Typological Peculiarities of the Georgian
Passive and the Information Structure
Rusudan Asatiani
Institute for Oriental Studies;
Tbilisi State University
Georgia
[email protected]
I. Introduction: setting a task

Within the theory of functional grammar the Passive Construction
(PC) is considered as a syntactic category: It is qualified as a
conversive one of the corresponding active construction where the
Patient is promoted to the subject position along the string of
hierarchically organized functional categories (S>DO>IO), while the
Agent is demoted and transformed into a prepositional phrase; so, it
does not represent a core argument defined by the verb valency any
more;

Yet, many languages present morphologically marked verb forms in
such conversive constructions (resp. PC) and, consequently, it is
possible to speak about the morphosyntactic category of passive
voice.

In Georgian there is a clear formal opposition between the active (first of
all, transitive) and the passive (first of all, active’s conversive) verb
forms that is represented by special morphosyntactic features (*terms
“active”/“passive” are used conventionally):
“Active”
“Passive”
-s
-a
Thematic marker
(in I-series tense
forms)
-eb-
-eb-, -ob-, -op-,
-av-, -am-, -i-, -Ø-;
Special markers
----
i-, e-, -d, -Ø-
S.3.SG suffix
(in present)
Subject case
NOM(in present)/
ERG(in Aorist)/
DAT(in Perfect)
NOM

But, none of the features can be regarded as a simple morphosyntactic
markers of the PC as far as they don’t exist just only in PC:




Main function of -s, -a suffixes is to mark S.3.SG and representing this
function they can be found in various cases: (1) -s is a marker of S.3.SG in
the forms of subjunctive mood of both, the passive and the active verbs; (2)
–s represents S.3.SG of some static verbs (so called, medio-passives); and
(3) -a can be the marker of active verbs S.3.SG in past tenses (Aorist,
Perfect); etc.;
Main function of -eb- is to mark dynamic verb forms and expressing this
function it exists with some active verbs as well;
Vowel prefixes are polyfunctional: in general, they represent derivational
changes of verb valency – either appearance or disappearance of verb
argument; e.g. -i- expresses such categories as subjunctive version,
reflexive, potentialis, deponences and has an additional function to form the
future tense of some medial (resp. medio-active) verbs;
Nominative is a case characteristic for the subjects of some intransitive (yet,
not passive) verbs and Ergative (or Dative) can be the subject marker for
any kind of intransitive (yet, active) verbs expressing active process.


Thus, we have two different formal models defined by the complex of
morphosyntactic features that represent active-passive opposition; yet,
the models can not be interpreted simply as far as
The Georgian morphosyntactically distingushed PC does not always
show a conversion of an active one and it actually expresses various
semantics:




Active semantics – dgeba ‘S/he is standing up’, ekačeba ‘S/he tugs hard at
smth./smb.’, ac’veba ‘S/he pushes smth./smb.’, etc.
Dynamic actions – tvreba ‘S/he gets drunk’, šreba ‘S/he dries’, tbeba ‘S/he
gets warm’, etc.
Potentialis – ič’meba ‘It can be eaten’, ismeba ‘It is drinkable’, ik’itxeba ‘It
can be read’, etc.
Reciprocals – etamašeba ‘S/he plays with sb.=They play together’,
ecek’veba ‘S/he dances with sb.=They dance together’, etc.

Tasks:
 What is the real function of the morphosyntactically differentiated
models?
 What is the actual semantics of Georgian so called PC?
 How does the Georgian PC fit to the universal functional
interpretation of PC?
 How can be qualified the Georgian PC?
 Is the Georgian PC defined functionally or semantically?
And so on.

It is clear that the issue certainly needs further investigations.
II. Theoretical approaches

Information Structure: According to the one theoretical approach
implemented in contemporary linguistics the active-passive functional
differences can be explained by the variety of information structures.

During the linguistic structuring of extra-linguistic situations some
languages conventionally conceptualize as the central part of
information either Agent or Patient. The first construction formally
emphasizes who is acting (Ag=focus), while the second emphasizes
what is done (P=focus). From the grammatical point of view,
conceptual foregrounding is represented by the unmarked, Nominative
case: In the nominative languages (in active constructions) it is the
Agent, who always stands in nominative, while in the ergative
languages it is the Patient (and not the Agent) who appears in
nominative.




