Äîêóìåíò âçÿò èç êýøà ïîèñêîâîé ìàøèíû. Àäðåñ îðèãèíàëüíîãî äîêóìåíòà : http://www.philol.msu.ru/~otipl/new/main/articles/daniel/DanielMaisakMerdanova%20AgulCausatives%20en.pdf Äàòà èçìåíåíèÿ: Wed Nov 4 19:01:12 2009 Äàòà èíäåêñèðîâàíèÿ: Thu Dec 31 23:46:09 2009 Êîäèðîâêà: Ïîèñêîâûå ñëîâà: sun dogs

Causatives in Agul
1 Introduction

1
Michael A. Daniel*, Timur A. Maisak, Solmaz R. Merdanova

This paper deals with causativization in Agul, a Lezgic language of Southern Dagestan, and is the first systematic account of causativization patterns in this language. It describes formal properties of the attested causativization patterns and reviews the semantics of Agul causatives and the contrasts between them. The structure of the paper is as follows. The present section includes a brief introduction to the sociolinguistic situation of Agul, its genetic affiliation and an overview of the relevant fragments of its grammar. Section 2 describes formal properties of causative verbs and constructions, starting from the productive pattern and proceeding to non-productive models. Section 3 describes semantic features of Agul causatives and contrasts between the attested patterns. Section 4 contains some typological background; against this background, we discuss what is common and what is uncommon about Agul causatives. Agul belongs to the Lezgic group of the Nakh-Daghestanian, North-Caucasian. Together with Lezgian and Tabassaran it forms the East Lezgic subgroup, as opposed to other subgroups such as Tsakhur - Rutul and Budukh - Kryts subgroups or `aberrant' members such as Archi and Udi. Agul is spoken in 15 villages of the Agul district (Agulskiy rayon ) and five villages of the Kurakh district in the south of the Dagestan Republic. The number of ethnic Agul in Russia is about 30,000 (28,297 according to the 2002 census, 23,314 ­ or 82% - of them living in Dagestan). Though the language is relatively well-populated, it has become written only in 1990; school education in Agul is only available in Agul district and only for the first two years of primary school. The vast majority are bilingual in Russian (the exception is young children before they go to school and some very old women). The Aguls living in Kurakh district (where the Lezgians are a majority) are also bilingual in Lezgian, and are often recorded officially as Lezgians. Agul has several dialects, most of them mutually understandable. This study is based on the Huppuq' dialect (spoken in a village on the north of the Kurakh district), the mother tongue of one of the authors of the present study2; her introspection is the source of our examples. So far no comprehensive general description of the Agul grammar has been compiled, although a considerable effort was made by (Shaumyan 1941; Magometov 1970), later continued by (Sulejmanov 1993; Tarlanov 1994). Recently a number of publications appeared dealing with various aspects of the Agul grammar specifically, in a more detailed way, e.g. (Maisak & Merdanova 2002; Ganenkov & Merdanova 2002; Dobrushina & Merdanova 2004; Maisak & Merdanova 2004; Ganenkov, Maisak & Merdanova to appear) and other; this paper continues this line. Agul nominal categories are typical of the Daghestanian, including ergative alignment and a rich set of spatial forms. The latter combine two categories, that of localization, or position of the trajector with respect to the landmark, and the direction of movement or
1

This article is based on two papers read in Kazan at LENCA-2 conference in June, 2004 and in Moscow at the Workshop on verbal derivation (the Institute of Oriental Studies, Russian Academy of Sciences) in May, 2005. * This author's work was partially supported by the RFBR (Russian Foundation for Basic Research), grant #0506-80351-a. 2 Solmaz Merdanova who lived in Huppuq' as a child is trilingual in Agul, Lezgian and Russian.

1


absence of movement (orientation). Examples are apudessive (`apud'+`essive': location near the landmark), apudelative (`apud'+`elative': movement away from location near the landmark) or superlative (`super'+`lative': movement onto the landmark), etc. Verbs may be derived and non-derived. We presently know of about 130 non-derived verbs, which is approximately ten percent of the verbal vocabulary. Derived verbs are typically produced by combining a noun or an adjective with a verb into a more or less closeknit compound or by prefixation. Prefixation is either locative, often irregular semantically and not fully productive (by means of one of a number of prefixes with more or less clear spatial origin) or refactive, which is productive and semantically regular (by means of prefix qa-/qu- `do again' or `cancel the result of the previous event'), for details see (Maisak & Merdanova 2002). Another important class of predicates is few stative verbs (as opposed to regular, dynamic verbs), the only verbal class that is not subject to regular causative derivation; they are discussed below in Section 2.2.3 in more detail. A rich set of TAM categories combines synthetism and analytism (using auxiliary verbs), but there is no person or number agreement of any kind on the verb; the category of noun class (gender) typical of most Lezgic and other Nakh-Daghestanian has been lost in Agul altogether. There is also no valency reducing derivation3, and there is only one major type of valency increasing derivation ­ causativization4.

2 Formal types
Agul has several devices of causativization (causativization patterns). Only one of them is productive: combination of the infinitive with the verb (a)q'as `do', or periphrastic `do'causatives, which are discussed in Section 2.1. Other, non-productive patterns include compound `do'-causatives, labiles and lexical causatives. These are discussed in Sections 2.2 through 2.4.

2.1 Periphrastic `do'-causatives
2.1.1 Lexical distribution
The only productive pattern of causativization is combining the infinitive of the lexical verb with (a)q'as `do'5 ­ periphrastic `do'-causatives (or simply periphrastic causatives) below. Periphrastic causatives are formed from intransitive (1, 2, 3), transitive (4) and ditransitive (5) verbs. Multiple periphrastic causatives (periphrastic causatives based on periphrastic causatives) are structurally possible (6), although the actual use of a more-thandouble causative is highly improbable (cf. Nedjalkov, Silnickij 1969; Kulikov 1993; Dixon 2000 on causative recursion; on structural availability of double causatives in Daghestanian see Kibrik 1996, Comrie 2000: 369, Ljutikova 2001: 393-394 for Godoberi, Tsez and Bagvalal, respectively).

However, see footnote 10 below on the assumed binominative construction; further study is needed.. There is only one, and rather peculiar, valency increasing pattern in addition to causativization, the verificational construction introducing the participant who verifies whether the situation described by the verb does actually take place (coded by ergative). For details on the latter see (Maisak & Merdanova 2004). 5 Apparently, the first vowel of the verb is always optional, both when it is used `lexically' and in causative constructions. In Agul, dropping the first (unstressed) vowel is also characteristic of some other verbs, as (a)Ras `say' (imperfective stem), (u)pune `said' (perfective stem), (a)lik'as `put on', (i)c'as `give' etc.
4

3

2


(1) intransitive (one argument) baw.a ]ynyK Rarx.a-s
mother(ERG) child(NOM) sleep.IPF-INF

q'.u-ne
do.PF-PFT

Mother made the child sleep. (e.g. put him/her to bed, or lulled him/her to sleep, etc.) (2) intransitive (two arguments) malla nesredin.a Pa{ah
Molla Nasreddin(ERG) king(NOM)

gada.ji-q
boy-POST

quX.a-s
believe.IPF-INF

q'.u-ne
do.PF-PFT

Molla Nasreddin made so that the king believed the boy. (e.g. confirmed the boy's words) (3) intransitive (experiencer verbs) baw.a-s agW.a-s q'.u-ne-wa
mother-DAT see.IPF-INF do.PF-PFT-Q

wun
you(ERG)

jarHun?
wound(NOM)

Why, you let your mother see the wound?! (the addressee was not supposed to let his/her mother see the wound not to make her upset) (4) transitive malla.ji gada.ji-w
priest(ERG) boy-APUD

q'ur>an
Koran(NOM)

ruX.a-s
read.IPF-INF

q'.a-a
do.IPF-PRS

The priest makes his son read the Koran. (5) ditransitive me ]q'aq'I-w
this niggard-APUD

malla nesredin.a-ra


kasib-ar.i-s
poor-PL-DAT

sadaq'a
sadaqa(NOM)

molla Nasredin(ERG)-and even

ic'.a-s
give.IPF-INF

q'.a-s-Tawa
do.IPF-INF-COP:NEG

Even Molla Nasreddin wouldn't make this niggard give alms to the poor. (6) double causative wun gi-w
you.Sg(ERG) that-APUD

dad.a-s
father-DAT

wuri

unx.a-s

q'.a-s
do.IPF-INF

q'.u-ne.
do.PF-PFT

all(NOM) hear.IPF-INF

You forced him to make Dad hear everything. (7) double causative Hadad.a zun gada.ji-w
grandfather(ERG) I(ERG) son-APUD

Habawa-s
grandmother-DAT

k'eD
letter(NOM)

lik'.a-s
write.IPF-INF

q'.a-s
do.IPF-INF

q'.u-ne
do.PF-PFT

Grandfather made me make my son write a letter to the grandmother. The only class of verbs that do not form periphrastic `do'-causatives are locative statives, including aa `be inside', aldea `be above', qaa `be behind' etc. (with various locative prefixes) and experiencer statives Kandea `love, want', Haa `know', iTaa `ache, be ill', gu[`aa `be afraid'. Statives have no infinitive. Experiencer statives produce inchoative `become'compounds and respective `do'-compounds as causatives (see below Section 2.2.3).

