+ All Categories
Home > Documents > Эвиденциальность в сванском языке

Эвиденциальность в сванском языке

Date post: 05-Mar-2023
Category:
Upload: rggu
View: 0 times
Download: 0 times
Share this document with a friend
27
EVIDENTIALITY, TRANSITIVITY AND SPLIT ERGATIVITY: EVIDENCE FROM SVAN Nina Sumbatova Institute of Oriental Studies, Moscow This article provides a description of the evidential forms in Svan (morphology, syntactic properties, semantic variants). The most interesting phenomenon concerning the evidential forms in Svan is their special case-marking patterns. The perfective evidentials form an “inversive”, or dative, construction (the subject in the dative case, the direct object in the nominative). A diachronic account for this may be the origin of the evidentials from resultative/perfect forms. Synchronically, syntactic peculiarities of the evidentials can be explained through the correlation of evidentiality and transitivity in the sense of Hopper and Thompson‟s transitivity hypothesis. 1. Introduction Evidentiality and related categories have been the subject of extensive and fruitful research for the last 1015 years (see, for example, Chafe & Nichols 1986; Guentchéva 1996). As DeLancey noticed in his recent paper, “the grammatical marking of evidentiality, long regarded as an exotic phenomenon found only in a few obscure languages, has in recent years come to be recognized as a widespread and significant typological parameter” (DeLancey 1997: 33). We even speak of an “Old World evidential belt” covering Turkish, Kartvelian, Bulgarian-Macedonian and Albanian. A part of this belt is the Kartvelian language family. Kartvelian verbal systems are well-known for their complexity and impressive syntactic properties. Their most striking characteristic is split ergativity. In the Kartvelian case split ergativity means that the choice of accusative/ergative/dative actant coding depends on the tense form of the verb. This paper deals with the most archaic language of this family, Svan 1 . Both Svan and the “main” language of the family – Georgian have an evidential “perfect” which together with the pluperfect and the subjunctive perfect constitutes the so- called “perfect” series of verbal tense/mood/aspect paradigms (in this series, the subject of transitive verbs is marked by the dative case, the direct object by the nominative). But only Svan has another evidential group of paradigms, an imperfective one, with nominative case marking. The narrow task of my paper is a description of Svan evidentials: their position in the verbal system as a whole; semantics and usage; syntactic properties of different evidential forms. But at the same time this paper is aimed at two remoter goals: suggesting and explaining a typologically correct hypothesis about the origin of the evidential forms and providing a functional explanation of their syntactic “inconsistencies” 2 . 2. Svan verbal system: necessary information Svan verbal morphology is complex and highly “synthetic”. It consists of 15 “tenses”, i.e. tense-mood-aspect paradigms (see Gudjejiani & Palmaitis 1986; Testelets 1989) 3 , or, as they are called in the Kartvelian tradition, “screeves” (see, e.g. Harris 1991: 4656) 4 . All the tenses can be characterized as either imperfective or perfective. Most
Transcript

EVIDENTIALITY, TRANSITIVITY AND SPLIT ERGATIVITY:

EVIDENCE FROM SVAN

Nina Sumbatova

Institute of Oriental Studies, Moscow

This article provides a description of the evidential forms in Svan (morphology, syntactic

properties, semantic variants). The most interesting phenomenon concerning the evidential forms

in Svan is their special case-marking patterns. The perfective evidentials form an “inversive”, or

dative, construction (the subject in the dative case, the direct object in the nominative). A

diachronic account for this may be the origin of the evidentials from resultative/perfect forms.

Synchronically, syntactic peculiarities of the evidentials can be explained through the correlation

of evidentiality and transitivity in the sense of Hopper and Thompson‟s transitivity hypothesis.

1. Introduction

Evidentiality and related categories have been the subject of extensive and fruitful

research for the last 10–15 years (see, for example, Chafe & Nichols 1986;

Guentchéva 1996). As DeLancey noticed in his recent paper, “the grammatical

marking of evidentiality, long regarded as an exotic phenomenon found only in a few

obscure languages, has in recent years come to be recognized as a widespread and

significant typological parameter” (DeLancey 1997: 33). We even speak of an “Old

World evidential belt” covering Turkish, Kartvelian, Bulgarian-Macedonian and

Albanian. A part of this belt is the Kartvelian language family.

Kartvelian verbal systems are well-known for their complexity and impressive

syntactic properties. Their most striking characteristic is split ergativity. In the

Kartvelian case split ergativity means that the choice of accusative/ergative/dative

actant coding depends on the tense form of the verb.

This paper deals with the most archaic language of this family, Svan1. Both

Svan and the “main” language of the family – Georgian – have an evidential “perfect”

which together with the pluperfect and the subjunctive perfect constitutes the so-

called “perfect” series of verbal tense/mood/aspect paradigms (in this series, the

subject of transitive verbs is marked by the dative case, the direct object by the

nominative). But only Svan has another evidential group of paradigms, an

imperfective one, with nominative case marking.

The narrow task of my paper is a description of Svan evidentials: their position

in the verbal system as a whole; semantics and usage; syntactic properties of different

evidential forms. But at the same time this paper is aimed at two remoter goals:

suggesting and explaining a typologically correct hypothesis about the origin of the

evidential forms and providing a functional explanation of their syntactic

“inconsistencies”2.

2. Svan verbal system: necessary information

Svan verbal morphology is complex and highly “synthetic”. It consists of 15 “tenses”,

i.e. tense-mood-aspect paradigms (see Gudjejiani & Palmaitis 1986; Testelets 1989)3,

or, as they are called in the Kartvelian tradition, “screeves” (see, e.g. Harris 1991: 46–

56)4. All the tenses can be characterized as either imperfective or perfective. Most

2 NINA SUMBATOVA

perfective forms are marked with a preverb5. The Svan verbal system is presented in

Table 1.

Table 1. Svan verbal system: tenses.6

Aspect Imperfective Perfective

Evidence

Time

Direct Indirect Direct Indirect

Present Present Narrative

present

Perfect

Past Imperfect Narrative

imperfect

Aorist Pluperfect

Subjunctive Subjunctive 1 Narrative

subjunctive

Subjunctive 2 Subjunctive

perfect

Future Imperfective

future

Perfective

future

Conditional Imperfective

conditional

Perfective

conditional

The tenses are grouped into three series that are distinguished according to the type of

actant coding. Three syntactic cases are used to code the nuclear participants of most

situations; their traditional7 labels are nominative, ergative and dative.

8 Most tenses

belong to the first series with nominative actant marking. The aorist and subjunctive 2

tenses form the second series, whose actant marking is most often considered as

ergative (in Table 1, the second series is shaded light-grey). The third series contains

three tenses (dark-grey in Table 1) and is dative, or inversive (for details see Sec.5).9

Thus, the indicative tenses are distributed as follows:

(1) a. Series I: present, imperfect, perfective and imperfective future,

narrative (evidential) present and imperfect;

b. Series II: aorist;

c. Series III: perfect and pluperfect.

One more verbal category to be mentioned here is the “object version”10

, i.e. a

valency-changing derivational category marking the addition of a dative actant to the

verbal frame. Below, if not noted otherwise, the forms of neutral version are meant.

3. Evidentials in Svan.

3.1. Evidential sub-system: remarks on the morphology and formal structure

The “evidential” component of the whole verbal system includes six tenses: these are

three imperfective paradigms labelled by Gudjedjiani & Palmaitis (1986) as

“narrative” and three perfective paradigms that are traditionally called (evidential)

perfect, pluperfect and subjunctive perfect. Below (3.1.2 and 3.2) I shall try to show

that in this case the term “perfect” is only historically appropriate.

EVIDENTIALITY, TRANSITIVITY AND SPLIT ERGATIVITY... 3

Table 2. Evidentials in Svan

Imperfective Perfective

Present Narrative present Perfect

Past Narrative imperfect Pluperfect

Subjunctive Narrative subjunctive Subjunctive perfect

The three imperfective and three perfective tenses show clear formal parallels:

both the “imperfective” and “perfective” column consist of a present, a past and a

subjunctive. The evidential tenses are the only paradigms that contain periphrastic

verb forms.

3.1.1. Imperfective evidentials

Usually imperfective evidentials denote imperfective (i.e. habitual or durative)

situations or events that were not directly observed by the speaker.