Patient’s foregrounding in the nominative languages, where agent is
conceptually highlighted part, can further be achieved by the changes
of functional roles.
In PC the Patient is functionally promoted and it is defined as the
Subject.
The term Subject actually denotes foregrounding of a certain part of
information to whom or what the information concerns. Active
construction shows Agent’s foregrounding (that means: Agent is the
Subject and, consequently, stands in nominative), while Passive
construction shows Patient’s foregrounding (that means: Patient is the
Subject and, consequently, stands in nominative).
Thus, syntactic opposition between active and passive constructions
can be provoked by different models of the information structuring.

The Georgian language shows split-ergativity and because of this the
restrictions of passivization reflect more complicated processes
defining the choices of either active or passive constructions.

Another device defining either appearance or disappearance of PC in
Georgian is a relatively free word order that makes possible to put the
focused patient in marked (mostly sentence initial and pre-verbal)
position without any kind of functional promotion and/or demotion
(resp. pasivization). Thus, PC is not the only means expressing the
functional foregrounding of patient.

Consequently, the role of PC in the process of patient functional
foregrounding in Georgian needs further investigations as well.

Methodology

Sentences raised in natural speech conversation are the most valuable for
seeking the main formal models of information structures. Stimulation of
such situations is possible by means of specially created experimental tasks.

Our empirical data is collected using the Questionnaire on Information
Structure (QUIS), which is being developed within the Sonderforschungsbereich 632 “Information Structure” at the University of Potsdam and the
Humboldt University Berlin (SKOPETEA & ALL 2006).

QUIS comprises a set of translation tasks and production experiments for
the collection of primary data.

The ‘production experiments’ contain a range of experimental settings that
introduce spontaneous expressions (e.g. picture descriptions, map tasks,
some plays and etc.).

For our goals the following experiments were especially interesting:

Description of the experimental task:
 The special experiment explores the interrelation between patient’s
animacy and agent’s visibility in the process of PC appearance.

It is assumed that, in general, the appearance of PC is more
probable:



with animate patient and less probable with inanimate one;
with the agent that is not identifiable and less probable with the agent
that is identifiable.
Logically possible all (four) different cases are presented in
pictures:




AC:the patient is animate: the agent is identifiable;
AD: the patient is animate: the agent is non-identifiable;
BC: the patient is inanimate: the agent is identifiable;
BD: the patient is inanimate: the agent is non-identifiable;

Pictures

Procedure:
 The instructor says:


You will be shown two scenes that belong together; that is, they
belong to the same story. Imagine that the first scene takes
place first and the second scene some times later, e.g. after
five minutes. Please give a short description of what is going
on in each scene.
The instructor shows the first picture to the informant and asks:
What is going on in this scene?
Then the instructor shows the second picture and asks:
What is going on in this scene?

Analysis

On the basis of the semi-spontaneous data, which is conducted during the 4
field sessions (16 informants, mostly students) using this QUIS experiment
task, PC in Georgian is not defined by the invisibility of agents and/or by the
animacy of patients and it does not always suppose the changes of syntactic
functions:
During the information structuring, when an invisible Agent
together with the animate Patient is presented in the situation and
PC might be the most appropriate construction (e.g. in languages
like English or German mostly PC is created by informants),
Georgian informants prefer to use active constructions with
uncertain subject represented either by S.3.PL suffixes, or by the
indefinite pronouns viɣac/raɣac ‘sb./smth.’, and marked word
order showing patient topicalization.
Examples:
(1) [BOtl-s]T
k’r-aven
pex-s
bottle-DAT
push-HAB-ACT.PRS.S.3.PL
foot-DAT
‘(They) are pushing the bottle.’
(2) [MA-s]T
viɣac
pex-s
u-rt’q’-am-s.
3.SG-DAT somebody foot-DAT [IO.3]OV-hit-TH-ACT.PRS.S.3.SG
‘Somebody is hitting him with foot.’

Results

Patient’s foregrounding is not expressed by PC in Georgian: there is no
functional foregrounding and the patient and/or the agent doesn’t change
their functional qualifications.

This is one more argument to interpret Georgian passive as a grammatical
category supposedly governed by semantic (or, more widely, by cognitive)
and not by syntactic features.

Consequently, it is necessary to find these semantic-cognitive features that
define the formal opposition between the active and the passive
morphologically distinguished models of formal representations.