2.1.2 Case assignment
As one can see in the examples (1) through (6), periphrastic causatives always code the Causer by ergative, while P is coded by nominative6. As to the Causee, in the examples above intransitives take Causee (intransitive Causee below) in nominative, while transitives mark
6

Below, capital letters mark morphosyntactic arguments (S, A, P); labels with the first capital letter stand for semantic roles and functions (Causee, Recipient etc.), while non-capitalized labels designate morphological case (nominative, ergative etc.).

3


the Causee (transitive Causee below) with apudessive. Although this is indeed the dominant pattern, the case may be assigned differently. First, transitive Causees may be ergative, as in the following example: (8) transitive Causee: ergative baw.a ru].a
mother(ERG) daughter(ERG)

jaK
meat(NOM)

eat.IPF-INF

q'.u-ne
do.PF-PFT

The mother made her daughter eat meat. Note that in (8) the Causee may also be apudessive; apudessive marking is even more expectable. Some intransitive Causees, conversely, may also be marked as apudessive: (9) intransitive Causee: apudessive dad.a u[i-n uQub-ar.i-l-di
father(ERG) REFL-GEN

gada.ji-w Xula-as
house-IN.ELAT

hi].a-s
flee.IPF-INF

q'.u-ne
do.PF-PFT

beating-PL-SUP-LAT son-APUD

Father's beating made the his run away from home. (lit. "by his beating father made son run away from home") Contrary to (8), in (10) apudessive marking is less expectable; normally, the Causee is marked by nominative. In other words both intransitive and transitive Causees may preserve their original nominative/ergative marking as former A or S or follow a causative-specific strategy and be marked by apudessive (with different preferences for intransitive and transitive Causees). These options pose obvious morphosyntactic problems concerning the syntactic structure of a causative predication (presence of two ergative arguments in (8) and absence of the nominative argument in (9)). Their morphosyntactic status will be discussed in Section 2.1.4. We will now focus on the variation of the case assignment for the Causee. Not all intransitives may take apudessive Causees. The first thing to note is that the availability of apudessive marking depends on the verb. Apudessive is possible with hi]as `flee', but impossible with iTarxas `become ill' or alurq'as `fall down'; cf. (11) and (12). (11) intransitive Causee: apudessive impossible [un ]ynyK (*]ynyK.i-w) iTa-r-x.a-s
you.pl(ERG) child(NOM) (child-APUD) ill-CMP-become.IPF-INF

q'.u-ne
do.PF-PFT

Your child fell ill because of you. (12) intransitive Causee: apudessive impossible [un ]ynyK (*]ynyK.i-w) alurq'.a-s
you.pl(ERG) child(NOM) (child-APUD) fall.IPF-INF

q'.u-ne
do.PF-PFT

You child fell down because of you. More generally, the condition licensing apudessive marking seems to be the control the Causee exerts over the situation. (Note again that this is a condition for availability of apudessive marking; the less marked option for these verbs remains nominative.) The Agul intransitives thus fall into two classes. This is a realization of the well-known typological distinction between patientive intransitives (apudessive marking unavailable) and agentive intransitives (apudessive marking available). Now consider experiencer verbs. In Agul they are typically intransitive, Experiencer being marked by dative, while Stimulus is nominative. Periphrastic causatives of these verbs preserve the case assignment of the lexical verb, adding ergative for the Causer. The dative Experiencer can not be coded as apudessive Causee; cf. (13), identical to (3), where apudessive is ungrammatical. One way to account for this is to say that the Experiencer has 4


no control over the situation in which (s)he participates; in other words, all experiencer verbs in Agul, as expected, are P-intransitives. (13) experiencer verb baw.a-s (*baw.a-w)
mother-DAT (*mother-APUD)

agW.a-s
see.IPF-INF

q'.u-ne-wa
do.IPF-PFT-Q

wun
you(ERG)

jarHun?
wound(NOM)

Why, you let your mother see the wound?! There is additional evidence for introducing the controllability parameter for intransitive Causees. The apudessive marking is combined mainly with human intransitive ausees, for non-human intransitive Causees it varies from acceptable or questionable (non-human animate intransitive Causee (14) and (15)) to highly questionable or ungrammatical (inanimate intransitive Causees in (16), where only the original, nominative marking is possible). (14) animate intransitive Causee dad.a u[i-n haraj-ar.i-ldi Hajwan.i-w (
father(ERG)
better

Hajwan)

hi].a-s
run.IPF-INF

q'.u-ne
do.IPF-PFT

REFL-GEN yell-PL-SUP.LAT horse-APUD (horse(NOM))

Father's beating made the cow run away. (15) animate intransitive Causee ? peB-er.i-w (+peB-er) dad.a
father(ERG) hen-PL-APUD

Qir.i-l-as

latk.i-na
chase_away.PF-CVB

(hen-PL(NOM)) perch-SUP-ELAT

be mud-INTER-LAT

go.IPF-INF

q'.a-a
do.IPF-PRS

Having chased the chickens away from their perch, father makes them walk in the mud. (16) inanimate intransitive Causee ]ynyK.i *tuP.u-w (+tuP)
child(ERG) ball-APUD ball(NOM)

be mud-INTER-ELAT bounce.IPF-INF

q'.a-a
do.IPF-PRS

The child makes the ball bounce in the mud. The situation with transitive verbs is similar but not identical. Here, apudessive marking seems to be equally available and even preferable (less marked) for human and non-human animate Causees, as in (17) and (18) (though original, ergative marking is also possible), but as with intransitive verbs, ungrammatical for inanimate Causees, as in (19) and (20), and even in (21), where the inanimate transitive Causee is metaphorically recategorized as animate by combining with the verb facas `catch'. (17) transitive human Causee baw.a ru].a-w
mother(ERG) daughter-APUD

jaK
meat(NOM)

eat.IPF-INF

q'.u-ne
do.PF-PFT

The mother made her daughter eat meat. (18) transitive animate Causee Qun]i Xurur.i-w
neighbour(ERG) dog:Pl-APUD

ze
my

Kel
lamb(NOM)

fac.a-s
catch.IPF-INF

q'.u-ne.
do.PF-PFT

My neighbor sied the dogs on the lamb. (19) transitive inanimate Causee ru].a raK da-qik'.i-na
girl(ERG) door(NOM) NEG-close.PF-CVB

kulak.i
wind(ERG)

(???kulak.i-w) raK
wind-APUD door(NOM)

daq.a-s
open.IPF-INF

q'.u-ne
do.PF-PFT

Because the girl did not close the door, the wind made it lock. (i.e. the door was locked by a blast of wind)

5


(20) transitive inanimate Causee gada.ji tuP.u (*tuP.u-w)
boy(ERG) ball(ERG) ball-APUD

BagW
mirror(NOM)

ar<.a-s
break.IPF-INF

q'.u-ne
do.PF-PFT

The boy broke the mirror with a ball (For instance, the boy kicked the ball and broke the mirror.7) (21) transitive inanimate Causee (metaphorically extended to animate) zun k'ur-ar.i c'aj (*c'i-w) fac.a-s q'.u-ne nafT
I(ERG) wood-PL(ERG) fire(NOM) (fire-APUD) catch.IPF-INF do.PF-PFT kerosine(NOM)

pour_in.PF-CVB

I made the wood catch fire by pouring some kerosene. There are thus the following hierarchies controlling availability of apudessive marking for the Causee: Verb class. Transitivity: {transitive preferable > A-intransitive available > P-intransitive
ungrammatical}

Causee type. Intransitive Causee: {human available > animate acceptable > inanimate ungrammatical} Transitive Causee: {animate preferable > inanimate ungrammatical} These three hierarchies may be summarized in the following Table 1. Table 1. Availability of apudessive marking P-intransitive A-intransitive transitive Human Available preferable Animate acceptable ungrammatical Inanimate This table shows two things. First, it is not clear why apudessive marking is preferable on transitive animate Causees but only available or acceptable with intransitive animate Causees. One solution could be that all transitive predicates are conceived as more controllable as compared to intransitive A-predicates. This is a plausible approach; however, in 2.1.5 we argue for a different solution. Second, the table shows that a straightforward lexical categorization of intransitives into agentive vs. patientive is not enough for Agul. To account for the availability of apudessive marking in terms of transitivity, we need to distinguish between patientive vs. agentive `intransitive situations' or at least between patientive vs. agentive intransitive constructions rather than just between patientive vs. agentive intransitive verbs. Indeed, whether apudessive is possible depends not only on the category of the verb but also on the category of the Causee. In other words, Table 1 illustrates Hopper and Thompson's claim that transitivity is not a lexical property of the verb but a cluster of properties (Hopper & Thompson 1980). To sum up, there are two options of case assignment for transitive and A-intransitive animate Causees. They may preserve the original S/A marking or to apply causative-specific marking. The semantic contrast between the two patterns is sometimes elusive, but it seems that an apudessive Causee has a reduced control over the event and often suggests a coercive type of causation. A more detailed discussion of the contrast is provided in Section 3.