(2) (a story about avalanches that had fallen down in the winter of 1986 –

1987 told by a young man from the village of Mulaxi)

amčikka mi mam xwardäs šwäns, mare kämumbwex

at.that.time I NEG be:IMPF:1SG Svania:DAT but tell:AOR:3PL

mäj xola dwrew lmär mulaxs i mtlijänd šwäns...

say bad time be:IEVID:3SG Mulaxi:DAT and whole Svania:DAT

„I was not in Svania at that time, but I heard (=they told) that it was11

a

bad time for Mulaxi and the whole of Svania‟.12

In Gudjejiani & Palmaitis 1986 this group of tenses is labelled “Narrative

Present/Past/Subjunctive”, obviously because it is often used in narrative discourse,

for example in folklore. The term “narrative” is not quite adequate, because the forms

in question are not always narrative and almost never present: usually they refer to the

past13

. Neither is this term traditional, because there is no corresponding tense in

Georgian, and there is no common tradition of labelling Svan forms. That is why I

shall keep using the term “imperfective evidential” (though it is a little awkward); if

not marked otherwise, the “present” tense is referred to.

Imperfective evidentials consist of a special “participle”14

(prefix lm- and in

most cases suffix -(w)in/ -ün) and the copula15

, which is present (1SG: xw-i, 2SG: x-i,

3SG: l-i, etc.), past (xw-äsw, x-äsw...) and subjunctive (xw-esw...) for the present, the

past and the subjunctive evidentials, respectively. Cf. the second person singular

evidential forms of the verb ligem „build‟:

(3) a. lm┘-gäm-└ün (xi)16

IEVID-build (be:PRS:2SG)

„you (SG) were/have been apparently building‟

b. lm┘-gäm-└ün xäsw

IEVID-build be:PAST:2SG

„you (SG) had apparently been building‟

c. lm┘-gäm-└ün xesw

IEVID-build be:SUBJ:2SG

„you (SG) would apparently have been building‟

4 NINA SUMBATOVA

In the present, the copula is optional in the third person plural forms (it can be

reduced to plural suffix -x):

(3) d. lm┘-gäm-└ün lix / lm┘-gäm-└ün-x

IEVID-build- be:3PL IEVID-build-PL

„they were/have apparently been building‟,

and is usually omitted in the third person singular:

(3) e. lm┘-gäm-└ün (*li)

IEVID-build (*be:PRS:3SG)

„he/she has apparently been building‟.

This means that the most common imperfective evidentials are not periphrastic any

longer.

Imperfective evidentials share the nominative case marking of all first series

tenses, but differ from the other tenses as to their verbal agreement. Generally, the

Svan verb agrees both with the subject and the object. There are two rows of

agreement markers, accordingly. There are suffixal and prefixal markers whose choice

is determined by a simple person hierarchy: 1, 2 > 3 (first and second person actants

are equally high and are both higher than any third person actant, see Kibrik 1996)17

.

But in periphrastic imperfective evidentials only the copula can agree. That is the

reason why for these only the subject agreement is possible, even if the verbs are

transitive (whereas all verbs possessing objects, direct and indirect, can agree both

with subject and object), cf.:

(4) a. si e lmbärwin xi

you he wash:IEVID be:PRS:2SG

„you were apparently washing him‟

(the copula xi marks the second person singular of the subject), but

b. e si lmbärwin

he you wash:IEVID

„he was apparently washing you‟ (no second person marker).

Cf., for example, a present non-evidential form:

(4) c. e si -abräli he you 2SG.O-wash:PRS

„he is washing you‟

(the prefix - marks the second person of the object).

3.1.2. Perfective evidentials

In Georgian grammatical tradition, the third series of tenses is called the “perfect

series”, and its three tenses are “(evidential) perfect”, “pluperfect” and “perfect

subjunctive”. The same terminology is usually applied to Svan. Though I am trying to

show here that, from a typological point of view, these tenses should not be treated as

a proper perfect18

, I shall still use the term “perfect” (more precisely, evidential

perfect) sometimes, as it has been done traditionally.

I mentioned above that the Svan “perfect” is essentially an evidential form. In

most cases the components of the “perfect” semantics, if present at all, are

EVIDENTIALITY, TRANSITIVITY AND SPLIT ERGATIVITY... 5

accompanied by an additional semantic burden: that of indirect evidence or mirativity

(see below): in the most neutral case, the perfect denotes a completive action that was

not observed by the speaker in the real world.

For perfective evidentials, there are two productive morphological patterns: the

perfect of “active” verbs (transitive verbs and intransitive verbs with an agent) is

marked by the suffix -a and the object version prefixes -i-/-o, cf. the forms of (5)

ligem „build‟:

(5) a. m-i-g-a

1SG.O-OV-build-PFT

„I have apparently built‟,

b. x-o-g-a / otga (morphonologically, *ad-x-o-g-a)

3SG.IO-OV-build-PFT / *PV-3SG.IO-OV-build-PFT

„he/she has apparently built‟.

Perfective evidentials of passive verbs are treated as periphrastic: they consist of the

passive participle with the confix l----e or the intransitive participle with the confix

me----e and the copula (present, past or subjunctive):

(6) a. algeli (morphonologically, *ad-l┘-g-el-└e li)

*PV-PASS.PART-build be:PRS:3SG

„(it) has apparently been built‟

b. ämqedeli (morphonologically, *ad-me┘-qed-el-└e li)

*PV-ITR.PART-come be:PRS:3SG

„(he) has come‟

c. algel-läsw

build:PASS.PART-be:PAST:3SG

„(it) had been built‟

d. algel-lesw

build:PASS.PART-be:SUBJ:3SG

„(it) would apparently have been built‟.

The periphrastic perfect shows a tendency similar to the one we observed for the

“present” imperfective evidentials: it tends to become synthetic; in the perfect, the

third person copula merges with the participle:

(6) a. algeli (see above)

e. alge(l)-lix

build:PASS.PART-be:PRS:3PL

„(they) have apparently been built‟.

Most perfects are marked with a preverb (here in all examples ad-)19

.

Perfective evidentials show dative marking of the “subject” for all transitive

and active intransitive verbs. The direct object of transitive verbs is marked by

nominative, as in the “ergative” series II, see Table 1 and (1b) above (the indirect

object either remains dative or becomes an oblique NP usually marked with the

postposition -d „for‟ or with the transformative case):

6 NINA SUMBATOVA

(7) medukän-d xäkw: alj-är-s eser lezweb-letre

dukhaner20

-ERG say:AOR this-PL-DAT said food-drink:NOM

loxwmamax i otšxamunax...

eat:PEVID:3SG.S/3PL.IO and spoil:PEVID:3SG.S/3PL.IO

„The dukhaner said: these (DAT) have eaten and spoiled (a lot of) food

and drink (NOM)...‟ (Gudjejiani & Palmaitis 1986: 124)

(8) ču lmbže lix, mäj kä rok ämt’q’wepeli ! PV astonished be:PL what:NOM devil:NOM said burst:PEVID:3SG

„(They) were astonished: what the hell happened there!‟ (lit. „what

devil (NOM) burst‟) (Gudjejiani & Palmaitis 1986:

115)

(9) ...loxt’ula, xola a, txer-äl eser

exclaim:PEVID:3SG.S/3SG.O bad self wolf-PL said

axč’wa dax...

beset:PEVID:3PL.S/3SG.O

„...(The archangel)... exclaimed “Poor me!”, wolves (NOM) had beset

(him)...‟ („beset‟ is an intransitive verb with an indirect object)

(Gudjejiani & Palmaitis 1986: 111)

(10) čottra al xälx säwj-är-d

recognize:PEVID:3SG.S/3SG.O this peolpe:NOM Kabardian-PL-TFRM

„(He) recognized this people to be Kabardians.‟ („recognize‟ is an

inversive21

verb) (Gudjejiani & Palmaitis 1986:

115)

As in other verbal forms, the dative NP controls object agreement irrespective of its

syntactic function (these in (7), omitted he/him in (9) and (10)). The nominative NP

controls subject agreement (food-drink in (7), devil in (8), archangel and wolves in

(9), people in (10)).

If the syntax of the third series were considered on its own, it could be called

ergative. Usually, the subject of an intransitive verb as well as the direct object of

transitives is coded as nominative (Sitr/O NOM), whereas the transitive subject

takes dative marking (Str DAT). However, as far as I know, nobody has ever used

this terminology. The reasons are quite obvious, though some of them are rather

extralinguistic. First, ergative status is ascribed to the second series, which uses a

distinct ergative case. Second, the case of Str in the third series (dative) is widely used

in Svan to mark another core NP, which makes it undesirable to treat it as a “second

ergative” case. Third, the case marking in the “perfect” series is in principle the same

as the marking of the so called “inversive” verbs across all series, which provides the

seemingly convenient term “inversive”, witness (10) and (11).

(11) eas xesmi her

he-DAT hear:PRS:3SG.O voice:NOM

„He (DAT) hears a voice (NOM)‟.

Nevertheless, I find it useful to note the principal similarity of the second and the

third series to the extent that their main typological characteristics (from the point of

EVIDENTIALITY, TRANSITIVITY AND SPLIT ERGATIVITY... 7

view of ergative/ nominative/ active structure) is principally the same, though the

concrete morphological cases marking Str are different22

.

