4. Morphosyntax and syntax of Udi


4.1 Basic aspect of relational behavior


The syntax of Udi is governed by at least three features: order of constituents, morphology, and referential tracking devices (discourse cohesion devices). With respect to the ranking of these features, the following hierarchy applies:


Morphology < discourse cohesion < word order 


These means that the bulk of syntactic organisation is encoded by morphology, which is (in parts) also reflected in discourse cohesion. Word order represents the least important feature of syntactic organisation. In consequence, every syntactic structure has to include morphological features. Additionally, morphology may react on certain certain semantic features of actancy (both split and fluid):


The relational primitives S (subjective), A (agentive), and O (objective) organize a syntactic structure in co-paradigmatization with the morphological resp. syntactic devices that encode these primitives:


Word order is coupled with:

            syntactic procedures (foregrounding/backgrounding);

            attention/information flow (AIF);

            pragmatic procedures;


Case marking is coupled with:

            syntactic features (foregrounding/backgrounding, syntactic pivot);

            semantic features (control);

            pragmatic features (referential indexing, pragmatic pivot);


Agreement is coupled with:

            semantic features (control);

            syntactic features (foregrounding/background, syntactic pivot);

            pragmatic featrues (focus marking);  


The relational behavior of the primitives is both accusative (word order, agreement, in parts case marking) and ergative (in parts case marking). As expected, the IO (indirect objective) domain is strongly coupled with the O-domain though the O domain shares important properties with the S-domain, too (see Schulze (in press) for a discussion of relational behavior):


                                       S                            LOC                     

                               A              O       







=>       A is coupled with S by agreement, case marking of personal pronouns, word order (AIF), foreground and pivot assignment;

=>       A is coupled with IO in endocentric A-splits (case marking, agreement, cf. 4.4.3);

=>       S is coupled with O by case marking (cf. 4.4.4);

=>       O is coupled with IO by case marking (cf. 4.4.5);

=>       A is coupled with the peripheric LOC-domain in exocentric A-splits (cf. 4.4.3);

=>       {O;IO} is coupled with the peripheric LOC-domain by polysemy and in exocentric O-splits (cf. 4.4.4).


The basic asymmetry between the A-domain (heavy actants > active force) and O (light actants > no active force) is generally maintained in Udi. As for as I know, there are no morphosyntactic means to encode counter-force with O-actants in the force dynamics schema: If an actant in O-function is considered to have (partial) control over the action exerted by an actant in A-function, or if O is higher in rank than A, no shift in case marking, agreement, or word order applies (except for verba sentiendi, see 4.4.3), cp.:


qos^i-n-en düs^man-ax bes-b-i-ne

army-sa-erg enemy-dat2 kill-lv-aor-3sg:a

‘The army has killed the enemy’.




qos^i-n-en sinamis^-ne-b-i düs^man-ax bes-b-esun-ax

army-sa-erg try-3sg:a-lv-aor enemy-dat2 kill-lv-inf-dat2

‘The army tried to kill the enemy.’


In the second example, the ‘enemy’ (in O-function) is thought to own some kind of counter-force which competes with the fature ‘force’ associated with the ‘army’ in A-function. The resulting reading is for instance ‘tries to V’. The example shows that strategies of marking counter-force are based on lexical representation rather than on variation in the marking of the A- resp. O-domains.


The same holds for modal structures which include the reduction of force features that are associated to the A-function. Such a modalization is either based on lexical structures (baksun A>IO in the sense of ‘can’) or on simple verbal modality (optative).


But note that so-called verba sentiendi do include aspects of variation in the force dynamics schema, cp. 4.4.3.



4.2 Preferred word order


The preferred clausal word order of Udi is accusative: The clause normally starts with an actant in S or A function, followed by (if given) actants in IO and O function, cp.:


S-V:                vi xunc^i ar-e-ne

                        your(sg) sister:abs come:past-perf-3sg:s

                        ‘Your sister has come’.


A-O-V:           vi xunc^i-n-en mal-l-ux tov-ne-d-i

                        your(sg.) sister-sa-erg goods-sa-dat2 sell-3sg:a-lv-aor

                        ‘Your sister sold the goods.’


A-IO-O-V:      vi xunc^-i-n-en za tängi-n-ax ta-ne-d-e

                        you(sg.) sister-sa-erg I:dat money-sa-dat2 give-3sg:a-lv-perf

                        ‘Your sister has given me the money.’


Peripheral elements often occupy a position between S and verb resp. between A and O which illustrates the relatively strong cohesion between an actant in O-function and the verb, cp.:


S-Per-V:         bez nana k’ua-ne ar-e

                        my mother:abs home-3sg:s come:past-perf

                        ‘My mother has come home.’


