the causer argument may or may not be introduced into the structure;
c. in adversity and desiderative causatives where an external argument is not
introduced, the nominative argument is often interpreted as affected by the event
expressed by the verb.
1.2.2 Argument structure of applicative constructions
Pylkkanen (2002) argues that the adversity causatives generally are not productive structures like
syntactic causatives in Japanese. Instead, they represent a variety of non-productive lexical causatives
whose meaning may be ambiguous between two interpretations as illustrated in (11). Adversity causatives
in Georgian and related languages do have a predictable meaning and they are productive, as will be shown
in Section 2.6.
This section introduces some core assumptions about the argument structure of
applicatives. Applicatives generally are defined as double object constructions in which
besides the lexical (VP) and event-introducing verbal projections (vP), which may
introduce arguments, the ‘additional’ functional head (Appl0) adds an ‘applied argument’.
Usually, semantic relations are established either between this applied argument and the
theme, or between the applied argument and an event described by the verb (Pylkkänen
2002, Cuervo 2003). The first relation, between a non-core applied argument and the
theme, cross-linguistically is conceptualized as that of a stative possessor, a recipient, or a
source (Cuervo 2003). These relations are referred to as the low applicative due to the
lower merging site of the Appl0 head, namely, below VP:
(13)a. Low Recipient applicative
Sheila baked John a cake.
b. Low Source applicative (Finnish)
Liisa.Nom sold
‘Lisa sold Matti’s house’. (Lit: ‘Lisa sold a house from Matti’.)
c. Low Possessor applicative (Spanish)
a Valeria
Pablo CL.DAT admires
‘Pablo admires Valeria’s patience.’ (Lit: Pablo admires Valeria the patience).
Although the meanings of these expressions are different from each other, the
structural relation between the applied argument and the theme is similar as shown in the
Low Applicative
Ext. Arg.
By contrast, the high Appl0 head takes the VP as a complement. In high applicatives,
the semantic relations are different: an applied argument may be experiencing or
undergoing the effects of an event expressed by the verb, but in no way does the applied
argument enter into a direct semantic relation with the theme and obtain the thematic
interpretations of benefactee, recipient, etc. Observe the examples and the syntax of this
relation in (15):
(15) High applicatives
a. Katonga
ya- kwant- i-dde
Past-hold- APPL-PAST
‘Katonga held the bag for Mukasa.’
b. n-a-y-lyi-`a
‘He is eating food for his wife.’
(Bresnan & Moshi 1993:49-50)
(Pylkkanen 2002: 25)
High Applicative
Applied Arg.
Cuervo (2003) argues that psychological verb constructions may also be interpreted as
high applicatives in which the high Appl0 head projects the dative experiencer argument,
and the latter is interpreted as ‘a Benefactee’ of an event:
A Daniela
‘Daniela likes cats.’
gustan los
like.PL the
A Daniela
Los gatos
(Cuervo 2003: 31)
The nominative theme (los gatos ‘cats’) is the subject of the stative predicate gustar
‘like’, while the Experiencer is external to the vP. The latter argument is argued to be
licensed by the high Appl0 head.
Notice that the number of arguments involved in applicative relations (high and low)
is essentially the same, but there is a semantic difference between them, which can be
attributed to the participation of an event argument. In high applicatives, this event
argument enters into a relation with the applied one, while in low applicatives, two DP
arguments, the applied and the theme are in close-knit relation with each other. As
mentioned above, the applied argument must be a recipient, source, or stative possessor
of the theme. By contrast, in high applicatives of psychological predicates shown in (16),
the argument structure contains one less phrasal argument since only the experiencer and
the theme arguments are projected, and not the external argument. In such structures, the
experiencer does not enter in any above-mentioned relations of low-applicatives. Chapter
3 addresses the question of whether there is an overt morphological realization of these
relationships in Georgian and related languages.
1.3 Variation of causatives and applicatives in terms of complement type
Besides the voice bundling parameter, causatives can also vary in the type of
complement that CAUSE takes (Pylkkänen 2002). Pylkkänen identifies three kinds of
complement that CAUSE may take: root, VP, and vP. Root-selecting CAUSE results in a
lexical causative, which may have idiosyncratic meaning and is unproductive relative to
syntactic causatives in a language like Japanese (Harley 2008). The VP and the vPselecting causative heads usually form a syntactic causative, which has non-idiosyncratic
meaning and is highly productive across various classes of verbs. The causative heads
that select the vP will be referred to as phase-selecting causatives following Chomsky
(1999, 2001), McGinnis (2000, 2001a, b), and Arad (2003, 2004) among others. Here are
the illustrations of these structures:
(17)a. ROOT-selecting CAUSE
b. Verb-selecting CAUSE
d. Phase-selecting CAUSE
The complement type of CAUSE has further consequences for the event composition
in a sentence. Root-selecting CAUSE behaves like a mono-eventive structure with
respect to the scope of adverbial because there is only one place in the clause where the
adverbial can attach: only VP can be modified in the root-selecting (lexical) causative.
The same prediction can be true for VP-selecting causatives. In the following English
causative of an inchoative verb, the manner adverbial reluctantly is expected to modify
both the causing and the resulting events regardless of whether it attaches to a position
higher or lower than the VP itself:
(18) VP-modifying adverbial:
a. Bill opened the door reluctantly.
b. Bill reluctantly opened the door. (Compare: The door opened reluctantly.)
In these sentences, the adverbial can not modify the causing and the resulting events
separately since zero-derived causatives of inchoatives in English do not allow such
separate modification of events: they are lexical causatives. Bi-eventive interpretations
are found with vP-selecting causatives (i.e. causatives in which CAUSE takes phasecomplete vPs as its complement). Adverbials can modify both the causing and the
resulting events individually depending on the attachment site (above or lower than vP).
Pylkkänen claims that vP-selecting causatives contain two vPs: the head of the first vP
expresses the causing event, and the head of the second vP, the resulting event.
Therefore, each of these events can be modified:
(19) The causing event modified
a. Sue reluctantly made John eat oysters.
b. Sue made John eat oysters reluctantly.
In (19a), reluctantly modifies the causing event and is interpreted as oriented toward
the external causer argument Sue, while in (19b), the same manner adverbial modifies the
resulting eating event, i.e., is oriented towards John. Thus, the adverbial test indicates bieventive properties of vP-selecting causatives. We will also use the test of depictive
modification as a diagnostic of mono- and bi-eventive properties of different types of
causatives. As is illustrated throughout Chapter 2, a depictive modifier can modify the
causee argument of the syntactic causative which matches its case marking with the
depictive while it is impossible to modify the causer argument from the same position
even when the case marking is matched with the relevant argument. The idea is that the
modification of the causer and the causee in syntactic causatives must be local and
presumably, these two arguments are projected in separate ‘clauses’. Therefore, the event
structure of syntactic causatives is more complex than that of the lexical one.
Applicatives also allow us to probe the mono- or bi-eventive properties of the
complements selected by the Appl0 head. High and low applicatives show a distinction in
terms of mono-/bi-eventiveness. Recall that in low applicatives, the low Appl0 head does
not take a verbal complement because it merges below VP licensing the theme and the
applied arguments as in (13)-(14). Since in these structures, the low Appl0 head does not
take a phase-complete complement, they are not subject to the same bi-eventive analysis
as vP-selecting syntactic causatives. Adverbials in low applicatives are expected to
modify higher verbal projections, such as vP and VP, and the structure is interpreted as
mono-eventive. For instance, in (20), the manner adverbial again modifies the baking
(20) Adverbial modification of low applicative
Sheila baked John a cake again.
Returning to high-applicatives, the Appl0 head that merges above VP does not
introduce a separate event into the semantics of complex predicates, i.e. it does not select
for vP complement as illustrated in (17c) for vP-selecting causatives. 4 Cross-
Here we follow Pylkkanen (2002) and McGinnis (2002, 2004) in which it is proposed that the high
applicative relation is established between an individual and an event due to the Appl0 head selecting for
the VP complement. See the tree in (15c) for details.
linguistically, vP is argued to introduce an event argument (Harley 1995, Pylkkanen
2002, Cuervo 2003 among others). One might argue that the event structure of high
applicatives cannot be as complex as that of vP-selecting syntactic causatives where two
events are introduced: the causing and the resulting one (i.e. two separate vPs). Since the
event structure of a high applicative is simpler than that of the vP-selecting causative, the
manner adverbial is expected to modify just the VP complement of the Appl0 head and
show the properties of mono-eventive structures:
(21) Modification of VP-high applicative in Georgian
‘The accident happened to Gia unexpectedly.’
mo- uxda
prev- APPLIC- happen
mo- uxda
prev- APPLIC- happened
‘The accident happened to Gia unexpectedly.’
In these examples, the interpretation of the scope of an adverbial can be the one where
unexpectedly modifies the event described by the verb. Both in (21a) and (21b), the scope
of the adverbial is the same since it modifies the same event of happening.
A detailed explanation of high applicative semantics with the verbs of happening will
be given in Section 3.8 of Chapter 3. The simpler event structure of high and low
applicatives and the size of the complement that both Appl0 heads take means that the
resulting structure will always show mono-eventive properties with respect to adverbial
1.4 Theoretical framework
1.4.0 Distributed Morphology (DM)
This work utilizes the framework of Distributed Morphology (DM) in the analysis of
the morphosyntax of complex predicates in South Caucasian languages. In this
framework (Halle & Marantz 1993, henceforth H&M), word and sentence formation are
viewed as a product of single derivational mechanism, and the morphology itself is
distributed among several components of the grammar. Word formation may happen at
any level where the head movement and merger of structurally adjacent heads create
complex heads (H&M 1993: 112). In DM, the terminal nodes in syntax from which
words are formed do not have any phonological features, and the nodes obtain those
features only at the level of Morphological Structure (MS) by the process called
Vocabulary Insertion. Vocabulary Items (henceforth, VIs) can be underspecified. For
example, the Vocabulary entry for the English verb sink is sufficiently underspecified so
that it can be inserted in either inchoative or causative contexts (Harley & Noyer 2000).
For vocabulary insertion to happen, it is enough that the feature bundle of the Vocabulary
Item (VI) does not conflict with the features of the terminal (H & M, 1993:121). It may
be that several VIs compete for the insertion in the same terminal node at the MS. The VI
that matches the most features of the node will be inserted 5.
The architecture of the grammar in DM is shown in (22):
Known as the ‘subset principle’ (H & M1993).
(22) (Harley & Noyer 1997:3)
Morphosyntactic features
[Det] [1st] [CAUSE] [+pst]
[root] [pl]
Syntactic operations
Merge, Move, Copy
Morphological operations
Logical Form (LF)
Phonological Form (PF)
(Insertion of Vocabulary Items,
Readjustment, phonological rules)
Vocabulary Items
[kæt]:root], [+count],
Conceptual Interface
[+animate], etc.
(Non-linguistics knowledge)
Cat: four legs, feline, pet, purrs,
In this model, it is assumed that roots are underived primitives in syntactic derivations,
i.e. they are ‘atomic’ non-compositional items. The syntax manipulates just placeholders
or roots, and these abstract elements receive phonological content when vocabulary
insertion applies.
Syntactic category in this system is derived from syntactic position. Roots are lexical
elements underspecified for lexical category (Arad 2003, 2005, Marantz 1997, 2000,
Embick 1998). They are marked with the symbol  and are combined with the categorydefining head (such as a adjectival, n nominal head, etc.).
It is of importance to note that according to DM model (Halle & Marantz 1993), the
ordering of constituents within words and the ordering of words within sentences do not
obey the same principles, such as ‘head’, ‘specifier, or ‘complement’ configurations of
syntactic building blocks, which trigger a certain ordering of affixes with respect to stems
and of phrases to syntactic heads. Consequently, different ordering of root/stems with
respect to affixes in various languages has little to do with the merging position of these
elements in syntax. The fact that in Georgian and related languages, the roots are located
in the fourth position of the verbal template followed by 4-5 affixes, and in Navajo, roots
come at the end of the template preceded by about 9 affixes are arbitrary and an affix’s
status as a prefix, suffix, or infix is independent of its syntactic role, although in the
default situation we do expect that the hierarchical morphological structure will mirror a
syntactic hierarchy from which it is derived as per the ‘Mirror Principle’ (Baker 1988) .
This idea will be expanded in Section 1.5 where the status of templates in the theory of
grammar is discussed.
One of the issues dealt within this study is the linearization of the morphemes in a
well-formed phonological word at morphological structure level (MS). As many argue
(Halle & Marantz 1993, Noyer 1997, Embick 2003, 2008 among others), linearization is
a property of phonological structure and in no way X0s get linearized in syntax. Further,
Noyer (1997) argues that “linearization applies in the first phase of Morphological
Structure, supplying adjacency relations to all constituents...” (Noyer 1997:38). He
claims that at MS two demands determine the mapping of the functional heads and their
morphosyntactic representations to phonological strings: (1) morphological wellformedness conditions on morphological words, and (2) two types of rules: readjustment
rules, which change just the phonological material and second, rules, which supply
phonological material (affixes) to stems, essentially discharging the features of
morphosyntactic representations. Some of these rules will appear instrumental for
understanding the morpheme ordering in complex words in this dissertation.
Thus, DM recognizes that at MS morphemes may be inserted to meet universal and/or
language-specific well-formedness conditions (Halle and Marantz 1993:115). An
example of such addition is when subject-verb agreement is implemented by adjoining an
Agr morpheme to the Tns node and then the features of the subject are copied onto the
Agr node. Such addition of terminal nodes at MS changes the number of terminal
elements that might find phonological realization and naturally, disrupt the isomorphism
between PF and syntax. Other processes that may disturb such a one-to-one relation
between terminal nodes in syntax and the elements at MS are: head-to-head movement of
a terminal element and adjunction to a terminal element in the tree; merger of structurally
adjacent nodes but keeping their identities-fusion, fission or impoverishment of certain
features at MS before the Vocabulary Insertion. For discussion on the exact nature of
these processes I will refer the reader to Halle and Marantz (1993). A detailed
explanation of the processes relevant to the analysis of the data in this study is provided
where relevant.
1.4.1 Chomsky’s (1999, 2001) notion of phases
The dissertation assumes the most recent version of the Minimalist Program as argued
in Chomsky (1999, 2001). In these versions of the program, the derivation of syntactic
expressions proceeds cyclically, phase by phase. Chomsky identifies two phrases as
being ‘strong’: vP and CP -- the verbal and the complementizer phrases which have a
‘propositional’ structure. The ‘strong’ phases are spelled out cyclically. The cyclicity of
syntactic derivations and spellout has natural implications for the derivations of words.
One goal of the research here is to explore to what extent the strong phases of syntactic
derivations (vP and CP) can be the sites of cyclic spellout, and more importantly, whether
the functional heads introducing non-core arguments to the argument structure of
causatives and applicatives can also head a phase. In particular, we will be looking at the
status of causative and applicative heads. The key evidence for the cyclicity of these
heads is supplied by the presence of contextual allomorphy in the morphemes inserted
into these heads (Embick 2009). The next section overviews Embick’s findings in this
1.4.2 Contextual allomorphy as evidence for phasehood (Embick 2009)
Embick (2009) claims that morphological operations-- which determine the
phonological form of morphemes — are constrained by the cyclic organization of the
grammar, i.e. by the local domains that are defined by syntax. He develops a localist
theory of allomorphy, in which linear adjacency and cyclic locality interact to produce a
theory of allomorphic interaction. Take, for example, contextually-defined allomorphy,
illustrated in (23) for English past tense suffixes. We can view this as the result of
Vocabulary Insertion assigning a phonological form to the morphemes computed in
syntax. VIs (Vocabulary Items) compete for insertion into a given node. The most
specific item, matching the node’s featural content, gives that node its phonological
(23) VIs for the English Past tense
-t  TPAST / ____ {LEAVE, BEND,…}
-  TPAST / ____ {HIT, SING, …} 6
-d  TPAST
(Embick 2009: 9)
As seen from these items, Roots like BEND and HIT ‘condition’ the insertion of –t
and - allomorphs. –d is a default elsewhere item for the [Past] and is only inserted when
the context for other morphemes is not defined. The cyclic domains of syntactic
computations may render the insertion of certain VIs possible as opposed to other VIs.
For example, the idea that roots are acategorial elements and that they merge with the
category-defining functional heads such as a, n is well-supported (Arad (2000), (2003),
Marantz (1997), (2000) among others)). When these category-defining heads merge with
roots, they are root-attached, i.e. they are in the inner domain. The same categorydefining head may be attached to a structure that has already been categorized by another
head. In such instances, the head which merges later is argued to be an outer domain
head. Embick & Marantz (2008) argue that both inner and outer domain heads that
categorize roots are cyclic, which means that when merged into a structure, they trigger
spellout, the operation that sends the part of the syntactic structure to the interface
We may argue that SING is a zero-derived past form much like as HIT since it does not change the
affix in the PAST, but rather the Root.
components, PF and LF. Observe these inner and outer domain category-defining heads
in the examples adapted from Embick (2009):
(24) v merged with  ROOT
(25) Two category-defining heads and ROOT
 root
In (24), the single category-defining head v merges with the root first defines a phase.
In (25), we have two domains because there are two category defining heads (a and v).
An important property of outer heads like a in (25) is that vocabulary insertion into this
position is not root-conditioned. By contrast, non-cyclic heads (like T0 in (26)) in the
outer domain allow contextually-defined allomorphy. For example, English past tense
suffixes –t/- still show the above-noted root-conditioned allomorphy even though the
tense morpheme is not root-attached, and it is in outer domain as seen in (26). It is argued
that the root first merges with the category-defining v and that T0 is introduced later in the
(26) English Past tense
(Embick 2009:10)
T0 is not a cyclic head as argued in Chomsky (1999, 2001) since it is not
phonologically independent and does not have a ‘propositional’ structure. The cyclic v
does not block T0 from showing contextual allomorphy conditioned by roots. Therefore,
Embick posits that there is a critical need to accommodate cases where the functional
heads attach higher than the cyclic category-defining heads yet still show root-determined
Another problem relates to the contextual allomorphy displayed by the nominalizer
category-defining head in English gerunds like John’s marrying of Kathy. In gerunds, the
nominalizer n morpheme spelled out as –ing attaches to the structure already verbalized
by v:
(27) Gerund derivation
[n, -ing]
[v, -]
(Embick 2009:11)
The VI –ing inserted into the outer cyclic head n shows no root-determined
allomorphy, although on the surface one might think that it is attached to the root.
Embick concludes that there are evidently no cases in which an Outer cyclic head shows
Root-determined allomorphy as opposed to non-cyclic ones.
Based on this evidence, he develops two theoretical generalizations that capture the
allomorphic variation of the English past tense morphemes and the lack of this
allomorphy with respect to the nominal category-defining heads (like –ing in (27)) in
gerunds. Here are these generalizations:
“a. ….  ] x] Z ]
Generalization 1: Non-cyclic Z may show contextual allomorphy determined by
, as long as x is not overt.
b….] x] y]
Generalization 2: Cyclic y may not show contextual allomorphy determined by
, even if x is not overt.” (Embick 2009: 11)
The idea behind these sequences of functional heads in (28a & b) is that outer noncyclic heads can see across an Inner cyclic node, but outer cyclic heads cannot. In (28),
lower case x, y are cyclic heads, upper-case Z is non-cyclic head, and  is an allomorphyconditioning element, i.e. Root.
Thus, in (28a), non-cyclic Z may be influenced by , if the intervening head x is not
phonologically realized. However, cyclic y may not be influenced by , because y will
trigger the spellout of the structure assembled by the point when y merges. Thus, the
phonological realization of x is irrelevant for the contextual allomorphy of the outer
cyclic head y.
Another conclusion drawn from Embick (2009) is that contextual allomorphy is
possible only with elements that are concatenated. However, in some cases, superficially
adjacent elements cannot influence each other allomorphically because they are not
active in the same PF cycle and are separated by a ‘strong’ phase (such as vP or CP as
argued in Chomsky (1999)). In the former case, the conclusion is straightforward: only
those morphemes that are adjacent can influence each other, while, in the latter case,
superficially adjacent elements may be separated by zero-realized cyclic heads that may
render the contextual allomorphy of the outer cyclic head impossible. These conclusions
will be taken into consideration when examining the data of causative and applicative
predicates in this work.
1.4.3 Kratzer (1994) on the VoiceP and external arguments
Since Kratzer (1994, 1996), it has been generally assumed that an external argument is
not a ‘true’ arguments of a verb, but is rather introduced by an inflectional head outside
of the VP shell (such as vP). Using some of the Neo-Davidsonian argument association
methods, she argues that external arguments have different semantics due to the way they
are introduced in the syntax of verbs. She specifically argues that the external argument is
introduced to the syntax by the separate functional projection such as VoiceP and they are
different from internal object arguments in that they never impose specific interpretations
on verbs as internal arguments do. The relevant evidence came from Marantz (1984) in
which the external argument was shown not to form idioms with V0 in contrast to the
arguments projected by VP itself as shown in the following:
(29) Idioms of English
a. throw support behind the candidate
b. throw a party
c. throw a fit (Marantz 1984 quoted from Kratzer 1994:113).
In these structures, while an internal argument may trigger a certain interpretation of
the verb, the external argument fails to do so. Following Marantz, Kratzer concludes that
the external argument is not a ‘true’ argument of the verb and develops this idea into a
coherent theory of VoiceP according to which the external argument is the argument of
Voice0 while the internal one is the argument of VP, specifically, the latter is generated in
the specifier of the VP as illustrated in the following:
(30) Mittie fed the dog.
the dog
(Kratzer 1994:121)
Furthermore, the internal argument in (29) and (30) first composes with the verb
enabling it to impose certain interpretations on the resulting expression, while the agent
argument composes with a separate light verb (or Voice) and then is conjoined with the
lower predicate via a process called Event Identification (EI). This process is basically
the conjunction operation applied to the separate functions followed by co-indexing of
the event variables. Co-indexing happens due to the shared <s, t> type functions in the
input functions as seen in (31). As the result of EI, the denotation of type
<e, <s, t>> , which includes Voice with the external argument, composes with the
denotation of the type <s, t> (including VP with internal argument) to produce a
function of the type <e, <s, t>> which is Voice’ 7 :
<e, <s, t>>
<s, t> ⇒
<e, <s, t>>
The following entities denote individuals (e), events (s) and truth values (t).
b. Brutus stabbed Caesar.
VoiceP x.e. [stabbing(e) & Agent (e, Brutus) & theme (e, Caesar]
Voice’ x.e. [stabbing(e) & Agent (e, x) & theme (e, Caesar)
(By event
x.e. [Agent (e, x)]
e.[stabbing (e) & theme (e, Caesar)
x.e. [stabbing(e) & theme(e, Caesar)
(Pylkkanen 2002: 13)
In the computation of the denotation of complex events the EI allows the same
existential operator to bind into the two event arguments. This kind of EI is assumed to
exist in causative and applicative constructions in this dissertation where Voice head also
introduces an event variable and the external argument to the event structure of these
predicates. Thus, the compositionality of the external argument with the VP is expected
due to this process.
1.4.4 Cuervo (2003) on different flavors of little vs
Cuervo (2003) posits three types of heads that may either themselves be the
complements of various functional heads (such as the ApplHIGH ), or take the Appl0LOW as
their complement:
(32) Three types of heads
a. Event-introducers: little v0s.
b. Argument introducers: Voice0 and Appl0
c. Roots
These three heads may introduce arguments in the syntax by the combination of the
lexical Root and the verbalizing head (little v). Roots combine with different kinds of
little v (Marantz 1991, Harley 1995) to build event predicates, which have different
meanings depending on the type or the ‘flavor’ of these little vs. Based on event structure
research, this work acknowledges that arguments are licensed either by an event predicate
or by an argument-introducing head (Voice0 or Appl0). Roots may also introduce
arguments semantically as their complements (Levin 1999, Kratzer 1996). In this work, I
will argue that these little v0 heads may also introduce arguments to the structure of
causative and other constructions. Specifically, they add causee arguments to the
structure of make X do V syntactic causatives.
Given these assumptions about how the argument structure is composed in the syntax,
Cuervo identifies the following types of little vs that define the meaning of predicates:
(33) Three types of v
a. vDO
c. vBE
Three types of events
dance, run
fall, go, die
like, admire
In this typology of event heads, vDO and vGO introduce dynamic events while vBE
introduces stative eventualities (Bach 1986), which are mainly represented by
psychological predicates and verbs like lie, hold, etc.
Usually vDO heads active structures and combines with Voice0, which introduces an
external argument (Kratzer 1994). It can be assumed that various activity verbs, both
transitive and intransitive (unergatives), are formed with this vDO.
Predicates of change are derived with vGO and include the verbs of movement and
happening. This kind of event does not license an external argument via the Voice0 head.
These are mainly change-of-state verbs, which include events denoting non-volitional
changes and they select for an object DP undergoing the change-of-state (ex. fall and
die). Thus, when vGO combines with roots of unaccusatives, this typically results in a
change-of-state interpretation for the theme argument.
In contrast to the two dynamic vs, vBE combines with Roots to express states and
Roots combining with this head license dative subject DPs in the specifier of VP (verbs
like matter and Spanish be useful). Stative psychological predicates are formed by this
kind of vBE.
Given these three main types of v, the syntactic composition of complex events such as
causatives, inchoatives, and applicatives is accounted for. I follow Cuervo’s taxonomy of
little v heads in chapters both on causatives and applicatives for simple events but
crucially, her decomposition of the causative head (CAUSE) into two different types of v
(vDO and vBE ) will not be accepted for the purposes of presentation in this chapter.
1.5 Status of templates and assumptions about the morphosyntax of Georgian and related
This section overviews some of the basic grammatical features of Georgian and
related languages based on the existing literature on these languages and the writer’s own
observations as a native speaker. Most of the data and functional analyses of morphemes
are drawn from Howard Aronson’s (1990) Georgian: A reader’s grammar, O. Kajaia’s
(2004) Mingrelian-Georgian Dictionary and V. Topuria’s monograph (1967) Svan
All three languages are moderately polysynthetic. The number of lexical and functional
morphemes marked on verbs can grow to up to 10-12 morphemes per word, but is more
typically 5-6. There is a great deal of fusion among different inflectional morphemes
such as between person and number agreement, series, screeve, 8 or mood markers, etc. on
both sides of stems, i.e. in the pre- and post-base positions. However, fusion does not
usually affect the argument structure-changing morphemes, which mark syntactic
alternations such as causative, applicative, etc.
The section extends the discussion on the structure of the Georgian verbal template
along with the language-specific restrictions on the realization of certain morphology in
the pre- and post-base positions. The status of verbal templates in the linguistic theory is
also discussed in light of DM and Baker’s Mirror Principle (1985) highlighting the fact
that the latter principle is often a default situation in the grammar and that templatic
constraints are essentially language-specific well-formedness rules as argued in Noyer
1.5.0 The structure of the verbal template
According to Aronson (1990), the Georgian verb template contains a maximum of
three positions for morphemes before the root and as many as 6-8 positions after the root.
Observe the following scheme of the Georgian verb:
Series and screeves are the terms drawn from the Kartvelian (South Caucasian) linguistics that stand for
the conjugation pattern of verbs in terms of tense, aspect and mood (TAM). There are three series in
Georgian and Svan, and four in Mingrelian. The screeve is a conjugation pattern for just one value of tense,
aspect and mood.
Table 1. The morphological template of the Georgian verb
Person agreement
(subject, object
3 Applicative,
Causative, Voice, etc.
4 Root
6 Causative, Passive, etc.
7 Screeve
8 Theme
9 Causative (screeve)
10 Person agreement/
Before explaining the distribution of these various morphemes in the template, note
that the status of a morpheme in this kind of template, i.e. whether it is a prefix, infix or
suffix, has no direct relation with the syntactic operations combining terminal nodes to
create words before the Vocabulary Insertion. In DM, the hierarchical location of affixes
in words is determined by the syntax but again there need not be a direct one-to-one
relation between the terminal elements in syntax and phonological exponents, neither
need the bracketing of the phonological pieces directly reflect the syntactic bracketing (H
& M 1993:11). As H & M show in this foundational article on DM, Baker’s Mirror
Principle arguing for the transparent relation between the checking of features in the
syntax and the ordering affixes in a word as a result of this checking is not consistently
accurate. In other words, the order of checked features does not always reflect the order
of affixation in the lexicon whereby the features of the innermost affix are checked first
and the subsequent features are reflected in word formation based on their order of
checking in the syntax. H & M reject this principle as too restrictive because natural
languages provide plentiful evidence against such a strict correspondence between the
checking of features on terminal nodes and their phonological realization in their
mapping to the Phonological Form (PF). Given this sometimes arbitrary relation between
the syntax of complex expressions and their phonological realization at PF one might
argue that the structure of the verbal template in a particular language(s) is purely
epiphenomenal and has little to do with syntactic derivations. Therefore, the order of
affixes with respect to Root and the position of the Root with respect to these morphemes
may be considered as the result of language-specific constraints on the morphological
well-formedness rather than the direct consequence of the order in which the checking of
these features proceeds. Thus, the position of morphemes in the verbal or nominal
templates need not be viewed as a direct consequence of their syntax.
In the Georgian verb template shown in Table 3 the order of three pre-base
morphemes is rigid, which means that in relevant syntactic contexts where all three
morphemes are realized, the ordering is the same: first preverbal prefixes are realized
followed by the person agreement markers, and finally, argument-structure changing
morphemes such as causative, applicative, voice, etc. The causative, applicative, or Voice
functional heads have multiple phonological exponents whose insertion depends on the
syntactic environment and the feature content on these nodes. Again the ordering of these
morphemes in the template above may have nothing to do with the order of checking of
relevant features in the syntax.
The preverbal prefixes mark both aspect and directionality and as such, they combine
the features of both inflectional and derivational morphology. These elements resemble
Russian preverbs, which, along with the perfective-imperfective distinction can mark
manner, saturation, or the intensity of an event expressed by the verb. In the latter
function, these preverbal elements are like the derivational morphology in that they
imbue a special meaning to the root, and are not productive. There are fifteen such
preverbal prefixes in Georgian and they are inserted into the initial slot of the template,
although we assume that the functional head which is realized with these prefixal
elements (which presumably is Asp0) does not check its features the latest assuming a
bottom-up derivation of words in the syntax.
The person agreement prefixes are spelled out in the second and the final positions of
the verbal template. A separate section below discusses which arguments trigger insertion
of these morphemes along with the case-marking of relevant DPs. Only one personnumber agreement morpheme appears in the pre-base position even when the verb is
transitive and contains two arguments. The reason for such restricted spellout of the
agreement morphemes will be explained below.
The third position, which immediately precedes the base (root) can be occupied by
various morphemes marking the argument structure of relevant verbs, including
causative, applicative, and voice morphemes. The VIs of one of these functional heads
are in competition for insertion in this position. Various post-syntactic rules are also
implicated in the spellout of these morphemes, to be discussed in Chapters 2-3 where
The order of morphemes in the post-base positions of the Georgian verb template is not
as rigid as in the pre-base environment, and some of the morphemes tend to reiterate
sometimes for no obvious semantic reason. The morphemes that can be repeated are the
thematic markers (especially one of the five suffixes –eb) and the syntactic causative
morpheme marking the make X do V causative alternation (-in). The iterated markers do
not usually contribute the same semantics twice, as in the case of reduplicated causative
suffix: the second –in does not introduce another causing event or causee argument to the
argument structure. Instead, the reduplicated causative morpheme may mark the
series/screeve or mood information as shown in the following Perfective series forms:
(34) The Perfective series verb
ga- m- eket- ebinebinoprev-1S-Voice- make- TH- CAUS- TH- CAUS- TAM‘I would have caused X to make Y.’
To the best of my knowledge the exact function of this iterated causative marker has
not been explained in traditional linguistic literature on Georgian and I provisionally
gloss this morpheme as the conjunctive perfect (screeve) marker.
Thematic markers glossed as TH in this dissertation are widely attested in the
morphology of various verbal and nominal classes cross-linguistically and they are
morphological well-formedness affixes not associated with any specific terminal node in
the syntax. As argued by Halle & Embick (2003), among others, they are dummy
elements and do not realize any particular functional head. In addition, I argue that they
‘signal’ the presence of certain functional morphology in individual languages. In
Georgian and related languages, the thematic marker (-eb) has been argued to indicate the
presence of the causative morpheme (Aronson 1990), although this relation is not always
clear. This will be illustrated in Chapter 2.
Among other morphemes inserted into various post-base positions, the morphemes
marking causative, voice, agreement, and case are the most important for understanding
further discussion in this work. The causative morpheme –in marks the causative head
added to unaccusative, transitive, or other classes of verbs. The morpheme marking the
passive alternation also shows up in the same post-base position in relevant syntactic
contexts. Finally, the person-number agreement morphemes inserted in the final position
may be fused with the series markers or be absent entirely.
1.5.1 Assumptions about language-specific templatic constraints
An essential note on Georgian verbal morphology is concerned with language-specific
constraints on the spellout of certain morphemes in the verbal template. Such constraints
limit the positions-of-exponence for the phonological pieces inserted into terminal nodes.
For example, only one person agreement morpheme can be inserted in the pre-base
position, as well as one argument structure-changing morpheme. I suggest that this type
of restriction is similar to the templatic structure discussed by Noyer (1997) in the Arabic
verb paradigm and in other Afro-Asiatic languages, whereby a single terminal node is
associated with two positions-of-exponence. This situation can result in ‘discontinuous
bleeding,’ in which the morpheme inserted in one position-of-exponence bleeds the
insertion of another morpheme in another position-of-exponence. Following Noyer, I
assume that in Georgian and related languages, the phi- and other substantive
morphosynatctic features are subject to arbitrarily-imposed constraints on the verbal
template. One such constraint limits the insertion of person features of a subject or an
object into the second and the final positions as shown in the following scheme:
(35) Restriction on person agreement morphemes:
base (root)
person agreement
person/number agreement
(subject, object…)
(subject, object…)
There is only one terminal node and two positions-of-exponence for the
person/number features of a subject and/or object even when both arguments are
referentially marked, i.e. represent 1-2 person and present in a sentence. Note that only
the features of referentially marked arguments (1-2 person, i.e. [+participant] or
[+PSE 9])) are realized in the mentioned positions. A separate post-base position of person
agreement morphemes also allows the spellout of the person features of 3rd person
arguments. The morphemes realizing these arguments are usually fused with the series
A similar constraint is operative with respect to argument structure-marking
morphemes such as causative, applicative, voice, etc. illustrated with the following
Causative, Passive, etc
Applicative, etc.
(36) ….
The numbers indicate the positions of these morphemes in the verbal template. Thus,
one position is available for the causative, applicative, and voice markers in different
This feature is posited in Halle (1997) for designating the Participants of Speech Events, i.e. the
referential arguments.
syntactic contexts. It is often the case that the Voice marker bleeds the causative or the
applicative marker in various syntactic contexts or vice versa. The relevant empirical
evidence and its analysis will be provided in the chapters dealing with these
constructions. The restricted spellout of these morphemes may be a part of general
economy strategies on the spellout of substantive morphosyntactic features.
Summing up this section, the following points should be taken into account when
analyzing the morphosyntax of causatives and applicatives: The structure of the verbal
template in Georgian may include a maximum of 10-12 morphemes, which are
asymmetrically linearized in the pre- and post-base positions: no more than three
morphemes can occur in the pre-base position, while the number of morphemes in the
post-base position is highly variable reaching a maximum of 6-8 morphemes. The
morphological realization of various morphemes is subject to arbitrarily imposed
restrictions that make only limited positions-of-exponence available for certain
morphemes. We assume that mismatches between syntax-semantics and the morphology
of complex expressions are due to not only various restrictions on the spellout of
morphemes but also to the readjustment rules operative in the post-syntactic component.