Among the six evidential forms, only “present” tenses are common in

modern Svan. “Past” forms tend to be substituted by the present:

(12) (ašxw-ägis lmzig ešxu mare. ea lmzeläl ltxwijarte, xočamd

xaxlena litwep i lumumgwin xoča c‟ignäls.)

atxad ea lmšijäl (läsw) usurman-äl-caxän.

before that:NOM fight:IEVID (be:PAST:3SG) muslim-PL-with

„(There was a man in a place. He used to go hunting, he could shoot

well and used to tell good fairy-tales.) Earlier he had been fighting

(IEVID:PAST) with the muslims.‟

Omitting the copula in the brackets (which is possible) makes the evidential

form “present”.

Subjunctives are used in certain types of complex sentences only:

(13) o tawrob äxčd al gwešs i ser

then governement:NOM intervene:AOR this affair:DAT and already

čwakwäc, e (hädi ši läjr) šwänjä muzeum-isga

decide:AOR that (Hadish:GEN book:NOM) Svania:GEN museum-in

alšxunel-lesw, i čwädjän mest‟ja-te

be.kept:PFT.SUBJ, and take:AOR Mestia-to

„Then the governement intervened into this affair, and it was decided

that the Book of Hadishi23

should be kept (subjunctive perfect) in the

museum of Svania, so it was taken to Mestia.‟

(Gudjejiani & Palmaitis 1986: 112)

Our further discussion will be limited to “present” evidentials.

3.2. Semantics of Svan evidentials

The main types of indirect evidence are “reported” and “inferring” evidence

(Willett 1988; Kozinceva 1994). Reported evidence can be further subdivided as to its

source (second-hand and third-hand evidence, hearsay, folklore, etc.). In turn,

inferences can be drawn from trivial and non-trivial results of events or from general

regularities met in the world. For Georgian, Willett describes the meaning of the

“perfect” as “non-narrative reported evidence and inference from results” (Willett

1988: 75). The meaning of Svan evidentials seems to be even more general: both

perfective and imperfective evidentials can be used in case of inference as well as

reported evidence of different types.

Evidentials are used, if the speaker wants to stress the fact that he did not

witness the situation/event in question and does not want to be held responsible for its

truth. It is even possible for the speaker not to believe it at all. See (14) as an

illustration.

(14) gärglix ere Dato kor-s lmgämün, mare

say:PRS:3PL that Dato: NOM house:DAT build:IEVID:3SG but

ala bäk li

that:NOM false be:PRS:3SG

„They say that Dato was building a house, but that is not true.‟

8 NINA SUMBATOVA

If the speaker just wants to tell a story, without any special stress on its “second-hand”

origin, he is not obliged to use evidentials, even if the story is obviously fantastical.

Thus, Gudjejiani & Palmaitis (1986: 132–137) adduce two variants of a “nonsense”

story taken from a collection of Svan texts (Svanuri Enis Krest‟omatia 1978). One of

the variants provides a number of evidential forms24

(though not in the whole text, see

example 16), while the other variant shows not a single evidential. For some narrative

texts one can observe an “evidential loss” effect: the speaker begins with evidentials,

but then “forgets” to use them and uses some neutral forms, for example the aorist or

the imperfect. Then, at the beginning of a new episode or after mentioning the source

of the story, the speaker may remember again that the story is not of his own and

return to evidentials. He may also use an evidential at the very end of the story to

remind the reader of its “reported” status.

The difference between imperfective and perfective forms is generally the

same in the evidential subsystem of the Svan verb as in other, non-evidential, tenses:

the imperfective covers durative (progressive) and habitual usages, whereas the

perfective is left for completive events. A few more difficult cases (perfect,

resultative, experiential) will be discussed presently.

3.2.1. Indirect evidence

The two main information sources of indirect evidence are inference and verbal

information. Svan does not keep apart these two principal sources (or any subtler

divisions) in its grammar. Both perfective and imperfective evidentials are possible in

either case.

3.2.1.1. Reported evidence

Verbal information can be obtained from another person, hearsay, in written texts,

folklore, etc.

(15) (following (2) in the same story)

xwae mus-i bädšw mankana-le lazelal mama

much snow-GEN because.of car-LOC place.to.go:NOM NEG

lmär i merme sopl-är-xän čušx-wš lmmäzx

be:IEVID:3SG and other village-PL-from foot-INST go:IEVID:3PL

žamišw-te lamarw-te

Zhamushi-to help-to

„Because of much snow, there were no passage for the trucks, and

(people) were walking from other villages to Zhamushi to help.‟

Since a natural source of evidentials is verbal information, they are quite usual in

folklore and any type of stories about remote past events.

(16) ašxw uwä p‟il-s qän lmgen i ašxw

one sea:GEN shore:LOC/DAT ox:NOM stand:IEVID:3SG and one

p‟il-xän sgwebne č‟šx-är xägnena i merme p‟il-xän –

shore-from front leg-PL stand:EVID:OV:3SG and other shore-from

wešgmeš...

back:NOM

EVIDENTIALITY, TRANSITIVITY AND SPLIT ERGATIVITY... 9

„An ox was standing at a sea shore, and his front legs were standing on

one shore, and his back (legs) – on another shore.‟

(Gudjejiani & Palmaitis 1986: 132)

(17) eč zaw lcobi gun lmär i kor girk‟id

that year:DAT flood:NOM-and many be:IEVID:3SG and house round

gim mäg katt’q’eca.

soil:NOM all:NOM wash.away:PEVID:3SG.S/3SG.O

„There were many floods that year, and they washed away all the soil

around the house‟. (Gudjejiani & Palmaitis 1986: 128)

3.2.1.2. Inference

Both perfective and imperfective evidentials are freely used in cases of inference as

well: the speaker infers the fact of a situation/event having or having not taken place

from some indirect data, and in the most usual case these data are results of the

situation or event. In the case of imperfective evidentiality, the inference is made from

a non-trivial, indirect result: since the forms in question are imperfective, the

component of an achieved (direct) result is not a part of their meaning:

(18) (about a person with red eyes)

ea lumgwänin

he/she cry:IEVID

„She/he has been crying.‟

(19) (said to somebody who looks tired and sleepy)

si des lumwi z-xi

you NEG sleep:IEVID-2SG

„You haven’t been sleeping (at night).‟

With perfectives, which refer to a completive event, the situation is different: if the

result is indirect, there is no striking difference from the imperfective forms:

(20) (approaching his house the speaker sees light in his windows and

guesses that his wife has probably come home)

Nona ämqedeli Nona:NOM come:PEVID:3SG

„Nona has come.‟

In case we are dealing with a trivial result (following directly from the semantics of

the verb), which is still present at the moment of the speech (otherwise no inference

could be made), we observe a kind of combination of evidentiality and a resultative

(perfect) meaning:

(21) a. (the speaker comes home and finds ready dinner on the stove)

xexw-s oxmara sädil

wife-DAT prepare:PEVID:3SG.S/3SG.O dinner:NOM

„The wife has (apparently) prepared dinner!‟

10 NINA SUMBATOVA

This sentence could not be used if the speaker had been sitting in the kitchen and

observing the process of preparing dinner – in that case he would have used a sentence

with an aorist:

(21) b. xexw-d anmare sädil

wife-ERG prepare:AOR:3SG.S/3SG.O dinner:NOM

„The wife has (already) prepared dinner!‟

Evidentials presuppose that the speaker does not possess direct knowledge of the

situation. Even if the speaker participated in the referential situation, but cannot

remember it himself and uses information about it obtained from some indirect

source, the perfect can be freely used. For example, the perfect tense can be used when

speaking about the speaker‟s early childhood that he “remembers” only through the

words of other, older people:

(22) xoxrob-ži mi lumgwänin-xwi;

childhood-in I cry:IEVID-1SG

mi ämbina ligrgäli do sg;

I begin:PEVID:1SG speak:MASD early

mišgu mänq‟wi sit‟q‟wa lmä r dede

my first word:NOM be:IEVID:3SG mummy.

„When I was a baby, I used to cry; I started talking early; my first word

was “mummy”.‟

The evidential is also possible when speaking about actions fulfilled by the speaker in

some unconscious state (being drunk, sleeping, unconscious, in hypnotic sleep, etc.):

(23) (said by a person who wakes up after being drunk and sees that the

door is broken)

mi leti q‟or čwamq’učura

I at.night door:NOM break:PEVID:1SG.O/3SG.O

„At night I (apparently) broke the door...‟

(24) (after being drunk, the speaker was told what he was doing in this

state)

mi lumira l-xwi i lmcral-xwi

I sing:IEVID-1SG and swear:IEVID-1SG

„I was singing and swearing.‟

3.2.2. Dreams

A dream is a case of direct evidence of some events that are later realized as being not

real: the speaker remembers what he has “seen”, but understands that this was not a

real-world event. It is interesting that in Svan only perfects, but no imperfective

evidentials can be used when retelling dreams.

