Deponent verbs in Georgian Kevin Tuite (Universit de Montral) In the summer of 2000, while on a research trip to Georgia, I came

across the following cartoon in a Tbilisi newspaper. Here is the text with a translation:

Waiter: supi ratom ipurtxebit? (Why do you keep spitting in the soup?) Customer: vsinav, cxelia tu ara. (Im testing if its hot or not.) Waiter: ??? Customer: emi coli qoveltvis egre amocmebs utos. (My wife always checks the iron like this.) Leaving aside the political incorrectness on several levels of the content of the cartoon, let us make use of it as a source of linguistic data. The verb in the first line, i-purtx-eb-it you [pl/polite] spit, is formally in the passive voice; its 3rd-person subject form would be ipurtx-eb-a. Its morphology contrasts with that of the transitive a-purtx-eb-s in exactly the same way as, say, the passive (kari) i--eb-a (the door) is opened is opposed to the active (kars) a-eb-s s/he opens (the door). If the meaning of a-purtx-eb-s is s/he spits, one would expect the first line of the above dialogue to mean something along the lines of Why are you being spit into the soup?, which is manifestly not the case. The Explanatory Dictionary of the Georgian Language (KEGL) glosses ipurtxeba spits continually, sprays spit all the time (from the mouth) (erttavad apurtxebs, cara-mara purtxs isvris (piridan)); according to Tschenklis dictionary, it means (dauernd) spucken. The near-synonymy of a formally passive verb with its corresponding transitive is not limited to this root in Georgian. The KEGL inventories 78 verbs with similar semantic profiles. Nor has the phenomenon gone unnoticed by grammarians. In his discussion of the uses of Georgian passive verb forms, Tschenkli (1958: 255-256) notes that passives die rein aktive Bedeutung besitz[en] can be employed as a type of iterative um eine gewohnheitsmssige, wiederholte Handlung oder auch eine dauernde Handlung im allgemeinen auszudrcken. Examples include i-qvedr-eb-a dauernd Vorwrfe machen, icoxn-eb-a ein Tier kut wieder, pflegt wiederzukauen. Some verbs with these features can even take direct objects: i-lev-a sb gives sthg, v-qv-eb-i ich erzhle etwas. Shanidze devotes a section of his monograph on Georgian morphology to what he calls deponents (deponensebi) (1953 366), defined as verbs which are passive in form and active in meaning. He contrasts such pairs as the deponent i-lev-a sb. gives sthg and its corresponding active a-lev-s sb. gives sthg to sb., and concludes that the primary function of deponents is the demotion of the object argument from the morphosyntactic structure of the sentence (tavidan airidos relacia anu mimarteba obiektisadmi). Shanidze makes the important observation that such verbs are neither new nor limited to colloquial Georgian. Several deponents are attested since the most ancient monuments of the Georgian language, from over 1500 years ago. The late Bessarion Jorbenadze touched upon deponents, which are formed with the version vowels i- and e-, in his book on the category of version (1983: 107-9, 136-41; see also

Deponent verbs in Georgian (Tuite) 11/9/05 page 2 Jorbenadze 1975, 1981). In his view, a verb form such as i-cer-eb-a is NEUTRAL with regard to active/passive meaning. The fundamental meaning of i-version (subjective version) is reflexivity (ukukcevitoba), which can be reflected by both transitive and intransitive constructions. It is the animacy of the subject which determines whether the verb is to be interpreted as a true passive (cerili icereba (viacis mier) a/the letter is being written (by sb)) or as a deponent ((viaca) cerils icereba (sb) is writing letters, a letter). The large number of deponents among e-prefixal passives is likewise attributed by him to the use of this prefix by both transitive and intransitive verbs, although few of the former are attested in Georgian. More recently, Aronson (1989, 1994) has commented on the difficulties posed by deponents for formbased theories of grammatical relations. The dative-case argument of a sentence such as cerils icereba should not be a direct object, since the verb form is intransitive by definition, yet there is nothing in the verb form (such as an objective-version vowel or an object-agreement prefix) which would enable one to categorize the argument cerils as an indirect object, either. In this paper I will attempt to expand upon the work of my distinguished predecessors concerning the morphosyntax and semantics of the so-called deponent verbs (henceforth DPs) of Georgian. To begin, let us consider the appropriateness of the term chosen by Shanidze to denote ipurtxeba and similar verbs. Deponent, of course, has been adopted from the vocabulary of traditional Latin grammar. Latin deponents correspond by and large to the media tantum of Greek and Sanscrit, and this lack of opposition to an active verb formed from the same stem is often cited in definitions. Another important feature is the hybrid paradigm of the Latin deponent, whence its name: It represents une catgorie de verbes actifs pour le sens, mais qui paraissent se dpouiller (deponere) de la forme active attendue, pour revtir la forme passive (Monteil 1970: 261). P. Flobert, for example, who devoted an important monograph to the Latin deponent, defined it as formellement un passif sans actif correspondant, qui complte son paradigme par recours lactif (Flobert 1967: xi). Floberts definition takes into account the fact that certain non-finite forms of Latin deponents are constructed like those of active verbs. As shown in tn the following table, the present and future participles of the deponent sequor I follow have been drawn from the active-voice paradigm:
ACTIVE PASSIVE DEPONENT

present indicative perfect present participle future participle present infinitive

pn I put posu pnns positrus pnere

pnor I am put positus sum positus pnendus pn

sequor I follow sectus sum sequns sectrus sequ

The above-cited definitions cannot, however, be transferred in their present form to the Georgian verbs Shanidze labels as deponents. Unlike Latin deponents, most Georgian DPs are contrasted with actives and often other verbal classes built from the same root. Although the conjugations of some Georgian DPs are filled out with formally active paradigms, this is not the case for most of them. The traditional characterization of deponents as passive in form but active in meaning, which Shanidze adopted for Georgian DPs, is rather vague as it stands. I believe it can be modified to accommodate the Georgian facts by specifying that active and passive refer to the deep-case or thematic roles subcategorized by the verb as well as to its morphology. Since it will be important to distinguish between these two levels of analysis in the discussion to follow, I will arbitrarily employ the term DIATHESIS to refer to the deep-case relation of the subject to the thematic frame of the verb, whereas VOICE refers to the morphology. A definition which takes into acount the relation between voice and diathesis has the advantage of avoiding impressionistic judgments of whether a given verb, taken in isolation, has active meaning or not. (Latin morior I die is commonly classified as a deponent, but some might dispute the semantic grounds for qualifying it as such). A true DP, as I define the term for the purposes of this paper, is A PASSIVE-VOICE VERB WHICH MAINTAINS THE SAME DIATHESIS AS THE CORRESPONDING ACTIVE. As a consequence, a Georgian verb can only be labelled as a DP if (i) it

Deponent verbs in Georgian (Tuite) 11/9/05 page 3 has so-called passive morphology; (ii) it is opposed to an active verb from the same stem; and (iii) the grammatical subjects of the passive and active verbs have the same thematic relation (i.e., agenthood) to their respective verbs. The hybridity which is fundamental to the notion of deponent, and which is expressed in the mixed complement of forms of a Latin deponent, appears in the relation between form and syntax in the case of Georgian (and in general, Kartvelian) DPs. The verb a-lev-s, like a not-inconsiderable number of Georgian actives, is in contrast with both DP and non-DP passives. The shift in thematic-role frame accompanying the latter, but not the former, is evident in the following examples: [ACTIVE] mdivan-i kal-s pul-s a-lev-s [secretary-NOM woman-DAT money-DAT O3gives-S3sg] the secretary gives money to the woman [DEPONENT PASSIVE] mdivan-i pul-s i-lev-a [secretary-NOM money-DAT gives-S3sg] the secretary is giving out money [NON-DEPONENT PASSIVE] kal-s pul-i e-lev-a [woman-DAT money-NOM O3-is.givenS3sg] money is given to the woman The criteria I adopt in this paper for the Kartvelian DPs would, of course, exclude precisely that type of verb labelled a deponent in Latin. Nonetheless, almost all of the verbs Shanidze lists as deponents would be admitted under my definition. I will begin with an inventory of the verb forms classified by Shanidze and others as DPs, and attempt to find some perspicuous criteria for subgrouping them. 1. DPs in Modern Georgian. Georgian verbs can be divided into four classes (sometimes called conjugations), according to the crosscutting criteria of (1) lexical aspect [atelic vs. telic, or the morphology of the future-tense form, which is strongly correlated with this distinction]; (2) the phenomenon of case-shift. The last-named term refers to the morphosyntactic properties of transitive and many intransitive verbs, which assign ergative case to their subjects in the so-called Series II paradigms (aorist and optative), and dative case in the Series III paradigms (perfect, pluperfect). MODERN GEORGIAN VERB CLASSES (CONJUGATIONS): case-shifting non-case-shifting [assigns ERG in Series II] [cannot assign ERG] FUTURE = PRESENT (CLASS 1) v-a-iv-eb (CLASS 2) m-iv-d-eb-a + PREVERB: I make sb/sthg go hungry I become hungry FUTURE: mo=v-a-iv-eb FUTURE mo=m-iv-d-eb-a FUTURE STEM (CLASS 3) v-imil-ob (CLASS 4) m-i-a PRESENT STEM: I go hungry I am hungry FUTURE: v-i-imil-eb FUTURE: m-e-i-eb-a Each of these classes corresponds to a voice category, as defined above (i.e. as a purely formal classification of Georgian verbs): Class 1 represents active voice, Class 2 passive, Class 3 active-atelic (or medioactive), and Class 4 stative (or mediopassive). Class 2, which comprises passive verbs, is divided into three formal subclasses, each of which includes absolute and relative forms. The latter have indirect-object-agreement prefixes, whereas the former do not. (a) suffixal passives (doniani vnebiti): These verbs form their stems by the addition of the suffix -d- (sometimes -(e)n- in Old Georgian) to the root. The root may be of nominal, adjectival or verbal origin, and the meaning of the suffixal-passive stem is generally inchoative: to become X (ga=citl-d-eb-a turns red < citel- red), to start X-ing (a=mer-d-eb-a begins singing < mer- sing), etc. The only example given by Shanidze of a suffixal passive with deponent meaning is da=h-pir-d-eb-a sb promises sthg to sb. The corresponding active, da=ipir-eb-s, however, has the sense sb receives a promise from sb (Tschenkli: j-m e. Versprechen abnehmen, j-n dazu bewegen et. zu tun). While the passive form is indeed rendered by active translation equivalents in western European languages (j-m et. versprechen,