Cognitive-semantic interpretation of active-passive
morphosyntactic oppositions: A continuum of active-passive
opposition


In many languages, like in Georgian, active-passive constructions do not
always express syntactically defined conversive forms and the passive
formal model is used to mark some other related constructions as well. In
general, there are languages where passive formal model marks reflexives
and reciprocals (e.g. Russian); in some languages it goes further and
expresses other grammatical relations as well; e.g. in Japanese it is the
formal representation for potential actions, polite constructions and,
moreover, plural forms. So, naturally, some attempts of new theoretical
approaches have been raised to explain such cases. One of such approaches
is SHIBATANI’s interpretation (SHIBATANI 1985).
SHIBATANI considers the active-passive opposition as a continuum where
polar dimensions fit in with the prototypical active and passive constructions
while non-polar, inter-medial cases share just only some semanticcategorical features of the categories which are characteristic for the
prototypical ones:
Prototypical active
AM


/
Medial forms
/
Prototypical passive
PM
Languages choose various strategies for formal representations of such non-polar
(let’s call them “Medial”) cases: they either create the new formal models or
choose from the existing ones the model that is conventionally regarded as the
most appropriate, more close according to certain semantic-categorical features
– either the active or the passive model.
In such cases, simple functional (resp. changing of syntactic functions) or
semantic (resp. defining active-passive semantics) interpretation of formal
models is much more complicated and sometimes impossible.
III. The Georgian data

Georgian active-passive continuum
Georgian active-passive opposition might be interpreted as a
continuum, where prototypical active corresponds to transitive active
constructions representing by the active model, while prototypical
passive defined by the patient’s foregrounding corresponds to active
construction’s conversive form representing by the passive model;

The medial forms grammaticalization process can be explained by the
following general cognitive tendency:
During the formal representation of medial forms Georgian
chooses either the active or the passive formal model. The
strategy of choice is defined by the specific, conventionally
accepted linguistic “decision”: which categorical-semantic features
of prototypical constructions are regarded as the central, main
ones.

For demonstrating such categorical-semantic features the following
linguistic empirical facts which are observed during the process of
formal representations of some intransitive medial forms must be
taken into account:
If a medial (resp. prototypically non-active and/or non-passive)
verb semantics tends toward an end (that is, it is semantically the
telic one), then a verb chooses the passive formal model of
representation; and if a medial verb semantics does not tend
toward an end (that is, it is semantically the atelic one) then a
verb chooses the active formal model of representation.

It is quite easy to give general formal interpretation of the fact:
If a verb with medial semantics can take just one preverb showing some
direction of action (sometimes creating the new semantics of a verb) in
future tense, then the verb has “passive form”.
Examples:
Compare –
 dgeba ‘S.3.SG is getting up’ : a-dgeba ‘S.3.SG will stand up’/gada-dgeba ‘S.3.SG
will stand elsewhere’/c’ar-dgeba ‘S.3.SG will step forward’ /čadgeba ‘S.3.SG
will stand in’;
 emaleba ‘S.3.SG is hiding from smth. or smb.’ : da-emaleba ‘S.3.SG will hide
from smth. or smb.’;
 ac’veba ‘S.3.SG is pressing down’ : mi-ac’veba ‘S.3.SG will push against
smth. or smb./da-ac’veba ‘S.3.SG will lie down on smth. or smb.’;
With –
 cxovrobs ‘S.3.SG lives’; pikrobs ‘S.3.SG thinks’; arsebobs ‘S.3.SG exists’;
k’ank’alebs ‘S.3.SG shivers’; goravs ‘S.3.SG rolls’; suntkavs ‘S.3.SG breathes’;
bč’obs ‘S.3.SG discusses’; brialebs ‘S.3.SG sparkles’; etc.)