With these observations as background, the next section overviews phi- and case
feature checking in syntax.
1.6 Case and agreement with various classes of verbs
The three languages of the South Caucasian language family display aspectuallyconditioned split ergativity instantiated by variation between nominative, ergative, and
dative cases on subjects, and dative/accusative/nominative/absolutive cases on objects,
across three series of transitive, and unergative classes of verbs. There are three such
series in Georgian: the Present, the Aorist, and the Perfective. The three series consist of
11 screeves. The screeves represent the conjugation pattern for one specific combination
of Tense, Aspect, and Mood. The three series differentially mark the case of subjects and
objects. In particular, the subjects of transitive and unergative verbs are assigned different
cases in the different series: nominative in the Present, ergative in the Aorist, and dative
in the Perfective. The following tables show the screeves in the three series and the
corresponding case marking of the arguments of transitive and unergative verbs:
Table 2. The structure of the series in Georgian
Present (S=Nom)
Aorist (S=Erg)
Perfective (S=Dative)
conjunctive present
conjunctive future
conjunctive past
conjunctive perfect
present perfect
Table 3. Case marking of transitive verb arguments
External argument
Present series
[Dative] 10
Aorist series
Perfective series
Note that the case marking of the subjects of unergative verbs is the same as that for
transitives, but objects are not projected in unergative structures.
The examples in (37) illustrate the case marking of subjects and objects in double
object constructions (DOCs) across the various series:
The brackets indicate the optionality of the projection of the Goal/Beneficiary argument in transitive
structures. However, the external and the theme/patient arguments are mandatory as indicated with the
absence of these brackets.
(37)a. Present series (active):
Nodar-Nom child-Dat
‘Nodar is presenting a candy to the child.’
b. Aorist Series (active):
‘Nodar presented a candy to the child.’
c. Perfective Series (active, irrealis):
bavshv-is-tvis t'k'bileul-i uchukebia.
child-Gen-for candy-Abs presented.Perf.
‘Nodar has (apparently) presented the candy to the child.’
Notice that in the Perfective series, the subject is assigned the dative case and the
benefactee object receives a postpositional genitive, which indicates the latter’s nonargumental position, as this argument no longer triggers the agreement with the verb (I
assume that this argument is in an adjunct position). Due to such differential case
marking of the subject, two agreement patterns emerge in Georgian. Nominative and
ergative subjects trigger basic agreement (Aronson 1990), which is realized with the ‘vset’ markers, while object agreement is realized with ‘m-set’ affixes. The terms ‘v-set’
and ‘m-set’ are based on the first person agreement markers of subjects and objects in
transitive contexts. These sets are shown in Table 6 that follows. V-set and m-set
agreement affixes cannot show up simultaneously in the same verb because of the abovementioned templatic restrictions 11.
I will refer a reader to Lomashvili & Harley (in press) for detailed explanation of the role of templates in
spellout of agreement morphology in Georgian.
When subjects of transitive and unergative verbs are assigned dative case, accusative
shows up on the theme argument 12. The agreement pattern with the dative subjects is
Inverse as it is realized by the ‘m-set’ agreement markers indexing subjects and ‘v-set’
affixes indexing Nom/Acc objects. The following table summarizes the ‘v-set’ and ‘mset’ affixes:
Table 4. Georgian agreement morphology
v-set markers
v-/x-s, -a, -o
v- -t
m-set markers
mg-, h-, sgv-
-en, -an, -nen, -n, -es
-, h-, s-
Both ‘m-set’ and ‘v-set’ paradigms are realized as prefixes when indexing the first and
the second person arguments. The inverse pattern shows up in all syntactic contexts
where subjects are assigned dative case. The relevant constructions, in addition to the
Perfective series forms of transitive and unergative verbs, are subject-experiencer
psychological predicates, two-place passives, adversity causatives, causatives of
internally-changed verbs, and various applicatives, etc. Here are some examples:
(38)a. The Present series:
‘Dato hates Vano.’
b. The Aorist series:
she- dzulda
Prev- hated.Aor.
‘Dato hated Vano.’
Notice that Acc, Nom, and Absolutive have syncretic surface realizations in Georgian, i.e. they are
marked with the morphological nominative marker.
c. The Perfective series:
she- dzulebia
Prev- hate.Perf.
‘Dato (apparently) hated Vano.’
In these examples, the third person dative subject triggers the ‘m-set’ agreement on
verbs which is spelled out as zero. The first person nominative theme object of
psychological verbs in (39) triggers ‘v-set’ agreement on verbs as illustrated in the
‘Dato missed me.’
mo- v- eprev- 1O- Voice-
However, when both dative and nominative arguments are 1-2 person, the agreement
pattern is somewhat more complex, i.e. both subjects and objects trigger agreement on
the morphologically complex verbs which consist of the main verb and the auxiliary BE.
The main verb carries the meaning of the entire predicate, while the auxiliary ‘AR’ (BE)
provides templatic extension for marking tense/aspect inflection along with the
nominative object agreement on the predicate. Thus, the secondary function of auxiliary
BE-support is to mark the agreement of nominative objects:
(40) Inverse agreement
m- iq’var- x- ar
1- love- 2- BE
'I love you.sg'
giq’var- v- ar
2love- 1- BE
‘You love me.’
In the inverse pattern, both simple and complex morphological forms of
psychological verbs have the same argument structure, and the agreement morphology
seems to be sensitive to the markedness of one or both verbal arguments. 13 Other
syntactic factors are also implicated in the spellout of subject and object person
agreement in these complex forms, which I will not discuss here for space reasons. Note
that tense/mood/aspect markers in these extended verbal units are kept intact, which
again shows that they are not fused with the person/number agreement morphemes.
Person agreement is not directly related to the main topic of this dissertation but it is
important for the analysis of various types of causative and applicative predicates because
of the way these constructions display variability in case marking of their main
arguments. This information is also instrumental in understanding the interaction between
the syntax-semantics and morphology of predicate constructions in this dissertation.
1.7 Outline of the Dissertation
Both Chapters 2 and 3 begin with an overview of the existing research on causative
and applicative constructions, and also discuss the same constructions in related
languages (Mingrelian and Svan). The limited data available on these languages suggest
that almost the same range of causative and applicative structures that are found in
Georgian are available in these languages as well.
Morphosyntactic properties of syntactic causatives derived from various classes of
verbs (inchoative, unergative, unaccusative, etc.) are the focus of Chapter 2. Two main
groups of syntactic causatives are identified: one taking inchoative verbs as an input to
the causative alternation (melt/melt type alternations) and another, transitive, unergative,
and other types of verbs resulting in make X do V alternations. Morphological distinctions
Markedrness of arguments in terms of person is defined whether these arguments are [+ participant] , i.e.
1-2 person, or [-participant], i.e. 3rd person. Marked arguments are usually [+participant].
between these two main alternations are argued to arise from their syntactic structure and
from the type of the complement that the causative head (CAUSE) takes in these types of
causative structures. This chapter also contains a section on adversity and malefactive
causatives, where the dative causee argument is the subject and no external argument is
projected by the Voice0 head. I claim that these dative subjects form a natural class with
the dative applied arguments of applicatives.
The morphosyntactic and semantic properties of high and low applicatives are
discussed in Chapter 3. First, I discuss low recipient and source applicatives of transitive
and other classes of verbs where the transfer-of-possession relation is established
between the external and applied arguments. This chapter also analyzes the stative
possession relation in low applicatives formed with psychological and existential
predicates. Two new types of applicatives are introduced in Chapter 3: those formed with
1) internally-caused change-of-state and 2) reflexive events. The applicative relation
between the external argument and the theme in reflexive applicatives is argued to be
established through the lower merging site of the reflexive Voice head, which ‘forces’ an
empty argument PRO to occur in the specifier of the lower Appl0 head which also
projects the accusative theme. The control relation between PRO and the external
argument along with the low Appl0 head is argued to be responsible for the transfer-ofpossession relation between the external argument and the theme. Chapter 3 also
discusses hybrid-type applicatives formed by four-place predicates in which high
applicative relation is established between the higher applied argument and an event
expressed by the verb, and low recipient relation is established between the applied
argument projected by the low Appl0 head and the theme. Some of these complex
structures are argued also to combine CAUSE with the lower PHAVE/PLOC head which
projects the Goal-Recipient and the theme arguments (Harley 2002, Jung & Miyagawa
2004) receiving either a ‘locative’ or ‘have’ interpretation. The study closes off with the
discussion of applicative structures in Mingrelian and Svan, which appear to display the
same range of applicative constructions as Georgian. Chapter 4 summarizes the basic
findings of the study concerning the interface of the argument structure and the
morphological realization of causative and applicative constructions.
2.0 Introduction
This chapter focuses on the causative alternation as it appears with various types of
verbs (inchoative, unergative, transitive, psychological, etc.) in Georgian and the related
languages Mingrelian and Svan. It specifically explores both lexical and syntactic
causatives to identify the morphosyntactic properties of these constructions that are
responsible for their productivity across various verb classes.
The main goal of the chapter is to analyze the interaction between the argument
structure of syntactic causatives and the morphological realization of the causative head.
As in many morphologically rich languages, it is expected that there are mismatches
between the syntax of these complex predicates and their morphological realization.
These mismatches will be accounted for in the Distributed Morphology framework as
indicated in the Introduction.
2.1 Organization and Outline of analysis
In section 2.2, I outline my basic assumptions relevant to this chapter. In section 2.3, I
discuss inchoative-causative alternations, which are morphologically realized with a-/ allomorphs. The novelty of the analysis presented in these sections is that the causative
head selects RootP (P) as its complement in the causatives of inchoative verbs. This
accounts for the relative productivity of the lexical causatives formed from inchoative
alternates and the root-conditioned allomorphy of the causative head. In section 2.4, the
morphosyntax of causatives formed from unergative verbs is discussed and the absence
of root-conditioned allomorphy of the causative head is accounted for by the size of the
complement that it selects (that is, vP). It is argued that in a simple X makes Y do V
causatives of unergative verbs such as sing, laugh ,etc. only one CAUSE is merged that
takes the aforementioned vP complement. However, when the CAUSE iterates, two
causing events are introduced into the structure and consequently, two causee arguments
are introduced to the structure of these complex causatives. Consequently, the
morphological realization of the two CAUSEs does reflect the syntax as two VIs (a- and
–in) are inserted into separate CAUSE heads. In Section 2.5, I analyze the causatives of
transitive verbs also as the result of a make X do V alternation and propose that in these
structures, two causative heads are merged and the lower causative head selects for the
argument-full vP complement. I propose that the presence and/or the lack of allomorphy
of the causative head in the alternates of transitive/unergative verbs on the one hand and
inchoatives on the other is a direct consequence of their syntax, namely, the size of the
complement that the causative head takes.
Beyond the analysis of the argument structure of various types of causatives, the
iteration of causative morphemes and their complex syntax is discussed in various
languages (Malagasy, Chemeheuvi, Georgian, Mingrelian and Svan). The conclusion
regarding the iteration of causative morphemes is that the reduplicated Vocabulary Items
(VIs) may not always mark the merger of two CAUSE heads in the same structure.
Instead, one CAUSE morpheme marks the causative event head while another may mark
some other functional head such as Aspect, Mood, or Tense.
Section 2.6 discusses the morphosyntax of adversity causatives whose syntax is argued
to combine the Voice0 head with a [NonActive] feature and CAUSE head under one
structure. The analysis of nonactive Voice in these structures is different from the
previous literature (such as Kratzer 1994) in that it allows the Voice head not to project
an external argument due to this [NonActive] feature. These structures project dative
causee and nominative affected arguments none of them introduced by CAUSE head. In
section 2.7, the event structure of ‘pretend-type’ complex verbal structures and their
causatives is analyzed. These types of complex predicates have not yet been analyzed
extensively in the literature. I propose that they combine a complex V0 head, including a
lexical head turned into the grammaticalized functional head ‘PretendP’, and CAUSE
composes with the vP and the VoiceP of a reflexive verb. Various event types of
psychological predicates and their differences in terms of causativization represent the
topic of section 2.8. Finally, in section 2.9, causative alternations of various types of
verbs are analyzed in Mingrelian and Svan.
2.2 Assumptions
Next, let us consider some basic assumptions and claims that have been made about
the structure of causatives in other languages. We start with the distinction between socalled lexical causatives and syntactic causatives. We follow the approach espoused in
Harley (2008) that both kinds of causative are formed in the syntax and differ only in the
depth of embedding of the structure. After examining this, we turn to the hypothesis that
causative morphology might be added at different ‘layers’ of the structure.
Harley (2008) (see also Miyagawa (1980), (1984), (1998), Jacobsen (1981), (1992)
and Matsumoto (2000)) discusses two main classes of causatives in Japanese: lexical and
syntactic causatives, seen in (4) and (5), where the syntactic causative has two different
Lexical causative
nio- ase-ta
‘Taro hinted at resignation.’ (Lit: ‘Taro made resignation smell’).
Syntactic (productive) causative
a. Make-causative
Youshi-A go-ase-PST
‘Hanako made Yoshi go.’
b. Let-causative
‘Hanako allowed Yoshi to go/Hanako had Yoshi go.’
(Miyagawa (1980, 1984), Jacobsen (1981, 1992), Matsumoto (2000)).
The differences between (1) and (2) are summarized in (3):
Lexical causatives
monoclausal in terms of case-marking (only a single nominative case is
possible on the Causer argument)
can have idiomatic interpretation
strong speaker sense of ‘listedness’, non-productivity, etc.
Syntactic causatives
biclausal by the tests involving scope, adverbial control, binding…
monoclausal by tests involving negative polarity, tense,...
(make-causative) mono-clausal in terms of case
causee must be animate/agentive
(Adapted from Harley
In the DM approach, vocabulary items are inserted on a competitive basis, determined
by the best fit to the features in a terminal node. In Harley’s approach, following the
proposal in Miyagawa 1998 the lexical causative head selects the root as its complement,
while in syntactic causatives, the same causative head takes an ‘argument-structurally
complete complement’ – roughly, a full vP. This is illustrated in the following structures:
(4)a. Lexical causative
b. Taro-ga
tenoeura-o kae-s(--).
‘Taro changes his attitude suddenly.’
(Harley 2008:31)
b. Syntactic causative
d. Taroo-ga
Hanako-D pizza-A
‘Taro made Hanako eat pizza.’
(Harley 2008:30)
Note that Taro is the causer argument both in lexical (4a & b) and syntactic (4c &d)
causatives but the former involves the causer combining with an idiosyncratic meaning of
the predicate while the meaning of the syntactic causative is predictable from the
combination of the causer argument with the causee and the event of causing.
An important consequence of Harley’s analysis is the difference between the lexical
causative head v0 and inchoative v0 head. Harley analyzes these items as interchangeable
rather than simultaneously present in the causatives of various verbs.
Proposals which are closely linked to Harley’s analysis on the distinction between
lexical and syntactic causatives are found in the work of Travis (2000), Pylkkanen (2002)
among others. These explorations make use of low and high attachment sites for
causative morphemes in various languages to account for cross-linguistic variation of
these structures. According to these explorations, lexical vs. syntactic use of causatives
result from the attachment site of the causative head either to a higher functional
projection (vP), or the lexical VP, and even lower projection (such as ROOT). As
Pylkkanen (2002) and Harley (1995, 2008) argue, the attachment of the causative head to
a higher projection (vP or VP) results in syntactic causatives, while the attachment of the
causative head to a ROOT results in lexical causatives.
Travis (2000) uses a Hale & Keyser-style (henceforth, H &K 1993) analysis with VPshells. She argues that in Tagalog and Malagasy, the attachment of the same causative
morpheme to different parts of the syntactic tree may result in different meanings and
productivity of causative structures. Pylkkanen (2002) and others build on this notion and
claim that the distinction between lexical and ‘syntactic’ causatives follows from the
‘size’ of the complement the CAUSE 14 takes: a causative head may select a categoryneutral ROOT, a VP, or the whole phase vP (which has an external argument in it). In
Travis’s analysis as in Harley’s, the causative heads that select the vP project an external
argument. They will be referred to as phase-selecting causatives following Chomsky
(1999 & 2001), McGinnis (2000, 2001a & b), and Arad (2003 & 2004) among others.
As noted above, in Harley’s (2008) trees above, the external argument is introduced in
the specifier of the CAUSE vP. However, following Kratzer (1996) Pylkkanen (2002)
argues that external arguments are instead syntactically introduced by a Voice0, separate
from the cause v0, and the variation in the causative heads bundling with Voice0 results in
important crosslinguistic differences in terms of productivity and meaning 15. Specifically,
she argued that in some languages the CAUSE v0 does not bundle with the Voice0 and
does not project an external argument. This happens in adversity causatives where the
CAUSE introduces the causing event but crucially does not the external argument. I will
adopt this assumption here. Below in section 2.6 I will show that in Georgian causative
structures, like in Finnish desiderative and Japanese adversity causatives (Pylkkanen
2002), the causative head may not project the external argument and therefore, other
functional heads such as Voice are argued to be responsible for introducing them.
2.3 Inchoative-causative alternation
2.3.0 Introduction
CAUSE stands for the causative head in Pylkkanen’s dissertation (2002). I will use this term for further
reference to the causative head.
See the discussion on the motivation of the Voice head in the Introduction.
As noted in the introduction, English has an inchoative/causative alternation (8), often
called the melt/melt alternation:
Ice melted.
Bill melted ice.
The same verb may participate in the alternation without a change in morphological
marking. In other languages, such as Georgian, the same alternation may entail a change
in the morphological shape of a verb.
2.3.1. Causatives of inchoative verbs
The inchoative-causative alternation in Georgian is marked with a- (7) or - (6).
Note that the morphological marking of inchoative causatives is distinct from that of
unergative-causative or transitive-causative alternations (which usually result in make X
do V meanings). The latter will be discussed in Sections 2.3 and 2.4.
-prefixed causatives formed from inchoative and reflexive verbs in Georgian:
da- imal- a
prev- intr- hid- Aor 16
‘X hid.’
damal- a
prev- - hid- Aor
‘Y hid X.’
da- ibana
prev- intr- wash- Aor
‘X bathed.’
daban- a
prev- - bathe- Aor
‘Y bathed X.’
da- iɣal- a
prev- intr- tire- Aor
daɣal- a
prev- - tired- Aor
The abbreviations used here are: prev- for the preverbal aspectual and directionality marker; intrIntransitivity marker in general, Aor- for the Aorist screeve morpheme; thmw- for thematic markers in
verbs; CAUS- for the causative marker;  stands for zero allomorph of the CAUSE, etc. The term screeve
in Kartvelian Linguistics stands for the verb conjugation paradigm in terms of Aspect, Tense, and Mood.
There are 11 such screeves in Georgian verbal morphology (Aronson 1990).
‘X got tired.’
Lit: ‘Y tired X’. 17
ga- t’q’- da
prev- break- intrans-Aor
‘X broke.’ 18
gat’ex- a
prev- - break- Aor
‘Y broke X.’
i- ch’r- eba
intr- cut- thmw- Aor
‘X got cut’.
gach’rPrev- - cut‘Y cut X.’
a-prefixed causatives formed from inchoative verbs
gai- ɣo
prev-intr- open- Aor
‘Y opened.’
prev- CAUS- open-Aor
‘X opened the Y’.
gamo- cxv- a
prev- bake- Aor
‘Y baked.’
gamo- acxo
prev- CAUS- bake- Aor
‘X baked Y.’
(6b) is a reflexive verb; (6a, c, & d) are inchoatives. The difference between (6) and (7)
is in the allomorphy of causative morpheme. Notice that the causative morpheme realized
as a-/- is root-adjacent, i.e. it linearly precedes the root. It might be argued that the
allomorphy of CAUSE is phonologically root-conditioned since a- occurs in vowel‘defective’ consonantal roots while - occurs in elsewhere contexts, i.e. it may occur
both with vowel-defective and consonantal roots as shown in (6e). The VIs inserted for
CAUSE with their conditioning environments can be shown in the following in which the
subscript notation [-] indicates a class of non-syllabic roots:
a-  CAUSE / [ _____ROOT-] 19
Note that the free English translation of (9c) is not as ‘natural’ as in other examples in (6). This is
because the inventory of verbs participating in the inchoative/causative alternation is language-specific.
It should be noted that (6a-b) are reflexive forms and they are marked with the same prefix i- as (6c-d),
which are inchoative. This prefix is uniformly present in all reflexives in Georgian such as body verbs but it
is not systematic in inchoatives though.
-  CAUSE / elsewhere
This sensitivity of the VI to the syllabicity of roots may be taken as evidence that the
causatives of inchoative verbs are “low attachment” in the sense of Pylkkanen (2002,
2008). This means that the causative head takes the root for its complement. This
alternation is fairly productive in Georgian and applies to a subset of monadic verbs
resulting in transitive structure. It has been already shown by many (cf. Pylkkanen (2002)
& Harley (2008)) that if the causative head takes the root as its complement, then the
resulting structure may have idiosyncratic meaning, and that it may acquire the property
of ‘listedness’, which is characteristic of lexical causatives in various languages. This
type of idiomatic interpretation may sometimes be present in lexical causatives but not
necessarily so (See Harley 2006 and Levin & Rappoport 1999 for further details).
Following Harley (2008) I assume that CAUSE merges in the same position as other
light verb elements (such as vBECOME in inchoatives, realized as i-/- in (6)-(7)). This
head takes RootP (represented as P) as its complement 20.
I propose that the derivation of causatives from inchoatives involves the root, which
is verbalized by the vP introducing causative meaning into the structure (CAUSE). The
allomorphy of the CAUSE head in such causatives may be determined by the Root, since
no overt material intervenes between them as seen in (9). As we have seen above, the
item a- realizes the CAUSE v0 when the Root is syllable-defective. Thus, based on this
It could be argued that in Georgian a- also realizes the functional head, which is responsible for the
projection of the theme argument. I would call the feature [TRANS] on vP that may be realized with the
same phonological exponent and this will be a primary exponence of a- . This proposal will be elaborated
below Section 2.4.
We will be referring to such events as mono-eventive further in this work because it does not combine
two independent events.
evidence and following Embick (2009)’s proposal on the phonological conditioning of
morphemes as indicating their non-cyclicity, I conclude that the CAUSE head is a noncyclic head. It is expected that in such reduced transitive structure, the head CAUSE may
still ‘see’ the Root and show the root-conditioned allomorphy.
Given these preliminaries, the structure of causatives resulting from
inchoative/causative alternations may be sketched in the following:
(9)a. tamar-ma
‘Tamar baked bread.’
gamo- acxPrev- CAUS- bake-
gamo-, etc.
-o, -a, -s (-features)
In this tree, a full specification of the functional heads merging in the derivation of
lexical causative is given. This derivation involves not only the causative head, which is
of primary concern in this work, but also the Voice, which introduces the external
argument. Cross-linguistically, Tense and Asp0 heads are introduced in various positions
of the left periphery of the clause. In Georgian and related languages where the Asp0 is
realized in the first position of the template one might argue that it must be the latest head
merging into the structure assuming bottom-up derivation of words and sentences in
syntax. However, this may not be true as it is evident that the morphemes realizing Tense
and Series features are in the word-final position fused with the person agreement
markers of non-local subjects and objects. As seen in (9), Tns merges later than the
CAUSE and other vPs. Thus, it may be assumed that the linearization of the VIs inserted
into the Tns and Asp0 heads may be subject to some post-syntactic rules and languagespecific well-formedness requirements. The exact nature of these rules is not of concern
here, but note that in the future discussions on the derivation of causative and applicative
structures, we will be assuming the existence of such rules and that they are applied either
before or after the Vocabulary Insertion (For example a metathesis rule applies after
vocabulary insertion, argued in Embick and Noyer (2001)). In positioning TP above the
AspP we follow Travis’s (2000) and others’ arguments based on the study of event
structure in various languages in which the category of Asp0 is closely associated with the
vP delimitation and presumably, it is merged above the latter projection regardless of its
relation with Voice.
As for the argument structure, causatives of inchoative verbs involve a causer
argument, which I assume is not projected by the CAUSE but by the Voice0 head
(following Kratzer (1994) and Pylkkanen (2002, 2008)). The theme argument is assumed
to be projected as the complement of the Root as argued by many in the literature.
The evidence that these structures exhibit a simple clausal structure may be seen from
the case-marking in (9). The causer is marked with ergative case just like the external
argument of a normal transitive verb while the theme is marked with
accusative/absolutive 21 case. The theme argument is projected by the lexical verbal head
and its content is sensitive to the selectional properties of this head. I assume that if the
structure contains just one clause, only one case-split argument (the causer) is possible in
such a clause:
Tamar-Erg baked-AOR
‘Tamar baked bread.’
Case assignment follows the ‘mono-clausal’/’mono-eventive’ pattern of transitive
structures. 22 Given the data of causatives in Georgian, it can also be suggested that the
external argument is introduced by the Voice0 rather than the CAUSE because the causer
never composes with the predicate to form idiomatic expressions. Also, if CAUSE
projects an external argument with nominative /ergative case, then, in adversity causative
structures, where no external argument is projected but the causative head is present,
there would be no way to explain why this head does not project the external argument.
We return to these questions below in section 2.6.
2.4 Unergative verbs of various types and their causatives
2.4.0 Hale & Keyser (1993, 2002) on unergatives and Noun-Incorporated vPs
There are several types of unergative structures in Georgian and their alternation with
causatives is unrestricted. The latter involves the addition of a causer argument and the
result is that the former external argument of an unergative verb becomes the causee
argument of a causative alternate. I assume that, as in other languages, unergative verbs,
Since Georgian and related languages display aspectually-conditioned split ergativity (Introduction), it is
assumed that the accusative object argument is marked with the absolutive case in the ergative pattern, i.e.
in the Aorist series.
A note on the choice of terms is due here. I use ‘clause’ to designate a phase-size (Chomsky 1999) unit of
syntactic structure, which can be either vP or CP. In this sense, the causatives of transitive verbs with make
X do V meaning may contain two ‘clauses’, i.e. two vPs as will be shown later in this chapter.
which largely belong to a denominal class such as laugh, cough, cry, etc., are derived
through incorporation of bare N0 in a ‘defective’ V0 head (Hale & Keyser 1993, 2002:
15-16). The operation is a type of External Merger (EM) of the sister nodes under
adjacency, which is termed conflation in H & K’s (2002:18). 23 The lexical head V0,
which may be empty or have a zero spellout, may conflate with the N or Adj projected as
its sister. The derivation of argument structure in these constructions proceeds according
to the principles of bare phrase structure (Chomsky 1995). H & K take Conflation as a
structural ‘pressure’ for the ‘empty’ V0 head to get conflated with its sister node, when
the latter is bare N or Adj. This theory of how argument structure of unergatives is built
resembles other types of incorporation analyses where the noun incorporates into the v0
head. 24
The derivation of an unergative verb begins with a category-neutral root (Marantz
1997) which contributes the phonological material (in H & K’s terms, the features
encoded in the phonological material of the morphemes is a p-signature) to the verb (H &
K 2002: 77). This root is also maximally a lexical projection and not an extended
projection in which the p-signature is passed onto the verb from a P0 head as in the
This dissertation crucially distinguishes Noun Incorporation (NI) into V0 head from the Adjective
Incorporation (AI). The basis for this comes from the fact that in Georgian NI-ed stems project external
arguments which are marked with three different cases while AI verbs project internal arguments, which
check just the nominative case across all three series of verbs like unaccusatives. This suggests that AI
(Adjective Incorporated) stems have derived subjects, which are VP-internal arguments rather than basegenerated as external arguments. This is why the causatives of NI and AI stems are treated separately in
this dissertation.
Besides the external argument, which is arguably projected by the Voice0, a Goal/Benefactor argument
may be added to the argument structure of such verbs that results in more complex argument structure.
(11) a. bottle the wine
the wine {P}
(H & K 2002:76)
The structure in (11) is derived through the conflation of {P} and {N} first and then
P ‘forces’ the merger of the specifier by selection, which is instantiated by the DP the
wine. After this merger {P} is conflated with the V0 with further Merge.
In Georgian, unergative verbs present a semantically wider variety than in English but
they rarely involve structures with location/locatum arguments like those in English
(bottle, saddle). As mentioned above, subjects are agentive and they check
nom/erg/dative cases against Voice across the present, aorist, and the perfective series
much like those of transitive verbs. Here are some examples:
Unergative verbs in Georgian
k’ivis 25
scream- TH- 3S/Ser
‘He/she screams.’
shpot- avs
agitate- TH- 3S/Ser
‘He/she is agitated.’
The main reason that I consider these verbs ‘unergative’ is that they have external arguments which are
assigned nom/erg/dative cases across the three series. However, one may argue that they are very much like
inchoative verbs shown in (6)-(7) participating in causative alternation. I consider these two classes
separately because first the subjects in (12) are case-alternating like external arguments and second, all
these verbs are clearly formed by NI, which is not present in inchoatives in (6)-(7). Also the causative
morpheme used in different in these forms.
q’vir- ishout- TH‘He/she shouts.’
cimcim- ebs
glitterTH- 3S/Ser
‘It is glittering.’
mɣer- is 26
sing- TH- 3S/Ser
‘He/she sings’.
Note that these verbs do not show any overt morphology marking intransitivity
(elsewhere expressed by the prefix i-) as in inchoative alternates shown above (see
Appendix A for details).
Locatum/location arguments do not merge in these structures either and this excludes
the merger of defective {P} in their derivation. It may be argued that the unergative verbs
in (12) are derived through root incorporation into an empty v0. Thus the structure is:
k’iv-i-s ‘X screams’
Note also that the verbs in (23) are marked with different theme suffixes (such as –eb, -i, -av), which is
primarily defined by the verb class. There is no principled reason why certain verbs show up with either of
these markers, since thematic suffixes are the formants inserted for morphological well-formedness.
As said above, the unergative/causative alternation may involve the causer and the
causee arguments with both the make X do V reading and the iterated causative Z makes
X cause Y to do V meaning. Here are examples of make X do V type alternation:
Make X do V causatives of unergative verbs
a. ak’ivl- ebs
CAUS- scream- TH 27- 3S/Ser
‘X makes Y scream.’
b. acimcim- eb- s
CAUS- glitterTH- 3S/Ser
‘X makes Y glitter.’
c. amɣerCAUS- sing‘X makes Y sing.’
eb- s
TH- 3S/Ser
d. acinebCAUS- laughTH‘X makes Y laugh’.
Notice that the morphological shape of these causatives is similar to the causative
alternates of inchoative verbs discussed above. The key differences are that only the
prefix a- is inserted in the CAUSE v0, never the - one, thematic marker –eb is added,
adjacent to the root. The productivity of this alternation suggests that CAUSE takes an
argument-full vP complement and that the head projects the causee argument in the
Recall that ‘thmw’ stands for the thematic marker in these verbs.
(15) The causative of unergative ‘a-kivl-eb-s’ X makes Y scream.
k’ivl ‘scream’
The causative is derived through root head-movement to higher functional heads such
as CAUSE, and Voice0. Two arguments are projected in this causative structure: the casesplit causer and the accusative causee. This case marking shows the mono-clausal pattern
and the causer is assigned the cases of an external argument across three series
(nom/erg/dative), while the causee is assigned a structural accusative.
I assume that the case-split argument checks its case against Voice0 head via spechead relation when it is marked with ergative and against Tns via Agree when it is
marked with nominative. The causee bearing the accusative case may be argued to be
checked against the little v0 head whose case probe can search down the structure without
incurring the minimality violations.
There is no root-conditioned allomorphy associated with the VI a- inserted in CAUSE
in these structures. The question is what are the insertion contexts of a- allomorph in
these unergatives? Recall that the insertion context for the insertion of a- was determined
by the syllabicity of roots in inchoatives (shown in (8)). Apparently, the syllabicity of
roots is not relevant in determining the allomorph of cause inserted in the causatives of
unergatives. I tentatively posit that the insertion of a- in unergative structures is sensitive
to the feature [+ trans] on the v0 head. Thus, this is a ‘different’ a- than the one inserted in
the CAUSE merging in the causative alternates of inchoative verbs:
(16) VIs for CAUSE in (14)
a. a-  CAUSE / ___ v0[TRANS]
a-  CAUSE / ___ [Root-]
c. -  CAUSE/ elsewhere
Thus, it may be argued that the VI inserted in CAUSE in these causatives is not
sensitive to root properties and the contextual allomorphy attested in the causatives of
inchoatives is not at issue here. Given this evidence, I designate the CAUSE as a cyclic in
these structures. Stemming from this, it can be argued that the cyclicity of a functional
head is relative concept depending on the size of the complement this functional head
takes. Again following Embick’s generalizations 1-2, I assume that only the CAUSE
head taking the Root complement is A non-cyclic head while the one taking the vP is
2.4.1 Adjective-Incorporated (AI) vPs and their causatives
In addition to NI (Noun Incorporation) and conflation, H&K (2002) discuss Adjective
Incorporation (henceforth, AI) structures that involve a root combining with ‘a nuclear
element’ functioning as a predicator. Such ‘predicators’ are like inflectional categories
which in some contexts may be ‘defective’ or phonologically unexpressed (H & K 2002:
54-55). In Georgian, adjective-incorporated structures are presumably formed with a
similar kind of functional element as shown in (17). The crucial property of these
intransitive predicates is that the v0 head does not project an external argument as in NI
structures and their highest argument projected in the specifier of VP is case-marked with
the nominative case across all series. This property makes AI structures look like
unaccusative verbs suggesting that their arguments will behave like derived subjects:
AI structures
a. c’q’ali
ga- mc’vanwater
prev- green‘Water became green.’
d - 28
b. xalxi
prev- calm.down- pass‘People calmed down.’
In these forms, the roots mc’van ‘green’ and mshvid ‘calm’ first combine with a (a
category-defining head) and then the resulting root is conflated with the ‘defective’ V0
head of the lexical projection. The latter head-moves to v0 which is realized with the VI –
d in all AI forms. I presume that this vP is the one introducing dynamic intransitive
events (such as vGO discussed in Cuervo (2003)).
The following structure shows the derivation of such Adjective-Incorporated verbs:
(18) The structure for AI verbs
a. c’q’ali ga- mc’van- da
water prev- green- pass- 3S/Aor
‘Water became green’.
The suffix –d is analyzed as the passive voice marker in the traditional grammar but this work analyzes it
as the VI inserted into v0 head of the unergative verbs generally.
AI structures in Georgian can alternate with causatives by adding a CAUSE head to
the functional layer of the clause imbuing a cause X to become Adj reading to the
resultant structure:
(19) AI intransitives
a. mo- tvinier- da
prev- tamed- pass- Aor
‘X was tamed.’
b. mo- atvinier- a
prev- CAUS- tamed- Aor
Lit: ‘X made Y become tame.’
c. ač’rel- d- a
prev- color- pass- Aor
‘It became multi-colored’.
d. aač’rel- a
prev- CAUS- color- Aor
Lit: ‘X made Y multi-colored.’
e. ga- lamazda
prev- beautiful- pass- Aor
‘It became beautiful.’
f. ga- alamaza
prev- CAUS- beautiful- Aor
Lit: ‘X made Y beautiful.’
g. mo- č’k’vian- da
prev- smartpass- Aor
‘He/she came to his/her senses’.
h. mo- ač’k’vian- a
prev- CAUS- smart- Aor
Lit: ‘X made Y sensible.’
i. ga- nat- da
prev- light- pass- Aor
‘It became day-like’/
‘It dawned’.
j. ga- anat- a
prev- CAUS- light- Aor
Lit: ‘X made Y lightened.’