Deponent verbs in Georgian (Tuite) 11/9/05 page 4 etc.), it shows the sort of shift of case relations, relative to the active formed from the same root, that is characteristic of the passive voice. In other words, its diathesis is consistent with its voice. (b) root passives (unino vnebiti): This is an archaic group of intransitive verbs, the stems of which are not derived from transitive verbs or other parts of speech. Some root passives are characterized by ablaut, which was probably a feature of the entire group in Proto-Kartvelian (Gamqrelidze and Maavariani 1965; Tuite 1998b). The only root passive with deponent meaning, again according to Shanidze, is mo=h-qv-eb-a sb recounts sthg, relative mo=u-qveb-a sb recounts sthg to sb. Here again, consideration of the thematic frame of the corresponding active-voice form (mo=a-qol-eb-s j-m et. erzhlen l[assen]) shows that the passive-voice morphology is accompanied by passive diathesis. (c) prefixal passives (iniani-eniani vnebiti): It appears that the only genuine DPs in Georgian, according to the definition given above, are of the prefixal class. I am stating this here as an empirical observation, but I hope to demonstrate below that restriction of deponents to the prefixal passive subclass is consistent with other properties of these verbs. To give an overview of the semantic range covered by Georgian DPs, I have divided the 78 i-prefixal DPs listed in the KEGL into three groups by the valence properties of the active verbs from which they are derived. Each group is further segmented by semantic features; I attach no great importance to these features, and no doubt other, equally valid, groupings could be arrived at. In certain groups, it should be pointed out, are verbs that look very much like DPs in several respects, but for which no active counterpart is known to exist. Such verbs can therefore not be definitively classed as DPs by the criteria I have proposed, but I have included them nonetheless, at the end of each subgrouping. Semantic classes of deponents. The format of the listings is: i-prefixed DP, e-prefixed DP [if any] (aorist, perfect; or P[resent] S[eries] O[nly]) English definition, adapted from KEGL [gloss, if available, from Tschenkli, < derivational information from Tschenkli: < T, KT, RM, MV = from transitive, causative-transitive, relative medioactive, (absolute) medioactive] TYPE I. base trivalent transitive, no relative DP The first group comprises verbs of giving and communicating, that is, fundamentally triactantial verbs signifying the transfer of something (object, money, speech) from a sender to a receiver. The DP form, as was illustrated above for the pair alevs ileva, has the same diathesis as its corresponding active, but with backgrounding of the indirect-object argument. The receiver is not expressed in the case frame of the DP; at the level of discursive interpretation, the receiver is either relatively irrelevant, or in some case assumed to be in the locus of the speaker (itqobineba, for instance, is glossed informs someone situated here [atqobinebs visme, aket mqops]). What appears to be direct-object argument, on the other hand, is frequently expressed. This noun phrase is assigned the dative case, as is normal for the direct object of a transitive verb in the present/future series, e.g. korispondenti iucqeba raionis axal ambebs the correspondent is reporting, reports (regularly) current news from the region. Although the argument denoting the thing, money or speech conveyed bears the earmarks of a direct object, the formal determination of its status, as noted by Aronson in the papers mentioned above, is problematic, since none of the apparently transitive DPs has Series II or III forms (which would permit one to observe the case shift diagnostic of a direct object in Georgian). The giving out of money, information or whatever is represented as a recurrent, on-going activity. None of these DPs appears in the perfective-aspect paradigms, i.e. they are what Tschenkli calls presentseries only verbs (nur Prsensreihe, abbreviated PSO in the lists below). Note that a handful of DPs are built on the causative stem of their active counterpart, although not in any systematic fashion (i-lev-a and i-lev-in-eb-a are synonymous). (A) GIVING ileva (PSO) gives to someone [(j-m) et. geben, < T alevs] ilevineba (PSO) = ileva

Deponent verbs in Georgian (Tuite) 11/9/05 page 5 (B) COMMUNICATION itvlevineba (PSO) tells someone (e.g. news) ircmuneba (PSO) tries to convince sb [versichern, beteuern (intr.), < T arcmunebs] iscavleba (PSO) teaches This is a true passive in modern Georgian, but Shanidze (1953 366) gives examples of its use as a transitive DP in the writings of King Vaxtang VI (early 18th c.): aka nadirobis rigs iscavleba Here he teaches the rules of hunting. itqvis, etqvis (PSO) will express in words [to sb] In Modern Georgian, these DPs function as the bivalent and trivalent futures of the highly suppletive verbs ambobs says and e-ubneb-a tells. This latter as well is a DP, with the same stem as the now-rare medioactive ubnob-s says, speaks. itqobineba (PSO) informs sb [(et.) mitteilen, melden, < KT atqobinebs] iucqeba (PSO) informs, tells sb [berichten, melden, kundgeben, < RM ucqis] icereba (PSO) writes to sb, informs sb through writing, letters [briefliche nachricht geben, schreiben, < T cers]; also (KEGL only) moicereba (cerils) writes (letters) [and sends them here]. TYPE II. base bivalent transitive, no relative DP Type II DPs are almost always intransitive, and based on transitive verbs of eating, gazing and body contact. What these DPs have in common is signification underlining the external contours of the action, as perceived by others or as reflective of the doers state of mind. Whereas the Class 1 verb coxn-i-s chews is normally said of ruminant animals, with a direct object describing the grass, cud, etc. being masticated, its DP i-coxn-eb-a is intransitive, and tends to be employed when the speaker wishes to emphasize the leisurely, repetitive nature of the chewing (when speaking of an animal), or to express a negative impression of a person making similar mouth movements (e.g. a child chewing gum). A sizeable proportion of the DPs of this and the following class have a distinctly expressive nature, a feature they share with certain types of medioactive (Class 3) verbs (Holisky 1981), but one that is commonly deployed to point out or sanction someones unseemly, puzzling, irritating or socially inappropriate behavior. For this reason, Type II and III DPs are not infrequently uttered in the 2nd person, either with the negative-imperative particle nu dont or in the exasperated-question format ras/rad i-X-eb-i?! Why do you [insist on, keep on] X-ing?!. The cartoon dialogue presented at the beginning of this paper is a case in point. Here are some further examples selected from the KEGL: ras irtqmevinebi?! Why do you keep on hitting?! exe, rogor dgas da iqiteba! Look how he stands there and gawks! kargia, nu icinglebi! Enough! Stop blubbering! (C) EATING iloneba (ilona, ) eats without pleasure, chews, ruminates [(mit zahnlosen Mund) kauen, appetitlos (fr)essen, < T lonis] isusneba (PSO) nibbles, eats a tiny bit (usually in secret) [naschen, < T susnis] ieeba (PSO) chews continually [(lngere Zeit) kauen, < T eavs] imurleba (PSO) eats with appetite [gierig essen, tchtig einhauen, < T murlavs] icoxneba/icoxnis (PSO) ruminates, eats in an ugly, unpleasant fashion [e. Tier kut wieder/ pflegt wiederzukauen; j-d kaut on et. herum, < T coxnis] [ga-]ibuskneba (PSO) eats with appetite (said of a child) [mit Appetit essen DP only] ilukmeba (PSO) eats with appetite, with enthusiasm [mit Appetit, gierig essen DP only] iqmindeba (PSO) eats greedily [(Kakhetian dialect)gierig essen (wie e. Ausgehungerter), DP only] (D) GAZE ibvireba (PSO) stares angrily, menacingly [grimmig/ wtend/ finster dreinschauen, < T3 ubvers] (0-/ga-/gamo-/gadmo-/da-/amo-)iqureba (PSO) gazes at sthg continually, looks at sthg [schauen, blicken, < T hqurebs]