A-Per-O-V:    bez baba-n sa me-n-en q’al-l-ux s/am-ne-p-i

                        my father-erg one knife-sa-erg>instr lamb-sa-dat2 slaughter-3sg:a-lv-aor

                        ‘My father slaughtered the lamb with a knife.’


The close coupling between a verbal structure and an AGR-marked constituent (in focus) often results in a secondary ordering procedure: The slot before the verb is often occupied by the focused constituent (if constituent focus applies), disregarding word class and relational behavior, cp.:




                        bez baba-n q’al-l-ux sa me-n-en-ne s/am-p-i

                        my father-erg lamb-sa-dat2 one knife-sa-erg>instr-3sg:a slaughter-lv-aor

                        ‘My father slaughtered the lamb with a KNIFE’.


Both before and after the core clause structure, there are functional slots that can be used for pragmatic purposes (cf. 4.10). These slots are another place favored by peripehric constituents such as locatives, cp.:


Per-A-O-V:    me s^ähär-ä mo-t’-g/-on täzä mal-ux tov-q’un-d-esa

                        prox town-dat1 prox-sa:obl-erg new goods-pl:abs sell-3pl:a-lv-pres

                        ‘In the city, they sell new goods.’


S-V-Per:         gädä ta-ne-sa me säs-n-a-laxo

                        boy:abs go-3sg:s-$:pres voice-sa-gen-pp(on)

                        ‘The boy follows this voice’ (lit. ‘The boy goes on this voice’).


A-O-V-Per:    [me-t’-in] sa dz^ähil g/ar-re biq’-sa xibq’o manat-al

                        [prox-sa:obl-erg] one young boy-3sg:a take-pres 60 rubel-super

                        ‘He hires a young boy for 60 rubels.’


The placement of localizing constituents in the final slot is comparatively rare with transitive structures. Yet, this slot is often used to signal (in parts contrastive) focus on constituents in S- or O-function, cp.:


A-IO-V-O:      pasc^’ag/-en k’ic’k’e g/ar-a ta-ne-st’a dz^ok’ k’odz^ tängä

                        king-erg little son-dat1 give-3sg:a-lv:pres separate house:abs money:abs

                        ‘The king gives [his] young son not only a house, but also money.’


Per-V-S          me-t’u-g/o-xol ta-q’un-sa q’eiri sövdäkär-ux-al

                        prox-sa:obl-pl-com go-3pl:s-$:pres other merchant-pl:abs-foc

                        ‘Some other merchants, too, go with them’


A-V-O ordering also allows clausal constituents to follow a verb:


A-V-O[cl]:      tuli-n-en bur-re-q-i bap-s-ax

                        young=dog-sa-erg start-3sg:a-lv-aor bark-inf-dat2

                        ‘The young dog began to bark.’


                        pasc^’ag/-un g/ar-en ex-ne axe.r is/a yaq’-en tag/-al adamar-ux te-q’un qai-bak-sa

                        king-gen son-erg say:pres-3sg:a finally close way-erg>instr go:fut-part:pres man-pl:abs neg-3pl:s return-lv:intr-pres

                        ‘The king’ son says [that] people who take the close way do NOT return.’                



4.3 The morphology of relational behavior


In unmarked structures with third person actants, case marking is ergative, whereas agreement is accusative:


AGR                S=A;O

CASE              S=O;A

















nana ta-ne-sa

mother:abs go-3sg:s-$:pres

‘Mother goes.’


nana-n s/um u-ne-k-sa

mother-erg bread eat-$-pres

‘Mother eats bread.’



Whereas there are no variations to be observed with AGR-strategies, case marking may reflect split patterns. This distribution indicates an underlying primacy of AGR which reveals that Udi has much more ‘accusative’ features than ‘ergative’ ones.



4.4 Split phenomena


All three relational primitives S, A, and O can be subjected to split strategies. Splits are always carried out with the help of case morphology and may be reflected in the agreement pattern. There are both true splits (subcategorizing the verbal domain) and fluid strategies (optional marking of actants). S and O are related to fluid strategies, whereas A is basically subjected to fixed splitting patterns:


S:        Fluid-S            endocentric

            Split-S             exocentric

A:        Split-A            endocentric / exocentric

O:        Fluid-O           exocentric


Splits may be endocentric (using the coding pattern of another relational primitive) or exocentric (using other coding patterns alien to the basic relational pattern) [see Schulze (in press) for details]. Endocentric shifts are called ‘motion’ and represented as follows (motion does not apply with O in Udi in a synchronic perspective):


Promotion:       S>A                [S behaves as if it were A]

                        S>IO               [S behaves as if it were IO]

Demotion:        A>S                [S behaves as if it were S]

                        A>IO              [A behaves as if it were IO]


Exocentric shifts are called usurpation: A relational primitive usurps another (basically locational) patterns:


Usurpation:      A>LOC           [A behaves as if it were localized]

                        O>LOC          [O behaves as if it were localized]


As far as I can see, there are no usurpation techniques related to the S-domain.