Again this alternation closely resembles the inchoative/causative alternation with its
morphological realization. Recall that the VI a- is inserted in two-place causative
alternates combining the causer and the theme arguments. AI verbs resemble inchoatives
in their argument structure, and their causatives are also similar to those of inchoatives.
The causer argument is added to the argument structure of unaccusative verbs and the
alternation is productive: causatives may be derived from any AI predicate. Based on the
argument structure of these causatives, it may be argued that as in lexical causatives of
inchoative verbs, CAUSE selects for the RootP (P) complement in these structures.
Therefore, it is expected that no causee argument is introduced by CAUSE or any other
functional head resulting in regular transitive frame as of inchoatives. This could be due
to the fact that the alternation is a part of a more general inchoative-causative one. Recall
that in these structures, the external argument does not cause another agentive argument
to perform a certain action, rather the causer performs the caused action itself. Therefore,
the head of the P does not project the causee argument. These selectional properties of
various verbal heads may be found in AI predicates:
(20) The causative of AI unergative verbs
a. bich’-eb- ma
xalx- i
boy- pl- Erg
people- Acc
‘Boys calmed people down.’
daamšvid- es.
prev- CAUS- calm- Aor/3S.pl
theme mshvid
xalx-i ‘calm’
As expected, there is no root-conditioned allomorphy associated with the CAUSE. The
exponent a- realizes CAUSE in all environments shown in (19). By hypothesis the
CAUSE can be realized with the a- in the contexts where the v0 head is specified for
[+ trans] feature. We assume that this feature is present on v0 merged in AI structures as
2.4.2 Syntactic and iterated causatives of unergative verbs
When the causatives in (14) further causativize and undergo X makes Y cause Z to do V
alternation, beyond the causer argument who is the initiator of the core event, two causee
arguments are projected and both are inherently agentive. Given this argument structure,
the insertion of the second VI for CAUSE –in shown in (21) is expected because the
structure becomes ditransitive projecting one causer and two causee arguments. I assume
that two CAUSE heads are merged in these structures introducing two separate causing
events. The case-marking of arguments is again ‘mono-eventive’ only the highest causer
argument bears the nom/erg/dative cases of an external argument. Of the two causees, the
higher one receives the structural dative case while the lower causee checks the
accusative case. Note also that the case of the higher causee argument does not change in
the Present and Aorist series since it is marked with the dative while the case of the lower
causee changes from the Present to the Aorist series which indicates that it is marked
with the accusative:
(21) Syntactic causative with X makes Y cause Z to do V meaning in the Aorist series
a. dato-m nino-s
daak’ivl- eb- ina
D-Erg N-Dat
N-Acc Prev- CAUS- scream- TH- CAUS- 3S/Ser
‘Dato made Nino cause Natia to scream.’
b. dato-m aleksi-s santeli
aacimcim- eb- ina
D-Erg A-Dat
candle-Acc Prev- CAUS- glitter- TH- CAUS- 3S/Ser
‘Dato made Alex cause candle to glitter.’
c. dato-m nana-s
aaq’vir- eb- ina
Prev-CAUS- shout- TH- CAUS- 3S/Ser
‘Dato made Nana cause Alex to shout.’
Thus, the syntactic structure of iterated causatives contains two causative heads which
introduce two separate causing events to the structure and I claim that the dative causee is
projected in the specifier of the CAUSE which is vP and its case is checked against the
head of this projection. I assume that the case-split argument checks its nominative case
against Tns and ergative case against Voice via government. The accusative case on the
lower causee argument may be checked against the same v0 which checks the dative case
on the higher causee assuming that the probe of the accusative case feature may be
checked first due to bottom-up derivation and when the dative causee is merged its dative
cause may be checked via spec-head relation. Observe these structural relations in the
following tree:
(22) The syntactic structure of iterated syntactic causatives
a-k’ivl-eb-in-eb-s ‘X makes Y cause Z to scream’
k’ivl ‘scream’
This analysis of the argument structure of causatives maintains Pylkkanen’s notion that
the CAUSE does not project the phrasal argument cross-linguistically but the event one.
As seen in (22), further causativization of make X do V causatives is possible for the
causatives formed from unergative verbs, but there is no iteration of the individual VIs
realizing the CAUSE. 29 Instead, the exponent –in presumably realizes the outer CAUSE
in these structures. At this point it may be suggested that the CAUSEs can be realized
with two different VIs in these structures. I tentatively suggest that the insertion context
for the –in allomorph of CAUSE is the feature [+ditrans] of the v0 head. Another point of
relevant interest is that when –in is realized in a given form, a- is also inserted into the
second CAUSE head. This type of mutual dependency between two morphemes may not
It will be shown below that the iteration of causative morphemes is possible only in the perfective series
verbs in Georgian and that the iterated suffix -in marks totally different functional category.
be an isolate fact of Georgian. It may be explained with the discontinuous feeding
mechanism when one morpheme ‘conditions’ the insertion of another into a different
functional head. Thus, the [+ ditrans] feature on v0 can be a contextual feature responsible
for the insertion of –in in the CAUSE and presumably, the insertion of a- can also be
sensitive to the same feature. Here is the list of VIs for CAUSE:
(23) Vocabulary items for CAUSE:
a. a-  CAUSE / ___ [Root-]
b. a-  CAUSE / ___ v0[TRANS], [DITRANS]
c. –in  CAUSE / ___ v0[DITRANS]
d. -  CAUSE/ elsewhere
Notice also that the thematic marker (-eb) can iterate in the causative shown in (22).
According to the morphological literature (Embick 1997, 2003, among others), thematic
markers are not usually associated with functional heads. Presumably, they are inserted
just due to well-formedness requirements, as dissociated morphemes. The iteration of one
such formant –eb in (22) may be accounted for with the series feature of the clause since
it occurs in the Present series while the verbs in (21), used in the Aorist series, lack these
iterated markers. Their distribution looks also to be associated with the causative marker
–in often preceding and following the latter. However, there is no fixed pattern of
correlation between this causative marker and the thematic syllable –eb. Below, this
correlation will be explored in greater detail.
It is notable that further causativization of the lexical causatives involving AI
structures is also possible resulting in the same X makes Y do V meaning. The
morphological realization of the resulting causative is very similar to those derived from
unergative verbs:
(24) Causatives of AI-incorporated verbs
a. mo- atvinier- eb- ina
prev- CAUS- tame- TH- CAUS- Aor
‘X made Y tame Z.’
b. ga- alamazeb- ina
prev- CAUS- beautiful- TH- CAUS-Aor
‘X made Y embellish Z.’
c. daamaxinj- eb- ina
prev- CAUS- uglyTH- CAUS- Aor
‘X made Y make Z ugly’.
d. ga- amc’van- eb- ina
prev- CAUS- greenTH- CAUS- Aor
‘X made Y make Z green.’
Recall from the discussion above that the VIs a- and -in are inserted into two CAUSEs
in unergative structures that add the second causee argument resulting in iterated
causative meaning make X cause Y to do V. Here only X make Y cause X V reading is
possible and still two CAUSEs merge introducing two causing events and two lower
arguments represent the causee and the theme. The insertion of the two VIs for CAUSE is
predicted given the ditranitive structure of these constructions. As noted in (23), –in is
inserted in the environment of v0 with the [+ ditrans] feature. I assume that the causatives
of AI predicates with X made Y to do V meaning have the following derivation:
(25) AI causatives
a. ga- alamazeb- ina
prev-CAUS- beautiful- TH- CAUS- 3S/Aor
‘X caused Y to make Z beautiful.’
Adjective-Incorporated causative with a make X do V reading should show the
complex internal structure on the syntactic tests. These tests can be the same as shown
with the similar causative structures of unaccusative verbs:
(26) VP-modifying adverbial test
a. dato
CAUS-tame-CAUS‘Dato makes students tame monkeys again.’
(27) Depictive modification test
a. dato
aɣeb- ineb- s
CAUS- paint- CAUS- TH- 3S
Lit: ‘Dato makes students merry open the house.’
b. dato
D-Nom students-Dat
aɣeb- ineb- s
CAUS- paint- CAUS- TH- 3S
c. ?? dato
D-Nom students-Dat
aɣeb- ineb- s
CAUS- paint- CAUS- TH- 3S
‘Dato makes students open the house merry.’
The VP-modifying adverbial isev ‘again’ can scope over the matrix (the causing event)
or the embedded (the caused event) depending on the attachment site. In (26a), isev
‘again’ modifies the caused event entailing meaning that students tame monkeys
repeatedly due to Dato’s initiation. In (26b), the meaning is different since the same
modifier scopes over the causee’s action, i.e. Dato repeatedly causes students to tame
monkeys. The causer Dato and the causee students may not be projected into the separate
clauses (CPs), but the adverbial scope facts suggest that the adverbial modifies two vPs,
which can be argued to be phases in the sense of Chomsky (1999) discussed in the
introduction of the dissertation.
As for the depictive modification test in (27), when the depictive modifies the causer
Dato it is more natural to use it more locally, i.e. before the VP merges as in (27b) rather
than at the end of the structure where the depictive of the lower causee is still fine (27a).
This indicates that depictive modifiers are sensitive to locality and we assume that the
depictives of the causer and the causee may be projected in different ‘clauses’ taking the
latter as a smaller structural unit than the clause with the full CP. Thus, we may assume
that bi-eventive analysis of these complex predicates is possible.
2.5 Transitive/causative alternation
2.5.0 Introduction
This section recapitulates some of the observations discussed in the previous parts,
applying them to transitive predicates. These are all syntactic causatives of make X do V
type, and the morphology is similar to those formed from other classes of verbs.
The rest of the section is organized as follows: in 2.5.1 the basic type of causative
structures derived from transitive verbs are discussed as well as their complex syntax;
section 2.5.2 present extensions of the analysis for iterated causatives.
2.5.1 X makes Y do V alternation
Transitive predicates allow a CAUSE and a VoiceP whose head projects an external
causer argument. The morphological shape of such causatives shows the insertion of the
two VIs a- and –in into the CAUSE head. The prefix a- is linearized before the root, as in
other kinds of causatives above. Here are some examples:
(28) Transitive
a. k’rep- s
pick- 3S/Ser
‘X picks it.’
b. ak’rep- ineb- s
CAUS- pick- CAUS- TH- 3S/Ser
‘X makes Y pick it.’
c. cvli- s
exchange- TH- 3S/Ser
‘X exchanges it.’
d. acvlevin- eb- s
CAUS- change- CAUS- TH- 3S/Ser
‘X makes Y exchange it.’
(29) a. apas- eb- s 30
CAUS- praise- TH- 3S/Ser
‘X praises it.’
b. apas- eb- ineb- s
CAUS-praise- TH- CAUS- TH- 3S/Ser
‘X makes Y praise it.’
c. ashen- ebCAUS- build- TH-
d. ashenCAUS- build-
eb- ineb- s
TH- CAUS- TH- 3S/Ser
Note that some of the transitive forms in (38) are marked with a- and some with - for transitivity. I
assume that a- is the VI inserted in the v0 of certain verbs, specifically, in those, which have –eb as a
thematic marker (38a & e), while - is an elsewhere item (38c & g). We are not concerned with the
realization of this v0, because the issue here is realization of the causative head in syntactic causatives of
the make X do V type.
‘X builds it.’
‘X makes Y build it.’
Notice that there are two different patterns of causative marking in (28) and (29). The
verbs pick and exchange have the zero phonological item for the CAUSE in their
transitive frame and then doubled causative marking of the syntactic causative of X
makes Z do V alternates, i.e. in (28a & b), both a- and –in are inserted for the CAUSE. I
may argue that in the causative alternate of these verbs only one CAUSE is realized with
two phonological exponents a-/-in since in their transitive frame in (28a & c), no
causative marker shows up. To explain this pattern of the realization of CAUSE, recall
from Section 2.3 that the null phonological item for CAUSE has elsewhere distribution
and it shows up both in syllable-defective and syllable-full environments. Thus, it may
be assumed that this null exponent of CAUSE is present in (28a & c). However, the
causatives of both sets of verbs are marked with the exponents a- and -in and I argue that
the stacking of the two morphemes indicates two CAUSEs both in (28) and (29).
The insertion of two VIs for CAUSE can again be explained with the conditioning
environments of a- and –in specified in (23). Both VIs can occur in the contexts of v0
bearing [+ ditrans] feature and here the structure satisfies this condition. However, the
suffix –in may or may not follow the thematic marker –eb, which is iterated in (28d) &
(29). The causative in (28d) is unusual in that it is marked with –evin rather than with -in.
I claim that the simple (-in) and the complex causative marker (-evin) are allomorphs of
CAUSE. This comes from the observation that diachronically the affix –ev is derived
from the thematic marker –eb. One might think that these simple and complex
allomorphs of CAUSE may be sensitive to the syllabic structure of roots they attach to
(as sketched in (30)):
a. –in  CAUSE [ ROOT[+ ] ____ ]
b. – evin  CAUSE [ ROOT [-] ___ ]
On the surface, it appears as if –in is inserted into CAUSE with the vowel-full roots,
while –evin, which contains the assimilated thematic marker, is inserted in syllabledefective roots. However, the data in (29) does not show that the post-base thematic
marker –eb is sensitive to the syllabicity of roots. The exact distribution of thematic
affixes is not a concern in this work and it may be explored in a separate paper in the
The relevant question, then, is why the thematic marker is ‘absent’ in (28b) and the
causative suffix –in appears immediately adjacent to the root. Notice that the root
contains the rhyme –ep, which one might think diachronically could also have been the
thematic marker –eb subjected to word-final devoicing of bilabial stops. However, this
hypothesis may not be true given the vowel alternating roots of this verb in the present
and the aorist series which are  ‘k’rep-k’rip ‘pick’ displaying the umlaut pattern of root
alternation. Based on this observation I suggest that v-k’rep cannot be analyzed as v-k’rep ‘I am picking’ in which the phonologically modified thematic marker –eb/-ep is
analyzed separately in the causative. I also argue that the absence of the iterated –eb in
(28) may NOT be taken as the lack of the second CAUSE in the causative alternates since
as we have argued in Section 2.3 the thematic marker –eb is associated with the series
features appearing in the present series present and future screeve forms in a number of
ergative and transitive verbs as shown in the following:
(31) Unergatives and transitives with –eb in the present and the future screeves
a. v- imɣer- eb
1S- voice- sing- TH
‘I will sing’.
b. v- icek’v- eb
1S- voice- dance- TH
‘I will dance.’
c. v- ak’et- eb
1S- CAUS- do- TH
‘I am doing it.’
d. v- ek’amat- eb- i
1S- voice- argue- TH- Tns
‘I am arguing’.
Based on these examples, it can be argued that the distribution of the thematic marker
–eb across various verb classes is not always linked to the presence of the CAUSE and it
shows up in elsewhere contexts both adjacent and distanced from the Root/base. The lack
of the thematic marker –eb in (28b) may be explained due to the structure of root itself. It
may be that a readjustment rule applies in the post-syntactic component and deletes this
affix in the environment of identical or near-identical phonological element like the –ep
in ‘k’rep’ ‘pick’. All other forms in (28)-(29) may be argued to contain –eb or its
modified phonological element –evin the latter being sensitive to the syllabic structure of
Now the derivation of these complex causative structures is shown in the following
(32) The causative of a transitive verb
a. ashen- eb- ineb- s
CAUS- build- TH- CAUS- TH- 3S
‘X makes Y build Z’.
In this structure, CAUSE selects for the vPDO complement. The external causer
argument is projected by the Voice and the vDO introduces the dative causee in the
causative alternate that receives agentive interpretation. The case checking pattern is
similar here as in other double-CAUSE structures above and we do not repeat it here. It is
expected that these causatives containing phase-complete vPs will behave like
syntactically complex structures on the adverbial scope and the depictive modification
(33) VP-modifying adverbials
a. giam
ča- ac’erina
prev-CAUS- record- CAUS- 3S
‘Gia made Eliso quickly record the whole concert.’
b. giam
In (33a), the adverbial modifier is agent-oriented, which may be interpreted as the
adverbial scoping over the causing event i.e. Gia quickly made Eliso record the concert,
while in (33b), the adverbial modifies the recording event, i.e. Gia caused Eliso to record
the concert quickly. Thus, the causative structures are derived in the syntax and word
derivation parallels the syntactic derivation.
The depictive modification test also allows us to diagnose the bi-eventive properties of
these causatives as shown in the following examples:
(34) Depictive modification of the causer and the causee
a. gia-m eliso-s
k’oncert’i cha- ac’er- ina
daɣlil- s.
G-Erg E-Dat
prev-CAUS- record- CAUS- 3S tired- Dat
‘Gia made Eliso record the concert tired.’
b. ?? gia-m eliso-s
G-Erg E-Dat
c. ?? gia-m
cha- ac’er- ina. 31
prev-CAUS-record- CAUS- 3S
cha- ac’er- ina
prev- CAUS- record- CAUS- 3S
As seen in (34), the depictive phrase ‘tired’ is associated with the causee Eliso rather
than with the causer Gia in (34a) because the depictive is marked with the dative case as
a result of the copying rule applied at the Morphological Structure to the case features of
the head noun (Halle 1991). This results in the agreement of the case features between the
depictive and the noun. The identical case marking of Eliso and the depictive in (34a) and
relative acceptability of such modification may be argued to be linked to the modification
It should be noted though that the ergative depictive is relatively fine immediately following Gia such as
in the following: gia-m daɣlil-ma Elisos k’oncert’i cha-a- c’er- in-a ‘Gia tired made Eliso record the
of the recording event by this depictive, while the relative unacceptability of the structure
in (34b &c ) may be due to the locality constraints on the position of the depictive and
modified element. Thus, it may be argued that the causer is not in the same domain as the
depictive phrase in (34c) making its depictive modification ungrammatical.
Having illustrated the complex syntactic structure of productive causatives in
Georgian, now I will turn to the iteration of various types of causatives before discussing
other types of causative alternations such as those of adversity and psych verbs.
2.5.2 Iterated causatives
Travis (2000) shows in both in Malagasy and Tagalog the causative morpheme can
iterate, provided there is an intervening morpheme. In these languages, the meaning of
such morphological idiosyncrasy is understood as ‘cause X make Y do V.’ This syntactic
causative arguably stands for two causative heads merged in the syntax. Travis suggests
that in such structures, the causative morpheme closest to the root is the lexical causative
while the others are productive causatives. This is because in both languages, the lexical
causative morpheme and the productive causative morpheme are the same (an- and pagrespectively) and in Tagalog, when the productive causative morpheme is added to the
lexical causative form the lexical causative morpheme disappears. Travis provides
further evidence that the iteration of causatives in Malagasy is attested when a lexical
causative turns into a productive syntactic causative. Each causative morpheme adds an
additional agent, so that one-place verb becomes two-place predicate, and a two place
predicate three place predicate, etc. Serratos (2008) also discusses the existence of such
causatives in Chemeheuvi and argues that the iteration of the causative morpheme is
possible only when a Low Attachment causative is further causativized resulting in the
structure X causes Y to make Z do V meaning:
(35) Chemeheuvi causatives
a. Nüü-k
1sg-cop 3sg-anim.vis-obl
‘I am making the girl sing.’
b. Ann
Ann-Nom John-obl refl-good-think-caus-mom-caus-past
‘Ann made John like her/himself.’
(Serratos, 2008:6)
Notice that stacking of causatives in (35a) is possible without an intervening
morpheme while in (35b), the morpheme –ngu interferes between these two suffixes.
This means that some languages iterate the same morpheme to mark X causes Y do V
syntactic causative while other languages (like Georgian) mark such structures with
different morphemes such as a-/-in, i.e. two VIs inserted into the CAUSE. This
morphological idiosyncrasy can be interpreted given the Late Insertion model of the DM
at the PF component of the grammar and the phonological rules operative at this level.
Topuria’s (1967) grammar of the Svan language shows the examples of causative
morpheme iteration:
(36)a. xat’x- unāl- une
return- CAUS- … - CAUS- tense
‘X makes Y return something to somebody.’
b. xa-k’r- unasg- une
open- CAUS-… - CAUS- tense
‘X makes Y close the door.’
Lashx dialect of Svan
(Topuria 1967)
These examples of the iterated causative –un suffix show similar distribution to those
shown in Travis (2000) and Serratos (2008). The first causative marker shows up in the
post-base position and the second after the intervening morphemes (-al and –asg
respectively). The first iteration is attested with the causative alternate of the transitive
verb to return whose root may also be used in an intransitive frame with the meaning ‘he
returned’. Related data in Georgian shows that both frames of the intransitive: ‘He
returned’ and ‘He returned X to Y’ are available with the same root. The difference is in
the Voice morphology: the transitive frame is marked with the low applicative prefix (i-)
while the intransitive one is marked with the suffix –d (often referred to as the passive
marker in the traditional grammar):
(37) ‘Return’ in Georgian
a. da- g- ibrun- a
prev- 2O- applic- return- Aor
‘X retuned Y to you’
b. dabrun- da
prev- return- pass - Aor
‘X returned.’
It is evident that the Svan causative in (34a) corresponds to Georgian (35a) and the two
iterated morphemes mark the CAUSE (and presumably the transitive v0). The Mingelian
data provided in Kajaia (2001) suggests that the iteration of causative morphemes is
possible in the present perfect (evidential) screeve of the perfective series:
(38) Mingrelian causatives
a. u-č’ar-ap-u-ap-u(n) ‘make X write.’ (perf.)
b. u-tol-ap-u-ap-u(n) ‘make X peel’ (perf.)
c. u-z-ap-u-ap-u(n)
‘make X mix dough’ (perf.)
Similar morphological structure is available in Georgian in the perfective screeves (in
the pluperfect and the conjunctive perfect) where evidentiality may also figure in the
interpretation of the Pluperfect forms. Observe the following forms:
(39) Conjunctive Perfect forms of causatives in Georgian
a. ga- m- ek’eteb- ineb- in32
prev-1S- recip - make- TH- CAUS- TH- CAUS‘I (apparently) would have made X do V’.
scrv- 3O 33
b. gam- erecx- ineb- inos
prev- 1S- recip- wash- CAUS- TH- CAUS- scrv- 3O
‘I (apparently) would have made X wash Y’
c. dam- erek’- ineb- inos
prev- 1S- recip- call- CAUS- TH- CAUS- scrv- 3O
‘I (apparently) would have made X to call Y’.
Notice that the difference between these Georgian and Mingrelian iterated causatives
and those found in Svan (in (36)) is that the iteration of the causative morpheme marks
the make X do V causative and that these are the structures where at least separate causer
and causee arguments are projected. It is also clear that, in Georgian and Mingrelian, the
iterated causative morpheme does not add any argument or another CAUSE to the
argument structure of these transitive verbs. It should stand for some other functional
category. I will attempt to identify this functional head shortly after listing some
morphological and syntactic idiosyncrasies of the Perfective series verbs in these
languages. Based on the preliminary evidence, the iteration of causative morphemes in
Georgian and Mingrelian marks just the causative alternates of transitive verbs which are
interpreted with make X do V meaning.
Some of these unusual characteristics of iterated causatives in the perfective relevant to
the analysis of double causative marking are:
Here the gloss ‘recip’ is provided for the reciprocal marker e-, which is multifunctional phonological
exponent realizing Voice0 and/or a little v0 in various verbs. The syncretism between these various
instances of e- will be analyzed in Section 2.8 dealing with the psychological verbs and their causatives in
I gloss the second causative suffix as CAUS both in Georgian and Mingrelian for the time being until we
come up with an explanation for what this morpheme stands for in these languages.
The prefix u- marks high applicative meaning in the perfective screeves
both in Mingrelian and Georgian because there is no recipient, source, or
possessor relation (characteristic of low applicatives) detected between the
applied argument and the theme in these structures.
The causer argument is assigned the dative case in these series and it
triggers the m-set agreement normally associated with objects of the
transitive verbs attested in the present and the aorist series 34;
The perfective series is the only tense/mood/aspect paradigm where
evidentiality plays a certain role in interpretation (Georgian- Aronson
1990, Mingrelian- Kajaia 2001, and in Svan --Topuria 1967, Sumbat’ova
1999). Evidentiality denotes the morphological marking of the type of
action which was not witnessed by the reporter of an event. Presumably
the reporter learned about the event by hearsay indicated by ‘apparently’
in (49).
In the perfective series, the reciprocal/reflexive prefix e- marks transitive
verbs, which is not expected because normally the transitivity marker is
the prefix a-.
Given these morphological and syntactic peculiarities of the Perfective series verbs, it
can be argued that the iterated causative morpheme in Georgian and Mingrelian may be
associated with two arguments, one who reports an event, and another, the causer of the
same event. In addition, I suggest that these two events are morphologically realized by
See the introduction for the definition of these sets of person agreement prefixes (‘m-set’ and ‘v-set’)
the same VI. In Svan, the iteration occurs in unaccusative verbs, which are syntactically
causativized. Presumably, the iteration was the reflex of syntactic causativization in Old
Svan and in Proto-Kartvelian. However, the absence of ‘double’ causative meaning in
Mingrelian and Georgian suggests that the reflex was lost sometime in the diachronic
development of the Proto-language into the daughter languages and the second exponent
of the causative morpheme was reanalyzed as a Perfective series marker. Thus, the
iteration of a causative morpheme in the Perfective series has no longer been associated
with the further causativization of the syntactic causative, but rather with some other
functional head, whose exact identity is not our concern here, in this work, and will be
explored in a separate article.
2.6 Adversity causatives in Georgian
Pylkkanen (2002) argues that the introduction of a new syntactic argument is not a
universal core property of causativization and that the basic distinguishing feature of
causative verbs from their non-causative alternates is a syntactically implicit event
argument ranging over the causing events. This means that all causative constructions
involve CAUSE which combines with non-causative predicates and introduces a causing
event to their semantics but does not necessarily introduce the external causer argument.
This structural configuration is arguably present in Japanese adversity and Finnish
desiderative causatives. Although both constructions lack the external argument, they
have causative meaning and they presumably have some implicit causer argument, which
is interpreted as an event itself. Due to such semantics of adversity causatives, Pylkkanen
(2002) arrives at two possibilities in terms of CAUSE and Voice bundling across
languages. In the languages like Japanese and Finnish, where the CAUSE does not
bundle with Voice, two configurations of causatives are possible:
(41)a. Causative with an external argument
X 35
b. Unaccusative causative
(Pylkkanen 2002: 90)
Pylkkanen’s main argument is that since some languages separate CAUSE from Voice,
the strongest theory would maintain this separation universally, so that the CAUSE
would never introduce an external argument. Stemming from these two possible
structures for causatives, she presents two opposing views on their semantics reported in
the literature. One view (Doron 1999) denies the existence of two event arguments and
relates the external argument to the caused event via a causer theta-role (Pylkkanen
2002:79). Another, more traditional, view recognizes a relation between two events in the
causative (Parsons 1990). Pylkkanen’s analysis is sympathetic to the latter. She refers to
causatives as bi-eventive (bi-clausal) predicates. Doron’s view interprets causatives based
on the thematic roles of arguments projected in these structures. Doron’s thematic role
view does not allow for the possibility of causatives without external arguments and
It is assumed that X is either DP or NP in these constructions.
identifies the introduction of causative meaning with the introduction of an external
argument. The bi-eventive analysis, by contrast, predicts the existence of causatives
without an external argument such as those attested in Japanese adversity and Finnish
desiderative causatives.
Pylkkanen (2002, 2008) specifically shows that, in Japanese adversity causatives, it is
possible to posit the existence of a causing event without relating any argument to it. Also
there is a clear morphological distinction between the adversity passive and adversity
causative structures. Here are some examples:
a. Adversity passive
musuko-ni sin-are-ta.
Taroo-Nom son-Dat
‘Taro’s son died on him.’
b. Adversity causative
‘Taro’s son died on him.’
(Pylkkanen 2002: 82)
Although the meanings of these two constructions resemble each other, they are not
identical. (42a) can be interpreted as Taro’s son died without any particular cause, while
(42b) attributes this death to some outside cause. In Georgian, adversity causatives with
similar meaning are productive and show similar semantic properties to their Japanese
counterparts in that the event expressed by the verb is caused by some external causer
argument which is not projected in the structure and the dative causee argument acts out
an event expressed by the verb:
(43) Adversity causatives in Georgian
a. m- e 36k’vl- evin 371S- Voice- kill- CAUS‘I am caused to kill X’.
eb- a
TH- 3O/Ser
b. metrevin1S- Voicedrag- CAUS‘I am caused to drag X.’
c. m- eglejin1S- Voice- tearCAUS‘I am caused to tear X’.
eb- a
TH- 3O/Ser
d. m- ebrdɣvevin1S- Voicetear.apart- CAUS‘I am caused to tear X apart.’
e. m- emasxara- v1S- Voicefoolinfn‘I am caused to fool X.’
f. m- elandʒ’ɣ- v1s- Voice- scold- infn‘I am caused to scold X.”
eb- a
TH- 3O/Ser
eb- ineb- a
TH- CAUS- TH- 3O/Ser
ineb- a
CAUS- TH- 3O/Ser
The lack of the root-conditioned allomorphy of -in in (43) is similar to the cases
previously discussed in this chapter. The idea is that CAUSE is an outer head and it is
insulated from the root with the ‘doubled’ layer of bracketing. This is because the
CAUSE in these structures selects for a vP complement whose head along with the
category-defining head v insulates the root from the CAUSE. Even though these two
heads are not phonologically realized, they still prevent Root-conditioned allomorphy of
the morpheme inserted into the CAUSE.
Note that I am glossing the prefix e- as Voice because in these structures, because it does not mark the
causative head. I assume that it is the part of the syncretism of Voice markers.
I am glossing –evin as a causative marker for convenience here assuming that –ev is a thematic suffix
presumably inserted into the Infl0 or T0. See discussion below.
Another question is why the VI for the CAUSE –in is inserted in these structures,
which are not obviously ditransitive (given the insertion context of this VI in (23)). The
answer can be found in the interaction between the morphemes realizing various
functional heads. As seen in (43), another exponent for the CAUSE, the affix a- is absent
and this could be due to the different argument structure of these structures from the
causatives discussed above. Since the causer argument is not projected this sets adversity
causatives apart from other syntactic causatives discussed so far. Following Embick
(1997) I argue that this is due to the [NonActive] feature on Voice and because of this
feature, the zero elsewhere item for CAUSE is realized in the pre-base position-ofexponence (the 3rd slot) and the exponent of NonActive Voice can be inserted
unobstructed. Since the affix –in is available as a primary exponent of CAUSE to be
discharged in the post-base position, it will be realized in this slot in the absence of a- in
the pre-base slot. This type of interaction between the exponents of Voice and the
causative heads can be argued to be an instance of discontinuous feeding (Noyer 1997).
As argued above, this mutual dependency of the VIs may be formalized one VI
‘conditioning’ the appearance of another VI in the same verb.
Also it is evident that –evin in the above adversity causatives is inserted in syllabledefective roots where the latter contains just consonant clusters. As argued above in
section 2.4, I assume that this –evin is a morphologically complex morpheme itself and
that it consists of two VIs ev+in that are inserted for the thematic and causative
morphemes respectively. As argued elsewhere in this chapter, CAUSE is an outer head
while the thematic markers are often placeholder affixes. The insertion of the VI into the
node immediately adjacent to the root can be root-conditioned as argued in Embick
(2009). I assume that this happens with the thematic affix –eb in the above adversity
causatives. Thus, -eb/-ev insertion will be subject to Root conditioned allomorphy.
(44) VIs for the non-cyclic head
-ev/-eb  / root[-]_____
-  /elsewhere
Now let’s expand a little bit on the prefix e- marking Voice in these structures. This
exponent is multifunctional in Georgian showing the instances of syncretism across
various classes of verbs. It can mark passive, certain two-place transitive verbs which I
call ‘reciprocals’, and reflexive structures as well as certain types of applicatives, which
will be discussed in detail in Chapter 3. (Also see Appendix 1 for detailed explanation of
this syncretism).
Here are some examples of the mentioned structures which show the syncretism of
Voice morphology with adversity causatives in (43):
(45) Verbs denoting reciprocal events 38
a. velap’arak’ebi
1S- voice- talk
‘I talk to X.’
b. v- ečxubebi
1S- voice- quarrel
‘I quarrel with X.’
c. v- ek’amat ebi
1S- voice- argue
‘I argue with X.’
(46) Passives
These two-place predicates are different from regular transitives in that they project non-case-alternating
nominative subjects and dative objects in Georgian. In English, these dynamic verbs also involve
prepositional objects, which sets them apart from transitive structures projecting accusative themes.
a. m- edʒ’ɣvneba
1S- voice- dedicate
‘I am dedicated X.’
b. m- eʒ’leva
1S- voice- give
‘I am given it.’
As seen above, e- shows up in a wide variety of structures including reciprocals and
passives among others. I argue that e- is an exponent of Voice in all these structures and
that Embick’s (1997) feature [NonActive] on Voice can account for the syncretism
displayed in these examples. One might argue that action sentences realized with this
morpheme are like deponents and their Voice head may be specified for the same feature
[NonActive] as in passives. How this feature is implemented at the PF blocking the
appearance of the VI realizing the CAUSE is discussed next.
The Voice head in reciprocals does not participate in any syntactic alternation in
syntax. Therefore, the voice morphology does not effect a syntactic change and this class
of verbs is expected to be inherently specified for [NonActive], which is realized with the
VI e- at the morphological structure (MS) 39.
In passives, Voice is also specified for the same [NonActive] feature. However, the
non-active morphology in passives effect the passive alternation and I assume that the
As a consequence of this feature on the Voice0 head, the latter does not project an external causer
argument. Rather, the DPs projected with this head are assigned nominative case across the series much
like those of unaccusative and inchoative verbs. They also behave like derived subjects since reciprocals
cannot alternate with passives as shown in (57):
(47) Passive of reciprocal events impossible:
* Dato
dalap’arak’ebul ikna Nanas mier.
D-Nom spoken
was N
‘Dato was talked to by Nana.’
feature [NonActive] is also present in syntax driving A-movement of an internal
argument to the subject position.
In adversity causatives, the feature [NonActive] may be also argued to be specified in
syntax since the argument structure of these constructions seems to be very much similar
to that of passive constructions: there is no overt causer argument projected by the Voice0
head and passive alternation of these structures is impossible as shown in (47). Although
the non-active morphology does not effect a syntactic alternation in these constructions,
the feature still has to be present in syntax to ensure that the external argument is not
projected by Voice0. This is why I suggest that the feature [NonActive] is present in the
syntax of all these constructions.
I hypothesized above that the null exponent for CAUSE in the pre-base position is due
to the feature content on the Voice head. The CAUSE is realized with the null exponent
in the pre-base position freeing up the slot for the VI of the Voice0. Thus, the
morphological shape of adversity causatives in Georgian is correlated with the argument
structure of the predicate and the feature content on the Voice0 head. Here is the structure
showing the derivation of adversity causatives:
(48) The structure of adversity causatives
m- ek’vl- evineba
1S- voice- kill- CAUS- TH- 3O
‘Something causes me to kill him/her.’
-in (evin)
k’vl ‘kill’
Note that the nominative argument is interpreted as an affected argument. The causer
argument is not overtly projected in these structures and it may be assumed to be implicit.
Arguably this affected argument marked with the nominative case checks its nominative
case against Tns since the dative case on the causee argument is inherent and does not
intervene between the Tns and the affected argument to check the nominative. The
CAUSE in these structures takes a vP complement rather than the VP as said above.