Deponent verbs in Georgian (Tuite) 11/9/05 page 6 (0-/ga-/gamo-/gadmo-/da-/e-/emo-/a-)ickireba (PSO) has his/her eye, gaze fixed (in a certain direction); gazes [blicken, schauen; dreinblicken, < RM sckeris] (0-/amo-/gamo-/e-/a-)ivriteba (PSO) peers into holes (and similar spaces); examines carefully, at length [durch e-e kleine ffnung gucken; verstohlen wohin blicken/ schauen; amoivriteba = (aus et. Tiefem) hervorgucken, < T vrets] iqiteba (PSO) looks, gazes, stares (toward somewhere, or out from somewhere) [wohin schauen; heraus- oder herein-gucken, < T (a)qets tvalebs die Augen aufreissen] ixedeba (PSO) has his/her eye, gaze fixed (in a certain direction, toward sthg, at sthg); gazes, stares [schauen, blicken, < T xedavs] (E) BODY-CENTERED ACTION ibertqeba (PSO) shakes off sthg (dust, water-drops, etc.) from oneself (amo-/ga-/da-/mo-/e-)itxupneba (PSO) smears sthg (e.g. cosmetics) on ones face in an unattractive manner [sich verschmieren/ schmutzig m./ beschmutzen, < T txupnis] ipudreba (PSO) puts powder on ones face [sich pudern, < T pudravs] ipxaneba (PSO) scratches (continually, all the time) [e-e Katze kratzt/ pflegt zu kratzen, < T pxanis] ipxoreba (a-/ga-ipxora, gapxorila) [turkey] ruffles its feathers; sb puffs oneself up [e. Tier plustert sich auf/ strubt die Federn od. das Fell; j-d blht sich auf, tut gekrnkt, zeigt seinen Unwillen, < T2 ipxoravs] ikaeba (PSO) puffs oneself up, puts on airs [(Imeretian dialect) sich aufblhen, wichtig machen, < T kaavs] ikekeba (PSO) scratches oneself; rummages (in search of sthg), pokes around (in sthg) [sich kratzen; herum-stochern, -stbern, -whlen, < T kekavs] ikekeba (PSO) [bird] pokes through its feathers with its beak, preens itself [et. (Vogel) whlt/ kratzt mit dem Schnabel in seinen Federn DP only] (F) AGGRESSION, CONTACT WITH OTHERS BODY (0-/mi-/mo-)ibrvis, ebrvis/ ebroleba (PSO) fights, struggles against sb [kmpfen, MV with RM ebrvis] ikbineba (PSO) bites sb or sthg; has the habit of biting [et. (z.B. Hund, Schlange, Mcke) beisset/ sticht (intr.), et. (z. B. Hund) ist bissig, et. (z. B. Nessel) brennt, < RM hkbens] ikvrevineba (PSO) hits (with the hands or feet) ikortneba (PSO) pecks sb or sthg [et. (Vogel) hackt (mit dem Schnabel) auf j-n/et., < T kortnis] ikocneba (PSO) kisses sb or sthg [kssen, < T kocnis] irtqmevineba (PSO) hits sb, beats irkineba, erkineba (PSO) butts sb or sthg with ones horns [e. Tier stsst mit den Hrnern/ ist stssig, < RM hrkens] iroleba, eroleba (PSO) butts with ones horns [e. Tier stsst mit den Hrnern/ ist stssig, < RM hrolavs < RM eris] itaneba, etaneba (PSO) [Pav. dialect] pursues, chases sb or sthg [j-n/et. verfolgen, zu fangen suchen; sich auf j-n/et. strzen, < T atans] ikmiteba (PSO) pinches sb or sthg [(pflegen zu) zwicken/ kneifen, < T kmets] ixvliteba (PSO) pokes sb or sthg; has the habit of poking [et. (Z. B. Dorn) sticht, < T xvlets] ixrikeba (PSO) searches sthg, rummages [herumstochern, < T xreks] icemeba, ecema (PSO) beats sb [Schlge verteilen, Prgel geben, < T scems]
TYPE III. base intransitive or semi-transitive (especially medioactive), includes relative DPs.

Type III DPs, like the Type II verbs just listed, are formed from verbs denoting behaviors (facial expressions, movements, speech acts) likely to attract attention for their appearance or appropriateness. Unlike Type II verbs, these are based on verbs that are always or usually intransitive; many of them are relative DPs, taking an indirect object denoting the person to whom the behavior in question is directed. The meaning difference between medioactives and the DPs derived from them is not always easy to specify, as I have found when asking native

Deponent verbs in Georgian (Tuite) 11/9/05 page 7 speakers about this matter. In general, the medioactive (e.g. qeps barks) has a less specific, unmarked meaning, whereas the DP is used to emphasize the duration or habitualness of the activity (iqepeba barks all the time, has the habit of barking [inappropriately, unnecessarily]). The comitatives (group J below) are for all intents and purposes homonymous with their Class 3 counterparts, save for the addition of an indirect object paradoxical as it might seem, here passive morphology is linked to an INCREASE of valence. (G) FACIAL iimeba (PSO) a smile comes over one, laughs slightly[lcheln, < RM himis] imieba, emieba (daimia, damiila) grimaces, twists the face into an expression of displeasure [das Gesicht verziehen, Grimassen schneiden, < T mes (pirisaxes)] irieba, erieba (gairia, gareila) laughs in an ugly, unseemly manner (with the teeth showing); grimaces [grinsen, Faxen m.; das Gesicht (vor Schmerzen) verzerren, < T res (kbilebs, pirisaxes)] icingleba / icindleba (PSO) weeps, sobs in an unattractive manner, with the nose running icremleba (PSO) sheds tears, weeps [Trnen vergiessen, weinen, < T cremlavs] itrikeba (PSO) laughs for no reason [ohne Grund, sinnlos lachen DP only] itxieba (PSO) laughs in an unseemly manner; grimaces [ohne Grund/ sinnlos lachen DP only] icirpleba (PSO) sobs in an ugly, unattractive manner, blubbers [flennen, plrren, DP and passive RP1 (mecirpleba tvalebi mir triefen die Augen)] (H) MOVEMENT, BEHAVIOR [ga-]izmoreba (PSO) stretches out ones limbs [sich strecken/ recken DP & IV1 (ga-/daazmorebs), i.e. indirect T] inazeba, enazeba (gainaza, ganazebula) acts cute [sich zieren/ genieren, schchtern tun, < MV nazobs] ipraneba, epraneba (gaiprana, gapranebula) flirts, acts cute, plays the coquette (to excess) [sich zieren, schntun, kokettieren, < T pranavs] ipurtxeba (PSO) spits continually, sprays spit all the time (from the mouth) [(Iterativ) (dauernd) spucken, < T apurtxebs] iliteba (ilita, ) to succeed at getting sthg (through cunning, deception) [(Imeretian dialect) durch List et. erlangen/ zu et. kommen, < T lets zerreissen, zerfetzen] imuneba (eimuna, emunula) moves clumsily, shrugs the shoulders [sich (schwerfllig) winden, < T2 imunis (mxrebs)] imuneba (eimuna, emunula) = imuneba ixikneba (PSO) writes, scribbles in an ugly fashion [stmperhaft schreiben, < T xiknis] (0-/da-/mi-/mo-)izlazneba (PSO) moves slowly, lazily [sich lssig/ trge/ langsam fortbewegen DP only] iareba (gaiara, gaarula) tries to do sthg with all ones force [(Imeretian dialect) alles Erdenkliche tun, mit aller Kraft et. versuchen, DP only] ixasreba (PSO) [waterfall, river ] sprays water drops, mist; foams [(Pav. dialect.) Wasser sprht Gischt, DP only] (I) SPEECH-ACT igineba (igina, ) pronounces a curse, curse-words [schimpfen (intr.), Schimpfwrter im Munde fhren, < T aginebs] ivedreba, evedreba (ivedra, ) addresses pleas to sb, implores [instndig um et. bitten/ flehen, < T avedrebs] ilaneba (ilana, ) pronounces insulting words; has the habit of insulting, cursing [schimpfen (intr.), Schimpfwrter im Munde fhren, < T lanavs] (0-/da-)iloceba (iloca, ) pronounces words of blessing; blesses, toasts sb [segnen, den Segen erteilen; segnende Worte sprechen, e-n Toast ausbringen, < T locavs] imadleba (PSO) reproach, remind sb of good deeds ones has performed [Dank (fr erwiesene Wohltaten) verlangen, < T4 amadlis] isaqvedureba, esaqvedureba (PSO) expresses reproach, blame; reproaches sb [Vorwrfe