4.4.1 The Silverstein Hierarchy in Udi


Synchronically, Udi conforms to the predications stemming from the hierarchic cluster {empathy<person<animacy}. The relevant section is represented in Udi by the feature ‘Speech Act Participant’ (rather than ‘person’). All speech act participants are neutralized in case marking with respect to S and A (see for case marking of personal pronouns):
















zu ta-s-sa (< ta-zu-sa)

I:abs go-1sg:s-$:pres

‘I go’


zu s/um u-zu-k-sa

I:abs bread:abs eat-1sg:a-$-pres

‘I eat bread.’


We have to deal with an accusative pattern here (which is matched by the agreement pattern resulting there from). But note that the use of the absolutve case with A cannot be regarded as an instance of demotion (A>S) as described above: The ‘accusativization’ of the pronominal paradigm does not include any further semantic or pragmatic shift in the actual language (historically speaking, such a shift may well have taken place, cf. Schulze 1998, 1999).


The S=A copuling of speech act participants coincides with the morphological marking of O if this function is covered by a speech act participant, cp.:


t’e adamar-en zax gölö ögmis^-ne-b-e

dist man-erg I:dat2 much praise-3sg:a-lv-perf

‘That man has praised me very much’.


Though occasionally an absolutive case form may be used, such as


zu un be?g/-sa-zu

I:abs you(sg.):abs see-pres-1sg:a

‘I observe you.’


such a ‘neutral’ type of case marking becomes more and more obsolete. In fact, Udi has generalized the marked variant of Fluid-O splitting (cf. 4.4.4).


The cut-off point in the hierarchy is indicated by SAP vs. nSAP. All nSAP basically conform an ergative case marking pattern with respect to S and A. If O is represented by a deictic element in anaphoric function, it normally behaves like a ‘personal pronoun’, especially if the antecedent in human or an animal ranked high in the animacy hierarchy, cf.


mo-n-o ta-ne-sa

prox-sa:abs-abs go-3sg:s-$:pres

‘(S)he (< this one) goes.’


s^e-t’-in mo-t’-ux (ref.: e?k’) a-ne-q’-sa

dist-sa:obl-erg prox-sa:obl-dat2 (ref.: ‘horse’) take-3sg:a-$-pres

‘(S)he takes it (ref.: horse)’.




s^e-t’-in mo-n-o (ref.: s/um) u-ne-k-sa

dist-sa:obl-erg prox-sa:abs-abs (ref.: ‘bread’) eat-3sg:a-$-pres

‘(S)he eats it (ref.: bread)’.




s^e-t’-in mo-t’-ux (ref.: s/um-ax) u-ne-k-sa

dist-sa:obl-erg prox-sa:obl-dat2 (ref: bread-dat2) eat-3sg:a-$-pres

‘(S)he eats it (ref.: the bread qualified before)’.



4.4.2 Fluid-S and S-split 


In the endocentric Fluid-S type of promotion, S is optionally marked by the ergative case (S behaves like A for case marking). Only nSAP can be involved. The speaker has the choice to mark S for a more (ergative) or less (absolutive) controlling behavior., cp.:


xinär axs/um-ne-xa

girl laugh-3sg:s-lv:res

‘The girl is laughing.’


xinär-en gölö axs/um-ne-xa

girl-erg much laugh-3sg:s-lv:pres

‘The girl is laughing very much.’


The second sentence informs the hearer that the speaker thinks the girl does its laughing intentiously. In other words: the ‘girl’ is thought to control the act of laughing. Note that in Udi the standard S-marking type (absolutive) represents the unmarked variant. Hence, we cannot conclude from a sentence like xinär gölö axs/umnexa that the girl is thought to unintentiously laugh. The senetnce can convey both the controlled and non-controlled meaning. However, the ergative marked pattern can only be used if the action is thought to be controlled.


Fluid-S marking seems (still) to be restricted to verbs of bodily action, especially with verbs like ‘laugh’, ‘weeep’, ‘cough’ etc. Verbs that share a state of ‘natural’ control such as ‘go’, ‘sit’ etc. are not (yet) involved in this splitting strategy.