As far as in Georgian preverbs have additional functions and they can express
Perfective-Imperfective aspect opposition and the Future Tense forms, it is
possible to reveal the semantic feature (resp. “completeness” of an action) which
governs the choice of passive formal model for some medial verbs and the above
given interpretation turns into the following semantically oriented interpretation:
If a medial verb with the concrete semantics implies the differences between
the imperfective (resp. incomplete) and the perfective (resp. complete)
aspect forms, then a medial verb is grammaticalized as a prototypical
passive and chooses the passive formal model.
Cognitively more predictable would be if such medial forms have chosen the
active formal model of representation as far as the perfective/imperfective aspect
is the characteristic category for active verbs, yet, if we take into account general
cognitive principles of formal markedness, it can be seen, that in Georgian
processes of linguistic structuring are defined by the following general tendency:
Non-prototypical passive (and it is non-prototypical because it can (like
an active one) differentiate completeness/incompleteness of an action), as
being cognitively marked, uses formally the most marked model (resp. the
passive formal model) of representation.
IV. Broadened continuum

Such cognitive-semantic interpretations of active-passive continuum could be
broadened comprising all spectrums of medial verb forms including so called
static passives and medio-passives, and the process of information structuring
can be reinterpreted as hierarchically organized one, where another opposition of
categories – “dynamic/static” – takes a distinctive role and is formally
grammaticalized according to the following restriction:
If medial verb form expresses static event, then a verb in present has
auxiliary conjugation.

That is, Georgian creates the new model (different from either active or passive
one) of formal representation with auxiliary conjugation:
Examples:
(1) me(1.SG) v(S.1)-dga(stand)-v(S.1)-ar(be.SG)
šen(2.SG) (S.2)dga(stand)-x(S.2)-ar(be.SG)
is(3.SG) dga(stand)-s(S.3.SG)
(2) me(1.SG) v(S.1)-gd(lie)-i-v(S.1)-ar(be.SG)
šen(2.SG) (S.2)gd(lie)-i-x(S.2)-ar(be.SG)
is(3.SG) gd(lie)-i-a(S.3.SG).

Such medial verbs fall into two subgroups following either the (1)-type of
conjugation (according to the Georgian grammatical tradition so called static
passives) or the (2)-type conjugation (so called medio-passives) distinguished by
S.3.SG suffix presented in presented tense:
Examples of the (1)-type with S.3.SG suffix -a:
gdia ‘S.3.SG lies strewn/thrown about’; q’ria ‘S.3.SG lie scattered/strewn’, penia ‘S.3.SG
is spread out’; k’idia ‘S.3.SG is hanging on’; c’eria ‘S.3.SG is written’; xat’ia ‘S.3.SG is
drawn’; abia ‘S.3.SG is tied (on)’, and etc.
Examples of the (2)-type with S.3.SG suffix -s:
dgas ‘S.3.SG stands’; c’evs ‘S.3.SG (smb.) lies’; zis ‘S.3.SG sits’; devs ‘S.3.SG (smth)
lies’; ɣirs ‘S.3.SG costs’, c’uxs ‘S.3.SG worries’; and etc.

The functional differences are more refined and the discovering of specific
semantic nuances defining the opposition needs more careful analysis. We can
suggest some formal testing expression:
If a verb creates correct phrase with the adverb tavad 'itself, personally'
(that is, the expressions like: tavad dgas, tavad c’evs, tavad c’uxs, and
etc. are correct), then it chooses the active model; if such a phrase is not
correct, the passive model of representation is chosen (that is, the
expressions like:*tavad gdia, *tavad kidia, *tavad c’eria and etc. are
unnatural or bad).

The testing adverb helps us to distinguish the feature: “personally,
according to the subjects will, controlled state, a state that is provoked
by the subject”. Let’s denote this feature by the term “Autotive” and
formulate the following tendency:
A verb conceptually close to autotive chooses the active model,
while verbs expressing a state that is not controlled or provoked
by the subject itself choose the passive model.

Thus, we can summarize all our discussion and suggest the dynamic
model which supposedly mirrors cognitive-semantic grounds of the
formal represetation of active-passive opposition including the medial
forms.
V. Hierarchically organized dynamic model

Linguistic representations of active, passive and medial verb forms can
be reinterpreted as a hierarchically organized realizations of cognitive
processes that define the choices of either the New (NM) or the Active
(AM) or the Passive (PM) formal models:

I stage: Prototypically active and prototypically passive relations are
represented by the main formal models: the active (resp. transitive,
showing agent’s foregrounding) and the passive (resp. conversive,
showing patient’s functional foregrounding) constructions;

II stage: Medial (non-prototypical) relations are marked according to
the two different strategies:
1. The new model is created;
2. Either active or passive models of representation have been
chosen.