(25) a. mi lämi snaw ere txere ämdagra

I dream:AOR:1SG that wolf:NOM kill:PEVID:3SG.S/1SG.O

„I dreamt that I killed a wolf.‟

b. ...ere mulax-ži-kanka xwiperialdäs (*lmperia l-xwi)

...that Mulaxi-upward-along fly:IMPF:1SG (*fly:IEVID:1SG)

„...that I was flying over Mulaxi.‟

EVIDENTIALITY, TRANSITIVITY AND SPLIT ERGATIVITY... 11

(26) ala-s läxi snaw, esnär taringzel eser ämqedeli this:DAT dream:AOR:3SG as-if archangel said come:PEVID:3SG

hädi š-žikante-ka i loxt’ula, xola a,

Hadishi-upward-through and exclaim:PEVID:3SG.S/3SG.O bad self

txer-äl eser axč’wadax...

wolf-PL said beset:PEVID: 3PL.S/3SG.O

„He dreamt as if the archangel had come through over Hadishi and

exclaimed “Poor me!”, wolves had beset him...‟

(Gudjejiani & Palmaitis 1986: 111)

“Imperfective” events in a dream are expressed by the imperfect (as in (25b).

In this case a semantic difference between imperfective evidentials and

perfective evidentials can be noticed: the latter points out that the situation has not

been attested by the speaker in reality. The imperfectives stress the fact that it is not

part of the “immediate” knowledge of the speaker, i.e. knowledge based on the

speaker’s own experience.

3.2.3. Mirative meaning

Mirativity is a semantic category expressing the opposition of new, surprising

information (situations for which the speaker provides “no psychological preparation”,

DeLancey 1997: 35) and the old, mastered, accepted one. Since mirativity is often

expressed by the same grammatical devices as evidentiality, it has often been treated

as a variety of indirect evidence, though in some cases mirativity markers are

combined with situations of the most direct evidence possible; DeLancey (1997: 39)

gives some examples from Hare (Athabaskan):

(27) Heee, guhde daweda! Ch‟ifi dach‟ida lo ! hey up.there sit guy sit

„Heey, he is sitting up there! The guy is sitting up there!‟

(said in a situation when the speaker suddenly notices it; lo is the mirativity

marker).

As DeLancey has shown in general, mirativity is an independent semantic category

that sometimes possesses grammatical marking of its own. In Kartvelian, as in some

other languages (Turkish, Korean, etc.), mirativity is bound together with

evidentiality.

A mirative component is an optional, but frequent part of many usages of

evidential forms. (28) and (29) are from Svan again.

(28) (The speaker sees the addressee‟s certificate of the institute

graduation:)

si institut igra-ja !

you instiutute:NOM finish:PEVID:3SG.S/2SG.O-EMPH

„You have graduated from an institute!‟

(29) (The speaker hears the addressee is playing the guitar)

isgowd xočamd oxwtorax gitara-ži

your:TFRM good:TFRM teach:PEVID:3SG.S/3PL.O guitar-up

12 NINA SUMBATOVA

lišwme

play:MSD:NOM

„You (apparently) have been well taught to play the guitar!‟

But there also exist certain situations when the mirative appears as a nuclear

component of the meaning of a verbal form. This happens when the speaker faces a

highly surprising situation – a situation that he did not, and could not, expect to come

about. This is probably the only case when the so called “narrative present” (in the

form of the imperfective evidential present) is not past, but really present, or at least

not time-bound (“general present”):

(30) (The speaker hears a friend of him is playing the guitar – the speaker

did not expect him to be able to)

si gitara-ži lumšwemin-xi !

you guitar-up play:IEVID-2SG

„You (surprisingly) can play the guitar!‟

This situation type (present and therefore not completive) can of course be marked by

imperfective evidentials only.

3.2.4. Mirative resultative and evidential perfect

In some cases, perfective evidential forms can express a resultative or perfect

meaning. This variant of meaning has already been touched upon (3.1.2).

The dimension of the perfect is covered in Svan by the aorist tense and the

evidential perfect. By using the perfect the speaker stresses the fact that he observes

the existing result of a completive event, but has not observed the process as such. The

perfect, thus, still retains its evidential meaning. The aorist, which is a simple form of

the past perfective, is used in neutral situations, including those when the speaker has

observed both the event and its result (21b).

The intransitive perfect can also be used as a resultative, but in this case it

stresses unexpectability and surprise being therefore mirative:

(31) (the speaker tries on a dress that she has not been wearing for a long

time and finds that now it is too loose)

mi čwamčxepel-xwi

I lose.weight:PEVID-1SG

„I (seem to) have lost weight!”

3.2.5. Experiential

Perfects (as in (32) and (33)) and sometimes also imperfective evidentials (34) can be

used with experiential meaning. Experiential meaning stresses the fact of the

situation/event being attested at some indefinite time in the past: the very action

having taken place is important, not its temporal reference or any accompanying

circumstances. Experiential meaning is the only variant of those we have treated here

that cannot be called evidential or mirative at all.

(32) a. si šomwale ima esa k‟ubdär ?

you ever eat:PFT: 3SG.S/2SG.O Q kubdar

„Have you ever eaten kubdar25

?‟

EVIDENTIALITY, TRANSITIVITY AND SPLIT ERGATIVITY... 13

b. mi dejšoma mima k‟ubdär.

I never eat:PFT:3SG.S/1SG.O kubdar

„I‟ve never eaten kubdar.‟

(33) mi al läjr (ješdin) ka26

mič’wdana.

I this book:NOM (ten.times) PV read:PFT:3SG.S/1SG.O

„I‟ve read this book (ten times).‟

(34) si moskow-s lmärd-xesa ?

you Moscow-DAT be:IEVID-2SG+Q

„Have you ever been to Moscow?‟

As illustrated in (32) and (33), only non-preverbials forms of the perfect are used with

experiential meaning. At the same time they were not possible with the pure evidential

meaning treated above (3.2.1 – 3.2.2).

It means that non-preverbial perfect forms are not evidential at all. In many

languages the experiential tends to be expressed by forms of the perfect (the simplest

example of it is English). That is why the experiential meaning of the Svan non-

preverbial perfects may be treated as a link between the traditional label of the forms

in question (“perfect”) and the most frequent semantic interpretation of the other

“perfect” forms (evidential).

Frequently, non-preverbial perfects refer to repetitive situations/events with

the same shade of experientiality. See (35)–(36).

(35) di-s gezläš-d xwäin xobca

mother-DAT child:GEN-for many.times promise:PFT:3SG.S/3SG.O

sačkwär

present:NOM

„Mother has many times promised her child a present‟ (but, quite

probably, didn‟t buy it).

(36) (said about a person who has many times, insistantly invited people to

his dukhan)

ala alj-är-s atmbažnx: dos eser

this:NOM this-PL-DAT surprise:AOR:3SG.S/3PL.O nobody said

xasma amži n, čik eser ka xoc’sa i atxe

hear:PRS:3SG/3SG thus time said PV invite:PFT:3SG.IO and now

pasw xešgwem

money:DAT request:PRS:3SG

„They were amazed at this: “Nobody has ever heard that – he has first

invited (people) and now requires money!”‟

(Gudjejiani & Palmaitis 1986: 124)

In such cases, the plurality of situations/events correlates with the absence of a

preverb, which is a usual signal of imperfectivity. At the same time, being a part of the

perfect series, the non-preverbial perfect refers to a completed situation/event.

Nevertheless, the non-preverbial perfect can not be called a true perfect (even if we

accept iterative perfects), since it does not imply the presence of the action‟s result.

14 NINA SUMBATOVA

The non-preverbial perfect of transitive verbs is no doubt an independent

tense. Both its morphological structure and its meaning are different from that of the

preverbial perfect. This is one more tense of the third series.

The intransitive non-preverbial perfect is a simple combination of the passive

participle and the copula – a syntactic construction with resultative meaning (see

section 4).

4. Historical connotations: evidentials as grammaticized resultatives

The typical historical sources of evidentials are perfects and, more generally,

resultative constructions (Willett 1988, Bybee & Dahl 1989). This development is

explained by the ambivalence of the perfect. On the one hand, the perfect refers to a

completed (hence, past) event. On the other hand, the perfect implies some state

resulting from completion of an action or an event. The first, actional, component of

the perfect is a source of changing perfects to perfectives and simple past forms.

Emphasizing a connection between an (existing) result and a (completed) past action

leads to evidentials with primarily an inferential meaning.