Deponent verbs in Georgian (Tuite) 11/9/05 page 8 m., < MV saqvedurobs] itqu(v)eba (PSO) says a lie; lies to sb [lgen, < MV tquis] ipiceba, epiceba (PSO) declares with an oath; swears an oath [(Iter.) schwren, versichern, beteuern , < T hpicavs] ikad(n)eba, ekadneba (PSO) boasts [(grosprecherisch) (mit et.) drohen, et. androhen; (mit et.) prahlen/ aufschneiden, < MV ikad(n)is] irineba, erineba (PSO) gives out a growling noise, growls [e. Tier (oder j-d) knurrt/ pflegt zu knurren, < MV rens] iqepeba (PSO) barks continously [e. Hund bellt (dauernd), < MV qeps] iqvedreba (PSO) says sthg with reproach; reproaches sb [dauernd Vorwrfe m., < T hqvedris] icqevleba (PSO) pronounces a curse; mentions with a curse; curses sb [fluchen, < T cqevlis] ixveceba, exveceba (PSO) asks for sthg imploringly [flehen, instndig bitten, < T axvecebs] imudareba, emudareba (imudara, ) addresses a plea to sb; implores [instndig bitten/ flehen, DP only] imukreba, emukreba (daimukra, damukrebula) pronounces a threat; threatens [(mit et.) drohen, Drohungen ausstossen, DP only] (J) COMITATIVE RELATIVE DEPONENT VERBS. This is a large, and in principle open, set of relative (e-prefixal) DPs formed from Class 3 (medioactive) verbs (Shanidze 1953 372). (According to Jorbenadzes estimate, such DPs might represent the majority of occurrences of eprefixal Class 2 verbs in Georgian (1975: 147)). The primary function of these DPs is to add an indirect-object argument denoting someone with/toward/against whom the action denoted by the medioactive is performed (Jorbenadze 1983: 95-6). With the exception of a handful of verbs listed above, comitatives do not have i-prefixal DP counterparts. Here are some representative examples: Class 3 medioactive verb v-ariq-ob I flirt, court v-tama-ob I play [intr. or tr. (e.g. football)] v-laparak-ob I speak v-kamat-ob I argue v-xub-ob I quarrel Class 2 comitative v-e-ariq-eb-i I flirt with sb, court sb v-e-tama-eb-i I play [sthg] with sb v-e-laparak-eb-i I speak with sb v-e-kamat-eb-i I argue with sb v-e-xub-eb-i I quarrel with sb

Comitatives have full paradigms, including Series II and III forms, although the latter are commonly borrowed from the paradigms of Class 1 verbs (e.g. v-e-tama-eb-i; AORIST v-e-tama-e; PRES. PERFECT m-i-tama-n-i-a (mastan)) (Tuite 1996). As I hope to have demonstrated, Georgians DPs form a coherent set in both formal and semantic terms. Formally, they are morphologically passive (Class 2) verbs with active diathesis. Semantically, they contrast with the Class 1 or Class 3 verbs formed from the same roots in one or more of the following ways: (i) DPs express repeated, habitual actions, sometimes with the implication that they are characteristic of the subject. (ii) Many DPs, those of Types II and III, signal a shift of focus from the end-point to the contours of action denoted by the verb (esp. its appearance, impression made on observers). (iii) Type I and II DPs are generally characterized by valence reduction relative to the corresponding Class 1, although the valency change is not accompanied by diathesis shift, as would be expected with a passive verb. In the case of Type I DPs, it is the INDIRECT object which is backgrounded, whereas the direct object may be expressed as a dative-case NP. Type II DPs are almost always intransitive, with backgrounding of the direct object. In this respect, Modern Georgian morphosyntax shares some features with the primary-object language type proposed by Dryer (1986) and Blansitt (1984). Such languages are characterized by morphosyntactic

Deponent verbs in Georgian (Tuite) 11/9/05 page 9 operations which treat notional indirect objects, and the notional direct objects of verbs that lack indirect objects, as a distinct grammatical relation (primary object). The primary object relation is accorded a greater morphosyntactic prominence than the secondary object relation, by which is meant the notional direct objects of verbs which also have indirect objects. (On the relevance of the primary-vs.-secondary object distinction to Modern Georgian, see Tuite (1998a: 21-22)). In other words, Type I and II Georgian DPs have the profile of a type of antipassive, albeit one that backgrounds the primary object rather than the direct object. Those DPs formed from verbs which are already intransitive (Type III DPs), as we have seen, do not reduce the valence, and some the comitatives actually increase it. (iv) DPs as a whole rarely appear in the perfective paradigms. The exceptions cluster in certain semantic subgroups of Type III: comitatives, some speech-act and behavior DPs. There is a strong correlation between relative (e-prefixal) DPs and the availability of perfective verb forms. The comitatives in particular, which function principally as the relative correlates of certain Class 3 verbs, all appear in Series II and III paradigms. 2. DPs in Old Georgian and in other Kartvelian languages. DPs are neither a recent innovation of Georgian, nor are they limited to that language. Several DPs are attested in Old Georgian literature (Imnaishvili 1979; Sarjveladze 1987), including the very oldest texts (those written in the so-called xanmeti dialect (Tuite 1990)). Among them are the very frequentlyused pair x-i-tqw-i-s sb says sthgand its relative form x-e-tqw-i-s sb says sthg to sb. Here are some further instances: vitarca x-i-sadil-n-es, xrkwa simon as they were dining, Simon said (Jn 21:15 Xanmeti Gospels (Kajaia 1984)) arca vsuldet da e=i-am-eb-od-i-t urtiertsa nor should we hate and consume each other (Mamata Scav. 268: 21; cited by Imnaishvili (1968: 51)) xolo ukuetu urtertas i-kbin-eb-od-i-t da e=i-am-eb-od-i-t ... But if ye bite and devour one another (Galatians 5:15) picxlad mo=i-gin-eb-od-a He was swearing hotly (Amiran-Darejaniani (XIth c.) 425) No investigation has heretofore been made, to my knowledge, of DPs in Zan (LazMingrelian) or Svan, but a cursory examination reveals that the matter merits a closer look. Qipshidze (1914/1994: 057) notes in passing that in Mingrelian sometimes one and the same form may have active and passing meanings, citing the example i-b-ar-u-ap-u-k, which, like its Georgian equivalent v-i-cer-eb-i, can mean either I am written or I am writing. On being presented with Georgian DPs, the handful of native Mingrelian speakers whom I asked readily supplied equivalents for at least some of them. One speaker, who spent much of his childhood in a linguistically conservative, monolingual environment, rejected some of the forms supplied by his wife as artificial morphological calques on the Georgian. Nonetheless, several Mingrelian DPs passed muster, e.g.: i-purtin-u-ap-u-(n) spits [Geo. ipurtxeba] i-gor-ap-u-(n) curses, swears [Geo. igineba] i-tqob-in-u-ap-u-(n) says, reports [Geo. itqobineba] I confirmed two of the above in Qipshidzes Mingrelian glossary (1914/1994), although the form he cited for the verb say is a near-cognate to Georgian [x]-i-tqw-i-s rather than itqobineba: v-i-gor-ap-u-k rugajus [= swear], menja rugajut (i.e. has both DP and passive readings) < active v-o-gor-an-k v-i-tq-u-u-k / i-b-tq-u-(u-)k skazhu (suppletive; Series II root = ragad-) If DPs seem to be less common in Mingrelian than Georgian, in Svan, by contrast, they

Deponent verbs in Georgian (Tuite) 11/9/05 page 10 appear to have found particularly fertile ground. As in Georgian, many Svan medioactives have e-prefixal DP comitatives, e.g. x-e-wd-il asks sb [Geo. ekitxeba], x-e-msaxwir serves sb [Geo. emsaxureba] (examples from Topuria (1967: 180-181)). DPs also abound in roughly the same semantic fields as in Georgian, e.g. speech acts, physical contact, gaze, etc. Here are a few of the dozens of DPs to be found in Lipartelianis dictionary of the Cholur subdialect of Svan (1994); I have chosen roots for which both absolute and relative Class 2 entries are given: i-nnl; x-e-nnl spreads gossip (about sb)(< nin tongue, with the vowel lengthening characteristic of denominal verb roots (Chumburidze 1981) + medioactive formant -l) i-ti:tanl has ones hands all over everything; x-e-ti:tanl tries to grab (sb, sthg) with ones hands(< ?twet hand) i-aurl; x-e-aural asks (sb) detailed questions about everything < medioactive verb aur-e investigates everything in detail i-pliel; x-e-pliel begs (sb) for sthg with the offer of bribes i-qioll; x-e-qioll stares (at sb) balefully, with half-closed eyes < ? qivr crosseyed or nearly-blind person This is clearly a large, and probably open, set of verbs. In a recent thesis on Svan stative verbs, Roena Chkadua [1999] described a further subtype of DPs linked to expressive statives, which are typically used in negative depictions of an individuals pose or behavior, e.g.:
STATIVE: RELATIVE STATIVE: DEPONENT:

qlip stands (amazed) with mouth open [pirdaebuli dgas] x--qlip stares with mouth open [pirdaebuli miserebia] i-qlip-an-l gapes with mouth open [pirdaebuli iqureba]

But the most striking evidence of the success of DPs in Svan is their penetration into the paradigms of active verbs. What is a lexical process elsewhere in Kartvelian has become inflectional in Svan. In the Lower Bal dialect of Svan a dialect characterized by numerous conservative morphological features DP verb forms are employed as the imperfective futures of Class 1 verbs. (Svan, unlike Modern Georgian, has distinct perfective and imperfective futuretense paradigms). The Lower Bal imperfective futures are formed by addition of the suffix un/wn- and the series marker -i, the same as that used with most types of passive verbs. The version vowels shift to i- for absolute and e- for relative verbs, as with ordinary prefixal passives (Chumburidze 1986: 167; Topuria 1967: 185): a-m-e mows (hay) > IMPERFECTIVE FUTURE i-m-un-i will be mowing (hay) x-o-m-e mows (hay) for sb > IMPERFECTIVE FUTURE x-e-m-un-i will be mowing for sb
PRESENT PRESENT