However, this does not amount to the bi-eventive analysis of these causatives because
arguably there is no causer argument projected in this structure and the causee and the
affected argument are processed as if they belong to the same ‘clause’. This can be shown
by the test of depictive modification in which the depictive phrase can modify the
affected argument when it attaches before the verb as shown in (49a) while the lower
modification of Dato is degraded as in (49b):
(49) Depictive modification test
a. dato-s ivane
D-Dat I-Nom
drunk-Nom Voice-kill-CAUS-
‘Dato is caused to kill Ivan drunk.’
b.?? dato-s ivane e-k’vlev-in-eba
D-Dat I-Nom Voice-kill-CAUS- drunk
‘Dato is caused to kill Ivan drunk’.
As seen from these structures, depictive modification of the affected argument is
possible only when the depictive phrase is local to this argument while the lower
modification of the same argument is degraded as seen in (49b). This illustrates that the
mono-eventive analysis of adversity causatives is an option for these types of structures
since there is just one attachment site for the depictive to occur. This could be taken as
evidence for mono-eventiveness of adversity causatives. Note also that the causee and
affected argument do not participate in a passive alternation and cannot occur with the
by- phrase:
(50) The passive and by- phrase of adversity causative impossible:
a. * iremi
monadires e-k’vl-evin-eb-a
Deer-Nom hunter-Dat CAUS.to.kill
‘The deer is caused to be killed by the hunter.’
b. * iremi
e-k’vl-evin-eb-a dato-s mier.
Deer-Nom hunter-Dat CAUS.to.kill
D-Gen by
‘The deer is caused to be killed by the hunter with Dato’s interference.’
As seen in (50), neither the nominative affected argument, nor the dative causee can
A- move to form the passive. Thus, adversity passives are different syntactically from
these adversity causatives.
The morphological similarity of adversity causatives with passives and reciprocals can
be due to the feature content on the Voice0 head. The hypothetical [NonActive] feature
may be responsible for the Voice syncretism in all these structures.
2.7 ‘Pretend’-type predicates and their causatives
2.7.0 Introduction
This section discusses causative alternation of the complex predicates formed by the
composition of the grammaticized light verb ‘pretend to be’ and the CAUSE resulting in
the syntactic causative with make X pretend to be Adj/N meaning. With the projection of
the external argument in the specifier of the VoiceP these structures resemble ‘reflexive’
unergative verbs, which are also formed via NI/AI (Noun Incorporation and Adjective
Incorporation, respectively). Similar constructions are attested in various languages
(Nahuatl, Hiaki) and they are considered as multi-headed complex verbs (Baker 1996).
As seen in (51), the morphologically complex structure contains the matrix and the
embedded verbs the latter being formed by the incorporation of a noun or an adjective
(dead) into some verbal head. Baker (1996) argues that alternatively the matrix verb may
have a ditransitivized reflexive form. In such cases the embedded verb is understood as
(51) Complex predicate in Nahuatl:
A:mo ni- c- nochīhua- ltoca
Not 1sS- 3sO- 1REFL- make- NONACT- consider
‘I don’t pretend to have made it.’ (Launey 1981: 269-271)
The meaning of ‘pretend to be’ may be also rendered with the grammaticalized suffix
‘consider’ as in the following structures:
(52) Complex predicates formed with compounding in Nahuatl:
ōni- c- mic- cātocaca
PAST-1sS- 3sO- die- PART- consider- PAST/PERF
‘I had believed him dead.’ (Launey 1981:269-271)
The matrix verb may involve the epistemic verbs such as mati ‘know, consider’, and
toca ‘believe, consider’.
In Georgian, the related constructions grammaticalize only ‘pretend’ and the
embedded verb is often derived via the incorporation of a noun or an adjective.
Interestingly enough, the embedded verb has the reflexive frame in Georgian.
2.7.1 Derivation of ‘pretend’-type complex verbs
The following examples show the morphological structure of ‘pretend’ verbs:
mo- v- imk’vdar- une
prev- 1S- REFL- deadpretend- Aor
‘I pretended to be dead.’
mo- v- imdʒ’inar- e
prev- 1S- REFL - sleeping- - pretend- Aor
‘I pretended to be sleeping.’
mo- v- igizhianprev- 1S- REFL- crazy.with‘I pretended to be crazy.’
mo- v- inaɣvlianprev- 1S- REFL- sad.with‘(I) pretended to be sad.’
mo- v- ik’at’une
prev- 1S- REFL- catpretend- Aor
‘(I) pretended to be like a cat.’
pretend- Aor
pretend- Aor
Notice that the allomorphy -un/- is associated with the ‘pretend’ part of this
predicate. The morpheme realizing ‘pretend’ is root-adjacent but its insertion may not be
subject to root-conditioned allomorphy. The affix un- for Pretend0 is inserted in the forms
with mono-syllabic complex roots, while the zero allomorph of ‘pretend’ shows up in
elsewhere contexts. Since the VIs for the Pretend0 are insulated from the Root with the
derivational affixes it may be that these derivational affixes block the insertion of the VI un for Pretend0. Based on the data above I tentatively posit the environments for the
insertion of VIs for Pretend0 as conditioned by the number of syllables in the Root +
Derivational affix complex. When this complex is just mono-syllabic, the insertion of –un
is not blocked, but when it exceeds this phonological shape, then the null exponent for
Pretend is realized. Here are these items:
(54) VIs for the ‘pretend’ head
a. -un  Pretend0 / root + Der. Affix = [] ______
b. -  Pretend0 / elsewhere
This evidence shows that the complex root containing the category-defining head may
be still visible at the point when Pretend0 merges as the allomorphy of the latter is rootconditioned. This grammaticalized head Pretend0 is thus a non-cyclic element that can
‘see’ the root upon the merger.
Notice also that the lexical V0 in these verbs is ‘complex’, in the sense that the root is
accompanied by category-defining affixes:
(55) Complex Vs of ‘pretend’-type predicates
a. m-k’vd-ar
b. m-dʒ’in-ar
‘Sleeping person’
c. gizh-ian
‘With crazy’
d. č’k’v-ian
‘With smart’
e. naɣvl-ian
In these complex words, the category-defining morphemes merging with roots also
have allomorphs (m--ar, -ian). Presumably, these derivational affixes categorize roots as
adjectives or nouns (Aronson 1990). Following Embick (2009), I assume that these
category-defining heads are cyclic but they may not prevent the root-conditioned
allomorphy of outer non-cyclic heads such as Pretend0. The relevant evidence will be
presented after the structure of these complex words in (53)–(54) is shown:
Complex Vs in ‘pretend’-type predicate
a. m-k’vd-ar-(i)
der-die-der 40- (Nom)
(57) Complex Vs in ‘pretend’-type predicate
a. gizh-ian- (i)
crazy-der (Nom)
Furthermore, I assume that the reflexive Voice is merged in these structures whose
head is realized with the prefix i-. This prefix also marks reflexive applicatives 41 and
there is a notable pattern of syncretism of this marker across various classes of verbs
including other reflexive verbs, which denote events directed to the body parts such as
The VIs inserted in the category-defining heads such as a or n are glossed as derivational because these
affixes are highly idiosyncratic occurring with the particular stems. There are many such affixes in
Georgian but the roots in ‘pretend-type’ predicates use only a few of them.
The derivation of the structures with reflexive applicative markers will be discussed in the next chapter.
comb, shave, passives, etc. 42 There is no evidence that root-conditioned allomorphy is
associated with this VI (the affix i-) though. This is perhaps due to the fact that Voice is
an outer cyclic head or it merges above vP whose head is cyclic. Thus, the lower portion
of the complex verb including grammaticalized ‘Pretend’ is not visible to the functional
heads merged later in the derivation. See also Appendix A for the syncretism associated
with the affix i- as the reflexive Voice marker.
The structure of ‘Pretend’ predicates may be sketched the following way:
(58) ‘Pretend’ -type predicate
a. mo- v- imk’vdar- une
prev-1S- REFL- deadpretend- Aor
‘I pretended to be dead.’
m-/-ar, -ian
I assume that the Root first is merged with n or a categorizing head and then this
structure is again categorized as a verb with the verbalizer v0. Apparently, the inner
category-defining head a or n is not a cyclic head and does not send the structure off for
LF/PF processing, since the functional head Pretend0 merged above these categorizing
I will not go into the details of this syncretism as it is irrelevant for the present purposes.
elements is still sensitive to the properties of Roots plus the derivational elements. Recall
that this was a property of non-cyclic heads as argued in the introduction in the
discussion on Embick (2009). Note that the affix realizing the reflexive Voice is inserted
into the slot of argument structure-changing morphemes and the verb incorporates into it
after the event-introducing v0 head which verbalizes the assembled structure.
2.7.2 Causative alternation of ‘pretend’-type verbs
CAUSE takes a complete vP, with all its arguments, as its complement. The
morphological shape of these causatives resembles those of make X do V syntactic
causatives of unergative and transitive verbs where the affixes a- and –in are present as
the phonological exponents of two CAUSEs since in addition to the transitive structure of
the Pretend-type predicate with its implicit ‘self’ argument, the causee argument is also
projected in this structure. I assume that this is expected given the [+ ditrans] feature on
v0 head. First observe the empirical base of causatives formed from such predicates:
(59) Causative alternates of ‘pretend’ type predicates
a. v- amk’vdar- uneb- ineb
1S- CAUS- deadpretend- TH- CAUS- TH
‘I caused X to pretend to be dead.’
b. vamdʒ’inar- eb- ineb
1S- CAUS- sleepingTH- CAUS- TH
‘I caused X to pretend to be asleep.’
c. v- agizhianeb- ineb
1S- CAUS- crazy.with- TH- CAUS- TH
‘I caused X to pretend to be crazy.’
d. v- ak’at’- uneb- ineb
1S- CAUS- cat- pretend- TH- CAUS- TH
‘I caused X to pretend to be a cat-like.’
The argument structure of ‘pretend’-type causatives includes the causer argument,
which is projected by the active Voice0 (marked with nom/erg/dative cases across the
series). The causee is presumably projected by the vP complement that the lower CAUSE
selects and is marked with the dative case. There is a sense that the anaphoric argument is
also projected in the structure, interpreted as self as in (53). As suggested in (23), the VIs
for CAUSE in such constructions containing both a- and –in exponents may be analyzed
as realizing two separate causative heads. The argument case-marking follows a monoclausal pattern. CAUSE is not subject to the root-conditioned allomorphy presumably
because it is insulated from the root with double-bracketing:
(60) Causative of ‘pretend’-type predicate
a. v- amk’vdar- uneb- ineb
1S- CAUS- deadpretend- TH- CAUS- TH
‘I cause X to pretend to be dead.’
 
a, n
Note that in (60b) the arrows do not indicate the order of linearization applying to the
morphemes realizing the relevant functional and lexical heads in this complex structure.
It is assumed that the post-syntactic pre-insertion rules modify this order which will not
be discussed here.
We can use adverbial, negation scope tests to establish the complex syntactic structure
of the causative forms derived from pretend-type predicates. These tests illustrate that
VP-modifying adverbials can scope over both the causing and the resulting event. A
negative marker should also do the same depending on the attachment site, low or high:
VP-modifying adverbial test
a. datom
mo- amk’vdarun- eb- ina
prev-CAUS-dead.pretend-TH- CAUS- 3S
‘Dato caused Gia to pretend to be dead again.’
b. datom
mo- amk’vdarun- eb- in- a
again prev-CAUS-dead.pretend-TH-CAUS- 3S
In (61a), the adverbial modifier scopes over the causing event carried out by Dato
while in (62b) it modifies the event carried out by the agentive causee Gia. This means
that the structure is internally complex and the adverbial ‘again’ (isev) modifies two
separate events. Thus the structure can be considered as bi-eventive. The following test
employing the placement of negative marker with respect to the causee argument also
shows the similar bi-eventive properties of these constructions:
(63) Negation test
a. dato-m gia-s tavi ar
mo- agizhianeb- ina.
D-Erg G-Dat self not
prev-CAUS-crazyCAUS‘Dato cause Gia not to pretend to be crazy.’
b. dato-m ar mo-a-gizhianeb- ina
D-Erg not -CAUS-crazy- CAUS‘Dato did not cause Gia to pretend to be crazy.’
The negative marker (ar) can either modify the causing event (the matrix verb) or the
caused event (the embedded one). Georgian is Low-attachment language for the NegP
and this means that Neg head cannot scope over the entire clause. We find two kinds of
readings for each of these structures: 1) Dato did not cause Gia to pretend as if he is
crazy, and 2) Dato caused Gia not to pretend to be crazy, I will conclude that the
causative structures (as evidenced with the Negation test) are complex syntactically and
that word formation parallels the syntactic derivation 43.
2.8 Psych verbs and causative alternation
2.8.0 The empirical base of state, dynamic passive, and activity psych verbs
In Georgian and the related languages Mingrelian and Svan, three types of psychverbs emerge in terms of morpho-semantics: state, dynamic passive, and activities
(following Aronson 1990). In what follows, I will assume that state psych verbs express
‘non-core’ events in the sense of Tenny (2000), which essentially are stative
eventualities, while dynamic passives and activity verbs express non-stative events (Bach
1986). These three classes show a differential ability to alternate with causatives and
passives, namely, only activity psych verbs can do both. Here are some examples:
(64) State psych verbs
a. m- iq’var- s
1S- applic - love- 3O
‘I love X.’
b. m- ʒ’ul- s
1S- hate- 3O
‘I hate X.’
It should be mentioned though that the Negation test is not as robust as other relevant tests used above
because Georgian is a non-strict Negative concord language (Haegemann 2006, Zejistra 2005) and in a
number of cases (such as one containing several negative words—nobody never cause nobody to do
nothing) the negative marker is not required to precede the verb and is dropped. Often such languages do
not have Negative Polarity Items (NPI) like English. Due to these properties it can be argued that the
negation test can work with the sentences without multiple negation words in Negative concord languages.
The applicative morpheme in this verb does not have the same semantics as it would have in the
transitive counterpart and in the traditional literature it is referred to as the version marker. In the following
transitive applicative structure I build the house for myself (a-v-i-shen-e) the prefix i- clearly has an
applicative meaning while in the above structure ‘I love’ it does not. However, we gloss this morpheme as
applicative and in Chapter 3 of this dissertation, I will argue that it realizes the high Appl0 head in
psychological predicates.
c. m- eshin- i1S- voice- afraid- TH‘I am afraid of X.’
d. mo- m- c’on- s 45
prev- 1S- like- 3O
‘I like X.’
e. m- ezizɣ- eb- a
1S- voice- loath- TH- 3O
‘I loath something.’
f. m- enat’r- eb- a
1S- voice- miss- TH- 3O
‘I miss X.’
g. m- examusheb- a
1S- voice- discomfort- TH- 3O
‘I am not comfortable with X.’
The state psych verbs in (64) project dative-marked experiencer arguments, which are
interpreted as non-voluntary subjects. Experiencer arguments trigger the inverse
agreement pattern (generally associated with dative subjects and spelled out as m-set
markers on verbs) 46. Next, consider dynamic passives:
(65) Dynamic passives
a. m- iq’var- deb- a
1S- applic- love- pass- TH- 3O
‘I am falling in love with X’.
b. m- dʒ’ul- deb- a
1S- hate
pass- TH- 3O
‘I am becoming hateful of X.’
Note that only (64d) is marked with the preverbal morpheme, which generally marks aspect on verbs.
The VI mo- does not have any function in this form because statives usually do not express achievements
or perfective events as argued by many (Tenny 2000). It is inserted for morphological well-formedness.
The realization of other morphemes in the set (e-, i- and - ) also depends on language-specific
readjustment rules operating in the post-syntactic component and we do not discuss them here as they are
irrelevant for present purposes.
See Introduction of the dissertation for discussion on the agreement patterns in Georgian.
c. m- zizɣdeb- a
1S- loath- pass- TH- 3O
‘I am becoming *loathful of X.’.
d. misc’ordeba
1S- applic- straighten- pass- TH- 3O
‘I enjoy X.’ 47
Dynamic passives also contain the inverse person agreement markers (m-set) which
index dative experiencers much like in states shown in (64). Another common syntactic
property that dynamic passives share with states is that they cannot alternate with
causatives, i.e. these structures cannot add the causer argument:
a. State verbs cannot causativize:
* m- aq’var- s
1S- CAUS- love- 3O
‘I am caused to love X’
b. Dynamic passives cannot causativize:
*m- aq’var- d- eb- a
1S- CAUS- lovepass- TH- 3O
‘I am caused to fall in love with X.’
It is also notable that states and dynamic passives have the same applicative marker
i-/u-, which is sensitive to the person feature of the dative experiencer: [+ participant]
experiencers are marked with the affix i- while [-participant] ones with u-:
(67) States
a. m- iq’var- s
1S- applic- love- 3O/pres
‘I love X.’
b. g- iq’var- s
2S- applic- love- 3O/pres
The form in (79d) is interpreted idiomatically, as the root has a different meaning (straighten) from the
resulting predicate (enjoy).
‘You love X.’
c. uq’varapplic- love‘He loves X.’
(68) Dynamic passives
a. m- iq’var- deb- a
1S- applic- love- PASS- TH- 3O/pres
‘I am falling in love with X.’
b. g- iq’var- deb- a
2S- applic- love- PASS- TH- 3O/pres
‘You are falling in love with X.’
c. uq’var- deb- a
applic- love- PASS- TH- 3O/pres
‘He is falling in love with X.’
Observe that the morphological shape of the activity verbs is quite different from the
above two classes. In what follows first I present the examples of activity psych verbs
followed by the analysis of the morphological make-up of these verbs:
(69) Activity psych verbs
a. v- iq’var- eb
1S- REFL- love- TH
‘I am loving X.’
b. v- idʒ’ul- eb
1S- REFL- hate- TH
‘I am hating X.’
c. v- ic’on- eb
1S- REFL- like- TH
‘I am liking X.’
d. v- izizɣ- eb
1S- REFL- loath- TH
‘I am loathing X.’.
e. v- inat’r- eb
1S- REFL- dream- TH
‘I will dream about X.’
Experiencer arguments of activities are marked with nom/erg/dative cases across the
series and the Case features are checked by the Voice0 head (and like external arguments
of transitive verbs they trigger the basic pattern of agreement that is spelled out as v-set).
Presumably, the surface position of experiencers in activities is different from that of the
dative arguments of dynamic passives and states. In the analysis of these classes, I will
argue below that the dative experiencers are projected by a different functional head than
the nom/erg/dative arguments of activities. Another difference between activities and the
other two classes is that the prefix i- in activities is not sensitive to the person feature of
the experiencer argument as in states and dynamic passives shown in (67)-(68).
Following Aronson (1990), I interpret this morpheme as the marker of the Reflexive
Voice0 head, which merges in various two-place reflexive verbs to express events
directed to one’s own body parts (shave, wash, comb, braid, etc.) and also, in some
applicative structures, to express the progressive possession meaning 48. In psychological
verbs this meaning is absent though. Here are the examples showing that the VI i-, which
is not sensitive to the person features of the external argument, is present in activity psych
(70) Activities
a. v- iq’var- eb
1S- REFL- love- TH
In this dissertation I follow Pylkkanen (2002) in classification of applicatives into high and low structures
based on the attachment site of the ApplP. See chapter 3 for detailed discussion of these and reflexive
‘I am loving X.’
b. iq’var- eb
REFL- love- TH
‘You are loving X.’
c. iq’var- eb- s
REFL- love- TH- 3O/pres
‘He/she is loving X.’
Another observation about the three types of psych predicates is that only activities
alternate with causatives and passives. They may have so called reflexive and nonreflexive causative alternates. The latter are interpreted as syntactic causatives with makes
X love… Y meaning in which CAUSE is marked with the a- and –in exponents, while
reflexive causatives are interpreted as make X love… self. These are marked just with the
affix a-. Observe the data in the following:
(71) Non-reflexive causatives of activity psych verbs
a. v- aq’var- eb- in1S- CAUS- love- TH- CAUS‘I caused X to love Y.’
b. v- adʒ’ul- ebineb
1S- CAUS- hate- TH- CAUS- TH
‘I caused X to hate Y.’
c. v- ac’oneb- ineb
1S- CAUS- like/approve- TH- CAUS- TH
‘I caused X to like/approve Y’.
d. vazizɣ- eb- ineb
1S- CAUS- loath- TH- CAUS- TH
‘I caused X to loath Y.’
(72) Reflexive causatives of activity psych verbs
a. v- aq’var- eb
1S- CAUS- love- TH
‘I am causing X to love (me).’
b. v- adʒ’ul- eb
1S- CAUS- hate- TH
‘I am causing X to hate (me).’
c. vac’oneb
1S- CAUS- like/approve- TH
‘I am causing X to like (me).’
d. v- azizɣ- eb
1S- CAUS- loath- TH
‘I causing X to loath (me).’
Notice that the morphological shape of reflexive causatives is similar to that of
causatives derived from inchoative verbs. Activities also alternate with passives as
expected and this is shown in the following example:
Passive of activities
she- q’var- eb- ul- i- a
prev- love- TH- der- TH- BE
‘The child is loved (by X).’
As seen in (71)-(72), activities alternate with passives and causatives. This syntactic
property can be associated with transitive eventful structures. Thus, activities crucially
differ from dynamic passives and states with their syntax and morphology as well.
It is also of interest to note that psych verbs in Svan come into two semantic varieties
of stative and activity verbs like in Georgian:
(74) State and activities in Svan
Lentex dialect
a. xalät’
‘he loves’
b. ilat’- une
applic-love- CAUS-tense
‘He is falling in love.’ (Topuria 1967: 235) 49
Note that the glossing of these forms is mine (L. L while the forms are taken from the mentioned source
When the activity psych verb is causativized the suffix –en/-ûn is added to the reduced
(75) Causatives of psych verbs:
a. ilat’- ûne
Upper Svan dialect
applic- love- CAUS- Tns
‘He is falling in love.’
b. xalat’- ûne
pers - CAUS- love- CAUS- Tns
‘He is causing X to love him.’
c. ilt’- une
intrans- love- CAUS- Tns
‘He is falling in love.’
d. xalt’eni
pers- CAUS- love- CAUS- TH
‘He is causing Y to love him.’
(75a & c) are the non-causative forms of two semantic types of psych verbs. (75b & d)
represent their causative counterparts. Note that the suffix –un/-en marks the reflexive
causative. Based on this limited data it is still possible to conclude that in Svan, the
causative homologue of Georgian –in/-evin is inserted even in activity psych verbs whose
meaning is not determined by the causative morpheme. 51
The morphosyntactic analysis of the three types of psych verbs is presented next.
2.8.1 Analysis of the morphosyntax of the three classes of psych verbs and their
We gloss person agreement markers in these verbs with the pers since their exact reference is not
important for our analysis here. See appendix 2 for exact meanings of these glosses.
I speculated that since only activity psych verbs alternate with causatives, in Proto-Kartvelian (the
mother tongue of Kartvelian languages) this –en/ûn could have been present and it may have been kept in
Svan because this is the most archaic language of the South Caucasian language family. The same
morpheme in Georgian has disappeared through diachronic development.
I propose that the morphosyntactic properties of the three classes of psych verbs may
be analyzed in terms of the feature content and the selectional properties of the Voice0
head, which selects different complements in these three structures.
In states and dynamic passives, dative experiencer arguments are interpreted as
involuntary, non-instigator subjects (Mithun 1986), which trigger the inverse agreement
(‘m-set’) on verbs illustrated in (64)-(65). These properties are presumably associated
with the lower structural position of dative arguments. Specifically, I argue that the
structural position of dative Experiencers in states and dynamic passives is the specifier
of high ApplP and this case is presumably checked against the Appl0. However, in
activities, the position of the external argument is in the specifier of the reflexive Voice0,
which is realized with the exponent i-.
The relevant question is which functional head in dynamic passives is marked with the
suffix –d. Following Embick (1997), I assume that this morpheme may be associated
with the [Passive] feature on the Voice0 head. Note also that states and dynamic passives
have dative subjects projected by the high Appl0 head. The high ApplP presumably is
selected by the Voice0 head which bears different feature than the Voice0 head in
activities, since the latter projects the external argument. Given these structural
differences, I argue that the Voice0 head in states and dynamic passives bears a feature
which I tentatively call [NonActive] which is the same feature as argued to exist in
adversity causatives above. NonActive Voice0 head selects for the ApplP. The
morphological realization of [NonActive] may be zero, since in the position where the
Voice0 is realized the applicative marker i-/u- shows up. I propose that the applicative
marker blocks the insertion of the VI for the non-active Voice both in states and dynamic
passives. Therefore, this non-active Voice feature may be realized with e- or null
elsewhere item.
As opposed to these two classes, I argue that the Voice0 in activities is specified for
the [+ Refl] feature, which is realized with the affix i- in all three persons. Observe the
feature content on Voice0 in all three structures:
The feature specification on Voice0:
a. States [+ NonActive]
b. Dynamic passives [+ NonActive] [Passive]
c. Activities [+ Refl]
It is evident that the feature [Passive] is realized in the post-base position because the
VI for this feature is not blocked by any other VI in this post-base slot. Another
conclusion that we can draw based on the morphological shape of these causatives is that
the VIs inserted into these different functional heads such as Voice and Appl0 realized in
the pre-base slot can be represented with a hierarchy within a rule block for the terminal
node (Halle & Marantz 1993 and others) in which the higher item with more specific
features wins over the lower listed items for the insertion in this slot. From the data
above, it was shown that the reflexive Voice beats the VI for the applicative in active
structures and the latter beats nonactive Voice exponent in states and dynamic passives.
Thus, the rule block for the realization of this slot in all three classes of psychological
predicates exhibits the following ordering:
(77) The rule block accounting for the insertion of VIs in the third slot
a. i-  VoiceREFL / Class 1 52
b. i-  Appl0 / [Dat] [+1], [+2]
c. u-  Appl0 / [Dat] [+3]
d. e-  VoiceNONACTIVE
e. -  elsewhere
Based on this rule block the items listed higher will be inserted first into the third
position of the template in these psych verbs and the higher listed item will win over the
lower ones.
Now the syntactic structure of states and dynamic passives may be presented in the
following tree in (86c):
(78) State
a. m- iq’var- s
1S-applic-love- 3O/pres
‘I love X.’
Dynamic passive
b. m- iq’var- deb- a
1S-applic- love- pass- TH- 3O/pres
‘I fall in love with X.’
c. The tree for states and dynamic passives
q’var ‘love’
I assume that the Class 1 is represented with the activities.
As opposed to states and dynamic passives, the reflexive Voice head in activities with
its [Refl] features projects an external experiencer/agent argument. This construction can
alternate with passives and causatives. The derivation of activity psych verbs can be
sketched in the following:
(79) Activities
a. v- iq’var- eb
1S- REFL- love- TH
‘I am loving X.’
q’var ‘love’
Now the analysis for non-reflexive and reflexive causatives of psych verbs is
presented. In reflexive causatives, the external argument causer and the theme are coreferential, while in non-reflexives, they are not. I argue that due to the presence of
reflexive Voice, an empty argument PRO is projected in the specifier of VP and it is
controlled from the external argument. The causer and the causee arguments are
projected due to Voice and the vP complement that the CAUSE selects for in these
structures. It is expected that the root-conditioned allomorphy associated with the
CAUSE may not be an option in these verbs:
(80) Reflexive causative of activities:
a. v- aq’var- eb
1S-CAUS- love- TH
‘I make X love me.’
external arg.
q’var ‘love’
The structure for the non-reflexive causative with no anaphoric causee argument may be
assumed to contain two CAUSEs and the lower one takes an argument-full vP as its
(81) Non-reflexive causative of the activity psych verb
a. v- aq’var- eb- ineb
1S- CAUS- love- TH- CAUS- TH
‘I make X fall in love with Y.’
q’var ‘love’
The structural difference between reflexive and non-reflexive causative structures is
that in the former, the theme and the causer external argument are co-referential and this
can be interpreted as transitive structure conditioning the insertion of only a- into the
CAUSE. The morphology therefore resembles that of causatives formed from inchoatives
and generally, two-place predicates as shown above. Non-reflexive causatives clearly are
ditransitive because co-reference between the causer and the theme is not present. It can
be argued that the morphological marking of these causatives again reflects the argument
structure of these predicates with two CAUSEs introduced in it.
Thus, the feature specifications on the Voice0 head and the ability of these
constructions to add the causer argument account for the absence of causative alternates
in states and dynamic passives. Since states are non-eventive predicates (Bach 1986),
they are unable to add the causer to the structure, while dynamic passives denote the type
of events that cannot add the causer due to the [passive] feature on Voice. Thus the
features of main functional projections above the root determine the argument structure
of these verbs and the morphological marking of these heads is consistent with these
2.9 Causative predicates in related languages (Mingrelian and Svan)
Mingrelian/Laz and Svan, the two languages spoken in the western part of Georgia,
display similar causative constructions to Georgian.
2.9.0 The causative alternation in Svan
The material for this section came from V. Topuria’s (1931, 1967) text on Svan.
Topuria classifies verbs into two general classes of intransitives and transitives. As the
author notes, most of the intransitives can be transitivized in all three dialects of Svan. I
will interpret this alternation broadly as a causative alternation or applicativization as
shown in the following:
a. xug (I am standing)
b. m-īg (it is standing for me)
c. m-ag (It is standing on me).
The transitive verbs with two different morphological shapes in (80) are attested in two
dialects: (80b) in Bal-Zemouri and (80c) in Lentex dialects. This alternation between
inchoative/transitive pairs presumably involves causation as well. Based on the forms in
(82)-(83), it is evident that inchoative-causative alternation is attested in Svan and is
marked with the elsewhere prefix a- much like in Georgian:
Zemo-Svan dialect
a. mi
open (it/them)
b. si
c. edʒ-a
open (it/them)
opens (it/them)
Lashx dialect
open (it/them)
Lentex dialect
open (it/them)
Notice that in the absence of preverbal morphemes (which in Svan marks the aspect)
only person agreement and valency-changing morphemes precede the verbal base in (78).
After excluding 1-2 person subject agreement morphemes the root (k’re) and the
causative marker a- can be separated from these forms.
Topuria notes that a more typical causative morpheme in Svan is the suffix –un. This
marks the syntactic causative of make X do V type:
Svan causatives
a. xašx- une
burn- CAUS- Tns
‘X causes Y to burn Z.’
b. xak’r- une
open- CAUS- Tns
‘X causes Y to open the door.’
c. xat’x- une
return- CAUS- Tns
‘X makes Y to return Z.’
This suffix is common in the Lashx dialect, while in the Zemo-Bal dialect al-ûn-e or
a-ûn-e and its phonological allomorphs may be inserted in CAUSE. The Georgian
causative prefix a- appears to be related to it. Some further evidence is presented next:
a. iûʒ- e
‘He is sleeping.’
b. aûʒune
CAUS- sleep- CAUS- Tns
*‘He sleeps X.’
c. izg- e
intrans-live- Tns
‘He lives.’
d. ä/azg- une
CAUS- live- CAUS- Tns
‘He/she settles X down.’
e. ičexû
‘(A cow) grazes.’
f. äčxune
CAUS- graze- CAUS- Tns
‘He/she lets (cows) to graze.’
g. ʒecx- n- i
‘X is awakening.’
h. äʒcxune
CAUS- wake- CAUS- Tns
‘X causes Y to awake.’
Note that the exponents ä(a)- and -un of the causative head is also found in Svan.
This is similar to Georgian a- and -evin(-in) marking in the causative alternates of
transitive verbs and ‘pretend’-type predicates. The causative of unergative verbs in Svan
also has the same marker, which corresponds to the projection of the causer and the
causee arguments yielding transitive structure. The same morphological reflex is present
in Georgian when complex bi-eventive structures are projected for the syntactic
causatives of various classes of verbs.
In the Russian synopsis of the Svan grammar, Topuria (1967) describes the causative
formation rule which entails –e /–i deletion with –un suffixation. The latter suffix can be
preceded by –al or –ä (in High Bahl dialect), which marks repeated events, and in Low
Bahl and Lent’ex dialects, by –al or –a, etc. (Topuria 1967:286). These affixes may be
followed by –ûn/-un. Sometimes these suffixes may be lost such as in the Low Bahl
dialect shown in (84b):
(86) Causatives of transitives in Svan
High Bahl
a. xamārā/ûne  xamaräûne
forces to prepare/cook something
Low Bahl
b. xamaralne
c. xamārune  xamārûne
The same –un marks the causatives of the unaccusative verbs such as sleep:
(87) Transitives in Svan
a. iûʒe  a-ûʒune
sleeps ‘X causes Y to sleep’
b. aq’ure  aq’ûne
‘X lies’ X lies Y down
It can be argued that in many respects the causatives of various classes of verbs in
Svan are very much similar to those in Georgian. The causatives of inchoative and
transitive verbs may be marked with the suffix –un which is the phonological homologue
of the Georgian affix –in. The limited data suggests that the variety of verb classes that
causativize in this language is almost the same as in Georgian as the limited data
2.9.1 Causative alternation in Mingrelian
Kajaia’s (2001) Mingrelian-Georgian dictionary is used as my primary source here.
This book also contains some discussion of causativization. Causatives in Mingrelian can
be formed from the Aorist screeve by adding the –ap-u suffix. Here are some examples
(Kajaia 2001:58):
Table 5. The transitive and causative predicates in Mingrelian
a. č’ar-u-a
To write
b. tol-u-a
c. zal-a
d. č’al-a
e. txu-al-a
To peel
To mix dough
To sew
To ask for
Aorist causative
‘Cause X to write’
o-č’ar-ap-(u)- u 
‘Cause X to peel’
o-tol-ap-u- u 
‘Cause X to mix’
o-z-ap-u- u 
‘Cause X to sew’
o-č’-ap-u- u 
‘Cause X to ask for’
o-txu-ap-u- u 
These examples show that in the Aorist screeve –apu is immediately adjacent to the
root which may or may not be reduced (such as in (86d)). Thus no morphology interferes
between the root and the causative suffix. I suggest that that the CAUSE takes the phase
complement (vP) in these causative structures of inchoatives and this explains the
absence of idiomatic readings for all these forms. The morphological structure of
causatives in these examples is almost the same as in corresponding Georgian
counterparts in the Aorist screeve. The only difference between the Georgian syntactic
causatives of make X do V type and the Mingrelian counterparts listed in Table 5 is the
absence of the homologous causative marker in Mingrelian. Compare the following
Georgian causatives to those above:
(88) Transitive verbs and their causative alternates in Georgian
Aorist causative
a. c’er-a
‘to write’
da- ac’er- ina
Prev-CAUS- write- CAUS- Aor
‘Cause X to write’
b. prckvn-a
‘to peel’
da- aprckvn- evin- a
Prev-CAUS- peel- CAUS-Aor
‘Cause X to peel’
c. zel-v-a
‘to mix the dough’
mo- azel- vina
Prev-CAUS-mix- INF- CAUS- Aor
‘Cause X to mix the dough’
d. k’er-v-a
‘to sew’
she- ak’er- vina
Prev- CAUS- sew- INF- CAUS-Aor
‘Cause X to sew’
e. txovn-a
‘to ask/request’
a- txovnina
Prev- request- CAUS- Aor
‘Cause X to request Z.’
In Georgian, both the prefix a- and the suffix –in/-evin are inserted as VIs for the
causative head in the syntactic causative. As seen in (88), the VIs inserted into the
CAUSE are not sensitive to the morphophonological properties of roots. This may be
expected given the uniform syntactic structure of these causatives across these languages.
As for the Mingrelian examples in Table 7, I attribute the loss of causative prefix to
diachronic development, assuming that in Proto-Kartvelian the prefix a- was present
based on the data of Georgian and Svan causatives. The empirical evidence from the
three languages shows the similar morphological shape of causative predicates and
almost all classes of verbs can participate in causative alternation. It thus seems
reasonable to assume that Mingrelian and Svan causatives show the same syntactic
properties in terms of bi-clausal/bi-eventive behavior. This requires further empirical
verification with native speakers.