A similar origin is traced for evidential perfects in Georgian (Natadze 1955,

Machavariani 1983). The (evidential) transitive perfect in Georgian originates from

the object version of the stative/resultative forms as a result of preverb adjustment and

structural reinterpretation. Cf. from modern Georgian (example from Machavariani

1983: 275):

(37) a. Bebia-s t‟axt‟-ze pardag-i ug-i-a

grandmother-DAT ottoman-on carpet-NOM spread-STAT-3SG

„Grandmother has a carpet spread on the ottoman.‟ (in this

sentence, the dative NP bebia-s „at grandmother‟s, to

grandmother‟ is added as a result of the object version

derivation)

b. Turme bebia-s es pardag-i tviton

apparently grandmother-DAT this carpet-NOM self

da-ug-i-a

PV-spread-PFT-3SG

„Grandmother has apparently spread this carpet herself.‟

The stative verb in (37a) differs morphologically from the evidential perfect in

sentence (37b) only by the presence of the preverb da-. Case marking is the same in

both sentences, but for (37b), the dative NP is the syntactic subject (which can be

established through the usual syntactic tests; see, e.g., Harris 1981).

Intransitive perfects in Georgian result from merging participles with the

auxiliary in resultative construction (Machavariani 1983: 275):

(38) da-mal-ul-i var da-v┘-mal-ul-└var

PV-hide-PASS.PART-NOM be:1SG PV-1SG-hide-PFT

„I am hidden‟ „I have apparently hidden (itr)‟

The development of evidentials in Svan should be similar. The resultative

origin of evidentials is supported by the morphological structure of evidentials as well

as their syntactic properties and semantics.

EVIDENTIALITY, TRANSITIVITY AND SPLIT ERGATIVITY... 15

Indeed, the usage of evidentials shows some remnants of the perfect semantics

(see sections 3.2.1 and 3.2.4). Another semantic argument connecting evidentials with

the perfect is the experiential meaning of non-preverbial perfects (see 3.2.5).

Turning to the morphological structure of the evidentials, we find some

structural indications such as the following ones:

(39) a. some of the perfect forms are periphrastic (or, at least, were

periphrastic in the recent past) consisting of a participle and the

copula; the participle of the intransitive perfect is a resultative

(past passive) participle;

b. the forms of the transitive perfect are connected with the object

version of stative verbs in the present (Natadze 1955; in Svan,

the stative differs from perfect by the presence of the suffix -a);

c. the pluperfect and subjunctive perfect show structural

parallelism with the forms of the intransitive aorist (they use

the aorist suffixes -än for the pluperfect and -en for the

subjunctive. See the forms of the verb „build‟ (40).

(40) a. atgen-än

build: PLUPFT:3SG

b. atgen-en-s

build:SUBJ.PFT-3SG (Natadze 1955)

The clearest parallels with resultative constructions, however, are syntactic. In modern

Svan there exist several types of resultative “participles”27

. The simplest case is the

“passive”, i.e. resultative participle of the type l----e referring to the object of a

completed transitive action (lrme „caught‟, lkwe „said‟). With the copula these

participles form a usual object resultative construction:

(41) lezweb lmare li

food prepare:PASS.PART be:PRS:3SG

„The food is prepared‟

(42) lerkwel lušq‟wed li

clothes wash.up:PASS.PART be:PRS:3SG

„The clothes are washed up‟, etc.

This construction is analogous to that in Georgian (see above in this section) and

structurally differs from clauses with an intransitive perfect only by the absence of a

preverb and a higher degree of the copula‟s independence (in nominal sentences

including the type of (41) and (42) the copula can neither be reduced nor omitted).

Furthermore, in Svan there exists a “secondary” resultative construction with

the object version of the verb „to be‟:

(43) mi lerkwel ču lušq‟wed m-ar

I clothes:NOM PV washed:PART 1SG-be:OV

„I (DAT) have the clothes (NOM) washed.‟ (Cf. German „Mir sind die

Kleider gewaschen.‟)

Constructions of the latter type have been called “secondary possessive resultatives”

by Nedjalkov and Yakhontov (1983: 25–26). These constructions often develop into

transitive perfects (Nedjalkov & Yakhontov 1983: 25–26) as a result of their syntactic

16 NINA SUMBATOVA

reinterpretation. It is very probable that the already existing “perfect” emerged in a

similar way.

On the other hand, participial constructions of this type are a common source

of arising ergativity. This is the case with Indo-Iranian split ergativity, for example,

where Old Persian constructions of the type

(44) mana kartam astiy

I:GEN done is

„I have (it) done‟

were reinterpreted in Middle Persian as „done by me‟ an then as „I did‟. Thus, the

indirect case of „I‟ originates from the genitive28

(Pirejko 1979). In many languages a

similar development creates ergative constructions where the ergative case derives

from the former genitive, dative, or locative. A prerequisite of this development is that

the language does not have a verb „have‟ (or, at least, does not use it in these

constructions). Recall that I stressed the principal similarity of “ergative” and “dative”

case marking in Kartvelian. In this case, the dative plays in Svan the role of a new

ergative.

As far as imperfective evidentials are concerned, they still consist of a

participle plus the copula. A natural presumption on the original meaning of the

corresponding participle will be that of subject resultative. This could also be a simple

explanation of their nominative syntax. If perfective evidentials can be compared with

resultative participles of the type l----e, imperfective evidentials are comparable to

another participle type: these are “resultative” participles29

(marked with the prefix

na-) with a rather difficult semantics that can be roughly formulated as “what is left

after the situation/event having taken place”. These derivates, thus, deal with indirect

results of the situation/event and in this sense are “imperfective” resultatives (cf.

3.1.1):

(45) a. ligwni na-gwän

„cry‟ „somebody who has traces of tears on his/her face‟;

b. liq‟wre nä-q‟win

„lie‟ „a place (a bed) where somebody has been lying (and

one can see this)‟.

In some cases, the difference between l----e and na- participles is very slight, but for

most transitive verbs it is clear:

(45) c. libeč‟k‟w lbč‟k‟we

„explode (tr)‟ „something that has been exploded‟

näbič‟k‟w

„remnants of the explosion‟.

It looks quite probable that the “old” resultatives in Svan drifted from

resultative/perfect meaning to expressing evidentiality. As a result, the vacant

“resultative” area has been occupied by the new resultatives, i.e. periphrastic

constructions with participles, which are to a certain degree structurally analogous to

the evidentials. In speaking of the Svan evidentials we therefore are dealing with very

old resultatives that are now forming quite new ergative constructions.

Wrapping up, in Svan we register three or even more different steps of

grammaticization of resultative constructions:

EVIDENTIALITY, TRANSITIVITY AND SPLIT ERGATIVITY... 17

(46)

less grammaticized

“true” resultative constructions

experiential (non-preverbial) perfect

perfective evidentials ── retaining a resultative meaning

── “pure” evidentials (narratives) and

imperfective evidentials

more grammaticized

5. Synchronic problems and explanations: inversion and transitivity

As already mentioned above, the grammaticization of perfective transitive evidentials

is accompanied by syntactic reanalysis: the former indirect object (dative) becomes the

subject, whereas the former subject comes to be the direct object. An analogous

development is observed for Georgian. But in Svan the syntactic changes are not so

crucial, since the formal properties of the subject and the direct object are not very

clearly expressed in this language at all. The status of the dative as “subject” is

supported by reflexivization rules in the first instance. See (47).

(47) ea-s otdagra miča txwim (*ea otdagra miča txwim-s)

he-DAT kill:PFT his head-DAT he:NOM ... ... head-DAT

„He has apparently killed himself (lit. „his head‟, NOM)‟.

However, other rules (such as causativization or conjunction reduction) show no

predominance of the putative subject. It is the role properties of the participants that

are immediately reflected by these syntactic rules. In previous work (Sumbatova 1993)

it was shown that looking at Svan we are dealing with a highly role-dominated (rather

than syntactically-driven) language. This characteristic of the language is to a certain

degree supported by its verbal classification, which can be characterized – with certain

reservations – as role-based. In this language the dative is not just a label of some

opaque syntactic relation: it has rather clear semantic contents expressing the

traditional roles for indirect objects (experiencer, benefactive, recipient: a person

affected by the situation or interested in its results). This is the usual semantic contents

of the object version dative30

.

Therefore, the very existence of the third series transitive verbs, where the

dative rather than a typically agentive case marks an agent, looks rather strange. I

believe that if this construction had violated the major structural principles of the

language to such a high degree, it would have made the evidential development of

resultatives impossible, or it would have changed its syntactic properties. That is why

we need to explain how it is possible for a dative to express agents in Svan.