An (pseudo)exocentric S-split can be observed with verbs denoting body sensations, such as ‘be cold’, ‘be hot’ etc. In this case, the actant in S-function behaves like an indirect objective (or locative). The agreement clitic copies this behavior. Note that S-splits are not restricted to nSAP but may apply to all ‘persons’, cp.:


va mi-va-b-sa

you(sg.):dat1 cold-2sg:io-lv-pres

‘You’re cold’


za q’e.?-za-bak-sa

I:dat1 afraid-1sg:io-lv:intrans-pres

‘I am scared’


ia va?-ia-bak-sa

we:dat1 faithful-1pl:io-lv:intrans-pres

‘We are faithful’.


Especially in the Nidzh dialect, the S-split tends to be leveled in favor of the canonical S-marking strategy (absolutive). This leveling happens in two steps: Some speakers only align the case marking pattern, others also change the agreement clitics (see Schulze in press b):


I.         un mi-va-b-sa

            you(sg.):abs cold-2sg:io-lv-pres

            ‘You’re cold.’


II.        un mi-n-b-sa

            you(sg.):abs cold-2sg:s-lv-pres

            ‘You’re cold.’ 


Obviously, the demotion of S to IO (S>IO) is related to the analogeous procedure with A-actants (A-split, see below). In consequence we have to describe this split as a technique that is co-paradimatized with the accusative behavior of S and A (hence S=A>IO). This split type is sometimes erroneously dscribed as ‘inversion’ (in the tradition of Relational Grammar). However, it should be noticed that the S=A-split in Udi does not involve any kind of ‘inverting’ whatever is thought to be an ‘underlying subject’. This split type is basically semantic in nature. It makes use of the metaphoric extension of the dative case function which now indicates affectnedness (< allative) rather than full control.



4.4.3 A-splits


If we neglect the type of A-split related to the Silverstein Hierarchy (cp. 4.4.1), we can describe two further types that include both SAP and nSAP. There is an endocentric split A>IO and an exocentric split A>LOC (rare).


As had been said in 4.4.2, the endocentric A-split is coupled with the S-split technique in intransitive structures. It is conditioned by the subcategorization of the verbal class which gives us the class of so-called verba sentiendi. A number of such verbs have ‘controlled’ correlates such as:


Controlled (A)

Non-controlled (A>IO)




to see



to want, love



to hear



to experience, know


Whereas ‘controlled’ verbs have the standard case patterns described for transitive structures (i.e., ergative case marking with nSAP in A-function), verbs with ‘demoted A’ call for an IO agreement marker in the verb. The (pro)nominal referent is normally encoded by the dative1, cp.:


gädi-n-en sa adamar be?-ne-g/-i

boy-sa-erg one man:abs see-3sg:a-$-aor

‘The boy saw (observed) a man.’




gädi-n-a sa adamar a-t’u-k-i

boy-sa-dat1 one man:abs see-3sg:io-$-aor

‘The boy saw (perceived) a man.’


As expected, such A-splits are restricted to human referents (or animals ranked high in the animacy hierarchy). Pending on dialectal criteria and individual idiosyncrasies, both the external referent and the internal agreement clitic may be aligned to the standard transitive coding paradigm (esp. in the dialectal variants of Nidzh). Thus we can observe the following variants:


gädi-n-en sa adamar a-t’u-k-i

boy-sa-erg one man see-3sg:io-$-i

‘The boy saw (perceived) a man.’


gädi-n-en sa amdar-e ak-i

boy-sa-erg one man-3sg:a see-aor

‘The boy saw a MAN’ [note that amdar is the Nidzh form of adamar; -e is the Nidzh clitic to encode 3sg:s=a]


The O-domain may be subjected to the Fluid-O split just as with standard transitive structures, cp.:


gädi-n-a sa adamar a-t’u-k-i

boy-sa-dat1 one man:abs see-3sg:io-$-aor

‘The boy saw (perceived) a man.’




gädi-n-a adamar-ax a-t’u-k-i

boy-sa-dat1 man-dat2 see-3sg:io-$-aor

‘The boy saw the/a man [already qualified before]’.  


The endocentric A-split is also used with the verb baksun ‘to be(come)’ in order to denote ‘ability’, cp.:


xinär-a furu-k’-esun ba-t’u-k-i

girl-dat1 walk-lv-inf2 be-3sg:io-$-aor

‘The girl could walk around’.


A doubled A-split is present e.g. in:


mo-t’u s^o-t’-ux ak’-sun ba-t’u-k-i

prox-sa:obl-dat1 dist-sa:obl-dat2 see-inf2 be-3sg:io-$-aor

‘(S)he could see it’. [But note that the verb be?g/sun ‘to observe’ is preferred in such a context]


With the verb tast’un ‘to give’ we may arrive at three datives:


za va manat-ax tast’un te-za bak-e

I:dat1 you(sg.):dat1 money-dat2 give:inf2 neg-1sg:io be-perf

‘I could NOT give you the money.’ 