2.10 Conclusions
The chapter showed empirical evidence of causative constructions in three related
languages: Georgian, Mingrelian, and Svan and argued that they display common
morphological properties of lexical and syntactic causatives formed from inchoative,
unergative, transitive and other classes of verbs. The inchoative/causative alternation is
both morphologically and syntactically distinct from the make X do V causative
alternation in these languages. This suggests that morphology is consistent with the
syntax of these constructions. Specifically, the prefix a- inserted into the causative head
in inchoative causatives ‘signals’ the feature [trans] of the vP and correspondingly, the
causer and theme or the causer and the causee arguments are projected into the structure,
while the additional marker for CAUSE -in is inserted in iterated causatives of
unergative, transitive and other types of verbs, which presumably aligns with the
ditransitive structure, combining the causer, the causee(s), and the theme arguments.
Table 8 below summarizes the morphological marking of these constructions by the class
of verbs:
Table 6. The morphological marking of causatives of various verb classes
aBake, open, break, etc.
Roll, cool, warm up, etc.
Sing, scream, cough,
‘Pretend’-type predicates
Psychological verbs
Psychological verbs
a-/ -in
Sow, make, eat, etc.
Pretend to die, pretend to
sleep, etc.
X makes Y love X,
X makes Y hate X, etc.
X makes Y love Z,
X makes Y hate Z, etc.
The distinct morphological shape of syntactic causatives is also associated with the
syntactic behavior of the mentioned causatives in terms of VP-modifying adverbials,
adjunct control and other syntactic tests, which show that make X do V causatives are
complex syntactically. Two different meanings are interpreted in the clauses where the
adverbial scopes over the event performed by the causer and the causee. The adverbials
in the causatives of inchoative verbs do not show such bi-eventive properties. Similarly,
the test of depictive modification applied to lexical and syntactic causatives has also
shown that the former have mono-eventive properties since only agentive modification is
allowed while X makes Y do V causatives display bi-eventive through the depictive
modification of both the causer and the causee arguments. The chapter has illustrated,
however, that the case-marking of the causer and the causee follows the mono-clausal
pattern and that these arguments are marked in a way that shows they are projected in the
same clause.
Another result of the research on causative constructions showed that the Vocabulary
Items for the causative head may be homophonous across different classes of verbs but
these items are different with their insertion contexts and distribution. For example, the
phonological exponent a- inserted into the CAUSE of lexical causatives is sensitive to the
root properties while the same phonological exponent in X makes Y do V causatives is
sensitive to the [trans] feature of v0 merged in these structures. The research has also
found that the double marking of the causative head in syntactic causatives is the natural
consequence of two CAUSE heads merged in these types of structures. Such transparent
correlation between the syntactic heads and their morphological realization is of interest
given a wide distribution of null exponents as default markers of different functional
heads in this language. These findings on the interaction of the functional morphology
with the argument structure and generally, the syntax of causative constructions will be
further examined in the larger context of applicative constructions in the three related
languages, which is explored next.
3.0 Introduction
The chapter explores the morphosyntactic properties of applicative constructions in
three languages: Georgian, Mingrelian, and Svan in which the Applicative head (Appl0)
is marked morphologically. The main goal of this research is to identify the syntactic and
morpho-phonological rules whose interaction determines the morphological shape and
associated meaning of these constructions. Since all three languages under investigation
are polysynthetic, a particular emphasis is placed on the morphological realization of
applicative heads in various syntactic contexts and the interaction of the applicative
morphology with other argument structure-changing morphemes.
3.1 Theoretical goals
3.1.0 Outline of the chapter
The goal of the chapter is to come up with a unified theory of how the meaning and
the morphological structure of applicatives are derived in syntax in Georgian, Mingrelian,
and Svan. The theoretical premise that underlies this goal is that the morphological
realization of the Appl0 head may not be uniform across various syntactic contexts and
may be subject to contextual allomorphy. Sometimes the applicative relations may be
expressed not by the specialized Appl0 head but by another functional projection. There
are also cases where the Appl0 head is not morphologically realized but the low
applicative meaning is still present 53. Therefore, this chapter aims to explore the role of
various rules and principles of Distributed Morphology (DM) with regard to the interface
between the syntactic-semantic and morphological components for a better understanding
of where the above relations between arguments come from in the absence of applicative
morphology. The ancillary goal of this research is to find out whether the grammar of
polysynthetic languages differentially marks high vs. low Appl0 heads.
Section 3. 2 analyzes low Recipient and Source applicatives along with low
applicatives of stative possession. In section 3.3 low applicatives of unaccusative and
inchoative verbs are discussed. Their morphological shape is accounted for based on a
proposal of feature-delinking in the Morphological Structure (MS), (Harley & Noyer
1997). The delinking of the person features on the Appl0 head is argued to stem from the
feature-geometry in the sense of Harley (1994) and universal feature hierarchy of Noyer
(1992). The upshot of this morphological process is that instead of the morpheme inserted
into the Appl0 head the morpheme realizing the Voice0 head shows up in the pre-base
position of the verbal template.
Section 3.4 discusses low applicatives formed from Noun- and Adjective-incorporated
verbs and their distinct morphological shape is accounted for referring to the feature
content on the Voice0 head and feature delinking on the Appl0 head. In section 3.5 a new
type of low applicative construction is analyzed as a reflexive structure. The novelty in
this analysis of these constructions is that the relation between the external argument and
the theme, which is very similar to the applicative, is expressed through the reflexive
See the differenced between high and low applicative structures and their semantics in the Introduction,
Section 1.4.
Voice head, which indicates the projection of an empty argument (PRO) in the specifier
of the ApplP. This empty argument c-commanded and bounded by the external argument
keeps the local low applicative relation with the theme. The structure thus contributes to
the Recipient relation between the external argument and the theme.
Section 3.6 analyzes the possessor dative constructions as low applicatives and it is
argued that Georgian and related languages provide the semantic interpretations of stative
possession in low applicatives through the little vBE head. The eventualities such as know,
approve, etc., which have activity frame in these languages with nominative/ergative
subjects, are capable of forming such relations. However, it is argued that psychological
verbs denoting states in these languages only form high applicatives. This is due to the
lower structural position of the dative experiencer argument itself, which prevents the
projection of another dative applied argument in the structure. Section 3.7 introduces
four-place complex predicates with location-as-object meanings that lexicalize a path
component in their semantics. These complex applicative structures are argued to
combine both high and low applicative heads since they show transfer-of-possession
through the Recipient argument and also, the high applicative relation between a
Benefactee argument and an event. Some of these complex predicates are argued to
combine the causative head with the postpositional locative phrase in which animacy of
the Goal argument is the key component. The following table summarizes low
applicative structures analyzed in this dissertation and the types of little vs associated
with their meanings:
Table 7. Low applicatives in Georgian
Meaning/ type of v0
Transitive (vDO )
Inchoative (vGO)
Unergative (vDO)
predicates of
activities (vDO)
applicative (vDO)
Dynamic Relation
Write, send, bake
Lose, hide
Static relation
Stative Possessor
Ripe, blossom
Like, approve,
adopt, etc.
Bake, build
Section 3.8 analyzes high applicatives formed from various classes of verbs such as
psychological, unaccusative, and existential predicates. The three morphosyntactic
classes of psychological verbs in Georgian discussed in Chapter 2 on causatives can
differentially express low and high applicative meaning. For example, activity psych
verbs can only express the low applicative relation, and the nom/erg/dative experiencers
projected in these active structures are interpreted as voluntary, instigating subjects
behaving like external arguments of transitive verbs. These active psychological verbs
add a dative applied argument, which is interpreted as the possessor of the theme
The dative subjects of dynamic passive and stative psych verbs form high applicative
relations as their dative arguments indirectly benefit from an event expressed by stative
Existential predicates such as ‘happen’, ‘occur’, etc. combine with the high
applicative head in Spanish (Cuervo 2003). In Section 3.8 high applicatives formed from
similar verbs are analyzed and also it is shown that they are morphologically realized the
same way as low applicatives. Unaccusative and unergative verbs like ‘arrive’, ‘run’, and
stative ‘hold’ may also form high applicatives in Georgian.
A special Section 3.9 is dedicated to applicative constructions in related languages:
Mingrelian and Svan, which show almost the same range of applicative meanings as
The following table summarizes the high applicative constructions and their
correlation with the event-introducing v heads:
Table 8. High applicatives in Georgian
Psychological verbs of state
Unaccusatives of happening
Stative relation
Like, love, hate
Dynamic relation
Happen, appear, occur
Walk, dance
Now I will review the proposals that are taken as the theoretical background for the
analysis of applicative structures in this dissertation.
3.1.1 McGinnis on applicative structures as phases
McGinnis’ (2004) proposal centers on the different syntactic properties of applicative
heads (high and low) that may affect not only the phrase structure but also the phasal
architecture of applicative constructions (Chomsky 1999, 2001). In her theory, low and
high applicative heads may differ in terms of their ability to demarcate the domain of the
phase as predicted by (1):
(1) The sister of VP heads a phase if an argument is generated in its specifier.
Based on this condition McGinnis derives the differing properties of high and low
applicatives in terms of their ability to form phases. Namely, the high applicative heads a
phase since it is a sister of VP and, in combination with VP, it assigns a theta-role to the
applied argument as seen in (2a). In contrast to the high applicative head, the low
applicative head is not a phase head because it is not a sister of VP. In low applicatives,
V0 may head a phase only in ‘special’ circumstances, i.e. when an argument is generated
in its specifier. The following structures adapted from McGinnis illustrate this ability of
Appl0 heads to head a phase:
(2)a. High applicative = a phase
b. Low applicative  phase
In McGinnis’ system, DO stands for Direct Object i.e. the theme argument and IO for
Indirect Object, i.e. the dative applied argument. As seen from these derivations, in the
low applicative construction, the Appl0 head does not demarcate the phase boundary in
the sense of Chomsky (2001), since the Appl0 is not a sister of VP. In contrast to low
applicatives, the high applicative head is a phase head because the Appl0 head is a sister
of VP and the IO, i.e. the applied argument, is projected in the specifier of ApplHIGH.
In light ofMcGinnis’ findings, the question arising with respect to applicatives is
whether the Appl0 heads in high and low structures are cyclic or not. The answer may be
given either by syntactic tests or by morphological evidence such as the presence of
Root-determined allomorphy of these heads. I follow Embick’s proposal (2009),
discussed in the introduction, that this kind of allomorphy can be a test determining the
cyclicity, in this case, of the Appl0 head. Specifically, if Root-conditioned allomorphy is
present both in high and low applicative heads, then these heads cannot be cyclic. If the
opposite evidence is found, i.e. the heads do not show allomorphy, the prediction will be
that these heads may be cyclic.
One caveat for the syntactic structure of applicatives should be kept in mind though.
In the low applicative, the Appl0 is introduced earlier in the derivation than the lexical V0
head itself. Due to this position of the Appl0 head, the expectation is that contextual
allomorphy of the Appl-low0 will be impossible, because at the spell-out, the lower head
cannot look ahead and ‘adjust’ to higher heads. Thus, the question about the cyclicity of
the low Appl0 head may not be the right one to ask with respect to low applicatives.
Another possible argument suggesting that the low applicative cannot be a cyclic
head has to do with the amount of structure projected at the point of the derivation when
this head is introduced. The low applicative relates two arguments: the theme and the
dative applied argument projected under ApplP, but there is just one DP (theme) to be
sent off to LF. Presumably, the structure of this size cannot satisfy the requirements of
phase size, which has been argued by Chomsky (1999) to be of the size of the transitive
The situation is slightly different with high applicative heads. Based on McGinnis’
hypothesis, in contrast to the low applicative, the high applicative head defines the phase
boundary. This means that when the high Appl0 head merges the lower structure will be
sent off for LF/PF processing, and Root-determined allomorphy of this head will not be
an option. This conclusion about the high applicative head will be examined in section
3.8 on high applicatives in this chapter. It should also be kept in mind that the categorydefining head (v), which has been argued by some morphologists to be a cyclic head
(Embick & Marantz 2008), may intervene between the Root and the high Appl0 head,
which would render the Root-determined allomorphy of the latter head impossible. This
is also expected in noun and adjective-incorporated unergative verbs where the categorydefining heads (n, a, etc.) may be overt and the morphemes realizing these heads would
possibly intervene between the Appl0 head and the Root rendering the applicative
morphemes resistant to root-conditioned allomorphy. Thus, the question whether the
Appl0 head is cyclic or not may not be answered based on the absence of the contextual
allomorphy of these heads.
Another related question regarding the syntactic structure of applicative predicates is
whether low or high applicative structures can be formed from verbs with various lexical
semantic properties. The relevant observations in the existing literature will be shown
3.1.2 Transitivity restrictions on high and low applicatives (Pylkkanen 2002)
Pylkkanen identifies several syntactic tests to distinguish between high and low
applicatives. One such test for low applicatives is that they cannot occur with completely
stative verbs such as ‘hold’. Here is the structure illustrating her point:
(3) High applicative of stative
* I held him a bag. (Pylkkanen 2002:24)
Since English is a language which selects only the low applicative head to express the
Recipient or the Source relations between the applied argument and the theme, the verbs
whose lexical semantics are interpreted as stative eventuality such as in (3) cannot form
Double Object Constructions (DOCs) with the mentioned thematic relations and
meanings. However, in Luganda, which is a high applicative language 54, stative events
can combine with the high Appl0 head:
(4) High applicative of stative verb
ya- kwant- idde
Past-hold- APPL- PAST
‘Katonga held the bag for Mukasa.’
(Pylkkanen 2002: 25)
Pylkkanen’s main argument to distinguish high from low applicative is the existence of
transfer-of-possession semantics between the applied argument and the theme. As seen in
above structures in (3)-(4), in no way does the bag become the possession of the applied
argument as the result of the event, although the high applied argument Mukasa benefits
from the event described by the verb.
Another syntactic restriction that Pylkkanen identifies is that only high applicatives can
combine with unergatives, while low applicatives cannot. The following structures
illustrate that applicative arguments are impossible in English; this is because English
only has low applicatives, which require an internal applied argument to combine with
the theme. Since unergatives lack internal applied arguments, low applicatives may not
occur with them, and high applicatives are unavailable in English:
(5) Applicative of unergative impossible in English:
a. *I ran him. (unergative) (Pylkkanen, 2002: 24)
[Intended meaning]: “I ran for him.”
b.*I walked him.
It should be noted though that Pylkkanen’s (2002) analysis of applicatives does not exclude the
possibility that there are languages which combine both high and low applicative heads and both types of
structures may exist in the same language (cf. Cuervo (2003).
[Intended meaning]: “I walked for him.”
The verbs in (5) cannot occur in DOCs in English to render the Benefactee reading for
the applied argument. This means that in no way can the applied argument can either be a
Recipient, or the Source. However, in the following Venda sentence, Pylkkanen
demonstrates that high applicative is possible with the unergative verb:
(6) Venda high applicative of Unergative verb
1SG-FUT-work-APPL-FV lady
‘I will work for the lady’.
(Pylkkanen 2002:25)
Based on this evidence, it can be argued that some languages can form high
applicatives with unergative verbs. Having established the contrast between high and low
applicatives in terms of the verb classes that these heads can combine with, Pylkkanen
also shows that thematic/semantic relations in an applicative sentence can also be used as
a diagnostic:
(7) Chaga high applicative
‘He is eating food for his wife.’
(Bresnan & Moshi 1993: 49-50)
In this Chaga example, food in no sense becomes a possession of wife, which can be
used as another diagnostic for the high applicative, in contrast to low applicatives which
entail a possession relation between the Theme and the applied argument. Thus,
Pylkkanen concludes that only high applicative heads can combine with unergative verbs
(as well as other verb classes) while low applicative heads must combine with various
transitive structures. The restriction on stative eventualities combining with low
applicative heads has been modified by Cuervo (2003) given empirical evidence from
Spanish which is discussed next.
3.1.3 Cuervo (2003) on low and high applicatives in Spanish
Cuervo (2003) identifies several types of high and low applicative constructions in
Spanish that allows the possibility of combining high and low applicative heads in the
same language. Low applicatives in this language are DOCs in which three different
types of relations can be established: recipient, source and stative possessor of the theme.
She additionally illustrates that the low applicative head can combine with stative
eventualities introduced by vBE, i.e. a stative event head, which, beyond the stative
meaning of an event itself, can show the affectedness of the subject argument as well.
Thus, the low applicative relation need not always express transfer-of-possession
between the two arguments but may also imply just ‘a static relation of possession
between two individuals’, as illustrated in the following examples:
(8) Spanish low possessor applicatives
a. Pablo le
a Valeria
Pablo CL.DAT admires
the patience.ACC
‘Pablo admires Valeria’s patience.’ (lit: Pablo admires Valeria the patience).
b. Pablo le
a Valeria.
CL.DAT. kissed the
‘Pablo kissed Valeria on the forehead.’ (lit: Pablo kissed Valeria the forehead).
A Valeria Appl0
la paciencia ‘patience’ (Adapted from Cuervo 2003:56)
(8a) involves the state psychological verb admira ‘admires’ along with the stative
possessor relation between the applied argument and the theme. However, (8b) is clearly
an active structure where the same stative relation is expressed between the applied
dative argument and the theme (Valeria and the forehead). This dissertation also shows
empirical evidence from South Caucasian languages in which non-transfer predicates can
be either stative (the Appl0 head combining with the stative verbs admire, have, see,
envy, etc.), or dynamic (such as kiss, wash, touch, etc). What unites these stative and
dynamic structures is their dative argument, interpreted as a stative possessor (or
location) of the theme. Thus, these low applicatives have the meaning of ‘at’ (possessor)
rather than ‘to’ (Recipient), or ‘from’ (Source) applicatives. The meaning of ‘at’
possessor applicatives is shown in the following formalism:
(9) Low –APPL-AT (Possessor Applicative)
x . y .  f e s,t e. f (e, x) & theme (e,x) & in-the-possession (x, y)
(Cuervo 2003:54)
The formalism illustrates that there are two variables in low applicatives: the applied
argument (y) and the theme (x), and the truth conditions of the low possessor applicative
depend on the calculus of these variables. The denotation of this expression shows that
possessor datives are instances of DOCs and the relation between the theme and the
applied argument includes the fact that the theme (x) must be in possession of an applied
argument (y).
The third type of applicative relation substantially widens the empirical base of
applicative constructions cross-linguistically and clearly teases apart three types of low
applicative meanings: Recipient (to), Source (from), and Possessor (at).
The conclusions drawn from the above research will be applied to the analysis of
applicative constructions in Georgian and related languages.
Now I introduce the common properties of dative arguments in three types of high
applicatives based on Cuervo (2003), which will be discussed in detail below:
(10)a. dative arguments are all licensed by the Appl0 head;
b. in some high applicatives, the dative argument is the highest argument in the structure
(such as psychological verbs), occupies the subject position, and exhibits all properties of
c. in Spanish, Experiencer subjects, like other dative subjects, never trigger agreement
with the verb. (Agreement is with the nominative object);
d. dative DPs are external to the vP that the Appl0 head takes as its complement.
Therefore, there is no predication relation between the dative argument and the stative
predicate, which means that only the stative theme and the verb enter into a predication
The properties in (10a-c) are descriptive statements and (10d) needs to be developed
in detail. The externality of dative experiencers from predication relations with
nominative objects results from the fact that most psychological verbs in Spanish can
appear without an experiencer. These are the following verbs: gustar ‘like’, importer
‘matter’, molestar ‘bother’, etc. These verbs are similar to the Italian piacere class, which
were analyzed in Belletti & Rizzi (1988, henceforth, BR) and were argued to have the
following structure:
a Gianni
A Gianni piace questo.
To G
pleases this.
In BR’s theory, the Experiencer argument Gianni asymmetrically c-commands the
theme (questo) but either one can occupy the preverbal subject position. The structure for
the high applicative of a psychological verb in Spanish has been presented in the
introduction of this dissertation in (16) and here we are repeating for convenience as (12):
(12) a. High applicative of psychological predicate
A Daniela
‘Daniela likes the cats.’
gustan los
like.PL the
a Daniela
los gatos
gust- (Cuervo 2003: 135)
It is evident that the nominative case is checked on the theme (los gatos) in a different
position than the accusative in low applicative constructions with the possessor datives.
The nominative argument is the subject of the vP ‘gustar’ (likes) rather than its object, as
Cuervo argues. The dative experiencer is ‘external’ to the predication relation because the
stative vP is the ‘first argument’ of the high App0 head that licenses the dative DP a
Daniela. Thus, the predication relation holds between the verb and the nominative DP
‘cats’, while dative argument is ‘external’ to this relation. This is one of the
characteristics of high applicatives according to Cuervo. It should be noted, though, that
the above structure is not ‘right’ in a sense that it allows the dative DP a Daniela to be
projected higher than the nominative theme object los gatos. This configuration could be
objected to on the grounds that it would involve a minimality violation for the purposes
of nominative case checking on the theme argument los gatos. This violation would not
occur, if the nominative argument is projected higher than ApplP itself by some other
functional head and the nominative case checked against Tns via Agree with this
argument. However, we are assuming that the inherent dative case of the experiencer
argument makes it invisible to the case-checking Agree relation, and hence the structure
does not incur a minimality violation for the nominative theme to check its case against
Tns via Agree.
These configurational properties of dative and nominative arguments in high
applicatives substantially differ from those of experiencer and theme arguments in low
possessor ‘AT’ applicatives. The contrast between low and high applicative meanings in
psychological predicates is shown in the following sentences adapted from Cuervo
(2003) again:
(13) Low applicative
A Laura
CL. Dat lacks
‘Laura is missing Pablo’s pen.’ 55
la birome
the pen
de Pablo
of Pablo
(14) High applicative
A Laura
Laura.dat CL.Dat likes
‘Laura likes the / Pablo’s pen.’
la birome
the pen
(de Pablo)
of Pablo
In (14), the dative and nominative arguments are in different domains and do not
relate to each other directly, as shown in the following tree:
In this sentence, the low applicative relation is established through the malefactive relation between the
applied argument Laura and Pablo’s pen as shown in (27).
A Laura
la birome
de Pablo
gust- ‘like’
(Cuervo 2003:144)
By contrast, in the existential construction (16), the relation between the theme and the
predicate is not direct, but the dative argument Laura is in direct relation with the theme
‘pen’. Syntactically, this relation exists due to the low applicative head projecting the
dative subject in the specifier and the complex theme as its complement:
(16) Low applicative (=15)
A Laura
la birome de Pablo
‘Pablo’s pen’
These two structures indicate the semantic and syntactic differences between high and
low applicatives combining with stative vPs. Some of these arguments underlie the
analyses of similar constructions in Georgian and related languages, in which the Appl0
heads are realized with several VIs, which is unlike Spanish where only the clitic le spells
out the Appl0 head. The interaction between the morpheme of the Appl0 head and other
argument-structure changing morphemes will be explored in each subsection along with
the syntactic and semantic relations between these arguments.
3.2 Low applicatives in Georgian
3.2.0 Introduction
As mentioned in the literature review, three types of low applicative relations are found
cross-linguistically: low recipient, source, and stative possession. An individual language
may select all three, two, one, or none of them. This section illustrates that Georgian
displays all three relations.
3.2.1 Low Recipient applicatives
Pylkkanen (2002) and Cuervo (2003) define low applicatives as involving a dynamic
transfer- of-possession, which can be literal (Maria gave John a book) or metaphorical
(Maria showed Mercedes a book). There can be two types of transfer: either a transfer
‘to’ or a transfer ‘from’. The difference between the two meanings is essentially
attributed to the two types of applicative heads, which define how the applied argument
will be interpreted: as the Recipient or the Source of the theme object. Thus, in the
following structures of Georgian the applied argument is interpreted as a Recipient:
(17)a. nino-m ek’a-s c’ign-i
E-Dat book-Acc
‘Nino bought Ek’a the book.
APPLIC- bought
b. lali-m
gamo- ucxo.
prev- APPLIC- baked 56
Recall that the abbreviations in this text stand for the following notions: ‘prev’ for preverbal marker of
aspect and directionality; APPLIC- for the applicative marker; numbers stand for the person agreement
affixes; Erg for ergative case, Dat for dative case and Acc for accusative case, etc.
‘Lali baked Gia the cake.’
The dative applied arguments in these structures express the animate entities who
‘receive’ the object denoted by the theme. There is a clear sense that dynamic transfer-ofpossession occurs between an external and an applied arguments. In order for such a
transfer to happen, the predicate must express a dynamic event of acquisition such as
(buy), or creation (bake), etc. Levin & Rappaport Hovav (1991, henceforth, L& RH)
argue that bake may denote either change-of-state or creation. With the former
interpretation bake can participate in the inchoative/causative alternation (the cake bakes/
Sue baked the cake) while with the latter meaning the same verb participates in the
benefactive alternation (bake a cake for someone/ bake someone a cake, L & RH 1991:
139). I claim that the verb in (17b) is one of creation, while that in (17a) is a change-ofstate. The products of these activities unambiguously belong to the class of dative applied
arguments and the structures are therefore interpreted as low applicatives. The relation is
applicative, if the arguments involved in applicative relations satisfy certain syntactic and
morphological requirements. These requirements are: the applied argument must be
assigned either inherent or structural dative case and projected in the specifier of the
In Georgian, the dative case of applied arguments in low applicatives must be
structural because it changes into another case when the verb is in the Perfective series,
specifically, into the postpositional genitive. How this dative and the accusative cases are
checked on these arguments is discussed after presenting the derivation of low applicative
structures in two series (Present and Perfective):
(18) a. lali-m
‘Lali baked Gia the cake.’
gamo- ucxo.
prev- APPLIC- baked 57
b. [VoiceP Lelam [voice’ [vP [VP [ApplP Gia-s [Appl’ namcxvari [Appl u-/i-]] cxo] v°] Voice°]
c. The Present and the Aorist series low applicative
v [-features]
cxo ‘bake’
Appl.arg. Appl’
Appl0 [-features] [Dat]
Recall that the abbreviations in this text stand for the following notions: prev for preverbal marker of
aspect and directionality; applic- for the applicative marker; numbers stand for the person agreement
affixes; Erg for Ergative case, Dat for Dative case and Acc for Accusative case, etc. Also see the appendix
1 for fuller list of these glosses.
d. The Perfective series low applicative
v0 [-features, Case]
cxo, ‘bake’
i-, u-
As seen in (18c), the external argument is projected by Voice0 and in the aorist series,
Voice0 checks the ergative case on this case-split external argument while in the Present
series, Tns checks the nominative case feature on the same argument. The dative and
accusative case features of two lower arguments in (18c) are checked as a bundle along
with the phi-features of these arguments. As argued above, the dative case of the applied
argument is structural and I assume that the accusative of the theme is also structural
since the latter is specified on the DP in a certain syntactic configuration. Assuming
bottom-up derivations for these constructions and the cyclicity of spellout of certain
functional heads, specifically, that of little v0 (Chomsky 1999), I claim that the phi- and
case feature of the Appl0 head is checked first against the applied dative argument via the
head-specifier relation. I assume that when the little v0 head is merged, the dative applied
argument is still ‘active’, so that it can still match its features against the probe of the v0
head which is looking down the structure for the goal and checks its case and phi-features
again against the dative argument. Note that the applied dative argument must be marked,
i.e. be specified for 1-2 persons in order to check the phi-features of the probe and get
cross-referenced on the verb with the ‘m-set’ agreement markers 58. Recall from the
Introduction that due to templatic constraints, only the most local marked argument to the
v0 or Tns0 heads can be cross-referenced in the verbal template. Thus, I assume that the
dative applied argument is privileged with respect to the person agreement marking than
the lower Theme argument due to locality to the functional head specified for m-set
agreement (v0 ). In this scenario, I am positing the possibility that one argument can
check the phi-features of two functional heads (v0 and Appl0 ) as long as these heads are
merged in the same phase, i.e. under the little v0. 59
The lower theme argument checks its accusative against the v0 via an Agree relation
presumably. As illustrated by Chomsky 1995 and Bobaljik and Branigan 2006 among
others, it is possible for one head to be specified for two different case features and
particularly, the little v0 head across various languages has been posited to be specified
for the dative and accusative case features. Due to locality, I suggest that the structurally
marked dative argument does not incur a minimality violation as it is already deactivated
by checking the phi- probe of v0 and this argument does not intervene between the theme
Thie discussion on how the cross-referencing of the arguments on verb happens is in the Introduction,
Section 1.8. The definitions for two types of agreement and two sets of the person agreement markers ‘vset’ and ‘the m-set’ is also in the same section.
This type of checking pattern is again supported by the morphological evidence, i.e. the
separate morphemes inserted into the Appl0 head and the person agreement presumably inserted
in the little v0 head as argued by Lomashvili & Harley (in press).
and the v0 head for the latter to check its case features. Thus, the checking of case
features in low applicatives may be accounted for with this kind of scenario. Further, we
will be assuming that the similar checking pattern exists for other applied arguments and
themes merged in DOCs.
Note that in the Perfective series, however, this mechanism is a little different because
the external argument is marked with the dative case and it absorbs the case feature of an
applied argument projected in the specifier of the ApplP. This dative case of the subject
is checked against the Appl0 head. Since the position of an applied argument is taken by
the dative subject, the former dative applied argument is projected as an adjunct marked
with the postpositional genitive. It can be argued that this structure is no longer the true
Double Object Construction (DOC) because the genitive applied argument is no longer
an argument of the Appl0 head. The position of the accusative argument stays the same
and presumably, its case is checked against the same functional head as in (18c), i.e. v0
Since the merging site of the Appl0 head is below the event-introducing functional
head (v), we assume that the dynamic transfer-of-possession is due to the Appl0 head in
this construction. The phonological exponents of the low Appl0 head in Georgian are iand u-, which are sensitive to the features of the syntactic environment, i.e. the person
features of the applied argument:
(19) VIs for the applicative head in low applicative:
a. i-  Applic0 low / [Dat] [+1] [+2] 60
b. u-  Applic0 low / [Dat] [+3]
d. -  Applic0 low/ Elsewhere
As we have seen in the above examples, the Vocabulary Items (VIs) realizing the
Appl0 head in these low applicatives are i- and u-, which are sensitive to the phi-features
of the dative applied argument. There is also - applicative marker in (19), which I
assume is inserted into the Appl0 head in elsewhere contexts. Observe the following
examples in which the elsewhere phonological exponent - is inserted into the Apll0
head and in some of these DOCs, a- shows up in the place of the applicative marker
arguably inserted in the CAUSE:
(20) Low Recipient applicatives:
a. deda-m
bavshveb-s velosiped-i
Mother-Erg kids-Dat
‘Mother presented the kids a bicycle.’
b. mama-m
Father-Erg D-Dat
‘Father showed Dato the picture.’
c. deda-m
‘Mother gave Nino a present.’
CAUS- presented 61
mi- s- c- a
prev- 3O- give- 3S
The number symbols stand for the person agreement, i.e. 1-2 person arguments check the relevant
features on the Appl0 head and i- is inserted while u- is inserted when the 3rd person DP checks the relevant
feature. We do not use Halle’s (1997) features [ PSE] (Participant of the Speech Event) or Noyer’s (1992)
[ Addressee] for the features of these arguments for the reasons that become clear later when the
impoverishment of these features is discussed in Sections 3.3. and 3.4.
The affix a- is tentatively glossed as the VI for CAUSE by assumption that the transitive structures
projecting external arguments may be generally interpreted as causative eventualities. We also assume that
this a- in Georgian and related languages is a multi-functional morpheme whose secondary exponence can
be something like TRANS head (Jelinek 1988), which projects the theme argument in transitive structures.
For the detailed analysis of secondary exponence see Noyer (1997).
A direct transfer-of-possession happens in (20a & c) while a metaphorical transfer in
(20b). Both low applicative meanings result from the attachment configuration. Note that
some of these verbs can in fact realize the allomorphy i-/u- in certain contexts. The verb
show ‘achvena’ in (20b) illustrates this:
(21) Low applicative of ‘show’
m- ichvena
1S- APPLIC- showed
‘He/she showed me X.’
b. v- uchvene
1S- APPLIC- showed
‘I showed him/her X.’
However, two other verbs a-chuka ‘present’ and mi-s-c-a ‘gave’ do not show this
allomorphy in any context:
(22) Applicative of present and give impossible:
a. *mo- m- isca
prev- 1O- APPLIC- gave
‘He/she gave me X.’
b.* m- ichuka
1O- APPLIC- presented
‘He/ she presented me X.’
It is evident that the causative structure in (20b) can undergo applicativization while
two other verbs (20a & c), which are semantically interpreted as the verbs of giving,
resist this alternation, i.e. in no frame are they capable of including the phonological
exponents i-/u- in their templates. The DOC data in (20) shows that the phonological
exponents a-, i-, u- and - markers may be realized in the same third position of a
ditransitive verb template. These items are inserted into different functional heads such as
Appl0 or CAUSE in specific environments bearing a different complex of contextual
features. The question is why some of these verbs ‘resist’ the insertion of i-/u- exponents
and the Appl0 head in some of these structures is realized with the null phonological
I argue that the verbs of giving are ‘pure’ ditransitives and by default, such structures
should deterministically project the applied argument along with two other main
arguments—an external and a theme. In other words, the roots of such pure ‘ditransitive’
verbs when projected require the functional apparatus, which projects all three
arguments. These types of structures may be considered as unmarked cases with respect
to the applicative relation. Here are the structures illustrating this point:
(23)a. John gave Mary a book. (Requires all three arguments)
b. *John gave a book.
(24)a. Mary baked John a cake. (Adds the applied argument)
b. Mary baked a cake.
As indicated in (23), give is by default a ditransitive verb and when projected it
requires all three arguments to be realized in the clause. Therefore, the transitive structure
with this verb is not good (23b). This verb is different from bake which can occur in the
monotransitive frame and can also add the applied argument to form DOC. Halle &
Marantz (1993: 133-134) argue that in certain environments UG may provide a zero
spell-out as a default phonological realization of a morpheme in an unmarked case. They
consider English present tense forms without a suffix as the Elsewhere case, where Tns is
realized by a default null suffix. 62 The conclusion is that the null phonological exponents
are provided by the UG for unmarked cases and the same explanation may be assumed
for the presence of null exponents for the Appl0 head in the verbs of giving since these
semantically are unmarked with respect to the applicative relation.
Note that the allomorphs of the Appl0 head are not phonological or prosodic contextdependent in the sense that they are not conditioned by the phonological or prosodic
properties of Roots. They differ in terms of their substantive morphosyntactic features,
such as [Dat] and phi-features, but they are not sensitive to the kinds of root-specific
contextual features which condition the allomorphy of CAUSE in Georgian (Chapter 2)
and the tense suffixes of English. Recall that the allomorphy of CAUSE in the
inchoative-causative alternations of melt/melt type was Root-conditioned, i.e. the
syllabicity of Roots was of importance in the insertion of causative markers. In English,
/t/ and /d/ are allomorphs of the Past Tense and they impose certain conditions on the
verbal stem, namely, that the former requires a voiceless stem-final consonant, while the
latter requires a stem ending on the voiced segment. Thus, the phonological exponents /t/
and /d/ are the instances of contextual allomorphy. Georgian applicative markers do not
impose such phonological or prosodic constraints on verbal stems but they do show verb
class-determined allomorphy as illustrated in the above discussion. This raises the
question of whether the Appl0 head is cyclic or not.