Before suggesting any ideas explaining the functioning of the dative case, we

have to return to the main types of actant marking in Svan (see Table 3). Traditionally,

four syntactic verbal classes are differentiated:

Class 1: transitive verbs (including ditransitives), e.g. lidgäri tr „kill‟, lihwdi „give‟;

Class 2: active intransitives (also called “medial” verbs) including bivalent verbs with

a dative object, e.g. ligrgäli „speak‟, libce „promise‟;

Class 3: monovalent and bivalent “passive” verbs (also called deponent verbs), e.g.

lidgäri itr „die‟, lihne „melt‟, lisgdi „look‟; and, finally,

Class 4: inversive verbs, e.g. lipeš „be tired‟, liter „see‟.

18 NINA SUMBATOVA

Each class has its own actant marking strategy in three verbal series. Witness

Table 3.

Table 3. Strategies of actant marking.

Series I II III

Class

1 NOM DAT (DAT) ERG NOM (DAT) DAT NOM (OBL)

2 NOM (DAT) ERG (DAT) DAT (OBL)

3 NOM (DAT)

4 DAT (NOM)

From Table 3 one can deduce that there exist two principal types of Svan verbs as far

as their syntactic properties are concerned: the verbs changing their case frames

(Classes 1 and 2) and those retaining the same actant marking throughout all series

(Classes 3 and 4). Generally speaking, agentive verbs (those with an agent-like

participant) are frame-changing, while non-agentive verbs are stable. This conclusion

is not without exceptions, though, since some verbs of class 3 (e.g., some formally

reciprocal verbs with the suffix -(j)el-, like liq’ärjel „fight‟ or limkaräl „compete‟) are

highly agentive. Nevertheless, there exists a strong correlation between the agentivity/

non-agentivity of a verb and its syntactic class.

If we take into account the case marking of core arguments only, we will be

able to differentiate five types of core arguments as shown in Table 4:

Table 4. Argument types.

Series I Series II Series III

Type 1 NOM ERG DAT

Type 2 DAT NOM NOM

Type 3 DAT DAT OBL

Type 4 DAT DAT DAT

Type 5 NOM NOM NOM

The first type of argument can be characterized as the most agent-like NP. It is the

first actant of transitive and agentive intransitive verbs (for example, verbs of speech

and sound production, verbs with an addressee like liqwämjel „thank‟, libce „promise‟,

limarZ wi „help‟, some verbs of motion). Let us label this type of argument

„effectors‟.

Arguments of type 2 are the most patient-like ones in the system. This group

includes the most effected participants of transitive verbs only; they are labelled

„patients‟.

Groups 3 and 4 are both semantically heterogeneous and, at the same time,

similar to each other. Group 3 includes indirect objects of ditransitive verbs („give‟),

recipients, all NPs being introduced into the verbal frame by the object version

derivation (most of them can be characterized as benefactives). Group 4 contains

mostly experiencers (which are typical of inversive class 4 verbs like liter „see‟, lisme

„hear‟, lipeš „be tired‟) as well as some benefactives and recipients. The common

property of these two groups is that the arguments are neither very agentive nor

patient-like. They neither effect, nor are they effected by other arguments. Rather, they

EVIDENTIALITY, TRANSITIVITY AND SPLIT ERGATIVITY... 19

are often affected: many of them are animated and even personal. Type 3 and 4

arguments we label here „goals‟.

Finally, group 5 includes, on the one hand, patients of intransitive verbs

(which are not strongly “patientive”, in the first place, since they do not undergo any

overt influence of another participant and change only because of their inner

properties or general laws of nature) and, on the other hand, participants that can be

characterized as stimuli and objects of perception and reflection (the names of

objects that are seen, felt, thought about). I would add here the single argument of

some states and properties expressed by a nominal predicate. This role is called theme.

The four semantic roles differentiated here can be ordered according to their

agentivity/effectedness (see also Foley & Van Valin 1984: 59):

(48)

Effectedness increasing Agentivity increasing

Patient Theme Goal Effector

In the series II and III, this scale is shared by different cases like this in (49ab):

(49a, b)

Effectedness increasing Agentivity increasing

Patient Theme Goal Effector

a. Ser.II

NOM DAT ERG

b. Ser.III ─

NOM DAT

It will be a commonplace remark if I say that case marking is determined by the

interaction of certain semantic and pragmatic factors. In Svan, semantics (and more

precisely, role semantics) of arguments is at least as important as their pragmatic

weight.

Semantic factors are principally sufficient to explain case marking in the

second and third series. If we assume that the ergative is a case that marks the highest

agentivity degree of a NP, whereas the nominative reflects the lowest degree of

agentivity and a high degree of effectedness, and the dative refers to the middle of this

scale, this will be enough to explain the marking in series II: only actors are agentive

enough to be marked by the ergative, whereas goals get a dative marking, and the

nominative covers themes and undergoers. From the three series of tenses, the second

series can be regarded as basic. This series separates three types of core arguments.

The case marking of the other series can be predicted from the case marking in the

second series (but not vice versa). Besides, the main tense of the second series, the

aorist, shows the highest degree of transitivity in the sense of Hopper & Thompson

(1980). It is the only past perfective verbal tense.

The average transitivity of the third series should be regarded as somewhat

lower, because its verbal forms are evidential or at least iterative (non-preverbial

perfect). Indirect evidence is not included in the list of Hopper and Thompson‟s

transitivity parameters. But it is obvious that evidentiality (the source of information

about some situation or event) is closely connected to the parameter of reality.

20 NINA SUMBATOVA

According to Hopper and Thompson (1980), the irreality of a situation decreases its

transitivity. Indirect evidence situations are treated as “less real” than “really unreal”

ones; the speaker does not affirm that they have not been attested in the real world,

but, at the same time, he cannot prove their truth because he has not witnessed them.

That is probably why even the actor in the third series is not agentive enough to be

marked by the ergative. In the third series, the actor conjoins the roles typical of dative

marking. Thus, the dative in the third series is not so extraordinarily unnatural as it

seemed at first sight.

Let us now turn to the first series as in (49c).

(49c)

Patient Theme Goal Effector

c. Ser.I ─

DAT NOM DAT NOM

To be sure, the description and explanation of the first series is not a part of our task.

However, since it seems to strongly undermine the general “role-expressing” tendency

of Svan, an explanation should at least be hinted at. I think that the first series (a

relatively young paradigm) illustrates the growing role of pragmatic factors. The

nominative tends to be the most pragmatically marked case. It is used in the “naming”

function. It is able to control both subject and object agreement under certain

circumstances. And, in most cases, it controls reflexivization as well. In the first

series, its pragmatic force comes to play a major role: the nominative is the case of the

first core argument, which could even be called a syntactic subject. The pragmatic

markedness of the actor in the first series is more important than its high agentivity.

This is even more plausible if we recall that the first series shows the lowest degree of

transitivity in general, because it contains only imperfective (atelic) and future tenses

including imperfective evidentials (not to mention the conditionals).

It is pragmatic factors that determine the choice of a dative NP in case of

conflict. In the third series (classes 1 and 2), it is both the logical subject and the

indirect object that can get dative marking. But the logical subject, being more

important pragmatically, occupies the dative position. This is why the semantically

close argument types 3 and 4 have been consistently distinguished in Series 3.

A similar situation prevails with the case marking of reciprocal verbs. All of

them belong to class 3 (intransitive passives) even if their symmetric participants are

highly agentive (like librgjel „wrestle‟, liq’häl „kiss‟). They are used either as

monovalent verbs with the single stable nominative actant (a plural or coordinative

NP) as in (50a);

(50) a. Vano i Givi ibrgjel-x

Vano:NOM and Givi:NOM wrestle:PRS-3PL

„Vano and Givi are wrestling‟;

or they are used as bivalent stable verbs with a nominative and a dative actant, like

(50) b. Vano Givi-s xebrgjel Vano:NOM Givi-DAT wrestle:PRS:OV:3SG

„Vano is wrestling with Givi‟.

EVIDENTIALITY, TRANSITIVITY AND SPLIT ERGATIVITY... 21

It is possible that the agentivity of symmetric counter-agents is never high enough to

“deserve” ergative marking. Thus, they are prototypically treated as goals. However,

at least one of them should be chosen as the subject and marked with the nominative.

If they are equally important for the speaker, a monovalent verb will be used.

Pragmatic factors are more consistent for verbal agreement. The logical subject

(the argument mentioned first in Table 3) controls subject agreement in all cases, but

with two major exceptions: inversive (mostly experiential) verbs in all series, and all

changing verbs in series 3. For these verbs, the logical subject (in the dative) controls

object agreement, whereas the nominative NP (if any) controls the subject agreement.

It is interesting that in series 3 the strict rules of agreement become much

looser (Kibrik 1996). Whereas the agreement prefixes follow the standard agreement

rules, the suffixes tend to erase the distinction between subjects and objects. Usually,

subjects are preferred suffix agreement controllers (see 3.1.1 and footnote 17). In the

third series, if the dative and the nominative actant are equally high in the hierarchy,

the choice of the agreement suffix is, at least in theory, controlled by the nominative

NP that is treated as a subject. But in many cases the dative actant optionally takes

control over the suffix. See again (7).