7

Note that in (20) instrumental marking (superlative) would be more natural for `ball'.

6


2.1.3 A note on apudelative
There is also a third option of case assignment for the Causee: apudelative; cf. (22). Note that apudelative is also used for Agent marking outside causative construction. It marks Agent in involuntary Agent constructions, available for A-intransitive predicates only (Ganenkov, Maisak & Merdanova, to appear); cf. (23). (22) apudelative: intransitive Causee ilde].i ru].a-f-as / ru].a-w
friend(ERG) girl-APUD-ELAT / girl-APUD

lak-ar
foot-PL(NOM)

kur.a-s
become_dirty.IPF-INF

q'.u-ne
do.PF-PFT

A friend made the girl make her feet dirty. (23) apudelative: involuntary Agent za-f-as (*za-w) lak-ar
I-APUD-ELAT (I-APUD) foot-PL(NOM)

kur.u-ne
become.dirty.PF-PFT

I made my feet dirty. (unintentionally) Although, unlike apudessive, apudelative is not a dedicated causative Agent marking, it has apparently been grammaticalized in this function. Indeed, when marking involuntary Agents, apudelative is limited to intransitive predicates, while in periphrastic causatives it may be used for marking Agent (Causee) in transitive contexts, too. (24) apudelative transitive Causee, involuntary Agent impossible a. baw.a gada.ji-f-as ]urpa mother(ERG) boy-APUD-Elat broth(NOM) eat.IPF-INF do.PF-PFT

Mother made the boy eat the soup. (e.g. by threats) b. *gada.ji-f-as
boy-APUD-Elat

]urpa
broth(NOM)

eat.PF-PFT

The boy ate the soup. (unintentionally) The semantic contrast between apudelative and apudessive Causees is minimal, if any, and is not yet adequately described. It seems that, at least in some contexts with nonhuman Aintransitive Causees, apudessive is more acceptable than apudelative, so apudelative may be even more sensitive to the Agent's animacy or control. (25) apudessive and apudelative Causee: contrast unclear baw.a kitan / ?kitan.i-w (*kitan.i-f-as) hi].a-s
mother(ERG) cat(NOM) / cat-APUD (cat-APUD-ELAT) flee.IPF-INF

q'.u-ne.
do.PF-PFT

Mother made the cat run away. But this contrast is extremely vague, and the apudelative construction is otherwise identical to the apudessive construction; below we use a cover term of `apud marking' of the Causee without distinguishing apudessive and apudelative Causees.

2.1.4 Syntax and morphosyntax
As we mentioned already in Section 2.1.2, some of the case marking patterns in periphrastic `do'-causatives pose problems as to the status of the construction and its argument structure. Cf. the possibility of two ergatives in (8), repeated here as (26). (26) two ergatives, two clauses? [baw.a [ru].a
mother(ERG) daughter(ERG)

jaK
meat(NOM)

[
eat.IPF-INF

VP2]S2

[q'.u-ne]

do.PF-PFT

VP1]S1

Mother fed meat to her daughter. 7


Combining two ergatives in one clause is not always ungrammatical in Agul; but the second of the ergatives must be used in a different function, e.g. instrumental or temporal. Here, however, both ergatives are obviously agentive, one marking the Agent of the causative situation, i.e. the Causer, and the other the Agent of the situation being caused, i.e. the Causee. The fact that both are marked by ergative could suggest that, syntactically, the combination of the infinitive of the lexical verb with `do' is in this case biclausal construction8. Assumedly, the Causer's ergative belongs to the main clause and is assigned by (a)q'as `do', while the Causee's ergative belongs to the subordinate clause and is assigned by the lexical verb. On the contrary, periphrastic causatives with transitive Causees marked by apud are naturally considered as monoclausal, with (a)q'as `do' being a causative auxiliary. Indeed, if we admit a biclausal structure for (27), which is the apud version of (26), it is unclear which of the verbs ascribes apudessive marking to the Causee. It is easier to admit the apudessive is ascribed by the construction on the whole. (27) ergative and apud, one clause? [baw.a ru].a-w Xap'a
mother(ERG) daughter-APUD gruel(NOM)

[ eat.IPF-INF

q'.u-ne]VP]
do.PF-PFT

S

Mother made her daughter eat gruel. The problem is that the apudessive is also possible in (28), repeated from (9). (28) apud in intransitive: failure of nominative requirement dad.a u[i-n uQub-ar.i-l-di gada.ji-w Xul.a-as hi].a-s
father(ERG) REFL-GEN beating.PL-SUP-LAT son-APUD house-IN.ELAT flee.IPF-INF

q'.u-ne
do.PF-PFT

Father's beating made the his run away from home. (lit. "by his beating father made son run away from home") This sentence contains no nominative argument, otherwise obligatory in Agul (nominative requirement)9. A syntactic solution that preserves the obligatory status of the nominative is to argue that (28) is biclausal, too, with the S-Causee raised to the main clause, assigned apudessive and coreferentially deleted from the subordinate clause. With this in mind, it is impossible to maintain that the presence of an apud Causee is an argument in favor of monoclausal interpretation. In other words, case marking patterns seem to support the biclausal interpretation of periphrastic causatives. Let us now consider other evidence. The first argument against biclausal interpretation of periphrastic `do'-causatives comes from a comparison with an indisputably polypredicative construction. The word order in Agul is rather free, so, although the forms of (a)q'as `do' prefer to stay in contact with the lexical infinitive, various material may come between them. However, if we compare how free is the word order in periphrastic causatives with constructions with e.g. the verb Hazur-xas `get ready, intend' (lit. "become ready"), the difference is obvious. The verb Hazur-xas can easily go to the position before the subordinate
An objection here could be that the causative construction formed from transitives is indeed the only context where two pretendents to the agentive ergative marking meet in Agul. The fact that two agentive ergatives are not met elsewhere in the language is then a trivial statement, therefore we can not dismiss the monoclausal interpretation basing on the occurrence of two agentive ergatives only. However, from other languages, we know that double ergative (or, for accusative languages, double nominative) construction is not attested with morphological causatives. Thus, typologically, presence of two A-marked NPs is a good argument in favor of biclausal interpretation. 9 Exceptions are extremely rare. One is an A-labile verb ruXas `read', which, with omitted nominative Patient, means `study'; the Agent may preserve ergative marking (alternatively, it may be marked by nominative): cf. gadaji kitab RuXaa `the boy (ERG) reads the book (NOM)' and gada (NOM) / gadaji (ERG) iDe ruXaa `the boy studies well'. Interestingly, the same A-lability pattern is observed in (Kibrik 1996) with this meaning in Godoberi.
8

8


clause, while the verb (a)q'as `do' placed in front of the predication it causativizes is extremely unnatural. (29) word order in a subordinate construction
+

dad
father(ERG)

]ynyK-ar
childPL(NOM)

Xul.a->
house-IN

at.a-s
leave.IPF-INF

Hazur-x.u-ne
ready-become.PF-PFT

Dad prepared to leave the children at home. (30) word order in a periphrastic causative dad.a
father(ERG)
???

]ynyK-ar
child-PL(NOM)

/

]ynyK-ar.i-w
child-PL-APUD

Xul.a->
house-IN

uq'.a-s
sit.IPF-INF

q'.u-ne
do.PF-PFT

Dad made the children stay at home. Second, consider the evidence from negative constructions, presented in Table 2. Table 2. Position of the negative category go.IPF-INF

q'.u-n-dawa
do.PF-PFT-NEG

go.IPF-INF

Hazur-x.u-n-dawa
ready-become.PF-PFT-NEG

did not let go ??? da-ilgW.a-s
NEG-remain.IPF-INF

q'.u-ne
do.PF-PFT

did not decide to go Hazur-x.u-ne da-ilgW.a-s
NEG-go.IPF-INF ready-become.PF-PFT

made not to go

decided against going

Any finite form of Hazurxas `plan to etc.' and (a)q'as `do' may form negatives, like the negative of perfective past in the Table 2 above. Agul also has negative infinitives with a prefix da-. This form is perfect in combination with the verb Hazur-xas `plan to etc.', but very odd in periphrastic causative. In other words, the negation only applies to the periphrastic causative as a whole, again indicating that lexical infinitive plus the verb (a)q'as `do' is not a free combination of two verbs. Negative causation, i.e. causing something not to happen, can not be expressed in Agul by a causative construction; subordination is necessary. (31) negative causation: subordination required dad.a ]ynyK-ar Xul.a-> da-ilgW.a-Xildi
father(ERG) child-PL(NOM) house-IN Neg-stay.IPF-PURP

q'.u-ne
do.PF-PFT

Father made children not to stay at home. Finally, the monoclausal nature of the periphrastic causatives is manifested in their interaction with adverbs. (32) adverbial scope me ]q'aq'i-w malla nesredin.a
this niggard-APUD Molla Nasreddin(ERG)

naq'
yesterday

kasib-ar.i-s
poor-PL-DAT

sadaq'a
sadaqa(NOM)

ic'.a-s
give.IPF-INF

q'.u-ne
do.PF-PFT

Yesterday, Molla Nasreddin made this niggard to give alms to the poor.