The way to determine the cyclicity of the applicative head in a given structure comes
from the allomorphy of this head across various verb classes. This proposal hinges on
The null suffix can also be homophonous, irregular and marked VI in other contexts, as it is the exponent
of [+past] tense in the verbs like drive and hit.
empirical evidence shown in (20) where Roots are concatenated with the applicative
morpheme in the template of a Georgian verb and no other morpheme intervenes between
them. I repeat these examples here for convenience:
(25) The position of the applicative morpheme with respect to Roots
a. gamo- ucxo
prev- APPLIC- bake
‘X baked someone a cake’
b. uchvena
APPLIC- showed
‘X showed someone Y.’
I argue that the following assumptions are of relevance to determining whether the
Appl0 head is cyclic or not:
(26) The assumptions regarding the Appl0 head allomorphy:
a. The allomorphs inserted in the Appl0 head may show the contextual allomorphy
sensitive to the phonological properties of Roots;
b. VI for the Appl0 head may be determined by the class of Root (the morpheme should
impose certain constraints on the class of Roots, such as unaccusative, unergative, etc.). 63
As seen in above examples, the VIs inserted into the Appl0 head showed sensitivity to
the person features of the dative applied argument and the null elsewhere item was
inserted into this head with the verbs of giving. This evidence shows that the low Appl0
head may not be cyclic since it shows verb-class determined allomorphic selection,
consistent with Embick’s proposal that contextual allomorphy can be conditioned by the
I assume that Roots are selected from the lexicon without categorical information (Marantz 1997, 2000)
and that their syntactic behavior can be largely predicted based on the category-defining heads that they
merge with.
class membership of Roots. The data from low Recipient applicatives demonstrate that
the VIs realizing the Appl0 head show such sensitivity to the aforementioned Root
3.2.2 Low Source applicatives
An applied dative argument can be licensed by the Appl0 head in the environment of
transfer predicates that express reverse directionality (Cuervo 2003: 51). This direction
may come from the applicative relation of Source when the Source is projected by the
Appl0 head in the specifier of the ApplP. The transfer-of-possession happens from the
Source to an external argument. Typically, the source is marked with the dative case and
the theme with the accusative case in such structures cross-linguistically. In Georgian, the
same cases are assigned to the applied argument and the theme. The source is interpreted
as a ‘possessor’ of the theme respectively:
(27) Low source applicative
a. erekle-m gia-s
‘Erekle hid the keys from Gia.’
da- umala.
prev- APPLIC- hid
b. dato-m
prev- APPLIC- lost
‘Dato lost Lia’s book.’ (lit: Dato lost a book from Lia’s possession).
c. nino-m
mak’a-s k’aba
prev- APPLIC- sold
‘Nino sold Mak’a’s dress.’ (lit: Nino sold a dress from Mak’a’s possession).
Notice that in all three examples, no direct transfer-of-possession occurs between the
dative and external arguments. However, the relation between the dative applied
argument and the theme is one of possession. This relation is affected by the event whose
endpoint marks the change of possession of the theme argument.
Note that English NPs which correspond to the dative arguments in Georgian are
assigned the genitive case in the translation, as they appear internal to the theme NP as
possessors of the NP. This could be due to language-specific constraints on the selection
of the Appl0 head by the verbal head in these structures. In Georgian, the assignment of
genitive case to the Source correlates with the disappearance of the applicative marker
from the verb. Note that in (26), the dative argument is interpreted as ‘affected’ since it
loses the possession of the theme in all three structures. The same ‘affected’ meaning is
not available for genitive possessors as shown in the following examples:
(28) Genitive possessors
a. erek’le-m gia-s
E- Erg
‘Erek’le hid Gia’s keys.’
b. dato-m
L-Gen book-Acc
‘Dato lost Lia’s book.’
c. nino-m ketevan-is
‘Nino sold Ketevan’s dress.’
da- mala.
prev- hid
da- k’arga.
prev- lost
ga- q’ida.
prev- sold
The disappearance of the applicative morpheme in (28) indicates that the Appl0 head
is not introduced in these structures and the applicative construction turns into the simple
transitive. The possessor relation between the genitive cased-marked DP and the theme is
kept. There is no relation between the possessor and the event.
Note that the morphological realization of low Source applicatives is i-/ u- as in low
Recipient structures. The morpheme realizing the Appl0 head is sensitive to case and the
phi- features of the dative argument ([Dat], [+1], [+2] and [+3]). These features on the
DP presumably check the relevant uninterpretable person and case features on the
functional head (Appl0). By assumption the case checking pattern of the lower dative and
the accusative theme arguments is the same here as in the low recipient applicatives
shown above. We do not repeat it here. By contrast to low Recipient applicatives, there is
no evidence of Root-determined contextual allomorphy of the Appl0 head in these low
Source constructions. I tentatively assume that low Appl0 head in these constructions is
not a cyclic head.
3.2.3. Low state applicatives: Possessor Datives (AT)
Recall that Spanish has low applicative structures formed by state and dynamic action
verbs which do not express transfer-of-possession either literally or metaphorically. The
following structures illustrate Cuervo’s point repeated here as (29):
(29) State and dynamic applicatives (at)
a. Pablo le
a Valeria
Pablo CL.DAT admires
‘Pablo admires Valeria’s patience.’ (lit: Pablo admires Valeria the patience).
b. Pablo le
la frente
a Valeria.
CL.DAT. kissed the forehead.ACC Valeria.Dat
‘Pablo kissed Valeria on the forehead.’ (lit: Pablo kissed Valeria the forehead).
(Cuervo 2003: 53)
There is no sense in which the dative argument ‘gets’ or ‘loses’ the theme in these
structures: simply, the applied argument is understood as the possessor (or the Location)
of the theme.
In Georgian, similar structures have slightly different meanings as showed in the
(30) Dative possessors (AT)
a. dato-m lia-s
L-Dat patience
‘Dato approved Lia’s patience.’
mo- uc’ona
prev- APPLIC- approved
(31) Datives of activities (AT)
lela-m bavshv-s loq’eb-i
da- uk’ocna.
check-Acc prev- APPLIC- kissed
‘Lela kissed a kid’s cheeks.’ (Lit: Lela kissed the kid on the cheeks)
Dative arguments are interpreted as stative or inalienable possessors of the same
possesseee. Patience and work are themes that are abstract entities cultivated by the
applied dative arguments. The event semantics of psychological verbs in these structures
can be interpreted as activities as opposed to states. As mentioned earlier in Section 2.8,
in Georgian, the psychological verb ‘like’ as well as love, hate, etc. comes in three
semantic and morphological shapes denoting state, dynamic passive, and active events.
The fact of interest is that only activity verbs project external arguments and can add the
dative applied argument via the Appl0 head. This ability of activity psych verbs to add
dative applied arguments correlates with low applicative meaning. In Spanish, as noted
by Cuervo, low applicative meaning (possessor AT) can be expressed by certain
psychological verbs denoting states, which project just the dative experiencer and the
theme. Indo-European languages, including Spanish, lack the morpho-semantic variety of
psychological verbs found in Georgian and related languages. The interpretation of
psychological verbs in these languages is mainly stative. In what follows I illustrate that
in Georgian, a low applicative possessor relation is expressed only in those structures
where the dative applied argument is added to the argument structures of active
psychological verbs. It should be kept in mind that the relation between the dative applied
argument and the theme is still one of stative possession as illustrated in (30)-(31).
The low applicative meaning expressed by the activity verb ‘kiss’ in (31) shows
inalienable possession of the theme by the dative argument. This possessor-possessee
relation introduces two variables in the structure: the theme and the possessor dative
argument. Here is the structure demonstrating this relation:
(32) Low possessor applicatives
a. dato-m lia-s
L-Dat patience
‘Dato approved Lia’s patience.’
mo- uc’ona
prev- APPLIC- approved
b. [vP Neli [v’ [VP [ApplP Kato [Appl’ shroma [Appl° u-]] pasa] [v° Ø]]
Dato-m VP
 c’on- ‘approve’
The morphological realization of the Appl0 head in this low possessor applicative is
the same as in other types of low applicatives: the Recipient and the Source as illustrated
above. The VIs inserted into the Appl0 head show the same kind of free allomorphy
which is sensitive to substantive features of the dative argument. Since the group of
predicates showing such possessor relations is not large, it is hard to test whether the VI
- is also available as an elsewhere item or not.
It can be argued that based on these structures, there is no enough evidence to
determine whether the Appl0 head is cyclic or not. The vocabulary items are inserted in
the contexts which are determined by the substantive morpho-syntactic features of dative
arguments. This kind of sensitivity may be taken as evidence for non-cyclic Appl0 head.
For the purposes of comparison of these structures with the genitive possessor
constructions, the following structures may shed light on the change in meaning and the
morphological spellout of verbs:
(33) Genitive stative possessors
a. neli-m
‘Neli approved Rusudan’s patience.’
b. deda-m
Mother-Erg kid-Gen
‘Mother kissed the kid’s cheeks.’
mo- ic’ona.
prev- REFL- approved
da- k’ocna.
prev- kiss
Along with the change in meaning, which is entailed as the result of the subtraction of
one variable (dative applied argument) from the corresponding applicative structure, the
verbal morphology also undergoes changes. The VI i- preceding the Root c’on- in (33a)
can be analyzed as a Reflexive marker for a number of reasons. This VI is not sensitive to
the phi-features of the applied argument and the series features of T0 or Infl0 like other
morphemes that realize event heads in the same position. The details of the insertion are
not relevant here, although they will be analyzed below, in Section 3.5 dealing with
reflexive applicatives. The following examples illustrate that the prefix i- is sensitive
neither to the person feature of an external experiencer argument, nor to that of the
(34) The psychological predicate and its morphology
a. mo- v- ic’one
prev-1S- REFL - like
‘I liked it.’
b. mo- ic’one
prev- REFL- liked
‘You liked it.’
c. mo- ic’ona
prev- REFL- liked
‘He/she liked it.’
Ignoring other morphemes in this paradigm, it is evident that the affix i- is not
sensitive to the person feature of any argument projected in the clause (See Appendix A
for detailed explanation of the homophony of this VI with other morphemes). The subject
is assigned nom/erg/dative cases like the external arguments of DOCs or transitive verbs
and the theme is assigned accusative. The checking pattern of these cases is the same as
in Recipient applicatives shown above. For the interpretation of meaning in these
structures only the theme argument is interpreted as a variable related to the event
calculus of the applicative phrase.
3.3 Low applicatives of unaccusative and inchoative verbs
3.3.0 Introduction
In Georgian unaccusative verbs, a low applicative relation can be expressed between
the dative subject argument and the theme projected by the internally caused change-ofstate verbs. Cuervo (2003) also discusses similar applicative constructions in Spanish but
none of her examples shows the Recipient relation between the dative applied argument
and the theme. Instead, in her analysis, the notion of affectedness better accounts for the
dative argument in the following applicatives whose event structure is argued to be
complex, as in other inchoative verbs:
(35) Affected applicatives formed from inchoatives
a. A Carolina
rompio el florero.
Carolina.Dat CL.Ref CL.Dat broke
the vase
‘The vase broke on Carolina.’ (Lit: ‘To Carolina broke the vase.’ )
b. [TP T [vP2 [vGO se] [ApplP [A Carolina] [Appl’ [Appl le] [vP1 vBE + root [el florero]]]]]]
A Carolina
vbe + Root
(Cuervo 2003: 112)
The structure of the affected applicative in (35) represents a complex event which can
be decomposed into the higher vGO, usually introducing change-of-state semantics in
inchoatives (fall, grow, break, etc.), and the lower vBE head, which expresses the resultant
state. The dative applied argument is sandwiched between these vPs, staying external to
vPGO. No external argument is licensed in these structures. In this sense, affected
applicatives are distinct from low applicatives and they are analyzed as a distinct kind of
relation. None of the low applicative relations (Recipient, Source, and the Possessor) is
present in the affected applicative illustrated above.
3.3.1 Empirical data and syntactic analysis
In what follows I argue that in applicatives of internally or externally-caused changeof-state verbs, as well as in certain inchoative and transitive verbs a low applicative
relation obtains between the dative subject and the theme with the same morphological
realization of all these diverse structures:
(36) Applicatives of unaccusative, inchoative and transitive verbs
a. m- epurčkn- eb- a 64
1S- voice- blossom- TH- 3O
‘It is blossomed for me. ( Lit: ‘I am blossomed it.’)
b. m- ezrd- eb- a
1S- voice- grow- TH- 3O
‘It is grown for me.’ (Lit: ‘I am grown it.’)
c. m- eɣeba
1S- voice- open- TH- 3O
‘It is opened for me.’ (Lit: ‘I am opened it.’)
d. m- ec’ereba
1S- voice- write- TH- 3O
‘It is written for me.’ (Lit: ‘I am written X.’)
e. m- exvneb- a
1S- voice- plow- TH- 3O
‘It is plowed for me.’ (Lit: ‘I am plowed it.’)
f. m- eteseb- a
1S- voice- sowed- TH- 3O
‘It is sowed for me.’ (‘I am sowed it.’)
The dative subject argument in these structures is interpreted as the Benefactee of an
event expressed by the verb. The theme argument undergoes internally or externallycaused change-of-state and, literally or metaphorically, the nominative theme argument
In these structures, the Dative subject is indexed with the m-set markers in verbs based on the locality
with the v0 head (see section 1.6 of Introduction for more details).
ends up in the possession of this higher dative argument. This meaning should involve
the low applicative head that relates the dative argument to the theme. Also, notice that
the morphological shape of these applicatives is uniform, although the event semantics of
these verbs is not the same. In (36a & b), the events are internally-caused changes-ofstate, while in (36c & f), the themes undergo externally caused changes-of-state.
Generally, the verbs in (36d-f) are transitive, as opposed to inchoatives and unaccusatives
in (36a-c). It may be suggested that the structures with the transitive verbs are passivized
versions of the corresponding DOCs in which the applied argument becomes the subject
of passive. The applicative in (36a) cannot appear in the transitive frame under any
argument-structure changing operation. Irrespective of these differences, the derivation of
all these events should involve the Appl0 head that relates the dative Benefactee argument
to the theme and this relation can be expressed locally by these arguments. Before
sketching the derivations for these structures, note that the event structure of the
applicatives involves both simple and complex events and some of these structures
should be analyzed individually. Following Cuervo (2003), I suggest that the internallychanged verb ‘blossom’ gaipurchkna consists of a simple event with the vPGO expressing
a change:
(37) The structure of a simple event of change (gaipurchkna)
ga- ipurchkna
prev-REFL- blossom
‘It blossomed.’
The structures in (36d, e & f) clearly represent the passives of ditransitive verbs such
as John sowed Alex the field in which Alex becomes the subject of the passive sentence
shown in (36f). Below it will be argued that this is supported with the non-active voice
morphology in these forms (the affix e-). The structure of a simple transitive verb
corresponding to the mentioned passive would be the following:
(38) The structure of an event in simple transitives:
Da- tesa
Prev- sowed
‘He sowed X.’
‘sow’ tes
However, the Roots like grow, open etc. that form the core of the applicatives in
(36b & c) can participate in the inchoative/causative alternation and in their inchoative
frame, they represent complex events combining two vPs in the derivation. It may be
argued that the underlying structure of inchoatives can be the same as in other languages.
Here is a structure of inchoative events adapted from Cuervo (2003):
(39) The structure of the inchoative = (36b&c)
(Adapted from Cuervo 2003:104)
Observe that vGO in these structures does not project an argument. Only the lower
stative head of vPBE does. The subject of inchoative verbs is assigned the morphological
nominative case across all three series. This case suggests that the subject in the
construction like (36b & c) is like the underlying object of the corresponding
transitive/causative structure, although it displays almost all of the properties of transitive
verb subjects. As seen from (37)-(39), the verbs in (36) have different underlying syntax,
namely the event-introducing heads (little vs) are different in them and in (39) the event
structure is complex as it combines two vs (vPBE and vPDO).
I propose that the applicatives of various event types, shown in (38)-(39), have one
common structural property: they introduce the ApplP whose head projects the dative
applied argument in the specifier of this projection. Since a Recipient relation exists
between the dative applied argument and the theme, structurally these arguments must be
in the same local domain and ideally would be related to each other by the Appl0 head
when the dative argument is projected as the specifier of the Appl0 head and the
nominative argument as the sister of this head. However, the analysis of other Recipient
low applicatives has shown that the theme argument was projected as the sister of the
Appl0 head (bearing the accusative case in Spanish and Georgian). This creates a problem
for the analysis of the nominative theme arguments in applicatives of inchoative, or
internally-caused change-of-state verbs. The nominative case-marked theme must be
projected as the subject of vP as shown in (39).
Dative arguments in these structures are also different from dative arguments in
DOCs. Recall from the introduction of the dissertation (Section 1.7, example (29)) that,
in the Perfective series, dative applied arguments are uniformly assigned the
postpositional genitive and the applied argument is demoted to an adjunct position due to
this change in case marking. Observe the difference of the case in the Perfective series
verbs and the dative subjects of inchoative verbs:
(40) a. Genitive applied arguments of DOC in the Perfective:
house-Acc Prev- APPLIC- build
‘Dato (apparently) has built the house for Gia.’
b. Dative Benefactor argument of unaccusatives in the Perfective series:
ga- purchknebia.
Dato-Dat flower-Nom
Prev- blossom
‘The flower has (apparently) blossomed for Dato.’
Thus, in DOCs, the Appl0 head checks the structural dative case of the applied
argument as opposed to internally-changed and other activities shown in (36) and (40b)
where the inherent dative case is marked on the Benefactee. Given these distinctions
between the dative cases on applied arguments in (36) and the applied argument in
DOCs, below I argue how structural and inherent cases on dative arguments are checked
without incurring minimality violations.
First let us analyze the properties of the theme arguments in applicatives of
inchoatives. As mentioned above, these nominative themes are underlyingly like
nominative subjects of unaccusative and inchoative verbs. However, these arguments
must be c-commanded by the dative Benefactees as illustrated by the native speaker
judgments. Therefore, the nominative theme which is a possessee of the dative argument
must be projected lower in the tree. I propose that the dative benefactee and theme
arguments in applicatives of inchoatives and internally-caused change-of-state verbs are
projected under the ApplP. I assume that case checking of the arguments may happen
from other functional heads rather than from those which license these arguments (such
as the Appl0 head may not check the nominative case of the theme, although the latter is
selected as its complement):
(41) a. m-e-teseba ‘It is sown for me’
b. m-e-ɣeba ‘It is opened for me.’
c.[VoiceP [TP [T’ [vPDO/GO [DPnom [vPBE [VP [ApplP DPdat Appl°] purchkn] vBE ] vDO/GO] T] e-]]
V0 purchkn,  tes, ‘blossom, ‘sow’
i-, uThe benefactee does not represent a barrier for the case checking relation between the
nominative theme and the T0 due to its inherent dative case. Thus, the dative case on
these higher Benefactee arguments is checked against the Appl0 head and the nominative
case on the theme against T0 via Agree.
3.3.2 Analysis of morphological marking
The morphological shape of the inchoative applicatives indicates that, instead of the
i-/u- and - exponents of the Appl0 head, the voice marker e- shows up in the pre-base
position of the argument structure-changing morphemes. Recall from Chapter 2 that the
same VI appears in contexts where the head of the VoiceP projection bears the
[NonActive] feature. I assume that this feature is also present on the Voice head in these
constructions, and consequently, Voice0 does not project the external argument in these
The question arises as to why the morpheme realizing the Voice0 shows up instead of
the VI of the Appl0 head. Recall from the introduction (Section 1.7) that in Georgian and
related languages, language-specific constraints are imposed arbitrarily on the verbal
template and they allow only one argument structure-affecting morpheme to be realized
in the pre- and the post-base positions. The following schema shows this constraint:
(42) … ___________ Root/Base __________ …
Antipass, etc.
These functional heads affecting the argument structure have just two positions-ofexponence in the verbal template, one in the pre-base and another in the post-base
position. In inchoative applicatives, only one VI marking either the Applicative or the
Voice head can receive a spell-out in the verbal template. The data in (36) shows that
only the VI e- realizing the Voice0 is inserted. The question is what privileges the VI of
Voice over the Applic0.
There can be two different answers to this question. The first could be provided by a
pre-insertion impoverishment rule, which changes the feature content on certain
functional heads (Halle & Marantz 1993, Halle 1997) bleeding the insertion of the
applicative morpheme. The second answer could be provided by the existence of a null
exponent for the Appl0 head in certain contexts. For the first scenario, I propose that an
impoverishment rule applies to the person features of the Appl0 head. The
impoverishment in this approach is represented as delinking, which means that the
delinking of certain features entails the delinking of features dependent on them. I follow
Noyer (1992) and Harley (1994) in positing this process and in recognizing that the
following hierarchy of features: 12 3 pl  dual  fem underlies the feature geometry
shown in (43). In Georgian, delinking of the feature [+1] entails delinking of [+2] and
consequently, of [+3] while the case feature [Dat] dominates the phi-features and it is
kept intact since the delinking applies just to person agreement features:
(43) Impoverishment as Delinking
[+ 2]
[+ 3]
As the result of this delinking, only the dative case feature will be left on the Appl0
head and the phonological exponent of Voice0 will be realized in the same position
The second solution is compatible with the syntactic structure that these verbs project
and the feature content on the heads that introduce arguments in these applicatives. As
argued above, the Voice cannot project an external argument due to the feature
[NonActive] and the dative Benefactee/Recipient has the same relation to the theme as
other dative applied arguments in DOCs. It could be that the feature content on the Voice
has some relevance for the phonological realization of the Appl0 head. A zero exponent
for Appl0 may be inserted in elsewhere contexts as argued in Section 3.2 on low
applicatives. Recall that the verbs of giving appear to realize a null exponent of the Appl0
head. Similarly, I may argue that the null elsewhere item is inserted into the Appl0 head
in structures where the Voice head bears this specific [NonActive] feature due to which it
cannot project an external argument. The presence of this null exponent does not prevent
the exponent of [NonActive] feature from being realized in the same position as the VI of
the Appl0 head.
As for the preference between these two analyses, it can be argued that the second is
more ‘economical’ because it does not need to posit the special rule for featural economy
as presented in the delinking analysis. Halle (1997), Siddiqi (2009) among many others
argue for both featural and exponential economy whereby UG restricts the featural
content on the nodes positing only those morphosyntactic features which are relevant for
the derivation of the structure. Siddiqi (2009) argues for the principle that minimizes
exponence, i.e. the derivation realizes “all the formal features of the derivation with the
fewest morphemes” (Siddiqi 2009:4). These are two different kinds of economy
principles that are active in vast majority of languages and I may also argue following
these accounts that the ‘Rule economy’ could be a better option for explaining the
appearance of the Voice morpheme in above applicatives.
3.4 Low applicatives of Noun/Adjective-incorporated predicates
The following set of verbs illustrates low applicatives of noun- or adjectiveincorporated (henceforth, NI and AI) verbs which project dative Benefactee subjects and
nominative theme arguments undergoing a change-of-state. These structures resemble the
applicatives in the above section in many ways:
(44) Applicatives of AI- and NI- incorporated verbs
a. m- imc’ip- deb- a
1S- APPLIC- ripe- pass- TH- 3O
‘It is ripened for me.’ (lit: ‘I am ripened.’) 65
b. m- iɤvin- deb- a
1S- APPLIC- wine- pass- TH- 3O
‘It is wined for me.’ (lit: ‘I am wined’)
c. m- ilp’eb- a
1S- APPLIC- decompose- TH- 3O
‘It is decomposed on me.’ (lit: ‘I am decomposed it.’)
d. misc’or- d- eb- a
1S- APPLIC- straight- pass- TH- 3O
‘I am satisfied with it.’ (lit: ‘I am straightened.’) 66
e. m- itvinier- deba
1S- APPLIC - tamepass- TH- 3O
‘It is tamed for me.’ (lit: ‘I am tamed X.’)
m- isxr- eb- a
1S- APPLIC- heal- TH- 3O
‘It is healed on me.’ (lit: ‘I am healed it.’)
Note the difference between the morphological shape of these applicatives and that
shown for the inchoative applicatives in (44). In these structures, the pre-base position of
the argument structure-changing morphemes is realized with i-/u- allomorphy, which is
sensitive to the person feature of the dative subject. This can be seen in the following
(45) Adj-incorporated applicatives
I assume that all these forms in English are impossible.
The latter verb has acquired idiomatic meaning recently—‘I like something.’
a. m- imc’ip- d- eb- a
1S- APPLIC- ripepass- TH- 3O
‘I am ripened’
b. g- imc’ip- deb- a
2S- APPLIC-ripepass- TH- 3O
‘You are ripened.’
c. umc’ip- deb- a
APPLIC- ripepass- TH- 3O
‘He/she is ripened.’
The Voice0 head in these structures is not active since no external argument is
projected into the clause. The applicative relation is interpreted as low because the theme
argument undergoing the change-of-state directly benefits or affects the subject
Another morpheme of interest in these structures is the post-base –d, which realizes
either the Voice0 head or some verbalizer functional head v0. I analyze this morpheme as
another VI of Voice0. The reason will become clear below in this section where the
interaction between the VIs realizing the Voice and Appl0 heads is analyzed. This -d
consistently shows up in Noun or Adjective-incorporated unergatives as well as the
dynamic passive psychological predicates:
(46) AI and NI dynamic events
a. mc’ip- dripe- pass‘It is ripened.’
b. tvinier- dtamepass‘It is being tamed.’
TH- 3S
eb- a
TH- 3S
Following Hale & Keyser’s theory of argument structure (2002) these forms may be
analyzed as Noun and Adjective-incorporated verbs in which the Roots are incorporated
into the verbalizing v0 head. Following Harley (1995), I assume that the verbalizing head
may be interpreted as BECOME. In Cuervo’s system, this v0 head may be interpreted as
the vGO of changes. The derivation of the applicative construction from these verbs will
involve a low Appl0 head projected below VP, and this VP is composed of the
acategorical Root first combining with a categorizing head of an adjective or a noun (a,
n) and then conflated with the empty V0 head, as shown in the following structure:
(47) Noun/Adj- incorporated applicative
Lit: ‘I am ripened.’
Voice0 [NonActive]
As seen from this structure, the dative argument is projected in the specifier of the
ApplP and the argument checks its inherent dative case against Appl0 head. The
nominative case of the lower theme argument is checked against Tns via Agree since
there is no structurally-marked argument intervening between this nominative theme and
the T0.
Next we address why the VI of the Appl0 head shows up in the pre-base position
rather than the Voice0 morpheme. Note that the Voice0 head specified for [NonActive]
feature was realized as e- in the applicatives of inchoatives, while in NI and AIapplicatives the Voice0 head is additionally specified for [+A] and [+N] features. It can
be assumed that Roots enter the derivation with the same features and VI for these
features is (-d). Thus, the following items can be inserted into the Voice0:
(48) VIs for Voice0:
a. -d  Voice0 [NonActive]/ Roots [+ N], [+ A]____
b. e-  Voice0 [NonActive]/ Roots (inchoative, etc.)_____
c. -  Voice0 [NonActive] / Elsewhere
The post-base position of the argument structure-changing morphemes will be filled
by –d due to the Elsewhere principle which privileges the VI with more specific feature
specifications compatible with the environment where it is inserted. Thus, I claim that
when –d is discharged in the post-base position and the pre-base position remains empty
for the VI realizing the Appl0 head, the VIs of this head will be inserted. This kind of
relation between the VIs realizing the Voice can be called as discontinuous feeding
(Noyer 1997) with which the two exponents of the same head do not bleed each other,
rather the insertion of one exponent such as that of Voice in the post-base position ‘feeds’
the pre-base position by insertion of the phonological exponent of the Appl0 head. This
happens with the VIs of the Voice0 and Appl0 head in above examples. Thus, DM
provides adequate explanation of the morphological shape of NI and AI applicatives.
3.5 Reflexive applicatives
3.5.0 Introduction
Georgian and related languages can express a benefactive relation between an external
argument and a theme with dynamic events. Verbs that participate in the paradigm are of
wide semantic variety such as creation (bake, cook, prepare, etc.), locative/locatum (sew,
stitch, load, etc.), etc. These are typical transitive structures projecting agentive external
arguments where the latter are assigned the nom/erg/dat cases across series. The
applicative relations of Recipient and Source are formed between the external argument
and the theme and this distinguishes them from regular transitive structures. Essentially,
the external argument acts like the dative applied argument in DOCs in many respects.
However, the relation between the dative applied argument and the theme in DOCs is one
of transfer-of-possession, while in these constructions, which I provisionally refer to as
reflexive applicatives, the theme is in the prospective possession relation with the
external argument. First, observe the examples:
(49) Reflexive applicatives:
a. lela
L-Nom food
‘Lela is preparing the food for herself.’
b. gela-m
prev- REFL- switch
‘Gela turn on the light for himself.’
c. aleksi-m
prev- REFL- spread
‘Alex spread the carpet for himself.’
d. me
av- ishene.
prev- 1S- REFL- built
‘I built the house for myself.’
e. nino-m
gamo- icxo.
prev- REFL- bake
‘Nino baked a cake for herself.’
f. shen
she- ic’vi
You-Erg potatoes
prev- REFL- fry
‘You fried potatoes for yourself.’
Notice that all these forms have the VI i- in the pre-base position. In contrast to the
applicative markers i-/u-, this morpheme is not sensitive to the person features of the
external argument that can be seen from (49d, e & f), where the subjects are of three
different persons: first, third, and second respectively. Below I argue that this i-realizes
the reflexive voice head rather than the Appl0.
Recall that the transfer-of-possession relation between the dative argument and the
theme was the defining structural feature of low applicatives in Pylkkanen (2002) and the
subsequent work. Therefore, it may be assumed that morphologically such a semantic
relation will have to be expressed differently than applicatives with transfer-of-possession
relation between the external and Theme arguments. External arguments and themes
should not form close-knit semantic relationships. Nevertheless, this appears to be
present here. First, note that in the following reflexive verbs, VI i- occupies the pre-base
position of the argument structure-changing morphemes:
(50) Reflexives in Georgian
a. ek’am tav-i
da- ibana.
E-Erg. head-Acc
prev- REFL-washed
‘Ek’a washed her head.’
b. dato-m p’ir-i
ga- ip’arsa.
face-Acc prev- REFL- shaved.
‘Dato shaved his face.’
These verbs denote dynamic events directed towards one’s own body parts. The
phonological exponenet i- shows the same properties as the one in the applicative
structures in (49). Specifically, this i- is not sensitive to the person features of the external
argument or to the series features as is detected with the same exponent of the Voice0
with unergative verbs. Below I argue that this i- is the exponent of the reflexive Voice:
(51) The VI i- in unergatives across series:
a. Present series
v- cek’v- av67
1S- dance- TH
‘I am dancing.’
b. Aorist series
v- icek’v- e
1S- VOICE- dance- Aor
‘I danced.’
c. Perfective series
m- icek’v- ia
1S-VOICE- dance- TH- 3O
‘I have (apparently) danced.’
Thus, I argue that the phi-feature insensitive i- realizes not the Appl0 head but rather
the reflexive Voice in (49). Note also that the exponent i- is not found in many transitive
structures as illustrated in the following:
(52) Transitive activities
a. v- ch’- am
1S- eat- TH
‘I eat X.’
b. v- c’er
1S- write
‘I write it.’
c. v- tamash- ob
1S- play- TH
Note that in (61a) the insertion of the reflexive morpheme is sensitive to the series features, i.e. in
[present] the default VI - is inserted into the Refl0.
‘I play X.’
This evidence will support the analysis of the morphosyntax of reflexive causatives
which is discussed next.
3.5.1 Analysis of morphosyntax of reflexive applicatives
I propose that both reflexive and reflexive applicative structures combine the same
reflexive Voice realized with the exponent i- in their derivation with the eventintroducing and lexical heads (vDO and V0). In reflexive applicatives, the ApplP also
merges. Presumably, the exponent of the reflexive Voice bleeds the affixes i-/u- of the
Summarizing what I have said above, the structure for the ‘pure’ reflexive events should
be the following:
(53) The structure for reflexives
a. da- ibana
prev-REFL- washed
‘He/she washed (his head).’
b. [VoiceP External Arg [vPDO [ReflP [VP theme  bana] [Refl° i] ] vDO] Voice°]
Tavi ‘head’
bana ‘wa
This configuration of the Voice head with respect to the theme does not necessarily
entail a close relation between the external argument and the theme but the import of this
structure will soon become clear when the structure of reflexive applicatives is analyzed
I conceptualize ‘pure’ reflexives much like high aplicative structures where the relation
obtains between the external argument and an event. Inalienable possession does not need
to be represented with the local relation between the arguments. Likewise, the
benefactive and prospective possession meaning between the external argument ands the
theme in reflexive applicatives can be attributed to the low attachment site of the Appl0
head, which relates the theme to the external argument through some syntactic
mechanism. In syntax, this relation is implemented in the following way: The reflexive
Voice head reflects that the Appl0 head has projected PRO in the specifier of the ApplP
rather than the full dative applied argument. This empty element is coindexed with the
external argument. The accusative theme can thus be selected by the Appl0 head:
(54)a. Gela-m
a- into.
prev- APPLIC- switch
‘Gela turned on the light for himself.’
nto ‘switch’
i-/uThe accusative case of the theme can be checked against vDO via Agree since no
other argument intervenes between the theme and this functional head to incur the
minimality violation. I assume that the case of the external argument is checked against
the Voice0 and Tns depending on the series features (Present or Aorist).
Thus, applicative relations can be expressed via the empty argument PRO and the
reflexive Voice head in these structures. The transfer-of-possession meaning between the
external argument and the theme can be obtained through the local relation between these
arguments. Language-specific templatic restrictions are responsible for bleeding the
exponent of the Appl0 head and leaving the reflexive Voice exponent in the surface
A somewhat similar proposal about the ReciprocalP has been advanced in Ndaiyragije
(2006) about Kirundi, a Bantu language. He argued that the ReciprocalP attached in low
and high applicative, yielded a Recipient meaning for the applied argument in these
We assume that the accusative argument does not need to move to the specifier of vP position in order to
check its case. This happens via Agree of its features with the head of vP.
structures. However, in Georgian, the morphological evidence does not allow us to argue
that the Appl0 head composes with the Reflexive like in the following Kirundi structure
in (64b):
(55) a. Kirundi applicative
bagabo ba-a- rungik-ir-ye
Those men
3p-PST-send-APPLIC-Asp children
‘Those men sent children money.’
b. Reciprocalized applicative
Abo bagabo
ba-a- rungik-ir-an-ye
Those men
Those men sent money to each other /people arb. (Ndaiyragije 2006: 288)
c. The structure for (55b)
As shown in these sentences, reciprocalization of an applicative structure results in
the complex morphological form which, along with the applicative marker, realizes the
Reciprocal0 head: the applied argument is no longer projected due to the reciprocal
marker –an. Thus, PRO is generated in the RecipP as the applied argument. This proposal
does not argue for high/low attachment of the Reciprocal head though.
The following section illustrates other types of low applicatives where the Appl0 head
has the interpretation ‘AT’ of stative possession.