(7) ... alj-är-s ... lezweb loxwmamax

... this-PL-DAT ... food:NOM eat:PEVID:3SG.S/3PL.IO

( morphonologically, *la-x-o-mam-a-x)

PV-3PL.IO-OV-eat-PFT-3PL.IO

„... these (DAT) have eaten (a lot of) food (NOM)...‟

The theoretically predicted form would be loxwmama with the zero sufix of the third

person singular subject. Analogous deviations are also observed in examples (29) and

(36). This tendency can be interpreted as a signal of the growing pragmatic role of the

“logical” subject (actor, effector) that is expressed by the dative in the third series.

6. Conclusions

In the foregoing sections, we have discussed different questions connected with the

evidential forms in Svan, including their formal morphology, syntactic patterns,

semantic variants as well as some points of their evolution.

Evidentials constitute a special subsystem of the Svan verbal system. More

than that, the evidential subsystem is more regular than the verbal system taken as a

whole (Tables 1 and 2). Most lines and both columns of Table 2 have their specific

semantic features and morphological markers (3.1).

Morphologically, the evidential subsystem looks like the newest development

of the verbal system. Thus, only the evidential segment of Table 1 contains

periphrastic verbal forms. However, at the same time one can observe a clear tendency

to avoid the periphrastic forms: “past” evidential forms tend to be substituted by the

“present”, while the “present” tends to become synthetic (the copula being omitted or

morphologically merged).

Evidentials in Svan have a rather general semantics (see a short summary of it

in Table 5 below): they cover both reported and inferred evidence as well as the

mirative. We have observed an unexpected difference between imperfective and

perfective evidentials: only the latter can be used to narrate dreams. We have observed

that only the preverbial perfects have an evidential semantics, whereas the non-

22 NINA SUMBATOVA

preverbial forms are mainly used with an experiential meaning, which has made us

distinguish the preverbial and non-preverbial perfect as two distinct tenses.

Table 5. Semantic variants of evidential and perfect forms.

Semantic variant Imperfective

evidential

Perfective evidential

(preverbial perfect)

(Non-preverbial

perfect)

Reported evidence + + –

Inferred evidence + + –

Inferred evidence +

perfect

– + –

Dreams – + –

Mirative (present) + – –

Mirative + inferred

evidence (past)

+ + –

Mirative + resultative – + –

Experiential – – +

The evidentials in Svan move along a typical grammaticization channel resultative

perfect inferred evidence (inferred and reported) indirect evidence. New

resultative constructions (Sec. 4) have come to replace old resultatives, which have

changed their original meaning. They are now probably at the beginning of such a

long path.

Imperfective and perfective evidentials in Svan are crucially different as to

their syntax: whereas the imperfectives belong to the first series of tenses with

nominative case marking, the perfective evidentials (series 3) show a specific case

marking pattern traditionally called dative, or inversive. This case marking pattern,

which is essentially equivalent to the ergative type, allows a diachronic explanation. If

we assume that the perfective (transitive) evidentials in Svan originate from a certain

type of resultative constructions, namely “secondary resultatives” (Sec. 4), this will

account for the dative case of the agent in the third series (inherited from the dative of

the benefactive in the resultative construction). With some reservations, imperfective

evidentials can be supposed to originate from subject resultatives.

A diachronic explanation of the syntactic peculiarities of the evidential forms

does not seem to be sufficient. It would be useful to find synchronic principles

explaining the numerous case marking patterns and the obscure verbal classifications

in Svan. I suggested that an explanation could be provided by the transitivity

hypothesis suggested by Hopper and Thompson (1980). Suppose that the three nuclear

cases in Svan correspond to certain degrees of the actant‟s agentivity (ERG > DAT >

NOM). This could account for case marking in the second and the third tense series.

The ergative case is used in the second series only, because the degree of transitivity

of the verb (and the agentivity of the first actant) is higher in the second series. The

first series is more pragmatically oriented: the subject is in the nominative case, which

is the pragmatically most important case in Svan (used as the case of nomination as

well). The high pragmatic status of the nominative case is supported by its

semantically motivated use in nominal and stative sentences.

Diachronically, the second series is the oldest, while the perfect series is the

youngest. This means that the syntactic and semantic development of the verbal

system in Svan follows a kind of spiral in the following sense. See (51a–e).

EVIDENTIALITY, TRANSITIVITY AND SPLIT ERGATIVITY... 23

(51) a. the oldest series 2 is typologically ergative (see footnotes 9 and

22); case marking and many syntactic processes are governed

by the role domination principle (Foley & Van Valin 1984);

b. series 1 is typologically accusative and more pragmatically

oriented;

c. series 3 shares its main typological characteristics with series 2

(though the concrete morphological devices are different: the

dative is used as the case of the transitive subject). However,

for series 3, we already observed that the role of pragmatic

factors is growing also (Sec.5).

NOTES

1. Svan is a language of the Kartvelian family spoken by more than 30 000 people in the North-

West of Georgia. Svan is traditionally divided into two main dialect clusters including four

dialects: Upper-Bal and Lower-Bal are the Upper-Svan dialects, Lentekh and Lashkh are

Lower-Svan. This paper deals with the most archaic Svan dialect, the Upper-Bal.

2. I would like to express my sincerest gratitude to David and Nona Kochkiani whose kindness,

fine feel for the language and love of their mother tongue, Svan, helped me to prepare this

paper.

I owe a lot to the editors of the present volume, Werner Abraham and Leonid Kulikov, who

did hard work to improve both the contents and the style of this paper. I am also grateful to

Nick Nichols and George Hewitt who read a preliminary version of this paper for their

important notes and useful criticism.

3. Below I shall try to show that there exist one more tense.

4. This word is a transliteration of the Georgian term mc’k’rivebi, which denotes a verbal

conjugation paradigm characterized by a certain tense, aspect, mood and evidentiality/non-

evidentiality. In this paper I used the term “tense” instead.

5. Preverbs originally had a directional meaning, but in most verbs they are desemanticized and

used just as grammatical markers of the perfective aspect.

6. The Svan verbal system is presented here after Gudjedjiani & Palmaitis (1986: 29–33, 65–91)

and Testelets (1989). Topuria (1985: 125-126) and, subsequently, Schmidt (1991: 525)

differentiate two forms of the evidential (imperfective) present, cf. (3f) xägmina (Evidential I)

and (3a) lmgämün (Evidential II) „(he) was apparently building‟. Both forms are covered by

the “Narrative Present” in Table 1 and are treated here as the object and neutral version of the

imperfective evidential present, respectively. At the same time, the descriptions by V.Topuria

and K.H.Schmidt lack past and subjunctive imperfective evidentials.

7. The “tradition” is the tradition in Georgian.

8. There is no accusative in Kartvelian: in nominative constructions its functions are covered by

the dative. Both the direct and the indirect object receive dative marking, cf. (52):

(52) Vano apxneg-s läjr-s xahwdi

Vano:NOM friend-DAT book-DAT give:PRS:3SG.S/3SG.O

„Vano gives/is giving (his) friend the book.

9. The words „nominative‟, „ergative” and „dative” are used here as commonly accepted and

rather convenient labels, though the real character of the corresponding tenses as nominative

(accusative), ergative or active remains to this day a subject of discussion. We shall not

consider the topic (see footnote 22).

10 . Another term for this category, more common outside Russia, is “applicative”. I keep using the

term “version” because it is traditional for Kartvelian studies.

11 . Italics in transcriptions, glosses, and translations correspond to evidential forms.

12. Svan possesses a rather intricate morphophonology. Explaining its details for each form would

extend thescope of this paper. That is why some morphemes and morphemic alterations

irrelevant for the goals of the present paper are not explained here. For abbreviations see the

list at the end.

24 NINA SUMBATOVA

13. Cf. the „Narrative Past”, which is a relative tense: it refers to an action preceding another past

action.

14. This formative is not really a participle because it is used only as a part of the verbal evidential

forms (no attributive or other independent usage is possible).

15 . For the forms of the object version, evidentials are synthetic (-in-a): cf.

(3f) x-ä-gm-in-a

„he was apparently building for him‟.

16. Lower halves of brackets mark two parts of a circumfix (glossed under its first part).

17. The choice of agreement markers is determined by the following rules (according to Kibrik

1996):

Prefix Suffix

S<O O O

S=O O S

S>O S S

This small table shows what actant controls the choice of the agreement morphemes. If, for

example, the subject is third person and the object is second person (the object is higher in the

hierarchy, S<O), then both the prefix and the suffix are taken from the object agreement row

and show agreement with the second person. If the subject is first preson and the object is

second person (equally high in the hierarchy, S=O), then the second person prefix of the object

row is also taken, but the suffix shows agreement with the first person of the subject, etc.