9


If the construction would contain two separate lexical verbs and consist of two separate clauses, one of the interpretations of this sentence would be that Molla Nasreddin yesterday talked to the niggard and convinced him to do the right thing ­ without the "sadaqa" being already distributed by the moment of speech. Another, on the contrary, would be that Molla talked to him few days ago, while the distribution took place yesterday. However, the Agul sentence may only mean that the causation and the following distribution of sadaqa took place yesterday, which means the scope of naq' `yesterday' can not be limited neither to the verb (a)q'as `do', nor to the infinitive ic'as `give'. The act of causation and the caused situation are conceived as one single event. Finally, only to repeat the argument already mentioned, the very fact that some periphrastic causatives allow apud marking of the Causee is an indication that something is going on between the two verbs, they form together some kind of construction ­ indeed, none of them may assign apud in isolation, and the Causee is never apud-marked except in periphrastic causatives. Probably, these tests do not exhaust evidence for monoclausal interpretation of periphrastic causatives in Agul, but they are already a strong counterevidence against a straightforwardly biclausal analysis. There is an obvious clash between these tests and morphosyntactic properties of the Causee exemplified in (26) (presence of two agentive ergatives) and (28) (the violation of the nominative requirement), both of which could be explained by adopting biclausal structure. The obvious solution is to admit that periphrastic `do'-causatives in Agul are intermediate between two separate clauses and one single clause, providing an example of clause union phenomenon (Noonan 1985)10. Importantly, however, there is an apparent correlation between accessibility of apud marking for the Causee and (in)transitivity of the non-causative verb. Apud seems to be the default, unmarked pattern for a transitive Causee, while ergative is a marked choice. On the other hand, apudessive marking is a marked ­ peripheral, though accessible ­ choice for an intransitive agentive Causee. In addition, apudessive non-human animate Causees are more acceptable for transitive than intransitive verbs. All this is probably an indication that the periphrastic causative gradually drifts towards straightforwardly monoclausal syntax.

2.1.5 Once again on case marking of the Causee
To sum up, we can isolate two different factors in availability of the original vs. apud marking of the Causee in periphrastic causatives. First, there is a semantic factor of control of the Causee in the situation being caused. If the Causee exerts some control, the choice between original and apud marking is available for some Causees. The second factor is that the periphrastic causative seems to be developing from a clause union structure into an auxiliary causative construction. This reduces the availability of original marking for a transitive Causee (because one clause would then have two agentive ergatives) and the availability of apud marking for an intransitive Causee (because that would leave us with a nominative-less clause). On the contrary, apud transitive Causee and nominative intransitive Causee are more than compatible with monoclausal syntax. These two

10

There is a strong analogy between the periphrastic `do'-causatives and the binominative construction typical of many Daghestanian languages. Daghestanian binominative constructions are converbs-plus-copula constructions that assign to the Agent nominative marking, resulting in two nominatives present in one construction. An example is dad XuPur uzaj aa `Dad (NOM) is ploughing the field (NOM)' (as answer to What does dad do?); their status may also be argued to be intermediate between bi- and monoclausal.

10


factors satisfactorily explain the distribution of case assignment in various contexts11. Cf. Table 3. Table 3. Factors of case assignment (black ­ original only; deep gray ­ apud questionable; light grey ­ apud available; white ­ apud preferable) Intr Tr
syntactic factor monoclausal tendency

original

apud

2.2 Compound `do'-causatives
A large number of causative verbs in Agul are formed by joining adjectival, nominal or verbal stem with (a)q'as `do' into a kind of `loose compound' (for its `looseness' see below Section 2.2.5). We will call these causatives compound causatives as opposed to periphrastic causatives, also formed with (a)q'as `do'.

2.2.1 Adjectives
A vast part of `do'-compounds are formed from adjectives which also form a `become'compound, an inchoative correlate to the causative one; cf. (33). (33) adjectival compound (Hazur-f `ready') a. ]ynyKar mexteb.i-s Hazur-q'e!
child-PL(NOM) school-DAT ready-do.IMP

Get children ready (to go) to school! b. ]ynyK-ar, mexteb.i-s Hazur-jux!
child-PL(NOM) school-DAT ready-become.IMP

Kids, get ready to (go to) school! Further examples of compounds derived from adjectives are given in Table 4. Table 4. Adjectives. Derivation of `do'- and `become'-compounds. Stem `become'-compound `do'-compound 1 ]ad-f `glad' ]ad-xas `become glad' ]ad-q'as `make glad' 2 Bysse-f `old' Bysse-xas `get old' Bysse-q'as `make old' 3 HyTe-f `sharp' HyTe-xas `become sharp' HyTe-q'as `sharpen'

2.2.2 Nouns and bound stems
Some nominal stems also form pairs of `become'- and `do'-compounds. The semantic of the compound is far from always predictable from the semantics of the noun; cf. (34) and (35).

11

Note there is no indication the ban on apudessive inanimate Causee is weaker for transitive than intransitive predicates; so the semantic factor apparently prevails.

11

apud availa ble

A-Intr or Tr

Inanimate Animate Human

semantic factor control

origin al only

P-Intr


(34) nominal compounds: transparent a. ]ur k'ildi XarD-x.u-ne
paint(NOM) fully expense-become.PF-PFT

There's no paint left (=The paint is fully spent) b. zun ]ur k'ildi XarD-q'.u-ne
I(ERG) paint(NOM) fully expense-do.PF-PFT

(I) spent all the paint. (35) nominal compounds: lexicalized a. wa-s baw.a un
you-DAT mother(ERG) sound

q'.a-a
make.IPF-PRS

Mother calls you (lit. "mother to you sound makes"). b. wa-s haraj-ar un x.u-ne-wa ?
you-DAT shout-PL sound become.PF-PFT-Q

Have you heard the yelling? These compounds are formed from the nominal stems HarD `expense' and un `sound'. The clause structure in (34) and (35) shows important differences. In (34) the nominal stem is integrated into the verbal stem so that the Patient position with nominative marking is free for ]ur `paint'. In (35) the nominal stem apparently occupies this position itself. This important parameter shows that the relation between the noun and the verb may be more and less loose. The same morphosyntactic difference is observed in e.g. [`ir-xas `become bad, unusable', [`irq'as `make bad, spoil' from [`ir `harm, effect of evil eye', which both take a nominative Patient, and Xabar-xas `become known' (), Xabar-q'as `inform' (), which take no nominative in addition to Xabar `news'. Some of the nouns only form `become'-compound (e.g. ['ymel-xas `become damp' from ['ymel `humidity') or `do'-compound (e.g. ]uRul-q'as `rat (on somebody)' from ]uRul `instance of ratting'), which means that the inchoative ~ causative correlation is not always present. The status of many stems is disputable, because they are not or almost not used in argument position. There is a certain `scale of boundness' of the assumedly nominal stem to the compound-forming verb. Outside compounds, the noun [`ir `harm' is used almost exclusively in a curse formula; cf. (36). (36) limited use outside compounds [`ir larH.u-raj wa-l
harm(NOM) fall.PF-OPT you.SG-SUPER

I wish bad luck strike you. Similarly, the nominal stem KyteH `end', present in KyteH-xas `end (intr)'and KyteHq'as `finish, end (tr)', is only used in isolation as a full utterance (meaning that's all or end of story). The stem gunt' used in verbs gunt'-xas `gather (intr)' and gunt'-q'as `gather (tr)' is identical to the noun gunt' which, in dialects other than Huppuq', means `heap, pile'. In Huppuq', however, this noun has developed into kunt', so that synchronically the stem of the assumedly nominal compound can not be identified with any noun. Finally, there are some clearly non-nominal, completely bound stems. Clear examples are predicative stems that have been adopted from Turkic or Russian, such as i]lemi]-q'as `use' and ba]lami]-q'as `begin' or -xas `become organized' ~ q'as `organize (tr)'. Note also that there is no necessary correlation between semantic transparency and morphosyntactic status of the nominal stem. Indeed, the relationship in (34), which is morph

12


syntactically more tight, is less lexicalized (more transparent) than the more loose relation in (35). The factor which seems much more powerful in that respect is the degree of boundness. The less independent a stem is, the more likely the compound will have a regular nominative argument. Thus, a highly frequent noun Xabar `news' excludes patientive nominative, while the rarely used [`ir `harm' requires it. The compounds whose stems are never used freely also can never be used without a nominative.