Again, the dialect of Nidzh tends to align case marking according to the standard transitive pattern, cp.:


zu bak-al-te-z ka-t’-u b-es

I.abs be-fut-neg-1sg:s>a med-sa:obl-dat1 make-inf1

‘I cannot do that’.


[Already in Vartashen, the external referent tends to be placed in the ergative].


Exocentric A-splits are rather rare in contemporary Udi. They are marked by the use of a locative case instead of the ergative, whereas the agreement pattern remains accusative (sometimes IO). The preferred locative case is the adessive (-Vst’a). Such A-splits generally denote the ability to do something and are strongly related to strategies to encode long distance possession, cp.:


zast’a pul-le

I:adess money-3sg:s

‘I have money’


zast’a mo-t’-ux as^-zu-b-sa [or: as^-za-b-sa; as^-ne-b-sa]

I:adess prox-sa:obl-dat2 work-1sg:a-lv-pres [or: work-1sg:io-lv-pres; work-3sg:a-lv-pres]

‘I can do it.’


The underlying possessive strategy becomes transparant in the following example (Nidzh):


bez sa penec’-en hikä-a b-es-bak-o

I:gen one plough-erg>instr what-3sg:q do-inf-be-opt

‘What can I do with one plough?’



4.4.4 Fluid-O


The Fluid-O strategies of Udi represent a kind of DOM (‘Differentiated Object(ive) Marking’) as known e.g. from the Turkic or Iranian languages. Generally, the speaker has the choice to mark a referent in O-function with respect to the feature cluster {[known], [specific], [referential], [high in animacy]}. This basically pragmatic routine is carried out with the help of the opposition absolutive vs. dative case. Referents that are thought to positively answer to the feature cluster mentioned above are marked by the dative case. The underlying strategy can be described as follows:




Degree of affectedness











The grammaticalization of O[+ref] (in the sense of the featuer cluster described above) stems from the allative reading of the dative case: An O-referent defined as [+ref] is only ‘approached’ by the referent in A-function, hence not completely included in the transitive act. With SAPs in O-function, we may also think of the feature counter-force as a relevant motivation. The absolutive case marks those referents in O-function which are thought to be completely localized ‘in’ the domain of A (see Schulze 1998 and Schulze (in press a) for a discussion of the localization schema of A). The underlying essive function of the absolutive case has survived in a number of East Caucasian languages, in parts also in Udi, cp.:


s^ü-n-e bi?g/ pasc^’ag/-un g/ar-i tuli-n-en bur-re-q-sa gölö ba?p-s-ax

night-sa-gen middle:abs king-gen son-gen young=dog-sa-erg start-3sg:a-$-pres much bark-inf-dat2

At midnight, the dog of the king’s son began to bark very much’.


The partial inclusion of O in the domain of A is optional espcially with inanimates, cp.:


zu s/um u-zu-k-sa

I:abs bread:abs eat-1sg:a-$-pres

‘I eat bread’




zu s/um-ax u-zu-k-sa

I:abs bread-dat2 eat-1sg:a-$-pres

‘I eat the bread’


Note that the English definite article only partially covers the function of the marked O in Udi. The higher the referent is in animacy the more likely is the use of the marked variant. The relevant scale is:


SAP < nSAP[pro]:SG < nSAP[pro]:PL < REF[+hum]:SG < REF[+hum]:PL < REF[+anim;-hum] < REF[-anim]


The more left an element in O-function is located on this scale the more likely it is that it is marked by the dative case.


In principle, the speaker has the choice between the two dative cases to mark a referent in O-function. In the dialect of Nidzh, the dative1 in generally preferred. In Vartashen, it is the dative2 which represents the standard solution. However, the dative1 may likewise be used, especially if two or more O-referents are linked together. In this case, the first O-referent is often marked by the datvie2, whereas the other referents are marked by the dative1, cp.:


t’eg/i zu baba-x-q’an nana be?g/-al-zu

today I:abs father-dat2-and mother:dat1 see-fut-1sg:a

‘Today, I will see father and mother.’


[t’eg/i zu babaxq’an nanax be?g/alzu is a parallel option].