18. According to Nedjalkov, “perfect” is understood as “an action (process, or state) in the past

which has continuing relevance for the present” (cf. Nedjalkov & Yakhontov 1983: 15)

19. From this point on „perfect‟ be understood to mean „perfect marked with a preverb‟, unless

noted otherwise.

20 . Dukhaner is a keeper of a dukhan, a little restaurant in Caucasus.

21 . In the Kartvelian tradition, the verbs that code their subjects with the dative case in all series

(the second actant, if any, receives a nominative marking) are called inversive. These are

mostly verbs of perception, cognition, physiological states, feelings, etc. (see section 5).

22. The similar situation in Georgian evoked a discussion on the real nature of the second tense

series: Harris (1981: 228–274) finds the principal features of an active typology, whereas

Hewitt (1987: 326–339) regards it as ergative. Whatever point of view may be true (for Svan, I

would side with Harris), no doubt the second and the third series are alike.

23. It goes about one of the old Georgian books, The Gospel of Hadishi, (897) .

24 . The beginning of this text is as follows: „An ox was standing by a sea, so that its front legs

were standing on one bank and its back legs on the other, so big was this ox. An eagle came

down from somewhere, seized the ox and carried it away... There was a goat somewhere. It

was so big that an ox herd with the herdsman all together hid under its beard to keep the rain

out in bad weather. The eagle sat down on the horns of this goat and ate the ox there...”

25 . Kubdar is a local pie.

26. This preverb is not part of the verbal form (which can be shown by way of

morphophonological rules).

27. Participles and deverbal nouns are highly lexicalized. That is why meanings of concrete

derivates can deviate quite far from the prototype.

28. See also the paper by Trask in the same book (Trask 1979), where this type of emergent

ergativity is analyzed on the basis of some broader typological data.

29. These derivates are traditionally called participles, but are usually used as deverbal nouns.

30. In Testelets 1984, its semantics is described through the notion of “involvedness”: the object

version dative is an actant that is involved in the situation less than the patient, but nevertheless

to a fairly comparable degree.

ABBREVIATIONS

AOR – aorist

DAT – dative

EMPH – emphatic particle

ERG – ergative

EVIDENTIALITY, TRANSITIVITY AND SPLIT ERGATIVITY... 25

EVID – evidential

GEN – genitive

IEVID – imperfective evidential

INST – instrumental

IO – indirect object

itr – intransitive

MSD – masdar

LOC – locative

NEG – negative

NOM – nominative

NP – noun phrase

O – object

OBL – oblique case (different from nominative, ergative, and dative)

OV – object version

PART – participle

PAST – past

PASS.PART – passive participle

PEVID – perfective evidential

PFT – perfect

PL – plural; 1PL, 2PL, 3PL – 1st, 2nd, 3d person plural

PLUPFT – pluperfect

PRS – present

PV – preverb

Q – interrogative particle

S – subject, Str – subject of a transitive verb, Sitr – subject of an intransitive verb

SG – singular; 1SG, 2SG, 3SG – 1st, 2nd, 3d person singular

STAT – stative

SUBJ – subjunctive

TFRM – transformative

tr – transitive

REFERENCES

Bybee, Joan and Östen Dahl. 1989. “The creation of tense and aspect systems in the

languages of the world”. Studies in language 13–1. 51–103.

Chafe, Wallace and Johanna Nichols (eds.). 1986. Evidentiality: the linguistic coding

of epistemology. Norwood, N.J.: Ablex.

DeLancey, Scott. 1997. “Mirativity: The grammatical marking of unexpected

information”. Linguistic Typology 1-1. 33–52.

Foley, William A. and Robert D. Van Valin. 1984. Functional Syntax and Universal

Grammar (Cambridge Studies in Linguistics 38). Cambridge: Cambridge

University Press.

Gudjejiani, Chato and Mykolas L. Palmaitis, 1986. Upper Svan: Grammar and texts

(Kalbotyra XXXVII (4)). Vilnius: Mokslas.

Guentchéva, Zlatka (ed.). 1996. L’énonciation médiatisée. Louvain–Paris: Peeters.

Harris, Alice C. 1981. Georgian Syntax: a Study in Relational Grammar (Cambridge

Studies in Linguistics 33). Cambridge: Cambridge University Press.

26 NINA SUMBATOVA

Harris, Alice C. 1991. “Overview on the History of the Kartvelian Languages”. In

Harris, A.C. (ed.), The Indigenous Languages of the Caucasus. Vol.1 – The

Kartvelian Languages, 7–84. Delmar, New York: Caravan Books.

Hewitt, B.George. 1987. “Georgian: ergative or active?” Lingua 71. 319-340.

Hopper, Paul J. and Sandra A.Thompson. 1980. “Transitivity in grammar and

discourse”. Language 56-2. 251–299

Kibrik, Andrej A. 1996. “Yazyk ne tak nelep, kak kazhetsya (lichno-chislovoe

soglasovanie v svanskom yazyke)”. [“The language is not as absurd as it seems

to be (person and number agreement in the Svan language)”]. Rusistika.

Slavistika. Indoevropeistika. Sbornik k 60-letiyu A.A.Zaliznyaka ed. by

A.A. Gippius, T.M. Nikolaeva and V.N. Toporov, 478–493. Moscow: Indrik.

[Russian, Slavic, Indoeuropean philology. Festschrift on the occasion of the

60th birthday of A.A. Zaliznyak ed. by A.A. Gippius, T.M. Nikolaeva and

V.N. Toporov, 478–493. Moscow: Indrik].

Kozinceva, Natalija A. 1994. “Kategorija evidencial‟nosti (problemy tipologičeskogo

analiza)”. [“The category of evidentiality: problems with the typological

analysis”]. Voprosy yazykoznaniya, 1994–3. 92–104.

Machavariani, Maja V. 1983. “Stative, resultative, passive and perfect in Georgian”.

In Nedjalkov, V.P. (ed.), Typology of Resultative Constructions, 259–276.

Amsterdam–Philadelphia: Benjamins.

Natadze, N.R. 1955. “K voprosu ob obrazovanii vremen i nakloneniy tret‟jej serii v

kartvel‟skix yazykax” [“On the formation of tenses and moods of the third

series in the Kartvelian languages.”]. Iberiysko-kavkazskoe yazykoznanie 7.

99-100.

Nedjalkov, Vladimir P. and Sergey Je.Yakhontov. 1983. In Nedjalkov, V.P. (ed.),

“The typology of resultative constructions”. Typology of Resultative

Constructions, 3–62. Amsterdam–Philadelphia: Benjamins.

Pirejko, L.A. 1979. “On the Genesis of the Ergative Construction in Indo-Iranian”. In

Plank, F. (ed.), Ergativity: Toward a Theory of Grammatical relations, 481–

488. London–NewYork: Academic Press.

Schmidt, Karl Horst. 1991. “Svan”. In Harris, A.C. (ed.), The Indigenous Languages

of the Caucasus. Vol.1 – The Kartvelian Languages, 473–556. Delmar, New

York: Caravan Books.

Sumbatova, N.R. 1993. “Causative constructions in Svan: further evidence for role

domination”. In Comrie, B. and Polinsky, M. (eds.), Causatives and

transitivity, 253–270. Amsterdam: Benjamins

Svanuri Enis Krest’omatia [A Reader Of The Svan Language]. 1978. (Dzveli kartuli

enis k‟atedris šromebi). Tbilisi: Metsniereba.

Testelets, Jakov G. 1984. “Objektnaja versiya v kartvel‟skix jazykax”. [“The object

version in the Kartvelian languages”]. In Lingvističeskie issledovanija.

Tipologija. Dialektologija. Etimilogija. Komparativistika,130–140. Moscow:

Nauka. [Studies in linguistics. Typology. Dialectology. Etymology.

Comparative linguistics].

Testelets, Jakov G. 1989. Svanskij jazyk. Materialy kavkazskoj lingvističeskoj

ekspedicii. Ms. [The Svan language. Materials of the Caucasian linguistic

expedition. Ms.]

Topuria, Varlam T. 1985. “Svanskij jazyk” (“The Svan language”). Ežegodnik

iberijsko-kavkazskogo jazykoznaniya XII, 100–147.

EVIDENTIALITY, TRANSITIVITY AND SPLIT ERGATIVITY... 27

Trask, R.L. 1979. “On the Origins of Ergativity”. In Plank F. (ed.), Ergativity: Toward

a Theory of Grammatical relations, 385–406. London–New York: Academic

Press.

Willett, Thomas. 1988. “A cross-linguistic survey of evidentiality”. Studies in

language 12–1. 51-98.


Recommended