2.2.3 Statives
As was mentioned above, the only class of predicates that can not form `do'-causatives are stative verbs. Statives are morphologically different from other Agul verbs in that they have reduced paradigm. They do not distinguish perfective and imperfective stems and do not form imperatives; one of the forms they lack is the infinitive, used in `do'-causatives. Instead, experiencer statives Haa `know' and iTaa `ache; be ill') form pairs of `become'- and `do'compounds from the stem suffixed with -r. (37) causative compound: stative wun ]ynyK iTa-r-q'.u-ne
you(ERG) child(NOM) be_ill-CMP-do-PF-PFT

The child fall ill because of you. The stative Kande-a `love, want' behaves differently; this verb also forms both `become'- and `do'-compounds, but in an irregular way (from the stem Kan- and without -r) ­ kan-xas `fall in love, start wanting', kan-q'as `make fall in love, want'. The stative gu[`aa `be afraid' does not form `become'- or `do'-compound; instead, these compounds are formed from the nominal stem gu[` `fear' (see above 6 in Table 4). Locative statives (aa `be inside' etc.) do not form `become'- or `do'-compounds at all. Compound derivation for experiencer statives is summarized in the following table. Table 5. Derivation of `become'- and `do'-compounds from statives.
1. 2. 3. 4. Stative iTa-a `ache; be ill' Ha-a `know' Kande-a `love, want' gu[`aa `be afraid' `become'-compound ITa-r-xas `start aching; fall ill' Ha-r-xas `learn' Kan-xas `start loving, wanting' `do'-compound iTa-r-q'as `feel pain; make ill' Ha-r-q'as `teach (smb smth); learn' Kan-q'as `make fall in love, make want' -

Note that the forms with causative morphology are not always straightforward causatives from a semantic point of view. Indeed, Ha-r-q'as in the second meaning `learn (by heart)' is not a causative of Ha-a `know', but its inchoative, thus being parallel to the same stative's `become'-compound. Irregular causatives are discussed in more detail in Section 2.5.

2.2.4 Non-Stative Verbs
The number of `do'-compounds where the lexical stem is a non-stative verbal stem is very limited. As statives, all these verbs form causatives from the imperfective stem (in -a-) suffixed with -r-; cf. (38). (38) causative compound: dynamic verb . ma]in Ruz.u-ne
car(NOM) stop(intr).PF-PFT

The car came to a halt.

13


b.

dad.a
father(ERG)

ma]in
car(NOM)

Ruz.a-r-q'.u-ne
stop(intr).Ipf-CMP-do.PF-PFT

Father stopped the car. The verb in (38b) is derived from the verb used in (38a) by adding suffix -r- and the verb (a)q'as `do' to the stem of the lexical verb; no correlative `become'-compound exists. So far, we are aware of only thirteen verbs that follow this pattern, including: Table 6. Compound derivation for non-stative verbs
1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. Original verb q'e]as `soak (intr), get wet' kysuQas `go stale (of bread)' ruQas `become dry' ruRas `become cold, cool (intr) down' uq'a-s `sit down' Ruza-s `stop (intr); wait' Rut'a-s `stand upright' aluQa-s `stick (e.g. of a stamp, intr)' kyQa-s `stick (e.g.of a hair, intr)' T'u]as `stir (intr), be loose (e.g. a tooth), move away; be displaced' agWa-s `see' Dik'a-s `find (occasionally)' Compound causative q'e]a-r-q'as `soak (tr), make wet' kysuQa-r-q'as `let go stale (of bread)' ruQa-r-q'as `make dry' ruRa-r-q'as `make cold, cool (tr) down' uq'a-r-q'as `make seated' Ruza-r-q'as `stop (tr), cause to wait' Rut'a-r-q'as `make upright' aluQa-r-q'as `stick (to smth, tr)' kyQa-r-q'as `stick (to smth, tr)' t'u]a-r-q'as `make move away; displace' agWa-r-q'as `show' Dik'a-r-q'as `find (intentionally, after looking for)'

The verbs that form compound causatives include physical processes (1 through 5), position verbs (including the verbs `stick to, be stuck to') and the verb `move' (6 through 11) and experiencer verbs (12 and 13). It seems that forming a compound causative is a lexical property rather than the property of the stem, because there are verbs that use the same stem as one of the verbs on Table 6 but do not form compound causatives. Examples are prefixed verbs q-uQas `be caught' (as of a dress occasionally caught by a nail in the wall; cf. (9) and (10) with the same stem), al-agWas `pretend to do something'. All other verbs on the table, however, are non-derived stems that do not combine with prefixes, so that evidence relevant for our claim is very limited. On the other hand, the regular refactive prefixation (qa-/qu- `do again'; see Section 1) preserves availability of compound causative; and the same is true of refactive prefixation with statives: q-agWas `see again' ~ q-agWarq'as `show again' (cf. 12 in Table 6); qa-iTaa `be ill again' ~ qa-iTarq'as `make ill again' (cf. 1 in Table 5).

2.2.5 Morphosyntax
We have now considered two types of `do'-causatives in Agul. Periphrastic `do'causatives are formed by combining the infinitive of the lexical verb with the forms of (a)q'as `do'. Compound `do'-causatives are similar in that the same forms are `suffixed' to the lexical stem (sometimes with -r- between them). These terms suggest the two patterns are clearly distinguished as syntactic vs. morphological. In fact, they are closer to each other than it might seem. We have discussed above that the (morpho)syntactic status of periphrastic `do'causatives seems to be that of clause-union, intermediate between mono- and biclausal, with (a)q'as `do' being very close to auxiliary. Although most of the `do'-compounds are more close-knit units than periphrastic causatives12, the lexical stem always preserves a certain degree of autonomy, which varies
12

It seems, for instance, that the variant with dropped initial vowel (q'as) is more natural with most of `do'compounds than the full variant (aq'as), while the two variants are equally natural for periphrastic causatives.

14


depending on the lexical item. Some compounds, especially those based on adjectives, are rather close to verbal phrases. For nominal `do'-compounds, looseness apparently differs depending on the degree of how bound/integrated into the verb the nominal stem is, as discussed in Section 2.2.2. But even in the case of the least loose `do'-compounds ­ those formed from statives and non-stative verbs and the compounds with bound stems ­ the stem may be separated from the form of (a)q'as `do' by other material, especially when the compound carries emphasis; the stem is fronted, as in (39), (40) and (41). Typically, the material that may come between the lexical stem and the conjugated stem are pronouns; in (39) also a particle. Note that the separation of the stem is possible even for the stems suffixed with -r- ((39) and (40)) and for bound stems in (41), though, distributionally, they do not occur outside compound constructions. (39) loose compounding: stem separation (stative) Ha-r ge-wur.i-s Kan-[i aq'e, Kan-[i
know-CMP that-PL-DAT want-COND do.IMP want-COND

m-aq'.a,
PROH-do.IPF

fira

degi]-x.a-s-Tawa

nothing change-become.IPF-INF-COP:NEG

Teach them or not, nothing will change. (lit. "teach them, teach them not...") (40) loose compounding: stem separation (dynamic verb) Ruza-r zun ge q'.a-s-e,
stand-CMP I(ERG) that(NOM) do.IPF-INF-COP

amma
but

mus
when

aq'.aj-[i,
do.IPF:PRS-COND

Ha-j-dewa.
know-CVB-COP:NEG

I'll stop him for sure, but I don't know when. (41) loose compounding: stem separation (nominal stem) gunt' gi aq'.u-ne, amma Xul.a-> qa[ix.i-n-dawa
heap that(NOM) do.PF-PFT but house-IN bring.PF-PFT-NEG

sara.
PTCL

He did gather it, but he did not bring it into the house. Thus, not only periphrastic causatives are close to analytical forms; compound causatives are also `loose' compounds. Both types of causatives belong to the same `typological stock' of the extremely widespread `do'-based causatives. They are most likely result of basically the same grammaticalization process that occurred in Agul twice, at different time. However, the two causatives, similar from the typological and diachronical points of view, are clearly distinct patterns. Together with labiles and lexical causatives discussed below in sections 2.3 and 2.4, compound causatives form a group of what we call below nonproductive causatives. Even the largest class of non-productive causatives, `do'-compounds based on adjectives, are lexical items rather than a fully productive category. All nonproductive causatives are opposed to periphrastic causatives in a uniform way. With nonproductive causatives an apud Causee is completely ungrammatical. All verbs that form nonproductive causatives are intransitive. All non-productive causatives form the same semantic opposition to the periphrastic causatives of the same verbs (roughly, that of direct vs. indirect causation; see Section 3). Another important point that concerns compound causative formation is the status of the suffix -r-. Within Agul, this suffix is certainly not a causative marker. Indeed, statives use this
This is, however, not true of all compounds; e.g. Din-aq'as `hide (tr)' is clearly preferred to Din-q'as (the same meaning).

15


suffix to form not only causative `do'-compounds, but also inchoative `become'-compounds, so it is a `compound forming' suffix (see Table 4); this is the reason why we gloss it as -CMP-, for `compounding', rather than -CAUS-). However, in some other Lezgian languages, a formally identical suffix is a non-productive causative suffix (Lezgian, Budukh) or a `redundant' transitivity marker (Lezgian); see (Klimov, Alekseev 1980: 189; Shejkhov 1980: 147-149; Haspelmath 1993a: 163-164, 358). It seems quite plausible that its generalization with statives in Agul is secondary, although further study is necessary.