4.4.5 The IO-domain


The IO-domain (indirect object(ive)) is normally covered by the dative case. Its basic locational (allative/essive) orientation conditions that metaphorical extensions (e.g. in the sense of a benefactive) rarely apply. Instead, the benefactive case is preferred in such constructions, cp.:


baba-n za sa k’odz^ ta-ne-d-i

father-erg I.dat1 one house:abs give-3sg:a-lv-aor

‘Father gave me a house.’




baba-n zenk’ena sa k’odz^ ser-ne-b-i

father-erg I:ben one house:abs build-3sg:a-lv-aor

‘Father built a house for me.’


Morphologically speaking, the IO-domain is strongly coupled with the O-domain: In fact, both the IO-domain and the O[+ref]-domain are marked by the dative cases, either dative1 or dative2, cp.:


tad-a zax sa ist’akan xe

give-opt:2sg:a I.dat2 one glas water:abs

‘Give me a glas of water!’


zax be?g/-a

I.dat2 see-opt:2sg:a

‘Look at me!’ (lit.: ‘see me!’)


In consequence, the Fluid-O type described in 4.4.4 can also be described as a shift from O to IO within the {O;IO}-cluster, cp.:

                          IO           O[+ref]



     O[-ref]         O


This analysis corresponds to the fact that the O[+ref] domain becomes more salient if the referent is marked by the feature [human]: The IO-domain is canonically restricted to the feature cluster {[human];[high in animacy]}.


Yet, today there seems to exist the tendency to split the two datives according to the functions O[+ref] and IO (especially in the dialect of Vartashen): the dative1 becomes more and more restricted to encode the IO-domain, whereas the dative2 basically covers the O-domain (together with its metaphorization e.g. in telic converbs). 



4.5 Grounding and passives


Quite in accordance with its massive ‘accusative’ architecture, Udi does not know antipassive strategies. Moreover, Udi is rather reluctant in formulating variations in the distribution of foreground and background. Normally, S and A are invariably located in the foreground, whereas O is locate in the background (of a scene). One standard way of manipulating such adistribution is the total backgrounding of the O-domain (which itself is a typical accusative feature), cp.:


adamar-g/-on gölö fi u?-q’u?n-g/-i

man-pl-erg much wine drink-3pl:a-$-aor

‘The men drank much wine.’


Total O-backgrounding:


me adamar-g/-on gölö-al u?-q’u?n-g/-i is/a finaxo-q’un bak-sa

prox man-pl-erg much-foc drink-3pl:a-$-aor now drunken(lit. ‘of the wine’)-3pl:s be-pres

‘These men have drunken to much - now they are drunk.’   


Partial O-backgrounding


Partial backgrounding of O is linked to incorporation strategies. An incorporated element in (former) O function looses its referential qualities as well as its relational behavior. In fact, an Inc-element behaves like an adverb, cp.:


s^o-n-or ic^ e?k’-ur-g/-on yaq’a-q’un-esa

dist-sa:abs-abs refl horse-pl-pl-erg>instr return-3pl:s-lv:pres

‘They return with their horses.’


Here, yaq’a is a dative1 (locative) of yaq’ denoting ‘on the way’. The concept ‘return’ is derived from the incorporation of the adverbial (< locative) structure yaq’a ‘on the way’ by adding esun ‘to go’ (as a light verb), lit.: ‘to go on the [same] way’. The lexical element yaq’ does not have any referential qualities anymore. The same is true for former referents in O-function, cp.:


s^o-t’-g/-on me-t’-a ait-ur-g/-ox imux-q’un-lax-i

dist-sa:obl-erg prox-sa:obl-gen word-pl-pl-dat2 ear-3pl:a-put-aor

‘They heard (listened to) his words.’


Historically speaking, the sentence means ‘they put the ear on (lit. to) his words’. However, the concept imux ‘ear’ no longer satisfies the O-function called for by the transitive valence of laxsun ‘to put’. imux is partially backgrounding by having lost all referential properties as well as the appropriate relational behavior. Such a backgrounding type is called ‘partial’ because the lexical structure is still present (contrary to total backgrounding as discussed above).


Udi backgrounding strategies are purely structural. There are no means to morphologically encode a backgrounded actant. This generalization does no longer hold for foregrounding strategies. Both autochthonous patterns and patterns derived from language contact (Armenian, Azeri) have helped Udi to develop emans to morphologically encode the manipulation of the foreground of a scene. Foregrounding may effect O in transitive structures as well as S in intransitive structures, cp.:




Case marking


Verbal morphology



O > S



-esun (light verb:intrans)

-baksun (in Inc-Structures)



S > S



-esun (light verb:intrans)



Foregrounding of O:


