The Georgian perfect tense series and

the Western European BE/HAVE auxiliary split

Steve Hewitt, s.hewitt@unesco.org
Bur. E-175, UNESCO, 7 place de Fontenoy, 75352 PARIS 07 SP, France
Verbs in the Georgian resultative-evidential perfect tense series (Series III) have what is known as inverted
marking, both on the arguments (subjects are in the dative; direct objects in the nominative) and on the verb
(indirect object person markers for the subject; subject person-markers for the direct object) in verb Classes 1
(transitive agentive) and 3 (unergative intransitive agentive). Class 4 (experiencer, psych affective) verbs
have inverted marking in all tense series, whereas Class 2 (unaccusative intransitive/passive patientive) verbs
have direct marking in all tense series, including perfect tense Series III, both on the argument(s) (subjects are
nominative; any indirect objects dative) and on the verb (in perfect tense Series III: subject person markers plus
passive participle plus forms of the verb BE). While such inversion is seen as natural for Class 4 verbs, which have
dative subjects in many languages (cf. French il me plat; German es gefllt mir; Russian mne nravits; Georgian
momcons; Hindi/Urdu mujh pasand hai), it is generally taken to be idiosyncratic in Georgian in the case of the
perfect tenses of agentive verb Classes 1 (transitive) and 3 (unergative). However, if I have is seen as TO.ME-THERE.IS
(and this is etymologically the case in, for instance, Breton), the dichotomy in the marking of the Georgian perfect
tenses closely parallels the BE/HAVE perfect auxiliary split found in a number of Western European languages
(including Breton), where, roughly speaking, BE is used with unaccusative intransitives (and passives) Georgian
Class 2, and HAVE with transitive verbs and with unergative intransitives Georgian Classes 1 and 3. The split in the
marking of the perfect tenses in Georgian is thus as natural as, and indeed analogous to, the Western European
perfect auxiliary split. (BEN benefactive; DIR direct person markers; LC link consonant; LV link vowel; OBL oblique
person markers; PP past/passive participle; PV perfective preverb; THS thematic stem)
Georgian
Class 1 transitive (agentive)
da- mi- xa- av- s
PV OBL.1SG BEN paint THS DIR.3SG
I have [apparently] painted it

liwed
m-eus
anea
painted.PP I.have [OBL.1SG-be.EXIST] it.M.OBJ
I have painted it

Class 3 unergative intransitive (agentive)

mi- mu- av- (n-) i- a
OBL.1SG BEN work THS LC LV DIR.3SG
I have [apparently] worked

laboured
m-eus
worked.PP I.have [OBL.1SG-be.EXIST]
I have worked

Class 2 unaccusative intransitive (patientive)

mi- vs- ul- v-ar
PV DIR.1SG go PP I.am [DIR.1SG-be]
I have [apparently] gone

aed on
gone I.am
I have gone

Class 2 passive (patientive)

da- vxa- ul- v-ar
PV DIR.1SG paint PP I.am [DIR.1SG-be]
I have [apparently] been painted

liwed
on (bed)
painted.PP I.am (been.PP)
I have been painted

Breton

The legendary complexity of the Georgian verbal system may be reduced to the following schematic outline:
Regular person-marking on verb
Case-arrays of arguments
Inverted person marking on verb
Case-array of arguments

N
E

DIRS subject
NOMS
ERGS
OBLIS inverted subject

DATS

(OBLDO direct object)

(OBLIO indirect object)
(DATDO)
(DATIO)
(NOMDO)
(DATIO)
(DIRIDO inverted direct object) [demoted ind. obj.]
(NOMDO)
(-tvisIO for).

Person-marking (regular/inverted) and case-arrays (N/E/D)

according to verb class/type and tense series
Verb
Class
2
1, 3
4

Type
patientive
agentive
affective

I Present/Future
reg.N

Tense series
II Aorist
reg.N
reg.E
inv.D

III Perfect
inv.D

There is thus a primary split between direct verbs and indirect (affective) verbs; such a split is quite widespread
across languages. Within the direct verbs, there is a further split between patientive verbs, which have regular
person-marking and the N (nominative) case-array in all tense series, and agentive verbs, which have regular
person-marking and N case-array in tense Series I (present, future, imperfect, conditional), but E (ergative) casearray in tense Series II (aorist); in tense Series III (perfect, pluperfect), they have inverted person-marking and the