In the other Svan dialects, the imperfective futures of Class 1 verbs retain the same version vowels as in the present tense, but the suffixes are essentially the same as in Lower Bal. The Upper Bal equivalents of the absolute and relative imperfective futures shown above would be a-m-un-i and x-o-m-un-i, respectively. Other clues in the morpho-logy, however, point to the DP origin of these forms. One, already mentioned, is the series marker -i, linked with passive voice (although some Class 1 verbs take it as well). Another is the Svan conditional mood, which, as in Georgian, is formed by adding the imperfect-tense endings to the future stem. In the case of the imperfective conditional, the suffix used is l-/ol, which is identical to the passiveimperfect formant. Compare the active imperfective conditional (Upper Bal a-qn-un-l, Lashx a-qn-n-l, Lentex a-qn-un-ol, Lower Bal [Becho subdialect] i-qn-un-l would be ploughing sthg), to the passive-imperfect (Upper Bal, Lashx i-qn-l-(da), Lentex i-qn-l-(da) sthg was being ploughed). A further clue, I believe, is the suffix un- and its variants (-wn-, -n-, -en-, -(i)n-). These are formally homophonous with the causative formant and its allomorphs, and furthermore the distribution of variants is highly similar:

Deponent verbs in Georgian (Tuite) 11/9/05 page 11 CLASS 1 present imprfv future causative present CLASS 3 present imprfv future causative present CLASS 4 present imprfv future causative present

a-m-e mows (hay) a-m-un-i will be mowing (hay) x--m-un-e makes sb mow (hay) ckl-i whimpers i-ckl-in-i will be whimpering a-ckwl-n-e < *a-ckl-in-e makes sb/sthg whimper x-a-x-a sb is named sthg x-e-x-(en)-i sb will be named sthg x-a-x-en-i names sb sthg

The hypothesis which comes most readily to mind juxtaposes the causative formant in the DP imperfective future of Svan to that spora-dically attested in Georgian DPs (i-lev-in-eb-a gives; i-kvr-ev-in-eb-a, i-rtqm-ev-in-eb-a hits). These otherwise unmotivated causative suffixes seem likely to have been added to the DPs morphological structure (independently? in Proto-Kartvelian?) to compensate for a perceived dissonance between the passive voice and the active diathesis. 3. The significance of deponents for Kartvelian historical morphology. Genuine DP verbs in Georgian, Svan and Zan belong to the formal category of prefixal Class 2 verbs; DPs as defined here do not appear in the root and suffixal Class 2 subgroups. One defining feature of prefixal Class 2 verbs is the presence of the preradical version vowels i- and e-. The Kartvelian category of version (subjective, objective, neutral and superessive) has been discussed in detail by a number of researchers, notably Boeder (1968), and there is no need to repeat the basic facts here. The presence of the version vowel i- in both subjective-version Class 1 verbs and in one subgroup of Class 2 verbs (prefixal passives) reminded Shanidze of the morphological similarities of middle and passive voice in Greek (1953: 317-8, 362-3): Greek looses for oneself; is loosed Geo. i-xsn-i-s (Class 1 subjective-version) looses, undoes ones sthg [e.g. button]; looses for oneself; aorist ga=i-xsn-a; Geo. i-xsn-eb-a (Class 2 prefixal passive) is loosed, undone; aorist ga=i-xsn-a Shanidze hypothesized that the Georgian prefixal passive originated from the subjectiveversion active even as the IE passive was born from the more ancient middle voice [see also Imnaishvili 1968; Schmidt 1962, 1965). I will not discuss this proposal here, as it would take us too far afield for the purposes of this paper. I will limit myself to a demonstration that the morphology of Proto-Kartvelian reflected UNDERLYING rather than surface transitivity. In other words, while the ancestors of Class1 and prefixal Class 2 verbs were not, as Shanidze surmised, contrasted as active and passive (as we commonly understand those terms), they did share a semantic structure we can qualify as TRANSITIVE, in that the grammar distinguished them in systematic ways from the various types of intransitive verbs. 3.1. Deep and surface transitivity. Let us begin with the formal subgroups of Class 2 verbs in Early Georgian (the language of the oldest known Georgian texts, 5th-8th c. AD; the two examples below are from the Xanmeti Gospels, c. 600, edited by Kajaia (1984)) . Of the three types (suffixal, root and prefixal), the last stands out from the others in the following respects: (i) Agreement in (e)n- with plural nominative-case NPs. Number agreement in (e)n- is restricted to certain types of finite verbs (Class 1 and prefixal Class 2) in certain forms (Series II, and those Series III forms which are based on Series II [aorist] stems). It does not occur with root

Deponent verbs in Georgian (Tuite) 11/9/05 page 12 or suffixal passives. In the following excerpt from the Old Georgian gospels, the active and prefixal passive forms of the verb meaning to castrate appear with the nominative-plural agreement suffix (e)n-; in the passive verb it marks the plurality of the surface subject, and in the active, that of the direct object: arian saurisni, romelni gamo=x-i-sauris-n-es kactagan; da arian saurisni, romelta gamo=i-sauris-n-es tavni twisni sasupevelisatwis cataysa (are eunuchs, who-NOM.PL weremade.eunuch-NOM.PL from-men; and are eunuchs, who-ERG.PL made.eunuch-NOM.PL themselves-NOM.PL for-kingdom of-heavens) there are eunuchs, who were made eunuchs by men; and there are eunuchs, who made themselves eunuchs for the kingdom of the heavens [Matthew 19: 12] gamo=x-i-sauris-n-es [preverb=O3-passive-EUNUCH-NOM.PL-S3pl] they were made eunuchs (I-PREFIXAL PASSIVE = eunoukhsthsan) gamo=i-sauris-n-es [preverb=SubVersion-EUNUCH-NOM.PL-S3pl] they made themselves eunuchs (ACTIVE, SUBJECTIVE VERSION = eunokhisan) (ii) Morphological bipersonalism. In the Georgian of the earliest texts, but not in later forms of the language, the 3rd-person object prefix (x-) could appear before vowels. In most instances the O3 prefix could be linked with an overt or presupposed dative-case argument. The most vexing exception to this correlation were the i-prefixal passives, which were syntactically monovalent, only governing a single nominative-case NP, but morphologically bipersonal, with both subject and object agreement markers in the verb (Jorbenadze 1975: 144-145; Tuite 1990): mi=x-i-qwan-a igi angeloztagan [preverb=O3-passive-TAKE-S3sg he:NOM angelGENpl-from) He was taken by the angels . [Luke 16:22] (iii) Transitive semantic structure. Prefixal passives, but not the other types, may appear with demoted agent phrases, marked by the postpositions mier by or gan from (as in the example just cited). In those cases where both prefixal and root or suffixal Class 2 stems can be derived from the same verb root the difference is easily discerned. One can say (data from Modern Georgian) deda xaapurs a-cxob-s Mother is baking the cheesebread, with a Class 1 verb; and xaapuri i-cxob-a dedis mier The cheesebread is being baked by mother, with an iprefixal passive; but the root passive from the same stem cannot cooccur with an agent phrase: xaapuri cxveb-a (*dedis mier) The cheesebread is baking (*by mother) (Harris 1985:60-61). I believe these three characteristics of prefixal passives are interrelated. The contrast between icxoba and cxveba is similar to the contrast between what are sometimes termed passive and unaccusative constructions in other languages. The semantic structures underlying the two constructions are quite different:
TRANSITIVE SEMANTIC STRUCTURE

(a1) active construction: agent subject (a2) passive construction: agent oblique NP

patient/theme direct object patient/theme subject patient/theme subject

INTRANSITIVE SEMANTIC STRUCTURE

(b) unaccusative construction:

The three morphological types of Class 2 verbs have equivalents in Mingrelian and Laz. Although the two Zan languages have neither 3rd-person object prefixes nor nominative-plural agreement, prefixal passives are distinguished from root and suffixal Class 2 verbs by an additional series marker -u-, the origin of which is not entirely clear (Danelia 1976):