The foregrounding of O leads to intransitive structures. O behaves like S both for case marking and agreement. The verb is marked by the light verb esun (< ‘to go’) which shows the suppletive paradigm discussed in 3.3.2. The AGR-slot for endoclitics is canceled in favor of the new slot between verbal stem and passive marker (which, in fact, plays the role of a derivational element). The refernt in A-function is automatically backgrounded. In most cases, backgrounding is complete, though occasionally the element in former A-function is retained in the periphery (marked by the ablative, probably copying foreign structures). This kind of passivization does not have a precise syntactic value (‘subject’-assignment). Rather, the motivation of Udi passives is based on pragmatic and in parts semantic aspects. The foregrounded element in former O-function is focused, the verbal relation acquires a more durative reading focusing on the process in which the foregrounded constituent is involved. As expected, passivization cancels all kinds of split strategies:


A:ERG - O:ABS V[simple]


bez xunc^i-n-en za xe ta-ne-d-i                                >          za xe tad-ec-i-ne

my sister-sa-erg I:dat1 water:abs give-3sg:a-lv-aor                  I:dat1 water:abs give-pass:past-aor-3sg:s

‘My sister gave me water.’                                                      ‘I was given water.’


A:ERG - O:DAT2 V[simple]


adamar-en t’e s/um-ax u-ne-k-sa                            >          t’e s/um uk-ne-sa

man-erg dist bread-dat2 eat-3sg:a-$-pres                               dist bread:abs eat-3sg:s-pass:pres

‘The man eats that bread.’                                                      ‘That bread is eaten’.


A:DAT1 - O:DAT2 V[simple]


xunc^i-n-a xod-d-ux a-t’u-k-i                                   >          xod ak’-ne-c-i

girl-sa-dat1 tree-sa-dat2 see-3sg:io-$-aor                               tree:abs see-3sg:s-passs:past-aor

‘The girl saw the tree.’                                                            ‘The tree had been seen’


A - O -V[Inc]


kala mus^-en k’uax xarab-ne-b-i                            >          k’odz^ xarab-ne-bak-i

big wind-erg house:dat2 destroy-3sg-lv:trans-aor                    house:abs destroy-3sg-lv:intrans-aor

‘The storm destroyed the house.’                                            ‘The house had been destroyed.’


The passive morphology of verbs often coincides with a slight shift in meaning, e.g. ak’sun ‘to see’ > ak’esun ‘to be seen’ > ‘to become apparant’. Passive verbs may again be transitivized with the help of a causative formation, e.g. ak’-es-d-esun (see-pass-caus-inf2) [> ak’est’un] meaning ‘to show oneself to someone’.



Foregrounding of S:


A remarkable feature of Udi is the possibility to mark intransitive verbs as ‘passive’. The effect of this strategy can be described as establishing a factitive (sometimes progressive) meaning from stative verbs. With stative verbs, the constituent in S-function is not explicitly placed in the foreground of a scene - its factitive variant now focuses on this allocation, cp.:


pasc^’ag/ ba-ne-k-e

king:abs be-3sg:s-$-perf

‘He was a king.’




pasc^’ag/ bak-ne-c-i

king:abs be-3sg:s-pass.past-aor

‘He became king’.


Most often, passives of intransitives are derived with the help of the passivized form of baksun ‘to be’ > bakesun ‘to become’. Yet, dynamic Inc-verbs based on just this light verb may substitute -baksun by -esun, e.g. gogin-baksun ‘to be green’ > gogin-esun ‘to become green’.



4.6. Causatives


Udi has two means to derive causatives from intransitive and transitive structures. The one is based on the morphological element -ev- which is inserted between the lexical base (often an old Inc-element) and the light verb, e.g. bat’k’esun ‘to perish’ > bat’evk’esun ‘to destroy’. However, this technique is no longer productive in actual Udi. Instead, analytic constructions are preferred that are based on the series of transitive light verbs described in 3.3. Intransitive Inc-verbs either exchange their light verb (esun/baksun > besun/pesun etc.) or add a transitive light verb to the masdar1, whereas transitive verbs simply add the appropriate light verb to the masdar1 -es, e.g.


goginbaksun   ‘to be green’    >          goginbesun                 ‘to make green’

ak’esun          ‘to be seen’      >          ak’est’esun                ‘to make o.s. seen’

besun              ‘to make’         >          besdesun                    ‘to have/let s.o. do’

k’alpesun       ‘to call’            >          k’alpest’esun             ‘to have/let s.o. call’


The light verb most often used to encode causative structures is -t’esun.


Udi does not differentiate degrees of controlhood with respect to the embedded S- or A-constituent. Hence, Udi causatives translate both English have and let causatives. Causatives create a new A-slot which causes the embedded a-argument to be demoted to O. It follows that the causative relation supercedes the embedded transitive relation: [A-O[>A-O]]. The primacy of the causative relation is also documented by the fact that it now is the causative A-constituent that shows agreement with the verb, cp.:


adamar-en zax kag/e.z-ax cam-p-es-t’-i-ne

man-erg 1:dat2 letter-dat2 write-lv-inf1-caus-aor-3sg:a

‘The man had me write a letter’.