2.2.6 `Do'-compounds: an overview
As we have seen, `do'-compounds, although by far less productive than `do'-causatives, play an important role in causative formation in Agul. These compounds are loose in the sense that in some contexts the lexical stem and `do' may be divided by some other material; however, such constructions are peripheral. The following table sums up properties of `do'compounds based on different lexical categories. Table 7. `Do'-compounds based on different lexical categories
productive compounding `become'-compound (inchoative correlate) nouns direct ± adjectives ± direct + statives ± (three statives) suffix -r+ non-stative verbs (dozen verbs) suffix -r-

As this table shows, the `do'-compounds of statives share properties both with `do'compounds of adjectives (they have the inchoative `become'-compound as a correlate) and of other verbs (they use -r- suffix for compounding), which corresponds to their mixed nature (intermediate part-of-speech status). The fact that statives and adjectives form inchoative ~ causative pairs in a regular way while other, non-stative verbs form only `do'-compounds, and only irregularly, is understandable. Adjectives and statives do not form periphrastic `do'causatives considered in 2.1; other verbs have a regular way to convey causative meaning. Both statives and qualificative adjectives denote a state; their `become'-compounds designate change of state (inchoative). Non-stative verbs may designate both the state and change of state by means of TAM marking system, so they do not require a `become'-compound as inchoatives; this is shown in Table 8. Table 8.
`become'-compound
change of state

stem stem
state

`do'-compound
cause to change the state

`do'-compound
cause to enter the state

enter the state

2.3 Labiles
Some verbs do not distinguish morphologically between non-causative and causative meanings; the only difference is the presence of the agentive argument in the latter case. These are labile verbs (syntagmatic conversive verbs in terms of (Nedyalkov & Silnickij 1969); P-labiles in terms of (Kibrik 1996); or ambitransitive verbs in terms of (Dixon 2000)). Cf. (42); in (b) the verb `break' is used transitively; in (a) the same verb is used intransitively, and the second core argument (ergative) is missing.

16


(42) labile: intransitive vs. transitive a. BagW ar<.u-ne
mirror(NOM) broke.PF-PFT

The mirror broke. b. BagW
mirror(NOM)

na
who(ERG)

ar<.u-ne ?
break.PF-PFT

Who broke the mirror? For the sake of brevity we will call occurrence of the labile in contexts like (a) intransitive labiles, and in contexts like (b), transitive labiles.

2.3.1 Intransitive labile or prodrop?
In a prodrop language like Agul, the surface structure does not always make it obvious whether we deal with an intransitive labile, as in (42a), or with a transitive verb with a prodropped or impersonal Agent, as in (43a) and (43b), respectively. (43) prodrop sal.a->
cattle_shed-IN

mal-ar
cattle-PL

ruK.a-a
butcher.IPF-PRS

I. `Elliptic' prodrop (Where is dad?) He's butchering cattle in the cattle-shed. II. `Impersonal' prodrop (What's going on?) They butch cattle in the cattle-shed (i.e. it's butchering time) Various tests can be used to distinguish between the two cases, including e.g. interpretations available for the imperative of the verb in question (Haspelmath 1993a, Kibrik 1996, Ljutikova 2001). For instance, the imperative of the Agul verb k'es `die ~ kill' may be interpreted both as `die!' and `kill!'. However, an imperative of the intransitive meaning of most if not all labile verb is problematic because they typically designate incontrollable states and processes (cf. Mirror, break!). Even in the case of the `die ~ kill' intransitive imperative is less natural (and probably more expressive). In some cases language specific tests can help13. In Agul, it is the availability of the involuntary Agent construction; (Haspelmath 1993a) uses the same test for Lezgian. This construction is only available to intransitive predicates; thus, if a verb combines both with regular Agent marked by ergative and involuntary Agent marked by apudessive, it means that the verb is a labile. (44) lability test: standard vs. involuntary Agent opposition . ru].a xed aTuz.u-ne
girl(ERG) water(NOM) pour_out.PF-PFT

The girl poured the water out. b. ru].a-f-as xed
girl-APUD-ELAT water(NOM)

aTuz.u-ne
pour_out.PF-PFT

The girl occasionally poured some water. Indeed, if the verb aTuzas `spill ~ pour' were an intransitive, it would not combine with ergative Agent; if it would be transitive, it would not combine with apudelative (involuntary) Agent. Thus, the verb is labile, used transitively in (a) and intransitively in (b). Lability tests are a useful formal means to prove the verb in question is labile. However, a native speaker of Agul (and probably more broadly of other Daghestanian languages) is always absolutely positive in answering the question whether the verb is strictly transitive (i.e. the omitted Agent is understood from the context) or labile (i.e. the situation is conceptualized
13

Thus, in Bagvalal, an Andic language of Dagestan, verbs distinguish between transitive and intransitive imperatives morphologically; the verbs that possess both are thus labile (Ljutikova 2001).

17


as Agent-less); cf. the appeal to the native perception of syntactic completeness in (Comrie 2000: 368, Note 6).

2.3.2 Semantics
The following labile verbs have been identified so far: (1) [`urXas `tear (paper, tissue ­ but not rope)'; (2) [`ut'as `crush (eggs, vegetables), crease (paper, clothes)'; (3) c'akas `become curved (as a person getting old); be or make deviant from the correct position (picture on the wall), uneven (drawing a line)' (transitive use is rare); (4) ar

(b) (c) (d) (e) (f) (g)

fire events (9 and 11) water processes and cooking verbs (12 through 14) `open', `close' and `lock' events (15, 16 and 17) dispersion verbs (18 through 22) birth and death events (23 and 24) and some additional verbal meanings that do not form any clear group (`cut', `be stretched ~ stretch', `drag', `spin').

The set of labile verbs, thought different in different languages, have some nucleus which recurs cross-linguistically. Investigating what kind of verbs may be labiles Martin Haspelmath comes to the following conclusion: "A verb meaning that refers to a change of state or a going on may appear in an inchoative/causative alternation unless the verb contains agent-oriented meaning components or other highly specific meaning components that make the spontaneous occurrence of the event extremely unlikely" (Haspelmath 1993b: 92-93). This definition is made somewhat more precise by specifying that events like `cook' or `boil' do require an Agent, but they require an Agent-Initiator who may then leave the process to its own. Thus, after being initiated by an Agent, the process develops autonomously (cf. e.g. Ljutikova 2002). In this respect, most of the Agul labiles are typical and do occur in other languages of the world, including verbs of deformation (`break') and destruction (`blow'), fire events, `boil', `melt', cooking verbs. The idea of spontaneous change of state or autonomous process is in fact so important that marginally some new intransitive labiles emerge, like in (45), which may occur among Aguls living in larger towns and is made possible by coming in use of washing machines that allow to conceptualize the process of washing linen as autonomous; same is probably true of qik'as `lock', though intransitive `lock' is less peripheral14. (45) occasional lability berHem ma]in.i->
shirt(NOM) machine-IN

wash.IPF-PRS

The shirt is being washed in the washing machine. However, some groups and verbs pose problems to Haspelmath's model. This is the case of at'as `cut', used intransitively in the following example (46) `cut': presence of Agent-oriented component of meaning ze t'ub at'.u-ne
my finger(NOM) cut.PF-PFT

I got my finger cut. (unintentionally) The verb `cut' is explicitly ruled out by Haspelmath as a candidate to lability because of the presence of an Agent-oriented component of meaning, some sort of sharp object used as an instrument. The transitive labile also has meanings `dig' and `saw', but these meanings are impossible for the intransitive labiles. Note that the presence of an instrument seems to be equally obligatory for `cut', `dig' and `saw'. The obvious reason is that intransitive at'as in (46) is used for an unintentional cutting, while unintentional digging or sawing are far less probable. In this case, the fact that the event is spontaneous overrules the presence of an Agent-oriented meaning component.

14

This seems to be the only evidence we have so far for Agul that a situation may be intermediate between strictly transitive and labile; more examples like that would probably corroborate (Kibrik 1996)'s claim that there may not be clear-cut borderline between agent-omitted transitive and intransitive use of labiles.