3.6 Possessor Datives as low applicatives
3.6.0 Introduction
Cuervo (2003) considers possessor dative arguments embedded under the stative
predicates (admirar ‘admire, envidiar ‘envy’, conocer ‘know’) or non-directional activity
verbs such as look ‘mira’r, ‘sostoner’ ‘hold’ as low possessor applicatives. In these
structures, the applied argument is licensed by a low applicative head and this argument
is related to the theme as a ‘stative’ possessor. As dative arguments in DOCs, these dative
arguments bear no direct relation to the event. The analysis of such low applicatives
presupposes the merger of a distinct type of Appl0 head that establishes a static relation
between the applied argument and the theme, but crucially not dynamic transfer-ofpossession. The syntactic structure of applicatives formed with such events results from
the merger of low applicative head with the Root that is embedded under the dynamic
event-introducing head vDO (wash, kiss), static event introducer vBE, or dynamic
unaccusative vGO head. With these structures we find the static possession between the
dative argument and the theme mediated through the configuration of the dative argument
as the specifier of ApplP and the theme as the complement of the Root. The following
structure illustrating these relations in possessor datives is adapted from Cuervo (2003):
(56) Structure for possessor datives
a. Voice P
b.[VoiceP DPsubj [ Voice [vP
(Cuervo 2003)
v [VP root [ApplP DPapplied [ Appl0 DPobj]]]]]]
The goal of this section is to present empirical evidence for this hypothesis. First,
stative psychological verbs with the event head vBE are discussed, and then those of nondirectional activity verbs (wash, kiss, hold, etc.)
3.6.1 Low applicatives of stative eventualities
Examining some stative verbs in Georgian such as love, know, etc, it appears that some
verbs fail to project dative applied arguments, even though genitive possessor arguments
are felicitous with those structures:
(57) Genitive possessors with stative events
a. dato-s
shurs ketevan-is
envies K-Gen
‘Dato envies Ketevan’s patience.’
b. datom icis
knows N-Gen
‘Dato knows Nino’s whereabouts.’
shurs ketevan-s
envies K-Dat
‘Dato envies Ketevan’s patience.’
d. deda-m
eliso- s k’aba
da- uc’una.
Mother-Erg E-Dat dress
prev- APPLIC- disapprove
‘Mother disapproved of Eliso’s dress.’
e. nino-m Salome-s p’erangi
mo- uc’ona
N-Erg S-Dat
prev- APPLIC- liked
‘Nino liked Salome’s shirt.’
The stative event shown in (57b) does not project the dative argument while the
psychological verbs in (57a, d & e) do. The following non-directional activity verbs also
project dative arguments denoting the possessor relation with their themes:
(58) Non-directional activities and statives
a. ekim-ma
doctor-Erg patien-Dat wound
‘A doctor cured a patient’s wound.’
b. ekim-ma
doctor-Erg patient-Dat
‘A doctor healed a patient’s wound.’
c. nana-m
‘Nana held Eka’s bag.’
mo- urchina 69.
prev- APPLIC- cure
APPLIC- healed
da- uchira
prev- APPLIC- hold
Note that in (58), both psychological and non-directional activity verbs form low
applicatives where the relation between the applied dative argument and the theme is one
of inalienable or alienable possession. The morphological realization of the Appl0 head is
i-/u- , which is sensitive to the phi-features of the dative possessor argument.
The relevant question might be why some psych verbs can form low applicative
relation while others cannot. Recall from the chapter two of this dissertation, that in
Georgian and related languages, three types of psych verbs were identified in terms of
Note that some of these verbs take the preverbal morphemes to mark perfective aspect, while some as in
(68b) does not. This property can be correlated with the lexical semantics of these verbs and since it is not
relevant for present purposes I will not pursue its analysis here.
their event semantics. They were called activities, state, and dynamic passives. Their
morphological idiosyncrasy correlated nicely with the syntactic and semantic properties
of experiencers in these contexts. The experiencers of activities were interpreted as
voluntary subjects while involuntary subjects appear in state and dynamic passives. Here
is an example of the verb ‘like’, which can express only two types of semantics:
(59) a. State
mo- m- c’ons
prev- 1S- like
‘I like X.’
b. Activities
v- ic’on- eb
1S- REFL- like- TH
‘I am liking X.’
Notice that this predicate does not form a dynamic passive. The prefix i- in activities
presumably marks the Refl0 head because it is not sensitive to the phi-features of both
arguments. The i-/u- allomorphs show up in the applicative structure when the dative
applied argument (in these structures, the possessor) is introduced by the Appl0 head.
Thus, the psychological verbs forming the low applicative relation in (57d & e) express
an activity rather than a state. The activity psychological verbs like, approve, etc. in their
active meaning can express a low applicative relation between the dative applied
argument and the theme. Crucially, this relation is that of static possession denoting AT,
but not Recipient (to) and Source (from). The stative counterparts of like and approve
cannot add the applied dative argument as shown in the following:
(60) State psychological predicates
a. me
‘I like Ketevan’s patience.’
b.* me
motmineba momc’ons.
I-Dat k-Dat
These sentences illustrate that state psychological predicates resist applicativization.
Low possessor applicatives (AT) of Spanish have corresponding Georgian counterparts
expressed by the psychological verbs of activities, which involve the active event head
vDO rather than the vBE head:
(61)a. v-
1S- APPLIC- approve- TH
Lit: ‘I approve it for somebody.’ 70
I argue that the reason why state verbs cannot add the applied arguments is that their
experiencer subjects are assigned an inherent dative case in the specifier of the ApplP.
Since the structural position of the dative experiencer argument is the spec of ApplP,
there is no place for the applied argument to be projected. This is a tentative conclusion
based on the structural positions of dative experiencers.
Note that this translation in English does not represent an applicative relation since the Benefactee
argument is marked with the preposition.
3.6.2 Other activity verbs in applicative constructions
Non-directional activity verbs such as cure, heal, and hold can also add dative
arguments and express a stative low applicative relation between the applied dative
argument and the theme. These dative arguments are added via an Appl0 head. I analyze
the structure of applicatives formed with these verbs below and argue that it includes the
vDO of activity event type above the ApplP, which relates the dative argument to the
theme. The meaning of this relation is interpreted as one of stative possession (AT) much
like in the activity psychological verbs shown in the preceding section. Crucially, these
non-directional activities also owe their semantics to vDO . The Roots like hold, heal, cure
can be embedded under this event head and receive activity reading. The subjects are
interpreted as voluntary agents and are assigned nom/erg/dative cases across series. The
dative applied argument can be added through the low Appl0 head and this argument
bears stative possession (AT) to the theme. The following derivation illustrates the
mentioned semantic and syntactic properties of low applicatives:
(62) The low applicative of ‘hold’
a. me
dato-s chanta
I-Erg. D-Dat. bag-Acc
‘I held Dato a bag.’
da- v- uch’ire
Prev- 1S- APPLIC- hold
‘me’ I
chir ‘hold’
‘chanta’ bag
I assume that the case checking relations in these structures is the same as in DOCs
where the Recipient and Source applied arguments where projected. However, in the
following example observe that the same verb hold, denoting a stative eventuality, cannot
add a dative applied argument and form the low applicative because the dative argument
is projected as a subject of such verb:
(63) Stative eventuality of hold impossible in an applicative construction
a.* me
‘I am holding Ketevan’s bag.’
b. me
I-Dat K-Gen
‘I am holding Ketevan’s bag.’
m- ich’iravs
1S- APPLIC- hold
m- ich’iravs
1S- APPLIC- hold
As seen in (63b), only the genitive possessor structure is available for the stative
interpretation of hold. However, the morpheme inserted into the pre-base position in
(63b) is sensitive to the phi-features of the dative subject and this prompts me to suggest
that the structure will be interpreted as a high applicative below.
Georgian fits the class of languages, which selects for a low applicative head and can
express Recipient, Source, and stative Possessor relations between the applied dative
argument and the theme. However, in contrast to Spanish, the low applicatives are not
available for stative eventualities, and even Roots, which typically show up in stative
contexts in other languages, in Georgian appear to form applicatives only with activity
semantics. Moreover, it has been shown that applicative relations can exist not only in
DOCs or psychological predicates, but in a wider variety of syntactic contexts, such as
inchoative, NI and AI environments, and reflexive Voice contexts. The morphological
realization of the Appl0 head is uniform across different constructions where Appl0 head
takes different types of vP complements.
3.7 Four-place predicates as hybrid type applicatives
3.7.0 Introduction
Four-place predicates present an interesting case of applicatives combining high and
low applicative relations in the same structure as well as high applicative relation with
path conflation. Some of these predicates project a location/locatum argument which
thematically is interpreted as a location or the stative possessor of the theme. First,
observe empirical facts:
(64) Four-place predicates in Georgian
a. man
‘He wrote a letter for me to Anzor.’
b. man
‘He/she stitched a button to the dress fro me.’
mim- ic’era.
prev- 1O- APPLIC- wrote
mi- m- ik’era
prev-1O-APPLIC- stitched
c. man
you-Dat article-Dat paragraph
‘He/she typed you a paragraph to the article.’
mi- g- ibech’da.
prev- 2O- APPLIC- typed
d. man me
‘He addressed them with the manifesto for me.’
mi- m- imarta.
prev- 1O- APPLIC- addressed
e. man shen
He you-Dat shirt-Dat
‘He glued you the picture to the shirt.’
mi- gic’eba.
prev- 2O- APPLIC- glued
f. man shen
‘He painted the picture for you on the wall.’
g. man
He/she you-Dat jacket-Dat price
Lit: ‘He lowered you the price of the jacket.’
mi- g- ixat’a.
prev- 2O- APPLIC- painted
da- gik’lo.
prev- 2O- APPLIC- dicrease
Notice that the two dative arguments in these structures have different meanings: the
higher one being interpreted as a Benefactee of an event performed by the Agent, while
the lower dative argument is interpreted either as a location (64b, c, e & f) or the
Recipient/Possessor of the theme (64a, d & g) 71. It is of interest to note that in Georgian,
the differential thematic interpretations of these lower dative arguments is associated with
the animacy of nominals projected in this low position. In (65), observe that animate
dative arguments alternate with the locatives marked with the postposition –ze ‘on’, while
inanimate datives with the nouns marked with the postposition –tvis ‘for’/’to’:
(65) Postposional marking of lower dative arguments associated with animacy:
a. man
mi- m- ic’era.
prev- 1O- APPLIC- wrote
The lower dative arguments in all these structures freely alternate with the postpositional adjuncts
denoting LOCATIONS or Recipients as shown in (61).
‘He wrote a letter for me to Anzor.’
b. man
‘He/she stitched a button to the dress fro me.’
mi- m- ik’era
prev-1O- APPLIC- stitched
c. man
you-Dat article-Dat.on paragraph
‘He/she typed you a paragraph to the article.’
mi- g- ibech’da.
prev- 2O- APPLIC- typed
d. man me
me-Dat them-Dat.to
‘He addressed them with the manifesto for me.’
e. man shen
He you-Dat shirt-Dat.on
‘He glued you the picture to the shirt.’
f. man shen
‘He painted the picture for you on the wall.’
g. man
He/she you-Dat jacket-Dat.on price
Lit: ‘He lowered you the price of the jacket.’
mi- m- imarta.
prev-1O- APPLIC- addressed
mi- gic’eba.
prev- 2O- APPLIC- glued
mi- g- ixat’a.
prev- 2O- APPLIC- painted
da- g- ik’lo.
prev- 2O- APPLIC- decrease
This difference in postpositional marking shows the correlation between postpositional
case marking with the animacy and thematic interpretations of these lower arguments:
animate arguments are interpreted as RECIPIENTS or Possessors (61g) as opposed to
inanimates, which are interpreted as LOCATIONS. This kind of correlation between the
case marking and semantic interpretations of arguments is not unique. Jelinek (1999) and
Jelinek & Carnie (2003) show that in Yaqui (Hiaki, an Uto-Aztecan language spoken in
Sonora, Mexico, and Arizona), the arguments of ditransitive verbs allow their internal
arguments are marked either with the accusative/dative or the accusative/accusative cases
depending on semantic interpretations of these arguments. Such verbs include ‘give’,
‘teach’, ‘borrow’, and ‘take’. Here is an example including ‘give’:
(66)a. ‘aapo Huan-tau ‘uka
he John-Dat
Det.Acc corn-Acc
‘He gave John the corn.’
b. ‘aapo Huan-ta ‘uka
John-Acc Det.Acc corn-Acc
‘He gave John the corn (as a gift).’
maka-k (Acc/Dat)
(Jelinek & Carnie 2003:273)
Jelinek & Carnie argue that the differential case marking on the goal DP induces
distinct semantics. Namely, the acc/dative pattern is similar to English to-dative
constructions with the location interpretation of dative argument, while the verbs that
require the double-accusative marking must have animate goals and the dative argument
be interpreted as ‘strongly affected’ by the action of the Agent. The “strongly affected”
meaning of John is actually being fed the corn.
A similar parallelism between the case marking in double-object constructions (DOCs)
and semantic interpretations of dative arguments is detected in Korean (Jung &
Miyagawa 2004, henceforth, J & M). J & M base their analysis of Korean ditransitive
structures on Harley’s (2002) proposal of give as a verb with two possible decomposition
scenarios: according to her, as argued in Section 3.2 of this dissertation, give in the DOC
John gave Mary the book has the meaning John CAUSED Mary to HAVE a book and this
interpretation is associated with two functional elements: CAUSE and PHAVE, the latter
being interpreted as HAVE 72. In its prepositional version, John gave a book to Mary the
Pylkkanen (2002) is not sympethatic with the small clause analysis of DOCs (Guéron, 1986, Hoekstra,
1988, Harley, 2000) which treat DOCs as types of causatives taking the Goal predicate as a complement.
PLOC is interpreted as LOCATION 73. The conclusion drawn about ditransitive structures
cross-linguistically is that double-object and to-dative constructions can be decomposed
into CAUSE and PHAVE /PLOC components in different languages such as English and
Korean. Additionally in Korean, like in the Hiaki examples above, there is a correlation
between the case-marking of goal arguments and their semantic interpretation. In DOCs,
Goals are marked with the accusative case creating double-accusative marking for
internal arguments. The accusative applied argument has obligatory specific
interpretation. J & M suggest that this ‘obligatory specificity’ is a corollary of double
accusative marking by CAUSE in these contexts. In locative structures, Goals are marked
with dative and they are not interpreted as specific. Observe the following structures
adapted from J & M (2004):
(67) Double Object Construction (DOC)
a. Mary-ka
John-ul chayk-ul
Mary-Nom J-Acc
‘Mary gave John a book.’
(Jung & Miyagawa 2004:17)
Pylkkanen argues that this type of analysis is good for give but not for other DOCs given contrasts like: #I
broke the vase but it didn’t break, or DOC: I wrote Sue the letter but she never got it, etc.
The analysis of HAVE as a preposition rather than a verb have been discussed by many authors
(Benveniste 1966, Freeze 1992, Gueron 1995, & Kayne 1993). The details of their analysis are irrelevant
(68) To-Locative construction
a. mary-ka
Mary-Nom John-Dat
‘Mary gave a book to John.’
As said above, no specific meaning is available for the Locative argument in (68).
These observations on the argument structure of ditransitives may be of relevance for
analyzing Georgian four-place predicates illustrated in (65).
3.7.1 Analysis of morphosyntax of four-place applicatives
As indicated above, the higher dative arguments are Benefactees. I interpret the relation
between this high dative argument and an event as a high applicative in which no
transfer-of-possession obtains between two arguments. Given these observations, I argue
that in Georgian four-place verbs, animacy is key to the Recipient/possessor versus
Location interpretations of lower dative arguments. Following Harley and J & M, I
propose that in the structures where the lower dative arguments are animates and marked
with the postposition –tan, these dative arguments are interpreted as Recipients much like
accusative arguments in Korean by PHAVE, while inanimate dative arguments
correspondingly marked with –ze postposition can be interpreted as Locations projected
by PLOC.
Also, as mentioned above, the higher dative argument in (65g) is interpreted as the
Benefactee of an event performed by an external argument. I suggest that the lower
dative argument in (65g) thematically can be interpreted as stative possessor of the
theme. I conceptualize the relation between this lower dative and the theme arguments
with the low applicative head introducing the possessor dative argument. The stative
possession relation may be introduced by the event head vBE (of stative possession) as
argued by Cuervo (2003) for Spanish psychological predicates and in Sections 3.6 and
3.8 of this dissertation for certain classes of Georgian verbs. The external argument of
this construction is projected by the Voice head and the case of this argument checked
against this head. Therefore, the structure in (61g) may be decomposed in the following
(69)a. man
He/she you-Dat jacket-Dat price
Lit: ‘He lowered you the price of the jacket.’
da- g- ik’lo.
prev- 2O- APPLIC- decrease
Man ‘he /she’
‘me’ I
i-, u-
kl ‘lower’
This is a hybrid type applicative structure combining two applicative heads: high and
low and two vPs one being a CAUSE and another, the stative event-introducer vBE. It is
the latter v0 head that is responsible for stative possession relation between the lower
dative argument and the theme. Another question to ask is which head is realized with
the i-/ u- affixes. In low applicatives, discussed so far, the low applicative head was
realized with this allomorphy. Now observe the following data, which shows that i-/u- is
sensitive to higher dative argument, i.e. Benefactee:
(70) The paradigm of 4-place predicate ‘write’
a. datom me
D-Erg me-Dat E-Dat
‘Dato wrote me the letter to Ek’a.’
mi- m- ic’era.
prev- 1O- APPLIC- wrote
b. datom shen
D-Erg you-Dat
‘Dato wrote you the letter to Ek’a.’
c. dato-m
‘Dato wrote him/her the letter to Ek’a.’
mi- g- ic’era.
prev- 2O- APPLIC- wrote
mi- uc’era.
prev- APPLIC- wrote
The VI for the Appl0 head is sensitive to the person feature of the higher dative
argument, i.e. the Benefactee projected by the high Appl0 head. Therefore, I assume that
only the high Appl0 head is realized in these constructions. This can be expected given
templatic constraints on the realization of argument structure-changing morphemes in the
Georgian verb template. Therefore, the insertion of the second set of i-/u- affixes is
prevented. The morphological realization of these hybrid type applicatives with two
Appl0 heads is no different from that of the low applicative head in DOCs.
The case marking of the four arguments in these structures is of interest because the
two dative cases checked on the Benefactee and Recipient arguments are structural as
they both change into the postpositional genitive or dative in the Perfective series. I
assume that these dative cases are checked against the applicative heads that introduce
them. The accusative case on the theme may be checked against v0 head the same way as
it was argued for the low Recipient and Source applicatives above.
Now the structures for the rest of four-place predicates will be analyzed. As
mentioned above, in (65a-f) the two dative arguments are projected the higher one
receiving the Benefactee thematic interpretation while the lower one either Recipient or
Location depending on the animacy. Following Harley (2002) and J & M (2004), I claim
that these relations in (66) can be decomposed in two types of PPs, which embed two
postpositional DPs in their specifiers one being interpreted as HAVE with the animate
DP and another as LOCATION with inanimate. Here is the structure showing these
interpretations of lower applied arguments:
‘me’ I
i-, u-
write, glue, etc.
The higher dative argument has high applicative relation with the event and it is
introduced by the Appl0 head merging above VP. The structural dative of this argument is
presumably checked against the Appl0 head. The lower applied argument is in the sister
position of the P head which may be considered as functional rather than lexical and
following various researchers (Hale & Keyser 2002 among them). The oblique
postpositional cases do not require checking and the accusative of the theme may be
checked against the v0 head via Agree because no other argument intervenes between
them to incur the minimality violation.
The morpheme order in these four-pace predicates is subject to linearization rules
which may be language-specific, governed by morphological well-formedness conditions
(as argued in the introduction). As shown in (71), only one high applicative head is
merged in these structures and its morphological exponent bleeds the exponent realizing
the CAUSE due to constraints imposed on the spellout of argument structure-changing
3.8 High Applicatives in Georgian
3.8.0 Introduction
This section on high applicatives focuses on dative arguments that do not relate to the
theme, but rather to an event expressed by the verb. Cross-linguistically, such structures
express three types of semantics (Cuervo 2003), which are distinguished by the type of
the vP complement that the Appl0 head takes: stative vPBE, a dynamic non-agentive vPGO
or a dynamic agentive vPDO. Languages differ as to how many of these heads may be
selected. They may select one, two or all three vP complements.
In 3.8.1 I develop an analysis of applicative constructions whose head selects for the
stative vPBE complement. The psychological state verbs like love, like, hate, etc. are
analyzed as high applicatives due to their morphological shape and the relation of the
dative argument with the event. The section does not focus on the activity psychological
verbs which have been shown above to take vDO and form low applicatives. In 3.8.2,
applicative structures taking vPGO, i.e. non-agentive dynamic events will be analyzed
moxda ‘happen’, chamouvida ‘arrive’, mouvida ‘occur to somebody,’ gamouvida
‘appear’, etc. Some of these structures are argued to show low applicative meaning
(‘arrive for somebody, got/receive something). Section 3.8.3 analyzes those dynamic
agentive events (walk, dance) that are argued to have high applicative meanings across
many languages (Pylkkanen 2002, Jeong 2007 among others). The semantic
interpretation of dative arguments varies in these structures, sometimes showing
benefactive, and sometimes malefactive meanings.
3.8.1. High applicatives with stative unaccusative predicates
In Georgian, the morphological shape of psychological predicates with stative
interpretations includes the applicative morpheme, which is sensitive to the person
feature of the dative Experiencer argument:
(72) State psychological verbs:
a. me
I-Dat flowers-Nom
‘I love flowers.’
m- iq’var- s.
1S- APPLIC- love- 3O
b. shen
q’vavilebi g- iq’var- s.
you-Dat flowers
2S- APPLIC- love- 3O
‘You love flowers.’
c. mas
q’vavilebi uq’var- s.
he/she-Dat flowers
APPLIC- love- 3O
‘He/she loves flowers.’
This paradigm shows that the allomorphy associated with the Appl0 head is sensitive
to the person features of dative experiencer arguments much like with the dative applied
arguments in DOCs. Dative arguments may also be interpreted as locatives or inalienable
possessors as shown in (73):
(73) Experiencer dative argument
a. m- inda
1S- APPLIC- want
‘I want X.’
b. g - inda
2- APPLIC- want
‘You want X.’
c. unda
APPLIC- want
‘He/she wants X.’
(74) Locative dative arguments:
a. me
m- ichans
I-Dat foot-Nom 1S- APPLIC- seen
Lit: ‘I have my foot seen.’ 74
b. shen
g- ichans
you-Dat foot
2S- APPLIC- seen
Lit: ‘You have your foot seen.’
c. mas
uhe/she-Dat foot
APPLICLit: ‘He/she has her foot seen.’
The dative argument in (73) is interpreted as an involuntary experiencer, while in (74)
as an inalienable possessor of the theme, although the English translation does not render
the meaning of this Georgian verb adequately. The interpretation is that the part of the
body or the property of the dative argument ‘is seen’ and the applied dative argument is
interpreted as a possessor of the property or the body part denoted by the theme. Observe
the metaphorical possessor meaning of this relation:
(75) ‘Seen’ locative
not APPLIC- have/seen
‘Nino’s patience does not have a boundary.’
(Lit: The boundary of Nino’s patience is not/cannot be seen.’
These free translations are awkward in English but this is only interpretations of these Georgian
In both (73) and (74), the dative DP is assumed to be introduced by the Appl0 head, but
there is a clear sense that the relation between the dative argument with the theme in love
and want is different from that in seen. In the former, the Experiencer argument does not
bear either Recipient, or Source or Possessor meaning, while in ‘seen’ the dative
argument is a location whose integral part is the theme. Thus, it can be argued that the
structures in (73)-(74) encode a high applicative relation between the dative argument
and an event while the applicative in (74) is interpreted as low because the relation of the
dative argument with the theme is one of part/whole or stative possession. I follow
Cuervo (2003) in these interpretations of the high vs. low distinction and propose that
locative/possessor datives can be analyzed as low applicatives where the head is the
complement of a stative vPBE. By contrast, the experiencer dative DPs are projected by a
high Appl0 head that takes a stative vPBE (i.e. vBE) as its complement. Thus, the structures
of high applicatives can be sketched the following way:
(76)a. k’ato-s uqvars/unda
APPLIC- loves/wants
‘k’ato loves/wants the ice-cream.’
-nda, q’var
want, love, etc.
In this configuration, the theme DP is in the specifier of the event head and is not
related to the higher dative argument through the Appl0 head 75. In the low applicative
with the stative predicate the configuration is different since the dative experiencer and
the theme are projected as the specifier and the complement of the Appl0 head. This
configuration ensures that the semantics of part/whole or possession between the theme
and the dative argument is syntactically established. Here is the structure:
(77) Low applicative of Locative
a. nino-s uchans
APPLIC- seen
‘Lit: ‘Nino’s foot is seen/appears.’
chans ‘seen’/’appear’
pexi ‘foot’
These structures show the necessity of careful elaboration of the diagnostic tests for
high and low applicatives when seeing such subtle distinctions in meaning in state
psychological and locative predicates. Structurally, dative arguments are very similar to
each other both in high and low positions because they are the highest arguments of the
clause and move to the subject position. The morphological properties of high
Note the position of Nom theme in this structures. I assume that this case cannot be assigned in headsister relation with V0. Therefore, the position of the Nom object can be in the spec of VP.
applicatives are no different from low ones because the head is also realized with the
allomorphy i-/u-. The evidence then for the different positions is purely semantic.
3.8.2 High applicatives of other unaccusative predicates
This section presents a range of high applicative constructions where the Appl0 head
takes a vGO complement (i.e. unaccusative verbs with change-of-state meaning or the
verbs of happening). In these structures, no direct relation exists between the dative
argument and the theme, but between the dative argument and an event such relation does
(78) High applicative above unaccusative verb
a. mak’a-s
‘Something bad happened to Mak’a.’
mo- uxda.
prev- APPLIC- happened
b. elene-s
gadmo- uvida
prev- APPLIC- flow.over
Lit: ‘Water overflew from the casserole on Elene.’
The dative arguments in these sentences are not in any way related to the themes:
Mak’a to something bad and Elene to water. Both applicatives have a malefactive
reading. The allomorphy i-/u- is again sensitive to the person feature of the dative
argument. High applicative meaning is encoded in the ApplP taking a dynamic vGO as its
complement and this syntax is associated with the relation between the dative argument
(subject) and an event adversely affecting this argument. In (78b), the Root of the motion
verb –vid ‘go’ expresses the movement of an inanimate entity from the designated
location. The dative argument does not bear the possession relation with the water in this
(79) High applicative of unaccusative verb:
a. mak’a-s
‘Something bad happened to Mak’a.’
mo- uxda.
prev- APPLIC- happened
mak’a vPGO
 xda, etc.
Something ‘ambavi’
The crucial aspect that distinguishes unaccusative verbs forming high applicatives
from those unaccusatives that form low ones is whether the dative argument gets/receives
the theme directly or metaphorically. If dative argument ‘receives’ the theme, then it
would form low applicative relation though. Compare:
(80) Applicatives with unaccusative verbs
a. nino-s mama
chamo- uvida
prev- APPLIC- arrived
‘Nino’s father arrived for Nino.’
b. tekle-s saxe-ze
‘Tekle got pimples on her face.’
gamo- uvida.
prev- APPLIC- came.out
Notice that ‘arrive’ may express a low applicative meaning by relating the two
individuals (Nino and the father) to each other as the recipient and theme. Crucially, both
arguments are projected under the vP so that the low Recipient relation between the
nominative theme (father) and the dative Recipient (Nino) argument. As I have argued
above in Sections 3.3. & 3.4, although the position of nominative theme is low, it checks
the case against T0 without incurring the minimality violations. The dative argument is
introduced by the Appl0 head checking its case as well:
(81) The structure for (80a)
vida ‘arrive’
i-/uSimilarly, the structure in (80b) is analyzed as a low applicative with the same vGO
taken as a complement in ‘overflow’, and in this structure, the Appl0 head relates the
dative and theme arguments as a part/whole.
Note that these distinctions in applicative meaning are not associated with a change
in morphology, namely in the marking of the high vs. low applicative heads. The general
picture of applicative morphology in Georgian and related languages is not complex, but
it still shows mismatches between the syntax-semantics and the morphological
components of the language. Both high and low applicative semantics can be obtained via
the same affixation or even with the zero applicative marker in elsewhere contexts. The
mismatch between these components of language can be due to many factors. One such
factor is language-specific constraints on realization of argument-structure changing
morphemes in the pre-base and post-base positions of the verbal template.
3.8.3 High applicatives with dynamic activity verbs
One of the tests to distinguish between high and low applicatives in Pylkkanen (2002,
2008) is whether the Appl0 head can combine with dynamic unergative verbs like walk,
dance, etc. The following sentence with the high applicative head projecting dative
benefactee argument illustrates that Luganda selects for the high applicative head:
(82) High applicative in Luganda:
ya- tambu- ledde
‘Mukasa walked for Katonga.’
The same diagnostics can be applied to Georgian dynamic action verbs such as dance,
ran, etc. and it appears that it is possible to form high applicative structures with dative
arguments as Benefactees. However, it is important to note that although these structures
contain canonically intransitive verbs, in the applied structures with these verbs, there is
in fact an accusative Theme present:
(83) High applicatives in Georgian
a. nik’o-m natia-s
‘Nik’o ran marathon for Natia.’
b. lado-m
mak’a-s mteli
‘Lado danced all night long for Maka.’
c. dato-m nadia-s
‘Dato rolled Nadia the ball.’
ga- urbina.
prev- APPLIC- ran
APPLIC- danced
ugor - av- a.
APPLIC- rolled- TH- 3O
In these constructions, the dative arguments can be interpreted as benefactees of the
events performed by the external arguments. More precisely, it is in the dative
argument’s interests that the action is completed by the external argument. Such
situations may well exist in the world but note the presence of theme arguments in some
of the above structures. Their presence is necessary in order for the structures in (83a &
c) to be well-formed. I would rather argue that in no sense the dative arguments ‘receive’
or ‘possess’ these themes. In Spanish, the constructions equivalent to (83a & b) are
considered low applicatives since the beneficiary is not directly related to the event but to
the theme. I depart from an analysis of these structures as low because the relation
between the dative argument and the theme does not fall into any of the low applicative
meanings such as recipient, source, or possessor. In (83a), Natia is a beneficiary of the
event performed by the external argument. Nik’o but crucially she does not relate to
‘marathon’ as a possessor, source, or as a recipient. The same can be said of Mak’a in
(83b). Even when the theme argument is restored in the structure—and it could be a
certain kind of dance such as polka, tango, or waltz—the relation between Mak’a and the
theme cannot be expressed. (83c) is a case in which the dative argument is interpreted as
the Benefactee. Crucially, Nadia is not a Recipient of the ball because the event is not
directed towards her in any sense, rather only for her benefit. Thus, again the relation is
between the external argument and the event.
Another suggestion for these constructions is that the change-of-state verbs whose
roots often form unergative structures (dance, scream, roll, etc.) may be used transitively
by adding themes and such transitives may involve the merger of the high applicative
head which relates the external argument to an event rather than two individuals—the
dative benefactee with the theme. If this analysis of above structures is tenable, I will
assume that high applicatives can be expressed by transitive verbs of change-of –state.
For the purposes of comparison with Spanish high applicatives, it should be noted that
dative arguments in this language are not overtly expressed in certain high applicative
structures and only the clitic indicates their presence. Cuervo uses the terms ‘datives of
interest’ or ‘ethical datives’ to refer to such implicit DPs in high applicatives. Analyzing
dative clitics as the spellout of the high applicative head, she argues that this head in
Spanish is in a way ‘defective’ because it does not project a specifier. In Georgian, there
is no evidence that the high applicative head is defective, since it can project the applied
Benefactee argument. This argument is related to the dynamic event itself, rather than to
the theme. The following sentences provide additional evidence for the high applicative
head combining with the vPDO as an introducer of a dynamic event of change:
(84) High applicatives
a. gia-m
G-Erg N-Dat
‘Gia played out the cards for Nana.’
ga- utamasha.
prev- APPLIC- played
gada- uch’ira
demonstrator-Dat belt
prev- APPLIC- flapped.
‘The policemen flapped the demostrator/protestor with the belt.’
In (84a), the applied argument Nana is in no way related to the theme argument,
although the cards are played by Gia to her benefit. The dative argument (Nana) does not
end up as a possessor or a ‘Benefactee’ of card play. Thus, the applicative relation is
interpreted as high. (84b) is a case of interest because the theme argument is an
instrument of the event completed by the external argument and it is in no way related to
the applied dative argument. These interpretations clearly fit the descriptions of high
(85) The high applicative of dynamic unergative verb in (83b):
kamari ‘belt’
flap, etc.
In the above examples, the high Appl0 head is realized with the same allomorphy
i-/u- as the low applicative. The vast majority of applicative meanings can be realized
with the i-/u- and the variability in marking is only detected in the psychological verbs
(love, hate, like, etc.) where the high applicative head can be realized with the i-/u- and
zero exponent.
3.9 Applicatives in related languages (Mingrelian and Svan)
3.9.0 Introduction
This section explores the morphosyntax of applicative constructions in Mingrelian and
Svan in sections 3.9.1 and 3.9.2 based on the existing literature on these languages. The
primary sources that I will refer to are Otar Kajaia’s (2001) “Mingrelian-Georgian
Dictionary” with a grammar synopsis of Mingrelian and Varlam Topuria’s (1967)
“Grammar of the Svan Language” written in a traditional grammar framework. The main
research questions with respect to applicative constructions in these languages are
primarily descriptive, i.e. whether they can express the same range of applicative
meanings as we have seen in Georgian and how the syntax-semantics of these structures
interacts with the morphology. I will be assuming the DM framework to account for
syntax-semantics and morphology mismatches in these languages.
3.9.1 Applicatives in Mingrelian
Kajaia’s grammatical note on Mingrelian outlines three types of applicative relations,
which he refers to as version, the term used in traditional Kartvelian Linguistics for the
applicative relation. One type of version, called “neutral”, corresponds to regular
transitive and DOCs that express no benefactor or possessor relation between the applied
argument and the theme, as well as the external argument and the theme. The second
type, termed “subjective”, in Kajaia’s system corresponds to reflexive applicatives. In
these, the theme object is in the prospective possession relation with the external
argument. The third constructions, “objective version” corresponds to high or low
applicatives in which either a dynamic transfer-of-possession or source relation obtains
between two individuals, or the external argument simply relates to the event without
specific possessor relation between it and the theme.
Both high and low applicatives of transitive verbs in Mingrelian can be expressed by
the same allomorphy i-/u- as in Georgian:
(86) Mingrelian applicatives
a. m- ibons
1O- APPLIC- wash
‘X is washing Y for me.’
b. g- ibons
2O- APPLIC- wash
‘X is washing Y for you.’
c. ubons
APPLIC- wash
‘He is washing X for him/her.’
(Kajaia, 2001: 57)
In these examples the i-/u- allomorphy is sensitive to the person feature of the
applied argument the same way as in Georgian. I assume that the VIs competing for the
insertion in the Appl0 head are the following:
(87) Vocabulary Items for the Appl0 head
a. i-  Appl0 / [dat] [+1] [+2] ____
b. u-  Appl0 / [dat] [+3] ____
c. -  Appl0 / elsewhere
Observe that the low applicative relation in DOCs has the same morphological
marking as in Georgian:
(88) Mingrelian applicatives
a. m- itasuns
1O- APPLIC- sow
‘X is sowing for me.’
‘X is sowing for him/her.’
b. m- ich’ans
1O- APPLIC- sew
‘X is sews for me.’
‘He/she sews for him/her.’
The person feature of the applied argument in low and high applicatives is crucial for
vocabulary insertion.
In applicatives formed from passivization of DOCs, both low and high applicative
meaning is expressed via the allomorphy i-/u- along with the passive suffix –d. In these
constructions, the movement of the theme argument to the subject position of passive is
impossible while that of the applied argument is acceptable:
(89) Mingrelian passive applicatives
a. mishkuron- du (n)
1S- APPLIC- builtpass- 3O
‘It is built for me’ (Lit: ‘I am built it.’)
b. ushkuron- du (n)
APPLIC- builtpass3O
‘It is built for him/her.’ (Lit: ‘He/she is built it.’)