Deponent verbs in Georgian (Tuite) 11/9/05 page 13 prefixal Class 2: i-kir-u-u-n [< *(x)-i-r-aw[?]-aw-n; cp Geo x-i-r-eb-i-s] is being cut root Class 2: tib-u-n [< *tp-aw-n; cp Geo tp-eb-i-s] becomes warm suffixal Class 2: in-d-u-n [< *siw-d-aw-n; cp Geo siv-d-eb-i-s] gets wet The double series markers of i-kir-u-u-n can be compared to the retention by Old Georgian and Svan prefixal passives of the series markers of their active counterparts, followed by a second series marker -i- < *-ej-, shared by all Class 2 verbs; e.g. x-i-mal-v-i-s hides, is being hidden, Class 1 x-mal-av-s hides sb/sthg; Svan x-e-b-m-i is bound to sthg, Class 1 xa-b-em binds sb/sthg to sthg (Topuria 1967: 181). (Svan has no suffixal passives. One of the allomorphs of the passive suffix might underlie either the Series I formant -en- of ablauting Class 2 verbs (deg-en-i it [e.g., fire] is going out), or the passive-aorist suffix -n- < ? *-en-a, or perhaps even both (Deeters 1930: 201-206; Tuite 1997). The semantic range covered by Georgian and Zan suffixal Class 2 verbs inchoative, change-of-state is in Svan entirely given over to the prefixal type; cp. Svan i-qwtil-i becomes, turns yellow [i-prefixal], and Geo. qvitl-d-eb-a, Mingrelian /viton-d-u-n [suffixal]). The Series III paradigms present perfect, pluperfect, perfect conjunctive of Class 1 verbs are, as has been noted many times, the result of the appropriation or recruitment of certain passive verb forms into the active conjugation (Deeters 1930: 166-191; Schmidt 1962, 1979; Harris 1985: 285-306). The forms in question are the prefixal passive and the passive of state, both of which, I argue, have transitive deep-case frames. The e-prefixal Class 2 aorist and optative are homophonous with the Georgian pluperfect and perfect conjunctive, respectively, of most Class 1 verbs. The suffix -n- used to form the Svan pluperfect and perfect conjunctive stems has likewise been appropriated from the morphology of the prefixal-passive aorist stem (cp. Class 2 prefixal aorist n-mr-n < *an=i-mr-n was prepared and the Class 1 pluperfect: ox-mr-n < *an=x-o-mr-n had prepared sthg). Furthermore, the Series III stems of ablauting verbs are marked by lengthened i-grade, the same as that of prefixal passives formed from the same roots: cp. Class 1 present perfect x-o-dg-a has put out, extinguished [e.g., fire]; and Class 2 prefixal-passive stem i-dg-i is extinguished (by sb). This lengthened-grade passive is semantically contrasted with the root Class 2 stem (deg-en-i is going out), in that it denotes an event or state with an underlying agent, i.e. it is transitive at the level of deep semantic structure (Tuite 1997; 1998b). The other source of Class 1 Series III morphology is the passive of state, from which are derived the present perfect stems of all Kartvelian languages. One morphological indicator of their underlying transitivity is the appearance of phantom 3rd-person object prefixes, comparable to those attested in Early Georgian i-prefixal passives, in both Georgian and Svan. Examples include Old Georgian h-g-i-e-s exists, and h-nt-i-e-n is lit, both of which are based on transitive stems (-g- set up, build; -nt- light), and Svan x--b is bound (Topuria 1967: 208); x-a-gp-a it is set, spread across (dagebulia, gadalilia (Cholur subdialect, Liparteliani 1994: 339)); x-a-riv-a a runny substance is spilled (txeli nivtiereba aris davrili (loc. cit., 344)). It is also worthy of mention that stative passives in all Kartvelian languages employ epassive stems for their future, aorist and optative paradigms. Phantom 3rd-person object prefixes in passives of transitive stems. -g- set up, build -nt- light -cer- write STATIVE PASSIVE x-g-i-e-s x-nt-i-e-s x-cer-i-e-s PREFIXAL PASSIVE x-i-g-eb-i-s x-i-nt-eb-i-s x-i-cer-eb-i-s The characteristics mentioned here support the hypothesis that prefixal passives shared certain morphosyntactic features with Class 1 verbs, which they did not share with other types of Class 2 verbs. This formal pattern, it seems to me, harks back to an earlier stage of Kartvelian verb morphology, which marked a fundamental distinction between underlyingly-transitive verbs and underlyingly-intransitive ones. This distinction cross-cut lexical aspect groups, and the current division of verbs by conjugation classes:

Deponent verbs in Georgian (Tuite) 11/9/05 page 14 underlying TRANSITIVE INTRANSITIVE valence: morphobipersonal morphology syntactic deponent reading available properties: [Geo.] ABS-PL agreement -(e)n [Zan] double series markers CLASS 1 (i) from transitive verb roots (ii) from intransitive roots (ablaut, derivation) (iii) denominal, etc. CLASS 2 i-/e-prefixal passives root intransitives (ablauting); suffixal (inchoatives, denominals in *-d-/-n-) CLASS 3 medioactives [possibly some atelic intransitives which later shifted to Class 3, e.g. Old Geo qiva crowed CLASS 4 passives of state mediopassives 3.2. The origins of the prefixal Class 2 Series I stem. In Proto-Kartvelian, the ancestors of Class 1 and Class 2 prefixal verbs were underlyingly transitive, and this was reflected in their morphology and syntax. The formal resemblance is particularly close for Class 1 subjectiveversion and Class 2 i-prefixal verbs. Shanidze wondered if the mixed active and passive nature of the Georgian DP could be a vestige of the origin of the i-prefixal passive from the active, in particular, the subjective-version stem (1953: 365; cp an earlier discussion in Shanidze 1925/1981)? I believe his hunch is basically correct, although the link between DPs and Class 1 subjective version is less direct than he appears to have thought, and the parallels with the IndoEuropean middle voice less evident. It has been pointed out on several occasions that the relation between the case-assignment properties of the Series I and Series II stems of Class 1 verbs goes back to an ancient antipassive transformation, which promoted the ergative-case agent NP of a transitive verb to nominative (absolutive) case, and demoted the patient NP to dative case (Aronson 1979; Boeder 1979; Harris 1985; Heath 1976; Palmaitis & Gujejiani 1986: 60). At the same time, the shift in diathesis between what are now the Class 1 subjective version and Class 2 i-prefixal passive was, as Shanidze argued, reflected in the syntax (case-assignment, also O3 agreement [with demoted agent? Tuite 1990]) rather than the morphology. The Series II stem -ibertq- permitted both active and passive readings without a change in its form (I am employing Early Georgian forms here, but assume that the situation was characteristic of at least a late stage of Proto-Kartvelian as well). Series I and II stems of neutral and subjective versions (ancestors of Class 1 and i-prefixal Class 2 verbs) in Proto-Kartvelian [Old Georgian forms shown] (and objective, superessive) version ( active voice)
NEUTRAL SUBJECTIVE

version ( middle voice) passive diathesis x-i-bertq-i-s [dust, snow] is shaken off x-i-bertq-(a)v-i-s [dust, snow] is being shaken off

SERIES II [permansive] SERIES I (antipassive) [present]

bertq-i-s shakes sthg [dust, snow] off x-bertq-av-s is shaking sthg off

active diathesis i-bertq-i-s shakes sthg off ones body x-i-bertq-av-s is shaking sthg off ones body

The nature of the semantic and formal contrasts among the above Series II stems was not, however, strictly mirrored in Series I. For the forms associated with active diathesis (-bertqand the active reading of -i-bertq- in the above diagram), the Series I stem ([i]-bertq-av-) represented the output of antipassivization, marked in the verb stem by the affixation of a series

Deponent verbs in Georgian (Tuite) 11/9/05 page 15 marker [SM]. In the case of passive diathesis, the situation was more complex. The constructions man is i-bertq-i-s s/he:ERG it:NOM shakes-off and is x-i-bertq-i-s (mis mier) it:NOM isshaken-off (by him/her) are related by a diathesis shift essentially unmarked in the verb morphology, a close parallel to the active-reflexive and passive readings of the IE middle (Kuryowicz 1964: 74). The ancestor of what is now the Series I i-prefixal passive stem was not derived from the Series II passive, but rather from the subjective version (middle), probably via its antipassive stem (-i-bertq- -i-bertq-av- -i-bertq-av-i-). The Zan and Georgian data point to Series I prefixal passive stems with two SMs, of which the second is either *-ej- (> Georgian -i-, Svan -i-, Zan -e-) or *-aw- (> Georgian -av-, Svan -a-, Zan -u(n)-). These latter stand apart from all other SMs in a number of respects, and for this reason I have elsewhere classed them together as so-called Group 1 SMs, distinguished from Group 2, which comprises all the rest (Tuite 1998b). As shown in the following table, the Group 1 SMs appear in a bewildering array of contexts: stative and atelic verbs, perman-sive-habitual and potentialis forms, and in the Series III paradigms of Class 1 and Class 3 verbs. They also are employed to form the Series I stems of certain Class 1 verbs, whereas most verbs of this class make use of one of the Group 2 SMs. Distribution of Group 1 and Group 2 series markers in the Kartvelian languages. (SV = principal marker; SV = frequent; sv = infrequent; = absent) GROUP 1 GROUP 2 series markers *-EJ *-AW *-AM/-EM *-EW OTHERS (Proto-Kartvelian, G. -iG. -avG.-am/em G. -eb/evG.-ob/op Georgian, Zan, Svan). Z. -eZ.-u(n)/(n) Z. -umZ. -ap/anZ. -u-an S. -iS. -aS. -emS. -eS. -sg/l Class 1, 3 Series I ge, SV GE, ZA ge, sv, za GE, SV, ZA ge, sv, za Class 1, 3 Series III GE GE, SV, ZA ge thematic passives of state GE, SV GE, SV, ZA permansive- habitual GE potentialis passive (GE, SV), ZA stem-final SM in Class 2 GE, SV, ZA (ge), ZA Series I stems In stative-passive verbs (e.g. Geo. x-tes-i-e-s [grain, crops] is sown; Zan b-u-n is tied < *x-b-aw-n) the Group 1 SMs are linked to resultativity and PASSIVE diathesis; in the context of Class 1 verbs such as Geo. v-txr-i , Svan xw--txr-i I dig sthg; or Geo. v-qn-av, Ming. b-xon-n-k I plough sthg, these same suffixes are associated with durative-linear aspect and ANTIPASSIVE voice (hence active diathesis), at least historically. The same paradoxical conjunction of seemingly contrastive aspectual and diathetic properties surfaces in the deponent and passive readings of Class 2 prefixal verbs. A form such as xibertqvis < *x-i-bertq-aw-ej-s could be read as either the passive OR the antipassive of the Class 1 subjective-version present xibertqavs; in the latter instance, xibertqvis would be in fact a double antipassive, since the Series I form xibertqavs was itself originally the antipassive of ibertqis. The Georgian, Svan and Zan DPs, with their characteristically antipassive traits of object backgrounding and durative-habitual aspect (Heath 1976; Cooreman 1994), have therefore a long pedigree in Kartvelian morphosyntax. The paradox that a single verb form can have both active and passive interpretations has its roots in the more fundamental paradox of the semantic range of the Group 1 SMs. 3.3. The special case of the e-prefixal conjugation. The case of e-prefixal Class 2 verbs, which in the modern Kartvelian languages represent the relative (indirect-object marking) forms of i-prefixal Class 2 verbs, is somewhat more complicated. Whereas there is near-consensus among Kartvelologists concerning the link between Class 1 subjective version and i-prefixal