19


It is also unclear how the lability of `birth' and `die' predicates is compatible with the idea that an intransitive labile must designate a spontaneous situation15. It is obvious that any child-bearing includes the mother participant (although she may be conceptualized as more or less agentive depending on the speaker and the language). Still, in Agul (47) is a perfectly complete sentence. (47) intransitive `birth' event ge jaXc'ur-pu is.a
that(NOM) forty-Ord year(ERG)

HuPuq'.a->
Huppuq'-IN

ruX.u-f-e
be.born.PF-NMLZ-COP

He was born in Huppuq' in 1940. The `die ~ kill' lability may seem less problematic as `die' does not necessarily involve an Agent (some deaths are natural), and thus seems similar to e.g. `break' or `go out (of a fire)'. However, an important point is that, in Agul, even an Agent-caused death may be described by an intransitive labile. Cf. (48), where the intransitive labile is used both in (a) which suggests death at the battlefield and (b) which may refer to a peaceful death at home. (48) intransitive `kill' event a. ze Hadad
my grandfather(NOM)

de
battle-IN

k'.i-f-e
die.PF-NMLZ-COP

My grandfather was killed during the war. b. ze Hadad de my grandfather(NOM) battle-GEN

waXT.una
time(ERG)

k'.i-f-e
die.PF-NMLZ-COP

My grandfather died during the war. This proves an important point. The relation between dying and killing events is more or less similar for speakers of different languages. It is true that dying can sometimes happen without killing (i.e. spontaneously), just as an object may break without any human Agent or at least without the Agent's intention to achieve that result. However, the verb `break' is a typical labile verb cross-linguistically, while `kill' and `die' are typically expressed by different lexical stems. To explain this, we must admit that Haspelmath's claim that to be designated by a labile verb a change of state (or a going on) must be conceived as occurring spontaneously is neither necessary nor sufficient. The important thing is that the event of such spontaneous change of state must be conceived as basically the same event as its non-spontaneous correlate, and this is language specific. Both in birth and killing events Agents are present in the real-world situation, but, in Agul, they may be absent from the frame of linguistic conceptualization, which is then Patient-focused. Cf. the transitive and intransitive labile ruXas in (49). (49) `be born': transitive and intransitive . zun gada ruX.u-ne
I(ERG) son(NOM) bear.PF-PFT

I gave birth to a son. (the labile is used transitively; the speaker is a woman) b. za-s gada ruX.u-ne
I-DAT son(NOM) bear.PF-PFT

A son was born to me. (the labile is used intransitively; the speaker may be a woman or a man) The situations described in (a) and (b) are very much the same, however the wording in (a) is avoided as too much physiological. This is a natural effect if we admit that the frame in (a) includes the mother and the process of childbearing she and the baby are involved in. In
15

Haspelmath is of course aware of the fact that the predicate `kill' is labile in e.g. Lezgian.

20


(47) above the focus is on the Patient; but if we have to put more emphasis on the Agent, transitive labile is used: (50) focus on the Agent: transitive childbearing gi wa-s gada-jar ruX.u-ne
that(ERG) you-DAT son-PL(NOM) bear.PF-PFT

, she gave you three sons!

16

Actually, the same is true of many classical labiles. Indeed, situations like cooking or frying or boiling require an Agent who intentionally starts the process. They may be used intransitively because the process itself does not require direct control by this Agent, so that the Agent, again, may be removed from the frame.

2.3.3 Labile derivation
An important question concerning the labiles is whether intransitive and transitive uses of the same labile verb are equally important or one of them may be considered as secondary (cf. Kibrik 1996 for Godoberi). R.M.W. Dixon notes that speakers of English consider spill, smash and extend as primarily transitive and only secondarily intransitive, while melt or dissolve or walk are primarily intransitive and only secondarily transitive (Dixon 2000). In other words, in English at least some P-labiles are `anticausative labiles' (transitive derives intransitive), while at least some other are `causative labiles' (intransitive derives transitive). This does not seem to apply to the bulk of labile verbs in Agul: native intuition does not discriminate the two meanings of e.g. ar 16 17

In the Caucasus, as in many other traditional cultures, boy is a much more wished newborn. Our causative and anticausative labiles are transitivity-increasing and transitivity-decreasing labiles in terms of (Kibrik 1996), respectively.

21


at'as `cut', `dig' and `saw'; however, the semantic connection between these three meanings seems more obvious. Examining further examples would probably uncover some additional intermediate cases. Another important typological issue is what happens to labiles that undergo morphological derivation (prefixation in case of Agul). May a labile non-derived verb become strictly transitive or intransitive through prefixation or, vice versa, a non-labile verb acquire lability? Productive and semantically regular refactive prefixation preserves lability: cf. ugas `burn (tr, intr)' ~ qa-ugas `burn again (tr, intr)'; this is natural because refactive only designates repetition of the situation without changing its concept in any significant way. On the contrary, locative prefixation is formally less productive and semantically more complex and may considerably modify the original (`non-prefixed') situation. Thus, we would expect that locative prefixes may both preserve or change (non)lability of the verbal stem. Most of the labile verbs do not combine with prefixes, so this expectation is only rarely verifiable. However, there are some examples that corroborate it. Labile al-darkas `spin, turn around' uses the bound stem also present in Ra-darkas `turn over (e.g. of the hay)' and ki-darkas `promenade' that are strictly transitive and intransitive, respectively. On the other hand, both ugas `burn' and al-ugas `burn from outside' are labile; this is possible because the two situations are semantically close and share lability-licensing properties. Cf. another prefixed verb from the same stem, k-ygas `aspire, burn with desire, be eager to' which is strictly intransitive (it takes clause complement).

2.4 Lexical causatives
Lexical causatives are pairs of morphologically unrelated verbs that are admitted to be in a causative relation. Whether they are or not is often hard to decide. An example is the English go ­ should either (or both) of the verbs send or lead be admitted to be its lexical causatives? The most important criterion for establishing lexical causatives is discussed e.g. in (Dixon 2000). If one meaning is cross-linguistically often derived from another by a clearly causative pattern, as in the case of Agul ada boy(NOM) REFL(NOM) go.PF-NMLZ-COP:NEG

dad.a
father(ERG)

hat.u-f-e
send.PF-NMLZ-COP

The boy didn't go there himself (all by his own wish), it was his father who sent (him). There are very few Agul verbs that seem to be lexical causatives: ket'as `wake up (intr)' ~ keRkas `wake up (tr)'; Raj]as `stand up' ~ RahadWas `raise'; kyrq'as `touched (of inanimate)' ~ kerHas `make touch' (as in `the dress's fringe touched the wall' ~ `I made the
18

In this verb, as with the other lexical causatives below, common material that may catch one's eye is prefixes; roots are different.

22


dress's fringe touch the wall');
2.5 Semantically irregular causatives
With lexical causatives discussed in the previous section, one verb is a semantic causative of another without being its causative morphologically. In Agul (as in many other languages) one sometimes encounters the converse situation. Two verbs are related to each other formally by a causative pattern, but their semantic relation is not causative, at least straightforwardly. In this section we will discuss all cases of what we consider to be semantically irregular causatives in Agul, including labiles ugas `itch, scratch' (also `burn') and at'as `be cut, cut'; `do'-compounds Harq'as `learn by heart' (from Haa `know' ~ Harxas `learn') and (qa)Dik'arq'as `find (intentionally, after looking for)' (from (qa)Dik'as `find (occasionally)'); and, finally, the special case of iTarq'as `ache; feel pain' (from iTaa `be wrong, be ill' ~ iTarxas `start being wrong; fall ill'). The verb ugas `burn', a labile, has two meanings, apparently secondary, `itch' and `scratch', intransitive and transitive respectively; cf. (52 a, b and c). (52) `itch' and `scratch' a. ze k'il
my head(NOM)

ug.a-a ug.a-a
itch.INF-PRS

itch.INF-PRS

My head itches. b. ]ynyK.i
child(ERG)

k'il
head(NOM)

The child is scratching his (the child's) head. c. zun ]ynyK.i-n k'il
I(ERG) hild-GEN head(NOM)

ug.a-a
itch.INF-PRS

I scratched the child's head. The meanings `itch' and `scratch' are obviously strongly related, but this relation is not simply causative. To scratch does not mean to cause to itch (whatever the wise might say20); scratching is first of all a natural reaction to itching, not its cause. From the point of view of the argument structure of the two predicates, `scratch' introduces a new argument corresponding to the participant who does the scratching (the Agent). However, the participant may scratch his or her own body part; this type of context is at least as natural as the other when the Scratcher and the Scratchee are not the same person. When the Agent scratches his/her own body part, he/she is not a new participant. The same participant has been necessarily present already in the `itching' situation as the possessor of the body part and thus, indirectly, as the Experiencer21. In other words, in `I scratch my foot' there is no new participant as compared to `my foot itches'. It is the possessor of the body part and simultaneously the Experiencer who is `promoted' to the Agent role.
19

Note that the verbs Xas `lead' and Rajkas `cause to go, make hurry' are not perceived as causatives of
23


This situation is similar to another labile, at'as `cut', whose intransitive use is exemplified above in (46) - `I got my finger cut' (lit. "my finger got cut") vs. `I (intentionally) cut my finger'. The only difference that the intransitive use here seems to be secondary, so that it is not the experiential possessor of the intransitive construction who acquires agentive status in the transitive construction but the Agent who looses his or her agentivity in the intransitive construction. However, in the case of `scratch', `The child is scratching his back' in (52b) may qualify simply as a version of `I am scratching the child's back' in (52c), the possessor being coreferential with the Agent-Causer rather than promoted to this role. A stative-based compound `do'-causative Harq'as (from Haa `know' ~ Harxas `come to know, learn'), in addition to a straight causative meanings `let know, inform; teach', has a meaning `learn by heart' (e.g. a poem). This case corroborates the hypothesis of Agentpromotion in a more unambiguous way. Learning a text by heart is obviously more transitive (+ control) situation than a typically experiential `know', and unlike the other, regular causative meanings (`let know' and `teach'), the meaning `learn by heart' has only one human participant. Similar are Dik'arq'as `find' and its refactive qaDik'arq'as `find something that has been lost' or `find again (for the second time)'. The original, non-causative (qa)Dik'as `find' may have one or two arguments. When the lost item is found all by itself, e.g. a person or animal comes back home after being missed for some time, or when the focus is not on the person who discovered the object but on the fact that the lost object has been found, the verb may have only one argument. In this situation, refactive qaDik'as `find again' is much more natural and means something like `to be there again; not to be lost anymore' or `come back again (of an animate)'; cf. (53 a, b). All uses of the non derived Dik'as `find' are perceived as elliptical of the dative Experiencer argument (cf. Comrie 2000: 368, Note 6 on perceived completeness of `find' predications in Tsez). (53) `find', one argument: refactive, inanimate and animate a. ze t'ubal qa-Dik'.i-ne
my ring(Nom) RE-find.PF-PFT

My ring is there again. (lit. "was found again") b. ]ynyK-ar qa-Dik'.i-ne RE-find-PF-P child-PL The children came back. (lit. "were found again")

FT

If, however, the person who discovered an object is in focus, it is coded by a dative22; cf. (54 a and b).