In these passive forms, the low transfer-of-possession relation between the applied
argument and the theme does not change from the corresponding active structure. I
suggest that the movement of the applied argument to the sentence-initial position is
unobstructed because, as was noted above, the applied argument is projected higher than
the theme, and asymmetrically c-commands the latter and the movement of the dative
argument to sentence-initial position does not incur any minimality violation. In addition,
the morphological marking reflects the changes in the syntactic structure as the suffix –d
is inserted into the Voice0 head, which bears the feature [Passive].
Recall that in Georgian the following applicatives formed from inchoative and
unaccusative verbs are common:
(90) Georgian applicatives of unaccusatives and inchoative verbs
a. m- itbeba
1S- APPLIC- warm
‘It is warming up for me.’ (Lit: ‘I am being warmed up X.’)
b. m- it’k’beba
1S- APPLIC- sweet
‘X is sweetened for me.’ (Lit: ‘I am being sweetened X.’)
The morphological marking of these applicatives is also sensitive to the person feature
of the dative applied argument, which is the subject of the construction. These structures
resemble reflexive applicatives, but are different from the latter in that they are formed
from internally- or externally-caused change-of-state verbs. Therefore, no external
argument is projected in these structures like in applicatives of internally-caused events
discussed in Section 3.3. Mingrelian allows similar applicative structures as illustrated in
the following:
(91) Mingrelian applicatives of change-of-state verbs
a. m- it’ibu (n)
1S- APPLIC- warm.up
‘It is warming up for me’
b. g- itibu(n)
2S- APPLIC- warm.up
‘It is warming up for you.’
The subject of these predicates is not an external argument in a crucial sense and it
must be marked with the dative case because it triggers the inverse agreement pattern
(‘m-set’) on verbs. I claim that the structures in (91) have a low applicative.
The passivized applicatives in Mingrelian do not presumably involve a full cycle, as
shown in the following high and low applicative structures:
(92) Low applicative in Mingrelian:
a. m- ishkurondu (n)
1s- APPLIC- built
‘I am built it.’
In this configuration of the argument structure, the semantic relation between the
applied dative argument and the theme is that of transfer-of-possession. The dative on the
applied argument is checked against the Appl head0. The accusative case of the theme is
presumably checked against vDO head.
Interestingly, the low applicatives of internally-changed verbs in Mingrelian are
marked with i-/u- allomorphy as in (93):
(93) Low applicative of internally-changed verbs in Mingrelian:
a. u-t’ibu (n)
‘It is warmed up for X.’(Lit: X is warmed up Y.’)
In both structures, the direct object undergoes the change-of-state and the applied
argument ‘benefits’ from this event. The applicative relation is low because the dative
argument is a Recipient of the theme, at least in a metaphorical sense.
There is evidence that Mingrelian also has e-prefixed low applicatives parallel to those
formed from internally-changed verbs in Georgian in Section 3.3. Recall that the
following forms were found in Georgian:
(94) Applicatives of internally changed verbs in Georgian
a. m- ezrdeba
1S- VOICE- grow
‘It is growing for me.’
b. m- eteseba
1S- VOICE- sow
‘It is sowed for me.’
In Mingrelian, a similar meaning is expressed via the prefix a-, which is inserted in the
pre-base position where various argument-structure changing morphemes are inserted.
Observe the following forms:
(95) Mingrelian applicatives of inchoative and transitive verbs
a. ardu-u (n)
VOICE - grow‘It is grown for him/her.’
b. atasu-u (n)
VOICE- sow
‘It is sown for him/her.’
c. achar-u (n)
VOICE- write
‘It is written for him/her.’
d. axacku-n
‘It is tilled for him/her.’
The prefix a- is inserted into the Voice0 rather than in the Appl0 head. As argued in
Section 3.3 for Georgian applicatives, the insertion of the VI in the Voice0 is due to the
impoverishment of the person features from the Appl0 head and the result is the
realization of the Voice0. Since the structures in (95b & c) resemble passivized
applicatives of transitive verbs, I assume that the external argument is suppressed and the
applied argument moves to the subject position.
The morphological marking of the Appl0 head in Mingrelian is in many respects similar
to that in Georgian:
(96) VIs for the Appl0 head (Mingrelian)
a. i-  Applic0 LOW / [Dat] [+2] [+1]
b. u-  Applic0LOW / [Dat] [+3]
c. -  Applic0 / elsewhere
Like in other languages in the family, only one morpheme can be spelled out for the
Appl0 and Voice0 heads.
The last type of applicative construction is derived from stative and unergative verbs,
and has high applicative meaning, since the applicative relation is obtained between an
applied argument and an event. First, observe the empirical base in the following:
(97) High applicatives formed from stative verbs in Mingrelian
a. udʒu(n)
‘He is lying for X.’ (Lit: X is laid Y.’)
b. udgu(n)
APPLIC- stand
‘It is standing for X.’ (Lit: X is stood Y.’)
c. uxe(n)
‘It is sitting for X.’ (Lit: X is sit Y.’)
These examples show that presumably, stative eventualities can also express an
applicative relation between the subject and event expressed through the theme
undergoing the state of lying, standing, sitting, etc. The Mingrelian forms in (97) have
parallel Georgian structures shown in the following:
(98) Georgian applicatives formed from statives:
a. uzis
‘It is standing for X.’ (Lit: X is stood Y.’)
b. udgas
APPLIC- stand
‘It is standing for X.’ (Lit: X is stood Y.’)
c. uc’evs
‘He is lying for X.’ (Lit: X is laid Y.’)
Note that both in Mingrelian and Georgian, the VI is inserted into the Appl0 head is
again sensitive to the person features of the dative argument, which is the subject in these
Unfortunately, there is little evidence for high applicatives formed from unergative
verbs such as dance, run, etc. in Mingrelian. The comprehensive study of applicatives in
this language should involve native speakers and detailed questionnaire of fieldwork
study designed to elicit a whole range of applicative relations found in Georgian in this
3.9.2 Applicatives in Svan
Svan applicatives are similar to Georgian and Mingrelian in that they can express both
the low and the high applicative relation and are morphologically realized via the i-/oallomorphy (corresponding to Georgian and Mingrelian i-/u-). The reflexive marker is the
same prefix i-, which is not sensitive to the person feature of the applied argument
(Topuria 1967). It may be assumed that this s- is a Voice marker inserted into the Voice
head due to the impoverishment of person features on the Appl0 head or the mentioned
discontinuous feeding mechanism between the VIs of the Voice and the Appl0 heads. An
alternative scenario for analyzing these argument structure-changing morphemes could
be that the fusion rule, which applies before the Vocabulary Insertion, combines the
morphemes of the Voice0 and Appl0 heads and the reflexive i- is inserted into this fused
head instead of the VIs realizing the Voice0 and Applicative heads beating out other
possible morphemes in a single competition. We will not pursue this scenario here but
this explanation may well be a case along with the impoverishment analysis proposed
Topuria also posits another type of applicative relation referred to as sazedao (literally
meaning ‘on the top’). This type of applicative (or “version”) indirectly expresses a
transfer-of-possession relation between the applied argument and the theme.
Morphologically realized as a prefix a-, these Recipient applicatives encode a close
spatial relation between two arguments without involving the postposional marking of the
applied argument, which is otherwise common on locative arguments:
(99) High Svan (a dialect) low applicatives
a. eǰa
‘He/she lends it to her/him.’
b. eǰa
x- ä-bdine
‘He /she kindles fire for him/her.’
Bal-Zemouri dialect of Svan
c. bepshû
‘A child feeds food to the hungry.’
Notice that in (99a) a transfer-of –possesssion relation obtains between the dative
argument and the theme. (99b) can be interpreted as a high applicative predicate because
the subject performs the event that the applied argument benefits from. However, the
dative argument is interpreted as a Recipient of the theme again and the structure can be
interpreted as a low applicative. In Georgian, the Appl0 head is realized via the i-/u-
allomorphy, while in Svan the VI is a- , which in Georgian has been argued to represent
the marker for CAUSE in various contexts.
The Georgian translation of (99c) looks like a causative predicate because the
causative marker a- is inserted into the third slot of the verbal template. Therefore, I
assume that the corresponding form in Svan is also causative marked with the prefix a- of
causative head. Many languages combine the causative heads with the event-introducing
vPs and ApplP to express the complex relations between the arguments in DOCs. Such
structures have not been discussed in this dissertation, but they are found crosslinguistically and this may be such an example. The meaning can be interpreted as ‘X
causes Y to become satisfied as the result of eating.’ Svan predicates in (99) can be
interpreted as low applicatives but their morphological realization with the default prefix
a- suggests that the prefix a- realizes CAUSE, rather than the Appl0 head, which I assume
is due to the discontinuous feeding between the VIs of these functional heads.
Now let’s look at the examples of i-/o- allomorphy of the Appl0 head in Svan:
(100) Svan low applicatives
He/she 3O- APPLIC- builds
‘He/she builds house for him/her.’
The prefix x- in the above example indexes the third person argument and o- realizes
the Appl0 head. The prefix x- indexes the applied argument because it is closer to the v0
High applicatives in Svan have the same morphological marking as low applicatives,
like in other Kartvelian languages. This is illustrated in the following:
(101) Svan high applicatives
a. x-o-xt’aûi
‘He is painting for X.’
b. x-o-ben
‘He is tying X for him/her.’
c. x-obne
‘He is beginning Y for X.’
Observe that all these forms in (101) contain the applied arguments in a direct or
indirect possessor relation to the theme and therefore, a low applicative interpretation is
obtained, i.e. when the subject and the event are related to each other but not two
individuals per se.
The last type of applicative relation that I will consider here is the one expressed by
various intransitive verbs. They may have both low and high semantics, much like eprefixed applicatives in Georgian derived from internally-changed verbs. In Svan, the
same prefix e- that shows up in internally-changed verbs and in passive forms where the
subject is suppressed and the applied argument moves to its position:
(102) Svan e– prefixed low applicatives
a. m- ext’aûi
1S- voice- paint
‘I am painted X.’
b. m- eben/mi
1S- voice- tie
‘I was tied for me.’
The applicatives in (102) contain implicit subjects and are derived through
passivization of active applicative structures of di-transitive verbs. However, the
following forms can be argued to express a high applicative relation between the subject
and a stative event:
(103) High applicatives of stative events in Svan
a. m-e- jshx- m- i
‘X is named for me.’ (lit: ‘I am named X.’)
b. x-e-g m- i
‘Something is standing for me.’ (lit: ‘I am stood X.’)
The following applicatives though are formed from internally-changed verbs and may
be interpreted as low applicatives:
(104) Svan low applicatives
a. ǰ- egûshi
2S- voice- grow
‘(Wings) grow on you.’ (Lit: ‘I am grown wings’)
b. m-e-dāʒsh-i-x
Lit: ‘I am blackened.’
The forms in (104) show the resultant state of some internal event that subjects
undergo. The structures are interpreted as reflexives, and I assume that the prefix e- is
supposedly realizing the Voice0 in these forms. There are a number of verbs that may be
interpreted as adversity causatives and show close resemblance to e-prefixed high
(105) High Svan (a dialect) applicatives
a. x- epxʒeni
3S- voice- scatter
‘Something was scattered on X.’
b. x- ogbeni
3S- APPLIC- dirty
‘He was dirtied.’
c. x- ekûceni
3S- voice- cut
‘He is cut.’
d. m- ekûreni
3S- voice- decompose
‘It is decomposed on him/her.’
e. x- otûpeni
3S- APPLIC- lose
‘It is lost on him/her.’
The sentences in (105) are interpreted either as high or low applicatives, where the
subjects are deep structure applied arguments and I assume that they are introduced by
the Appl0 head giving them Source, Recipient or Benefactee interpretation. The latter
arguments may have benefactive/malefactive interpretations.
3.10 Conclusions
In conclusion, it can be argued that applicatives in the Kartvelian languages can be
both low and high, and their morphosyntax is very similar to each other in a way that
suggests the unity of functional projections in the Proto-Kartvelian. The applicative heads
of both low and high varieties show the same morphological realization in all three
languages, much like causatives. There are instances of transparadigmatic syncretisms
across voice, applicative, and causative morphemes, which are not explored here in detail
but will be considered in future research that can be conducted on voice syncretisms in
these languages. Here it can be definitively argued that applicatives derived from
intransitive verbs are morphologically distinct than those derived from simple transitive
verbs. High/low semantics is not distinguished in the morphological component as is
expected from the disjunction between grammar components where syntactic, semantic,
and morphological information is processed.
Causative and applicative structures in polysynthetic languages represent both
syntactically and morphologically complex structures in which the syntax/semantics and
the morphological realization of the functional heads may not be always straightforward.
The study has attempted to show that the interaction between these components of
grammar in lexical and syntactic causatives is largely determined by the variation in the
complement size that the CAUSE takes in these constructions. Empirical evidence
presented in the study has shown that morphological realization of the causative head is
determined by whether the CAUSE takes RootP or vP complement. Given this
observation, we conclude that the realization of the syntax at Morphological Structure
(MS) is not entirely random but rather quite predictable in three languages: Georgian,
Mingrelian and Svan.
The analysis of applicative structures also showed similar evidence with respect to the
interaction between the argument structure and the morphological realization of
applicative heads. It became evident that various Vocabulary Items (VIs) are sensitive to
the morphosyntactic features of applied arguments, i.e. those arguments, which are
definitive for this type of construction. As argued in the literature on applicatives, the
applied argument is the one which is added to the argument structure of transitive verbs
to form a Double Object Construction (DOC) in which applicative relations can be
established between this non-core applied argument and an event or the theme argument.
Again empirical data from Georgian and related languages illustrated the fact that the
addition of an applied argument to a transitive structure results in distinct morphological
realization of predicates as opposed to regular transitive structures. Specifically, the latter
realize CAUSE in the third slot of the verbal template while applicatives in DOCs realize
the Appl0 head instead of the CAUSE. This unambiguously illustrates the transparent
interaction between the syntax of applicatives and their morphological realization.
Moreover, in so called reflexive applicatives when the applicative relation such as
Recipient or Benefactee is established between the external argument and the theme, the
morphological realization again ‘responds’ to the distinct syntax of such constructions. In
these structures, a reflexive Voice head is merged and its phonological exponent is
different from the one inserted into the Appl0 head in DOCs. Similarly, in the
applicatives formed from internally- and externally-caused events, the morphological
realization of the structure is different allowing the exponent of the Voice head to get
realized, rather than the exponent of the Appl0 head through the mechanism of
discontinuous feeding. Such an interaction between the syntax and morphology of these
constructions is argued to result from various morphological rules such as discontinuous
feeding or impoverishment of certain features on terminal nodes in the post-syntactic
component of the grammar. Overall, evidence shows that in order for distinct applicative
relations to get established between two individuals (low applicative) or an individual and
an event (high applicative) it is not necessary to have distinct morphology.
Future research on these constructions should include the interaction between
causative and applicative structures in complex constructions involving applicatives as
complements of syntactic causatives or vice versa. The interaction between the
morphology and syntax in these complex embedding structures will shed light at the
precise nature of language-specific morphological well-formedness constraints which
preclude the realization of multiple exponents for the whole range of functional heads
merging in these complex structures. We hope that this work is a modest starting point
for the study of more complex predicate structures in these types of languages.
Homophony across phonological exponents of various morphemes in Georgian
I Phonological exponent a1. a- realizes CAUSE in lexical causatives formed from inchoative verbs such as in the
CAUS- bakes
‘He’she bakes it.’
2. The same exponent realizes CAUSE in syntactic causatives of X makes Y do V type:
CAUS- scream
‘He/she makes X scream.’
3. The exponent realizes the aspectual marker in past and future tense verbs denoting the
endpoint of an event:
prev- lended
‘He/she lended X it.’
4. The same phonological exponent at the end of the template marks non-local subject or
object and in this position it is fused with the series marker:
utxr- a
told- 3S/Ser
‘He/she told X.’
Here are the insertion contexts of these VIs:
a. a-  CAUSE / a. a-  CAUSE / ___ [Root-]
b. a-  CAUSE / ___ v0[TRANS], [DITRANS]
c. a-  Asp0 [Accomplishment]
d.-a  Tns0 + [Person]
II Phonological Exponent i-
1. The exponent i- realizes the Appl0 head in an environment of the local dative applied
arguments [1-2] person:
m- icxobs
1O- APPLIC- bakes
‘He/she bakes me it.’
2. The same exponent realizes the Reflexive Voice in reflexive verbs such as in the
REFL- shaves
‘He/she shaves.’
3. The same exponent realizes the reflexive Voice in reflexive applicatives:
v- icxob
1S- REFL- bake
‘I bake it for myself.’
4. The same exponent is inserted as a thematic marker in a set of verbs as in the
prckvn- is
PeelTH- 3S/Ser
‘He/she peels it.’
5. The same exponent realizes an intransitivity marker, which is referred to as a passive
voice marker in the traditional linguistics literature:
intrans- bite
‘He/she bites.’
Here are the environments for the insertion of these VIs:
 Appl / [Dat] [+1] [+3]
 VoiceREFL / reflexives, applicatives
 TH
III The phonological exponent e-
1. The phonological exponent e- is inserted in the [NonActive] Voice of adversity
m- ek’vlevineba
1S- Voice- kill
‘Something causes me to kill.’
2. The same expoenent realizes the Voice in the applicatives of internally and
externally caused verbs:
m- epurčkneba
1S- Voice- blossom
‘I am blossomed it.’
3. The same exponent marks the Voice in verbs denoting reciprocal action such as:
v- elaparak’ebi
1S- Voice- talk
‘I am talking to somebody.’
4. The same exponent realizes the Aorist marker fused with the [1-2] person subjects
or objects:
Daviban- e
Washed- 1O/Aor
‘I washed it.’
5. The same exponent realizes the passive Voice:
m- edz’ɣvneba
1S- Voice- dedicate
‘I am dedicated it.’
Here are the insertion contexts of these VIs:
e-  Voice[NONACTIVE]/ causatives, applicatives
e-  Voice [PASSIVE]
e-  Voice [ACTIVE] /activities
e-  Tense [AORIST] + -features
These are main types of syncretisms attested across various functional morphemes in
Anderson, Stephen. 1992. A-Morphous Morphology. Cambridge: Cambridge University
Arad, Maya. 2003. Locality constraints on the interpretation of roots: the Case of Hebrew
Denominal verbs. Natural Language and Linguistic Theory 21, 737-778.
Aronoff, Marc. 1976. Word-formation in Generative Grammar. Cambridge, MA: MIT
Aronoff, Marc. 1994. Morphology by Itself: Stems and Inflectional Classes. Cambridge,
Mass: MIT Press.
Aronson, Howard. 1990. Georgian: A Reader’s Grammar. Columbus, Ohio: Slavica
Bach, Emmon. 1986.. The Algebra of Events. Linguistics and Philosophy 9: 5-16.
[Reprinted in Paul Portner and Barbara H. Partee, eds. 2002. Formal Semantics: the
Essential Readings. Oxford: Blackwell]. [Reprinted in The Language of Time: a Reader.
2005. Eds., Inderjeet Mani, James Pustejovsky, and Rob Gaizauskas, Oxford: Oxford
University Press, 61-69.]
Baker, Mark. 1985. The mirror principle and morphosyntactic explanation, Linguistic
Inquiry 16, 373-416.
Baker, Mark. 1988. Incorporation: a Theory of Grammatical Function Changing.
Chicago: University of Chicago Press.
Baker, Marc.1996. The Polysynthesis Parameter. New York, Oxford: Oxford University
Baker, Marc & Chris Collins. 2006. Linkers and the internal structure of vP, Natural
Language and Linguistic Teory 24: 307-354.
Bejar, Susana. 2000. Structural markedness in formal features: deriving interpretability.
Revue Qucbecoice de Linguistique 28.1, 47-72.
Bejar, Susana. 2002. Locality, Cyclicity and Markedness in Georgian Verbal
morphology. Ms., University of Toronto.
Bejar, Susana. 2003. Phi-syntax: a Theory of Agreement. Ph.D dissertation. University of
Bejar, Susana & Rezak, Milan. 2003. Cyclic agree. Lisbon Workshop on Agreement.
Univesidade Nova de Lisboa.
Belletti, Adrianna & Luigi Rizzi. 1988. Psych-verbs and Theta-theory. Natural Language
and Linguistic Theory 6, 291-352.
Bobaljik, Jonathan. 1993. On Ergativity and Ergative Unergatives. MIT Working Papers
in Linguistics 19. Ed. C. Phillips. Papers on Case and Agreement II. 45-88. Cambridge,
MA: MIT Press.
Bobaljik, Jason & Phil Branigan. 2006. Accentric agreement and multiple case checking.
In Ergativity, Emerging Issues, Eds. A. Johns et al. Dordrecht, Netherlands: Springer.
Bonet, Eulalia. 1991. Morphology After Syntax: Pronominal Clitics in Romance
languages. Ph.D Dissertation. MIT.
Borer, Hagit. 1994. The projection of arguments. In Occasional Papers in Linguistics 17,
ed. E.Benedicto and J.Runner, 19-48. University of Massachussetts, Amherst: GSLA.
Bresnan, Joan and Lioba Moshi. 1990. Object Asymmetries in Comparative Bantu
Syntax, Linguistic Inquiry 21, 147-185.
Butt, Miriam. 1995. The structure of complex predicates in Urdu. Stanford, CA: CSLI
Butt, Miriam & Wilhelm Geuder. 2001. On the semi(lexical) status of light verbs. In
Norbert Corver and Henk van Rjiemsdijk (eds.) Semi-lexical categories: the function of
content words and the content of function words. Mouton, 323-369.
Carmack, Stephen. 1997. Blocking in Georgian Verbal morphology. Language, 73(2):
Chomsky, Noam. 1972. Remarks on Nominalization. In Studies on semantics in
generative grammar. 11-61.The Hague: Mouton.
Chomsky, Noam.1995. The Minimalist Program. Cambridge, MA: The MIT Press.
Chomsky, Noam. 1998. Minimalist Inquiries: the framework. In Step by step, ed. R.
Martin, D. Michaels, and J. Uriagereka, 89-155. Cambridge, MA: MIT Press.
Chomsky, Noam. 1999. Derivation by phase. Ken Hale: A life in Language, ed. M.
Kenstowicz, 1-52. Cambridge, MA: MIT Press.
Complex Predicates. 1997. (eds.) A. Alsina, J.Bresnan and P.Sells. Stanford: CSLI
Cuervo, Maria. 2003. Datives at Large. Ph.D. Dissertation. MIT.
Davidson, Donald. 1967. The logical Form of Action Sentences. In N. Rescher (ed.), The
Logic of Decision and Action. 81-95. University of Pittsburg Press, Pittsburgh.
Davies, Henry and Hamida Demirdache, 2000. On Lexical Verb Meanings: Evidence
from Salish. 97-138. Stanford, California: CSLI Publications, Center for the Study of
Language and Information.
Dowty, David. 1991. Thematic proto-roles and argument selection. Language 67 (3):
Embick, David. 1998. Voice and the Interfaces of Syntax. Ph.D dissertation, MIT.
Embick, David and Morris Halle. 2005. On the status of stems in Morphological theory.
In T. Geerts and H. Jacobs (eds.). Proceedings of Going Romance 2003. Amsterdam,
NY: John Benjamins Publishing company.
Embick, David & Alec Marantz. 2006. Architecture and Blocking. MS, University of
Pennsylvania and MIT.
Embick, David. 2009. Localism versus Globalism in Morphology and Phonology. MS.,
University of Pennsylvania.
Embick, David. 2007. Linearization and Local Dislocation: Derivational mechanics and
interactions. Linguistic Analysis 3, 303-336.
Embick, David and Rolf Noyer. 2007. Distributed Morphology and the SyntaxMorphology Interface. In G. Ramchand and C. Reiss (eds.).The Oxford Handbook of
Linguistic Interfaces. Oxford, New-York: Oxford University Press. 290-322.
Embick, David.1997. Voice systems and the Syntax/Morphology Interface. In MITWPL,
H. Harley (ed.) The Proceedings of the Penn/MIT Workshop on Aspect, Argument
Structure, and Events, May 1997, 1-32.
Embick, David and Rolf Noyer. 2001. Movement operations after syntax, Linguistic
Inquiry 32/4, 555-595.
Ergativity: Emerging Issues. eds. A. Johns, D. Massam, and J. Ndayiragije. Dordrecht,
Netherlands: Springer.
Folli, Rafaella, Heidi Harley, and Simin Karimi. 2005. Determinants of event structure in
Persian complex predicates, Lingua 115.10, 1365-1401.
Gleitman, Lyla. 1990. Structural sources of Verb meaning. Language Acquisition 1: 3-55.
Ghomeshi, Jila and Diane Massam.1994. Lexical/syntactic relations without projection.
Linguistic analysis 24 (3-4), 175-217.
Grimshaw, Jane and Armin Mester. 1988. Light verbs and theta-marking. Linguistic
Inquiry 19 (2), 205-232.
Gurevich, Olga. 2006. Constructional Morphology: The Georgian Version. Ph.D.
Dissertation, UC Berkeley.
Hale, Ken. 2001. Navajo verb stem position and the bipartite structure of the Navajo
conjunct sector. Linguistic Inquiry 32.4, 678-693.
Hale, Ken and Samuel Jay Keyser. 1993. On Argument Structure and the Lexical
Expression of Syntactic Relations. In The View from Building 20: Essays In Linguistics
In Honor of Sylvain Bromberger, ed. K. Hale and S.J. Keyser, 53-109. Cambridge, MA:
MIT Press.
Hale, Ken and Samuel Jay Keyser. 2002. Prolegomenon to a theory of argument
structure. Cambridge, MA: MIT press.
Halle, Morris. 1997. Distributed Morphology: Impoverishment and Fission. In MIT
Working Papers in Linguistics 30. 425-449. Cambridge, MA: MIT press.
Halle, Moris and Alec Marantz. 1993. Distributed morphology and the pieces of
inflection. In The View from building 20, ed. Kenneth Hale and Samuel Jay Keyser, 53110. Cambridge, MA: MIT press.
Halle, Moris & Alec Marantz, 1994. Some Key features of Distributed Morphology. In
MIT Working Papers in Linguistics 21, 275-288. Papers on Phonology and Morphology.
Cambridge, MA: MIT Press.
Harley, Heidi, 1994. Hug a tree: Deriving morphosyntactic feature hierarchy. MIT
Working Papers in Linguistics 21, Papers in Phonology and morphology. Eds. A. Carnie,
H. Harley with T.Bures.
Harley, Heidi and Rolf Noyer. 1999. Distributed Morphology, GLOT International 4:3-9.
Harley, Heidi and Noyer, Rolf. 1998. Licensing in non-lexicalist lexicon:
Nominalizations, vocabulary items, and the Encyclopedia. In Papers from the
UPenn/MIT Roundtable on argument structure and aspect. Ed. Heidi Harley, 175-200.
MIT Working Papers in Linguistics 32, Cambridge, Massachusetts: MIT press.
Harley, Heidi, 1995. Subjects, events and licensing. Ph.D. dissertation. MIT.
Harley, Heidi. 2002. Possession and the Double Object Construction. In Linguistic
Variation Yearbook 2, 29-68. Amsterdam: John Benjamins Publishing Company.
Harley, Heidi. 2008. On the causative construction. In S. Miagawa and M.Saito (eds.)
Oxford Handbook of Japanese Linguistics, New-York, Oxford: Oxford University Press.
Harley, Heidi. 2002. Minimal(ish) Linking Theory. A talk given at the Maryland
Minimalist Mayfest, Maryland: May 16, 2002.
Harris, Alice. 1981. Georgian Syntax. London: Cambridge University Press
Horwath, Julia and Tal Siloni (forthcoming). Causatives Distributed, Ms., Tel Aviv
Jackendoff, Ray. 1990. Semantic structures. Cambridge, MA: MIT Press.
Jacobsen, W.M. 1992. The transitive structure of events in Japanese. In Studies in
Japanese linguistics I. Tokyo: Kurosio publishers.
Jelinek, Eloise.1999. Accent in Yaqui. With Richard Demers and Fernando Escalante.
International Journal of American Linguistics.
Jelinek, Aloise & Andrew Carnie. 2003. Argument hierarchies and the mapping
principle. In Formal Approaches to function in grammar: In honor of Eloise Jelinek. Eds.
Andrew Carnnie, Heidi Harley & Mary Willie. 265–96. Amsterdam: John Benjamins
Publishing Company.
Jeong, Youngmi. 2007. Applicatives: Structure and interpretation from a minimalist
perspective. Amsterdam, Netherlands: Linguistik Aktuell / Linguistics Today.
Jung, Yeun-Jin & Shigeru Miyagawa. 2004. Decomposing Ditransitive verbs. In
Proceedings of SICGG, Summer 2004, 101-120.
Kajaia, Otar. 2001. Megrul-Kartuli leksikoni. [Mingrelian-Georgian Dictionary]. Tbilisi,
Georgia: Nekeri.
Karimi, Simin.1997. Persian complex verbs: idiomatic or compositional. Lexicology 3
(2), 273-318.
Karimi-Doostan, Mohhamad-Reza. 1997. Light verbs constructions in Persian. The Ph.D
dissertation. University of Essex.
King, Halloway Tracy. 1994. SpecAgrP and Case: Evidence from Georgian. MIT
Working Papers in Linguistics 22, ed. H. Harley & C. Phillips, 91-110. Cambridge, MA:
MIT Press.
Kratzer, Angelika. 1996. Severing the External Argument from Its Verb. In Phrase
Structure and Lexicon. Eds. Johan Rooryck & Laurie Zaring, 109-138. Dordrecht:
Kuroda, S.-Y. 1965. Causative forms in Japanese. Foundations of language 1. 30-50.
Kuroda, S.-Y. 1993. Lexical and Productive causatives in Japanese: an examination of
the theory of paradigmatic structure. Journal of Japanese Linguistics 5, 1-82.
Larson, Richard. 1988 (1988). On the Double Object Construction. Linguistic Inquiry
19.3, 335-391.
Levin, Beth and Malka Rappaport Hovav. 1995. Unaccusativity. At the Syntax-Lexical
Interface.Cambridge, Massachusetts: MIT press.
Lieber, R. 1992. Deconstructing Morphology: Word Formation in Syntactic Theory.
Chicago: University of Chicago Press.
Lomashvili, Leila & Heidi Harley. (In press). Phases and templates in Georgian
Manning, Christopher. 1996. Ergativity. Stanford, CA: CSLI Publications.
Marantz, Alec. 1992. Case and Licensing. ESCOL 91: 234-53.
Marantz, Alec. 1984. On the Nature of Grammatical Relations. Cambridge,
Massachusetts: MIT Press.
Marantz, Alec. 1997. No escape from syntax: Don't try morphological analysis in the
privacy of your own lexicon. In Penn Working Papers in Linguistics 4, ed. A. Dimitriadis
et al. 201-225. Philadelphia, Pennsylvania.
Massam, Dianne. 2001. Pseudo-incorporation in Niuean. Natural Language and
Linguistic theory 19: 153-197.
Mederdoomian, Karine. 2001. Event structure and Complex Predicates in Persian.
Canadian Journal of Linguistics 46 (1/2): 97-125.
Mithun, Marianne. 1984. The Evolution of Noun Incorporation. Language, Vol.60, No.4,
Miyagawa, Shigeru. 1999. Causatives. (ed.) Natsuko Tsujimura, Handbook of Japanese
Linguistics, 236-268. Blackwell.
Miyagawa, Shigeru. 1980. Complex verbs and lexicon. The Ph.D. dissertation, University
of Arizona.
McGinnis, Martha.1997. Case and Locality in L-Syntax: evidence from Georgian. MIT
Working Papers in Linguistics 32, Ed. H. Harley, 1-20. Cambridge, MA: The MIT press.
McGinnis, Martha. 2001. Variation in the phase structure of applicatives. In Linguistic
Variation Yearbook 1: 105-146.
McGinnis, Martha (with Dona Gerdts). 2004. A phase-theoretic analysis of Kinyarwanda
multiple applicatives. In Proceedings of the 2003 CLA Annual Conference. Eds. Sophie
Burelle and Stanca Somesfalean, 154-165.
Mohanan, Tara. 1997. Multodimensionality of representation: NV complex predicates in
Hindi. In Alsina et al. (eds). 431-471.
Ndayiragije, Juvenal. 2006. The Ergativity Parameter. The view from antipassive. In
Ergativity: Emerging Issues. eds. A. Johns, D. Massam, and J. Ndayiragije. Dordrecht,
Netherlands: Springer.
Nash, Lea.1994. On BE and HAVE in Georgian. In MITWPL 22, 153-171. eds. H.
Noyer, Rolf. 1997. Features, Positions and Affixes in Autonomous Morphological
Structure. New-York: Garland.
Pesetsky, David. 1989. Language-Particular Processes and the Earliness Principle. Ms,
MIT. Available online at
Pesetsky, David.1995. Zero syntax. Cambridge, MA: The MIT press.
Pinker, Stephen. 1989. Learnability and Cognition: The Acquisition of Argument
Structure. MIT Press.
Pustejovsky, James. 1995. The generative lexicon. Cambridge, MA: The MIT press.
Pylkkanen, Liina. 2002. Introducing arguments. Ph.D Dissertation, MIT.
Pylkkanen, Liina. 2008. Introducing arguments. Linguistic Inquiry Monographs.
Cambridge, MA: The MIT Press.
Radford, Andrew. 2004. Minimalist syntax. Exploring the Structure of English.
Cambridge: Cambridge University Press.
Rezak, Milan. 2004. Elements of Cyclic syntax: Agree and merge. The Ph.D dissertation,
University of Toronto.
Ritter, Elisabeth and Sara Rosen. 1998. Delimiting events in syntax. In The Projection of
Arguments, ed. Miriam Butt and Wilhelm Geuder, 135-164. Stanford, California: CSLI
Sadock, Jerrold. 1980. Noun incorporation in Greenlandic. Language 56. 300-319.
Serratos, Angelina. 2008. Roots, Functional Heads, and Phases in Chemehuevi. Ph.D.
dissertation. University of Arizona.
Siddiqi, Daniel. 2009. Syntax within the Word. Economy, allomorphy, and argument
selection in Distributed Morphology. Linguistik Aktuell series. Amsterdam: John
Benjamins Publishing Company.
Silkirk, Elisabeth. 1982. The Syntax of Words. Cambridge, MA: MIT Press.
Stewart, Thomas. 2001. Georgian Agreement without Extrinsic Ordering. In The Ohio
State University Working Papers in Linguistics, vol. 56, 107–133.
Stump, Gregory T. 2001. Inflectional Morphology: A Theory of Paradigm Structure
[Cambridge Studies in Linguistics 93]. Cambridge: CUP.
Topuria, Varlam. 1967. Svanuri Ena. [The Svan Language]. Tbilisi: Tbilisi University
Travis, Lisa. 2000. Event structure in Syntax. In Events and Grammatical Objects. (eds.)
Carol Tenny and J.Pustejovsky. Stanford: CSLI Publications.
Van Gelderen, Elly. 2004. Grammaticalization as Economy. [Linguistik Aktuell 71].
Amsterdam: John Benjamins.
Vendler, Zeno. 1967. Verbs and times. In Linguistics and Philosophy, ed. Z. Vendler.
Ithaca: Cornell University Press.
Woolford, Elen. 2006. Case and agreement mismatches. Agreement Systems, (ed.)
C.Boeckx. 299-316. Amsterdam, Philadelphia: John Benjamins B.V.
Zwicky, Arnold. 1985. Heads. Journal of Linguistics 21: 1-30.
Was this manual useful for you? yes no
Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Download PDF