The demotion of the embedded A-constituent to superficial O is also documented by the shift in case marking: The embedded ‘agent’ is normally encoded by the dative2 (in rare cases, the absolutive may be used, too).


Causatives of passives often result in medial structures or verbal reflexivity, cp.:


s^e-t’-in xalx-n-u ak’-es-ne-d-i

dist-sa:obl-erg people-sa-dat1 see-pass:inf1-3sg:a-caus-aor

‘He showed himself to the people.’



4.6 Reflexivity


The reflexive pronoun ic^ (see covers a wide range of functions related to what is generally called ‘reflexivity’. The head of the reflexive pronoun can be


a) clausal internal


düs^man-en ic^en-ic^-ux bes-b-i-ne

enemy-erg refl:erg-refl-dat2 kill-lv-aor-3sg:a

‘The enemy killed himself.’


b) in long distance:


nut’ u?g/-a?l-le fi-n-ax-q’an sik’er-ax, va? ive?l elmug/-oxo-al tam-bak-al-le hala ic/ nana bukun-e

[Lk 1:15]

not drink-fut-3sg:a wine-sa-dat2-and siker-dat2 and holy ghost-abl-foc fill-lv-fut-3sg:s instantly refl mother:gen womb-dat1

‘He will not drink wine and siker and his mother will be instantly filled in [her] womb by the holy ghost.’


c) S=A in emphasis:


va? evaxte s^o-n-o bai-ne-c-i k’ick’e gämi-n-a, dz^in bu-o-t’-in tavaxq’a-ne-b-esa-i s^o-t’-ux, te ic^-al bak-a-ne-i s^o-t’-xol [Mk 5:18]

and when dist-sa:abs-abs go=in-3sg:s-lv:past-aor small vessel-sa-dat1 dzhin:abs be-[abs]-sa:obl-erg pleed-3sg:a-lv-pres-past dist-sa:obl-dat2 that refl-foc be-opt-3sg:s-past dist-sa:obl-com

And when he came into the ship, he that had been possessed with the dzhin prayed him that HE might be with him’.


d) Empathic :


etärte p-e-ne beš baba-g/-o Avraam-a va? ic^ dz^ins-n-u hammas^alug/-a [Lk 1:55]

as say-perf-3sg:a our father-pl-dat1 Abraham-dat1 and his seed-sa-dat1 eternity-dat1

‘As he said to our fathers, to Abraham and his [Abraham’s] seed for ever.’


The reflexive pronoun can be used in attributive function (conveying a possessive meaning) or as an independent constituent. In the latter case, the pronoun is often doubled just as in many other East Caucasian languages: the case marked reflexive pronoun is preceded by the pronoun in the ergative case, e.g.:


adamar-a ic^-en-ic^-ux bu-t’u-q’-i

man-dat1 refl-erg-refl-dat2 love-3sg:io-$-aor

‘The man loved himself.’


The distribution of simple and reduplicated reflexive pronoun roughly correponds to close and long distance reflexivity, cp.:


Close reflexivity

Long distance reflexivity

ic^-en ic^ + CASE

ic^ + CASE


Note that the ergative case cannot be motivated by the assumed A-trigger, that is *[A:erg ic^en] [ic^-CASE] as it is sometimes described in the literature. Doubling also takes place with intranstives, cp.:


har pasc^’ag/lug/ ic^-en ic^-bos^ dz^ok’-bak-al-o amc’i-ne bak-o [Mt 12 :25]

each kingdom:abs refl-erg refl-pp(in) separate-lv-part:pres-abs empty-3sg:s be-opt

‘Every kingdom that desintergrates shall become empty (i.e., be in desolation).’


Occasionally, the short reflexive can be used in close reference, too, cp.:


s^et’abaxt’inte s^in-te ic^-ux ala-ne-b-sa, s^o-n-o bak-al-le oq'alu; amma s^in-te ic^-ux  oq’alu-ne-b-esa, s^o-n-o

alalu-bak-al-le [Mt 23:12]

because who:erg-sub refl-dat2 high-3sg:a-lv-rpes dist-sa:abs-abs be-fut-3sg:s but who:erg-sub refl-dat2 down-3sg:a-lv-pres dist-sa:abs-abs high-lv-fut-3sg:s

‘Because whosoever exalts himself shall be abased; and he that abases himself shall be exalted.’