The strategies of choices are defined by the specific cognitive
processes and semantic features. First of all, the feature “Dynamic Static” plays a decisive role:
Verbs expressing “Static” states are marked according to the 1strategy and the new model of conjugation with the auxiliary verb ‘to be’ is
chosen, while verbs expressing “Dynamic” action choose either active or
passive formal model of representation (2-strategy).

III stage: For “Dynamic” subgroup further choices are defined by the
semantic feature telicity:
Telic medial verbs choose passive formal model of representation, while
atelic medial verbs – the active model of representation.
For “Static” subgroup further choices are defined by the semantic
feature “Autotive”:
Verbs denoting static states that are more or less controlled by the subject
itself have the same S.3.SG ending in present tense as the active ones,
while all others choose the same S.3.SG suffix as the passive ones.

Thus, on the basis of formal and semantic-functional analysis of
passive, active and medial verb forms it is possible to suggest the
cognitive generative model that supposedly mirrors the hierarchically
organized processes of grammaticalization:
prototypes
prototypical
active
AM
non-prototypes
prototypical
passive
PM
dynamic
(thematic markers)
telic
+preverb
PM
Active transitive
/ Conversive passive /
Dynamic passive
atelic
-preverb
AM
/
static
(auxiliary conjugation)
NM
+autotive
(-s)
-autotive
(-a)
Medio-Active / Medio-passive / Static passive

Examples of medial verbs:




(1)-type medial verbs: dgeba ‘S.3.SG is standing up’, šreba ‘S.3.SG
becomes dry’, k’vdeba ‘S.3.SG dies’, xmeba ‘S.3.SG dries out’, tetrdeba
‘S.3.SG turns white’, k’acdeba ‘S.3.SG becomes man’, iq’epeba ‘S.3.SG
barks’, igineba ‘S.3.SG is sworn at’, c’veba ‘S.3.SG lies down’, tvreba
‘S.3.SG gets drunk’, etc.
(2)-type medial verbs: cxovrobs ‘S.3.SG lives’, pikrobs ‘S.3.SG
thinks’, arsebobs ‘S.3.SG exists’, k’ank’alebs ‘S.3.SG shivers’, goravs
‘S.3.SG rolls’, suntkav ‘S.3.SG breathes’, bč’obs ‘S.3.SG discusses’,
brialebs ‘S.3.SG sparkles’, etc.
(3)-type medial verbs: dgas ‘S.3.SG stands’, c’evs ‘S.3.SG (smb.) lies’,
zis ‘S.3.SG sits’, devs ‘S.3.SG (smth.) lies’, ɣirs ‘S.3.SG costs’, c’uxs
‘S.3.SG worries’, and etc.
(4)-type medial verbs: gdia ‘S.3.SG lies strewn/thrown about’, q’ria
‘S.3.SG lie scattered/strewn a lot of smth./smb.’, penia ‘S.3.SG is spread
out’, k’idia ‘S.3.SG is hanging on’, c’eria ‘S.3.SG is written’, xat’ia
‘S.3.SG is drawn’, abia ‘S.3.SG is tied’, and etc.
VI. Some notes and conclusions

We suppose that representing continuum of active-passive opposition
and the dynamic hierarchically organized cognitive model explain the
complex processes that define the choices of either the active or the
passive formal models of representation for the non-prototypical
medial forms in Georgian.

Efficiency of such approach confirms once more that Georgian
morphological passive doesn’t always represent the syntactic changes
implying by the information structuring, namely by the patient’s
foregrounding.

Because of these peculiarities morphologically represented passive
verb forms create an opposition with the syntactic passive that is
formed by the periphrastic constructions:
{Passive Participle + auxiliary verb q’opna ‘to be’}
Examples:

dac’erili=a ‘written+is’,
dac’eril ikna ‘written+was’
dac’eril ikneba ‘written+will be’
Main function of this opposition is to formalize the functional differences between
syntactically defined and semantically defined passive constructions:
Periphrastic, analytical passive represents functional changes (resp.
patient’s functional foregrounding) of semantic roles (Patient => Subject,
Agent => Prepositional phrase), while synthetic, morphological passive can
represent semantically passive (resp. prototypically inactive, yet, dynamic
and telic) constructions.
Even in case when an active verb has not morphologically opposed
passive, it still has periphrastically opposed conversive form:
Examples:





ik’vlevs ‘(S)he researches smth.’ : gamok’vleulia ‘Smth. is researched’ (yet,
*ik’vleveba);
c’armoadgens ‘(S)he presents smth.’ : c’armodgenilia ‘Smth. is
presented’ (yet, *c’armoidgineba);
arčevs ‘(S)he chooses smth./smb.’ : arčeulia ‘Smth./smb. is
chosen’ (yet, *irčeva), and etc.
It can be concluded that Georgian analytical, periphrastic passive
corresponds to the PC existing in some Indo-European languages (it is
syntactically defined), while synthetic, morphological passive has
different functional loading and represents mostly semantically (and
not syntactically) defined peculiar forms.
References