Deponent verbs in Georgian (Tuite) 11/9/05 page 16 Class 2, the prehistory of the e-prefixal conjugation remains obscure. In all Kartvelian languages this version vowel is linked to the presence of a dative-case-marked indirect object (Jorbenadze 1983: 105). In Zan the e-prefixal conjugation is limited to Class 2 verbs. This also true for the most part in Georgian; from the earliest texts onward, however, a handful of undeniably Class 1 or Class 3 verbs are attested with e-prefixal stems (note the imperfect stem formants -d- and -id-, rather than the -od- diagnostic of Class 2 verbs). Here are some examples from the Early Georgian (Xanmeti) gospels: saxelsa missa carmartni x-e-s-v-id-en and the Gentiles shall hope in his name [Mt 12: 21] eri igi x-e-i-eb-d-a mas the people were seeking him [Lk 4: 42] x-e-kicx-ev-d-e-n mas da tanon igi they shall mock him and torment him [Mk 10: 34] puri weni samaradisoj mo=m-e-c wen des give us this day our daily bread [Mt 6: 11] Shanidze (1953: 365) interpreted these facts as evidence that the e-prefixal passive originated from an e-prefixal active of much wider distribution in prehistoric Georgian, even as the i-prefixal passive originated from the i-prefixal active (i.e. Class 1 subjective version). Jorbenadze (1975: 149) added to the evidence cited by Shanidze the high number of e-prefixal DPs, especially the comitatives (group J) in support of the hypothesis that evidence that epassives originally were associated with active meaning, and only later shifted toward passivediathesis contexts. If, as Shanidze postulated, the e-prefixal conjugation was once characteristic of Class 1 verbs, and furthermore was in roughly the same relation to the e-prefixal Class 2 verbs as the Class 1 and Class 2 i-prefixal conjugations were to each other, i.e. something approximating the range of middle voice in the classical Indo-European languages, what would their range have been? Perhaps a clue can be discerned in the scanty evidence of the use of the e-prefix outside of Class 2. In addition to the Xanmeti examples cited above, the Old Georgian corpus includes the verbs e-glov-da, mourned for sb, e-tq-eb-da bewailed sb, e-r-d-a obeyed sb [imperfects shown, since many of these verbs are not attested in the present tense]. What these verbs have in common is atelic lexical aspect many of them are Class 3 (activity) verbs in Modern Georgian and a seeming incompatibility with Series II paradigms, which in Old Georgian marked punctilear aspect. (For example, the Series I conjunctive-future x-e-kicx-ev-d-e-n, in the passage from the Gospel of Mark quoted above, occurs within a long sequence of Series II optativefuture verb forms). The verb to give in Georgian is triply anomalous, since (1) it is a Class 1 verb with eversion; (2) the e-version is employed in Series II, and only there; (3) it is limited to 1st and 2ndperson subject forms (S1sg mi=xw-e-c I gave it to sb, S2sg mi=x-e-c you gave it to sb; but S3sg mi=x-c-a). The last-named anomaly is not limited to this verb, and indeed finds an echo in Svan, where distinct vowel-syncopation patterns in the 1st and 2nd-singular aorist and imperfect lead me to reconstruct a forward displacement of the accent in the S1 and S2 forms of the past indicative paradigms for certain ancient verb types in Proto-Kartvelian (Tuite 1997: 9; in press). This ancient mobile accent might enable us to explain away the problematic version vowel in mi=xw-e-c, which furthermore is not attested in verbs formed from the cognate root -- of Zan. In the Old Georgian imperfect of verbs taking the SMs av and -am, syncopation of the vowel of the imperfect stem-formant id- is accompanied by umlaut of the /a/ in the preceding syllable (Kiknadze 1947). It might have been the case that a similar phenomenon produced the eversion vowel of the S1 and S2 aorist of c- give; i.e. an ancient prefixal a- (marker of the socalled superessive version) was raised to e- under the influence of a syncopated S1/2 aorist suffix i: Old Georgian verb forms with mobile stress in the past-indicative paradigms. IMPERFECT (prs. x-kl-av-s kills) AORIST (prs. x-[a]-c-em-s gives) S1sg x-w-kl-ev-d *x-w-kl-v-i-d-i x-w-e-c *x-w--c-i S2sg x-kl-ev-d you were killing x-e-c you gave S3sg x-kl-v-id-a *x-kl-av-d-a x-c-a *x-a-c-

Deponent verbs in Georgian (Tuite) 11/9/05 page 17 With the give verb set aside, all attestations of e-version with non-Class-2 verbs in Old Georgian are limited to the Series I stems of certain atelic-activity verbs (i.e. Class 3). Here again, data from Svan prove useful. In the Upper Bal dialect, the imperfect stem of prefixal Class 2 verbs comprises the formant l-, e.g. x-e-x-i sbs sthg burns, impf. x-e-x-l-[da]. A number of e-prefixal verbs with meanings comparable to those of Georgian and Svan Class 3 verbs, though, form their imperfects without the distinctive l- formant; indeed, their imperfect stems are identical to those of regular Class 1 verbs: x-e-l-e guards, impf. x-e-l-a; x-e-lde shepherds sthg [flocks], impf. x-e-ld-da; x-e-gwem asks sb for sthg, impf. x-e-gom-da (Topuria 1967: 182-3). There is no formal reason for not classifying these as Class 3 or Class 1 verbs. I conclude from these facts that e-version, like i-version, was once common to active and passive verbs, and that its meaning did not make reference to diathesis as such. If ProtoKartvelian i-version was, as Shanidze argued many decades ago, a sort of middle voice permitting active and passive readings, then e-version would have been its relative counterpart. What semantic range would have been encompassed by the [+ indirect object] form of subjective version in Proto-Kartvelian is difficult, if not impossible, to ascertain from our present perspective. The Old Georgian and Svan data point to its being focussed around, and perhaps restricted to, active atelic (medioactive) verbs. A few of the Old Georgian e-prefixal actives have attested i-prefixal counterparts, which might preserve some vestige of the meaning opposition between absolute and relative forms of this verb type, e.g. i-glov-da, mourned, e-glov-da, mourned for sb. In any event, the close association between i- and e-version (especially the latter) and the medioactive verb class is clearly very old in Kartvelian (note that the translation equivalents of many Georgian Class 3 verbs were media tantum in the classical IE languages, even as many other Indo-European middle-voice verbs overlap semantically with i-version Class 1 (Gonda 1960; Jannaris 1968: 284-6; Hatcher 1942/1973: 21)). The apparent unavailability of eversion for Class 1 verbs may stem from aspectual factors, or perhaps competition from the objective version marker -u-. In the following diagram, I present my reconstruction of the distribution of i- and e-version forms in Proto-Kartvelian (only present-tense forms are shown, and the examples come from Old Georgian): Absolutive and relative forms of certain underlyingly-transitive verb types in ProtoKartvelian.
NEUTRAL VERSION SUBJECTIVE VERSION

ABSOLUTE (no indirect object) RELATIVE (+ indirect object)

active (Class 1) x-cer-s writes sthg x-u-cer-s writes sthg for sb

active (Class 1) x-i-cer-s writes sthg for oneself

medioactive (Class 3) x-i-glov-s mourns

passive/deponent (Class 2) x-i-cer-eb-i-s is writing; is written

x-e-glov-s x-e-cer-eb-i-s mourns sb (writes to sb [unattested]); is written for sb

In the final analysis, much of the bafflement evoked by deponents and subjective version (or middle voice) is due to linguists reluctance to admit that morphologically-rich languages, the grammars of which accord clear and distinctive marking to a variety of semantic categories, could tolerate, as it were, ambiguity in subject-verb relat ions (what I have been calling diathesis). What is considered normal in morphologically-impoverished languages such as English consider the active, passive and reflexive readings of wash in the sentences He washes the shirt, The shirt washes easily, He never washes with soap seems positively perverse in Kartvelian or ancient Indo-European. Like beauty, however, paradoxes only exist in the eye of the beholder.