D (dative) case-array, like the indirect, Class 4 verbs in all tense series. In the direct verbs, person-marking and
case-array in the aorist and perfect tense series appear to be linked, with reg.N for both with patientive verbs;
with agentive verbs, reg.E in the aorist implies inv.D in the perfect, and vice versa.
Alice Harris, Diachronic Syntax: The Kartvelian case (Academic Press, New York, 1986, ch. 13), maintains that the
original Kartvelian verb system had Series II tenses with E case-array, from which Series I tenses with N case-array
developed by a process of intransitivization and object demotion. Series III tenses with inverted person-marking
and D case-array appear to be a more recent (barely prehistoric) development for the perfect tense series of
agentive verbs. At first resultative, the evidential value (at least in the affirmative) is probably a later accretion,
possibly an areal feature (cf. Turkish -mi- forms). Damana Melikivili, Kartuli zmnis ulebis sisema [Conjugation
system of the Georgian verb] (Logos resi, Tbilisi, 2001, p.22), suggests that the perfect tense arose out of a passive
of state: miceria cerili mir steht ein Brief geschrieben ~ a letter stands/is written for me ~ I have a letter written;
with the addition of a perfective prefix, the construction became an active perfect: damiceria cerili I have written
a/the letter. The negative aorist indicates a conscious intent (control/volitionality): tamarma ar dacera cerili
TamarERG did not [intentionally] write the letter, whereas the negative perfect indicates a denial of intent
(control/volitionality): tamars ar dauceria cerili TamarDAT has not written / did not [failed to] write the letter.
Italian is the Western European language with split auxiliarity which has the broadest use of BE as a perfect
auxiliary for unaccusatives. A sample of Italian verbs and their Georgian equivalents shows a better than 90%
correlation between Italian verbs taking essere be (unaccusative intransitive, reflexive, passive) as their perfect
auxiliary and Georgian Class 2 verbs (with direct marking plus BE in the perfect series) on the one hand, and
Italian verbs with avere have (transitive, unergative intransitive) as their perfect auxiliary and Georgian Class 1
and 3 verbs (with indirect marking plus BE ( HAVE) in the perfect series) on the other hand.
In Antonella Soraces Auxiliary Selection Hierarchy (ASH) (Gradients in auxiliary selection with intransitive
verbs, Language 76/4: 859-90, 2000): selects BE (least variation): change of location > change of state >
continuation of state > existence of state > uncontrolled process >controlled process (motional) > controlled
process (non-motional): selects HAVE (least variation). In both Georgian and Italian, the split comes low down,
between existence of state and uncontrolled process (motional), with Georgian Class 2 verbs and Italian BE
verbs for the all former, and Georgian Class 3 verbs and Italian HAVE intransitives for the latter.
The difference between unergative and unaccusative intransitives has been described from an abstract
syntactic point of view as that between a deep-structure subject: John works unergative NP [VP V], and a deepstructure object: John returns unaccusative [VP V NP]. It is natural for underlying subjects to have more
control/volitionality than underlying objects. According to David Dowty, Thematic proto-roles and argument
selection (Language 67/3: 547-619, 1991), prototypical Agents (sentient, volitional, stationary, causally effective,
etc.) correspond to the subjects of HAVE verbs, whereas prototypical Patients (inanimate, affected by the event,
undergo a change of state or location, measure out the event i.e. incremental themes) correspond to the
subjects of BE verbs. This works well for Georgian Classes 1 and 3 (agentive > reg.E in the aorist; > inv.D in the
perfect tense series) on the one hand, and Class 2 (patientive > reg.N in both the aorist and the perfect tense series)
on the other.
The Georgian verb classes are defined primarily by their morphological characteristics. True enough, the broad
grammatical characterizations (1 transitive agentive); 2 intransitive unaccusative/passive patientive;
3 intransitive unergative agentive; 4 experiencer affective) cluster very strongly with these morphological
classes, but not absolutely so. Alice Harris, Georgian Syntax: A study in relational grammar (Cambridge University Press,
1981, p.246), notes certain discrepancies concerning the control/volitionality of the subject: for Standard Georgian
Class 3 verb cqalma icveta macivridan waterERG dripped from the fridge the ergative subject in the aorist is often
replaced with a nominative, reflecting the lack of control: cqali icveta macivridan waterNOM dripped from the fridge;
conversely numerous dialects have for Standard Georgian Class 2 verb with a nominative subject in the aorist aci
cavida the manNOM went off an ergative subject, reflecting the effective control: acma cavida the manERG went off.
However, this never, as far as I am aware, triggers a corresponding change of person-marking or case-array in the
perfect tenses. Regular person-marking may correlate with either the N or the E case-array; it is therefore easy for
the change to be made, if warranted by the semantics of the verb. On the other hand, inverted person-marking
always corresponds to the D case-array. The morphology of the verb class thus prevents the corresponding change
from being made in the perfect tense series: inv.-D *> reg.-N; reg.-N *> inv.-D.
In conclusion, by refusing to accept as idiosyncratic, and instead reflecting on a functional explanation for the
rather neglected person-marking and case-array split in the perfect series between agentive and patientive verbs,
we may have shed some light on the overall functioning of the Georgian verb system: the Georgian ergative case
used for the subject of agentive verbs in the aorist tense series appears to imply an overt assertion of
control/volitionality; the dative case used for the subject in all tense series of indirect affective verbs and in the
perfect tense series of agentive verbs implies a denial of control/volitionality; the nominative case used in the
present/future tense series of agentive verbs and in all tense series, including the perfect, of patientive verbs,
implies no particular salience of the control/volitionality parameter. For agentive verbs, the control/volitionality
parameter is thus salient, and either asserted (ERG for the aorist tense series) or denied (DAT for the perfect tense
series); for patientive verbs, it is not salient for any of the tense series (NOM).
These observations all serve to underscore the importance of Georgian for theoretical work not only on the
unaccusative/unergative distinction, but also, by morphological analogy at least, on split auxiliarity.