ASATIANI, RUSUDAN (1982): mart’ivi c’inadadebis t’ip’ologiuri analizi. tanamedrove kartuli
salit’erat’uro enis masalaze (Typology of simple sentence. On the data of modern literary
Georgian). Tbilisi: mecniereba; (2001): Conceptual Structure of Reflexive and Middle, in:
Proceedings of 4th International Symposium on Language, Logic and Computation,
Amsterdam: ILLC scientific publications, 5-16; (2007): inpormaciis st’rukt’urirebis sint’aksuri
modelebi kartulshi, in: semiot’ik’a-II. Tbilisi: universali, 3-13; (2007): The Main Devices of
Foregrounding in the Information Structure of Georgian Sentences. in: Proceedings of Tbilisi
Symposium on Language, Logic and Computation - 2005. Amsterdam: Springer. 21-31;
CHAFE, WALLACE L. (1971): Meaning and the structure of language. Chicago and London:
Chicago Un. Press.
CHIKOBAVA, ARNOLD (1968): mart’ivi c’inadadebis p’roblema kartulshi (The Problem of the
Simple Sentence in Georgian). I. Tbilisi: mecniereba.
DAVITIANI, AK’AK’I (1973): kartuli enis sint’aksi. I. mart’ivi c’inadadeba(Syntax of the Georgian
Language.I. Syntax). Tbilisi: ganatleba.
DIXON, ROBERT M.W. (1979): Ergativity. Cambridge: Cambridge Un. Press.
ENUKIDZE, LEILA (1981): c’inadadebis akt’ualuri danac’evreba da misi mimarteba sint’aksuri da
semant’ik’uri analizis tanamedrove metodebtan (Actual Parsing of a Sentence and Its Relation
to Contemporary Methods of Syntactic and Semantic Analysis, in: tanamedrove zogadi
enatmecnierebis sak’itxebi VI. Tbilisi: enatmecnierebis inst’it’ut’i, 24-35.








HARRIS, ALICE (1998): Georgian Syntax: A Study in Relational Grammar. Cambridge etc.:
Cambridge University Press; (2000): Word Order Harmonies and Word Order Change In
Georgian, in: Sornicola R, Poppe E., Haley A. (eds), Stability, Variation and Change of WordOrder Patterns over time. Amsterdam/Philadelphia: Benjamins, 133-163.
HEWITT, GEORGE (1995): Georgian: A Structural Reference Grammar. Amsterdam, Philadelphia:
Benjamins.
KVACHADZE, Levan (1996): tanamedrove kartuli enis sint’aksi Tbilisi: rubik’oni.
KIBRIK, ALEXANDER (1997): Beyond Subject and Object: Toward a Comprehensive Relational
Typology. Linguistic Typology. I. Berlin-New York: Mouton de Gruyter, 279-346.
SHANIDZE, AKAKI (1948): kartuli enis gramat’ik’a II. sint’aksi (Grammar of the Georgian
Language II. Syntax). Tbilisi: tsu gamomcemloba; (1973): kartuli enis gramat’ik’is sapudzvlebi
(Principles of the Georgian Language Grammar). Tbilisi: tsu gamomcemloba.
SHIBATANI, MASAYOSHI (1985) Passives and Related constructions: A prototype Analysis.
Language, Vol. 61, No4. Kobe: Kobe Un. Press.
SKOPETEAS, STAVROS & ALL (2006): Questionnaire on Information Structure (QUIS).
Interdisciplinary Studies on Information Structure 4. Working Papers of the SFB 632, Potsdam:
Universitätsverlag Potsdam.
TUITE, KEVIN (1998): Kartvelian morphosyntax. Munich: Lincom Europa.
Thank you!
Descargar

Gender Differences and Consonants' Distribution in