Deponent verbs in Georgian (Tuite) 11/9/05 page 18 REFERENCES. Aronson, Howard. 1979. Towards a typology of transitivity: the strange case of the Georgian subject, Chicago Linguistics Society parasession on the elements: 297-306. Aronson, Howard. 1982. On the status of version as a grammatical category in Georgian, Folia Slavica 5: 66-80. Aronson, Howard I. 1989. Inflection vs. derivation in Georgian conjugation, Non-Slavic languages of the USSR: linguistic studies, H. Aronson (ed.), 1-19. Chicago: Chicago Linguistic Society. Aronson, Howard I. 1994. Indirect objects and datives in Georgian, NSL 7: Linguistic studies in the non-Slavic languages of the Commonwealth of Independent States and the Baltic Republics, H. Aronson (ed.), 1-14. Chicago: Chicago Linguistic Society. Blansitt, Edward L. 1984. Dechticaetiative and dative, Objects: towards a theory of grammatical relations, F. Plank (ed.), 127- 150. Orlando: Academic. Boeder, Winfried. 1968. ber die Versionen des georgischen Verbs, Folia Linguistica 2: 82-151. Boeder, Winfried. 1979. Ergative syntax and morphology in language change: the South Caucasian languages. F. Plank (ed.) Ergativity: towards a theory of grammatical relations, pp. 435-480. Orlando: Academic Press Chkadua, Roena. 1999. statikur zmnata strukturul-semantikuri analizi zemosvanuri. (The structural-semantic analysis of Upper Svan static verbs). Rsum of candidate thesis, A. Chikobava Linguistic Institute, Tbilisi. Chumburidze, Zurab. 1981. nasaxelari zmnebi svanuri (Denominal verbs in Svan), Akaki anies (To Akaki Shanidze), Sh. Dzidziguri (ed.), 145-154. Tbilisi: Tbilisi State University Press. Chumburidze, Zurab. 1986. mqopadi kartvelur enebi (The future tense in the Kartvelian languages). Tbilisi: Tbilisi State University Press. Cooreman, Ann. 1994. A functional typology of antipassives, Voice: form and function, Barbara Fox & Paul Hopper, eds., 49-88. Amsterdam: John Benjamins. Danelia, Korneli. 1976. vnebiti gvaris carmoebisatvis kolxuri (On the formation of the passive voice in Colchian [Zan]). veli kartuli enis katedris romebi 19: 163-174. Deeters, Gerhard. 1930. Das kharthwelische Verbum: vergleichende Darstellung des Verbalbaus der sdkaukasischen Sprachen. Leipzig: Markert und Petters. Dryer, Matthew. 1986. Primary objects, secondary objects and antidative, Language 62: 808845. Fhnrich, Heinz & Zurab Sarjveladze. 1990. kartvelur enata etimologiuri leksikoni. (Etymological dictionary of the Kartvelian languages.) Tbilisi: Tbilisi University Press. Flobert, Pierre. 1968. Dponent et passif et italique et en celtique, BSLP 63: x-xiv. Gamqrelidze, T. and G. Maavariani. 1965. sonantta sistema da ablauti kartvelur enebi: saerto-kartveluri strukturis tipolo-gia. (The system of sonants and ablaut in the Kartvelian lan-guages: a typology of Common Kartvelian). Tbilisi: Mecniereba. Gogolashvili, Giorgi; Kvantaliani, Cicino; and Shengelia, Dali. 1989. kartuli enis zmnuri pueebis leksikoni. (Dictionary of Georgian verb roots). Tbilisi: Mecniereba. Gonda, J. 1960. Reflections on the Indo-European medium, Lingua 9: 30-67, 175-193. Harris, Alice C. 1985. Diachronic syntax: the Kartvelian case (Syntax and syntax 18). New York: Academic Press. Hatcher, Anna Granville. 1942/1973. Reflexive verbs: Latin, Old French, Modern French. New York: Johnson Reprint Corp. (originally published in the collection John Hopkins studies in Romance literatures and languages, v.43. Heath, Jeffrey. 1976. Antipassivization: a functional typology, BLS 2: 202-211. Imnaishvili, Gr. 1979. vnebiti gvari, carmoeba da punkciebi tanamedrove saliteraturo kartuli (The passive voice: form and functions in Modern Literary Georgian). Iberiulkavkasiuri enatmecniereba 21: 22-56. Imnaishvili, Ivane. 1968. vnebiti gvaris zmnata taviseburebani vel kartuli (Characteristics of passive-voice verbs in Old Georgian). veli kartuli enis katedris romebi 11: 27-54.

Deponent verbs in Georgian (Tuite) 11/9/05 page 19 Jannaris, Antonius. 1968. A historical Greek grammar. Hildesheim: Olms. Jorbenadze, Bessarion. 1975. zmnis gvaris pormata carmoebisa da punckiis sakitxebi kartuli (The issue of the derivation and functions of the voice forms of the verb in Georgian). Tbilisi: Tbilisi University Press. Jorbenadze, Bessarion. 1981. Zum Princip der Opposition von Aktiv- und Passivformen im Georgischen, Georgica 4: 79-82. Jorbenadze, Bessarion. 1983. zmnis xmovanprepiksuli carmoeba kartuli (Vowel-prefix derivation of the verb in Georgian). Tbilisi: Tbilisi University Press. Kiknadze, L. 1947. pirveli seriis mckrivta carmoebastan dakavirebuli sakitxebi (Questions concerning the formation of the paradigms of Series I). Tbilisi universitetis romebi XXXb: 317-343. Kuryowicz, Jerzy. 1964. The inflectional categories of Indo-European. Heidelberg: Carl Winter Liparteliani, Aslan. 1994. Svanur-kartuli leksikoni (oluruli kilo). (Svan-Georgian lexicon: Cholur dialect). Privately-financed publication. Monteil, Pierre. 1970. Elements de phonetique et de morphologie du latin. Paris: Nathan. Qipidze, Iosep. 1914/1994. Grammatika mingrelskago (iverskago) jazyka s xrestomatieju i slovarem. St. Petersburg: Imper. Akad. Nauk (reprinted 1994 by Tbilisi State Univ. Press). Sarjveladze, Zurab. 1987. deponensur zmnata esaxeb vel kartuli (About deponent verbs in Old Georgian). Akaki Shanidze 100: saiubileo krebuli, E. Khintibidze (ed.), 102-105. Schmidt, Karl Horst. 1962. Zum Passivum im Georgischen und in indogermanischen Sprachen, Bedi Kartlisa 13-14: 116-126. Schmidt, Karl Horst. 1965. Indogermanisches Medium und Sataviso im Georgischen, Bedi Kartlisa 19-20:129-135. Schmidt, K. H. 1979. Die vorgeschichtliche Grundlagen der Kategorie Perfekt im Indogermanischen und Sdkaukasi-schen, Arnold ikobavas (dabadebis 80 clistavisadmi mivnili krebuli), Sh. Dzidziguri (ed.), 76-83. Tbilisi: Mecniereba. Shanidze, Akaki. 1925/1981. kartuli zmnis sakcevi ([The category of] version of the Georgian verb), reprinted in txzulebani, II. kartuli enis strukturisa da istoriis sakitxebi (Collected works, II. Issues in the structure and history of the Georgian language), 323-376. Tbilisi: Mecniereba. Shanidze, Akaki. 1953. kartuli gramatikis sapuvlebi, I: morpologia (The fundamentals of Georgian grammar I: morphology). Tbilisi: TSUG. Topuria, Varlam. 1967. svanuri ena, 1: zmna. (The Svan language, 1: Verb.) Tbilisi: Mecniereba. Tschenkli, Kita. 1958. Einfhrung in die georgische Sprache. Zrich: Amirani Verlag. Tschenkli, Kita. 1960-1974. Georgisch-deutsches Wrterbuch. Zrich: Amirani Verlag. Tuite, K. 1990. Das Prfix x- im Frhgeorgischen, Georgica 13: 34-61. Tuite, K. 1992. The category of number in Common Kartvelian, The Non-Slavic Languages of the USSR: Linguistic Studies, H. Aronson (ed.), 245-283. Chicago Linguistics Society. Tuite, K. 1994. Aorist and pseudo-aorist for Svan atelic verbs, NSL 7: Linguistic studies in the non-Slavic languages of the Commonwealth of Independent States and the Baltic Republics, H. Aronson (ed.), 319-340. Chicago, Chicago Linguistics Society. Tuite, K. 1996. Paradigm recruitment in Georgian, NSL 8: Linguistic Studies In The Non-Slavic Languages of the Commonwealth of Independent States and the Baltic Republics, H. Aronson (ed.), 375-387. Chicago Linguistics Society. Tuite, K. 1997. Svan. (Languages of the World / Materials, vol. 139). Mnchen: LINCOM Europa. Tuite, K. 1998a. Kartvelian Morphosyntax. Number agreement and morphosyntactic orientation in the South Caucasian languages. (Studies in Caucasian Linguistics, 12). Mnchen: LINCOM Europa. Tuite, K. 1998b. Note de terrain sur le verbe svane, Langue et langues. Mlanges Albert Maniet. Y. Duhoux (ed.), 267-89. Cahiers de lInstitut de Linguistique de Louvain. Tuite, K. in press. Kartvelian series markers, to appear in Current trends in Caucasian, East European and Inner Asian linguistics: Papers in honor of Howard I. Aronson, D. A. Holisky and K. Tuite (eds). Amsterdam: John Benjamins.