22

Although the verb `find' can be used with or without dative argument, the dative here is a core argument, Experiencer, rather than a peripheral benefactive argument. Indeed, a benefactive dative (dadas `for (my) father') may be introduced in (54), even though it is more natural with the causative Dik'arq'as (cf. 55 below). In (Ljutikova 2001) similar uses of `find' in Bagvalal, Avar-Andic, Nakh-Daghestanian are qualified as noncanonical lability with two alternative case-assignment patterns and . Without arguing for or against grouping Dik'as `find' with other, canonical labiles discussed above in 2.3, we would like to indicate that the pattern of conceptualization here is in a way similar to that of e.g. `die' ~ `kill'. The monovalent Dik'as `find' may refer to both situations when the needed item is found all by itself (53b) or when it is irrelevant who is the person who found it (53a); cf. the intransitive labile k'es `die, perish' that may refer either to a natural death, as in (48b), or to a death in a battle when it is irrelevant who is the

24


(54) `find', two arguments: refactive and non-derived a. za-s sad qa-Dik'.i-ne
I-DAT one(NOM) RE-find.PF-PFT

I found another one. (e.g. one more bullet, coin or mushroom) b. za-s pul Dik'.i-ne
I-DAT money(NOM) find.PF-PFT

I found money. (e.g. occasionally found money someone put in a cache, or it suddenly occurred to me from whom I could lend it) In (54 b) the verb Dik'as `find' refers to a situation when someone unexpectedly finds the lost or hidden money. Its causative correlate Dik'arq'as `find' designates situation of intentional discovery, i.e. the act of finding as a result of a purposeful search. (55) intentional `find' zun dad.a-s pul
I(ERG) father-DAT money(NOM)

Dik'.a-r-q'.u-ne
find.IPF-CMP-do.PF-PFT

I found money for my dad. (56) `find': occasional vs. intentional a. za-s gugaj qa-Dik'.i-ne
I-DAT doll(NOM) RE-find.PF-PFT

I walked across my lost doll. b. zun gugaj
I(ERG) doll(NOM)

qa-Dik'.a-r-q'.u-ne
RE-find.PF-CMP-do.PF-PFT

I (finally) found my doll. (i.e. after looking for it) Note that this semantic shift from `find occasionally; walk across' to `find as a result of search' under causativization occurs elsewhere in Daghestanian; cf. (Kibrik 1996) for Godoberi, (Comrie 2000: 368) for Tsez, (Ljutikova 2001) for Bagvalal. Finally, consider a special case of another stative-based `do'-compound iTarq'as `ache; make ill'. Stative iTaa means `be ill' ­ of a person, as in (57a), or specifically of a recurring painful sensation or disease localized in his or her body part, as in (57b). Its `become'compound iTarxas conveys respective inchoative meanings (`fall ill (of a person), start being wrong (of a body part)'). (57) `ache', non-causative a. zun iTa-a I am sick. b. ze
I(NOM) be_ill-PRS

Xil
hand(NOM)

iTa-a
be_ill-PRS

my

My hand is hurting me. (i.e. aches sometimes; lit. "my hand is ill") In addition to the regular causative meaning `make ill' (indirect causation as in `The child fell ill because of you'), the `do'-causative iTarq'as also has meanings of `ache (of a body part)', `feel pain (of a person)', with an unexpected dative argument marking that breaks two morphosyntactic rules at the same time (first, there is no nominative required by any Agul
person who did the killing, as in (48a). When the second participant moves into the focus, the second argument ­ dative in the case of Dik'as, ergative in the case of k'es ­ is introduced.

25


verb; second, there is no ergative required by the causative morphology); cf. (58 a, b). Ergative Agent may be added to both sentences with the effect of `you're hurting my hand' and `you hurt me', respectively; cf. (58c). However, the ergative argument is clearly optional, so the verb is in a way labile. (58) `ache', morphologically causative a. ze Xil.i-s iTa-r-q'.a-a
my hand-DAT be.ill-CMP-do.IPF-PRS

My hand aches. b. za-s
I-DAT

iTa-r-q'.a-a
be.ill-CMP-do.IPF-PRS

I feel pain; it hurts. c. wun za-s
you(ERG) I-DAT

iTa-r-q'.a-a
be.ill-CMP-do.IPF-PRS

You're hurting me. We have no satisfactory explanation for the semantic effect the causative morphology takes with this verb. To sum up, the semantic irregularity of Harq'as `learn by heart' and Dik'arq'as `find' is clearly produced by agentivization of the original Experiencer (promoting the would-be Causee, Experiencer, into Agent's role). The rearrangement of the semantic roles and arguments with labiles at'as `cut, be cut' and ugas `itch, scratch', at least in some contexts, may be explained by the same process (ugas also manifests additionally irregular semantic development). The irregularity of another stative-based causative, iTarq'as `ache, hurt', on the contrary, does not match any other irregular pattern we are aware of for Agul. All of the original, non-causative situations have at least one property in common. With `know', `find', `itch', `be cut (of a body part)' and `be ill' the only human participant, the would-be Causee, is an Experiencer who has no control over the situation. Moreover, with exception of `know', the original situation suggests no default way of direct causation that some other experiential situations suggest (cf. `know' ~ `learn', `see' ~ `show'). Of course, indirect (non-default) causation is always available, but this tends to be associated with periphrastic `do'-causatives (see discussion in Section 3). In other words, these are situations for which there is no natural way to introduce an additional agentive participant, which is the basic function of causativization, while the non-periphrastic causatives are available. The language is forced to re-conceptualize the original situation. In doing so, it tries to keep as close to the prototype of the causativization as possible. The result is that some of the `derived' situations do feature a new Agent without adding a new participant; they merely change the role of the original human participant from Experiencer to Agent23. The causativization of iTaa `be ill' into iTarq'as `ache' is the case where this mechanism of finding an alternative causative-like interpretation apparently fails24.

23

This also applies to at'as `be cut', only the direction of the derivation changes. In other words, not the original Experiencer (possessor of the body part) is moved into the Agent slot by causativization, but rather the former Agent is moved into the Experiencer slot by decausativization. 24 Reinterpretation of an existing participant instead of introduction of a new one under causativization is also attested in Godoberi (Kibrik 1996). However, in Godoberi this applies to causatives from transitive verbs and results in intensification of the verbal meaning, assumedly caused by the Agent's increased agentivity (`X splashed water' causativizes into `X splashed water repeatedly'). Interestingly, the causative of the meaning `be ill, hurt' is also reported to be idiosyncratic (though not as irregular as in Agul; it simply develops an additional intensifying meaning: `cause to be ill' develops into `beat up').

26


3 Semantics and contrasts
The only fully productive pattern of causativization in Agul is periphrastic `do'causatives. Periphrastic causatives cover a wide range of causative meanings, including direct and indirect, coercive (`forced to') and non-curative (`did not prevent') and other types of causation. Below are various examples; some of them repeat the examples in Section 2.1. (59) direct causation ]ynyK ket'.a-s
child(NOM) wake_up.IPF-INF

q'e.
do.IMP

Wake up the child. (60) direct causation baw.a ]ynyK
mother(ERG) child(NOM)

Rarx.a-s
sleep.IPF-INF

q'.u-ne
do.PF-PFT

Mother made the child fall asleep. (e.g. put him/her to bed, or lulled him/her to sleep, etc.) (61) indirect causation we dallaj-ar.i
your talking-PL(ERG)

baw

Rarx.a-s

q'.u-ne.

mother(NOM) fall.asleep.IPF-INF do.PF-PFT

Your (endless) conversations made Mom fall asleep. (62) coercive malla.ji
molla(ERG)

gada.ji-w
boy-APUD

q'ur>an
Koran(NOM)

ruX.a-s
read.IPF-INF

q'.a-a.
do.IPF-PRS

The priest forces the boy read the Koran. (63) indirect (non-curative) baw.a-s agW.a-s q'.u-ne-wa
mother-DAT see.IPF-INF do.PF-PFT-Q

wun
you(ERG)

jarHun?
wound(NOM)

Why, you let your mother see the wound?! (the addressee was not supposed to let his/her mother see the wound not to upset her) (64) indirect (assistive) baw.a gad.a
mother(ERG) boy(ERG)

hi].a-s
flee.IPF-INF

q'.u-ne.
do.PF-PFT

Because of what mother did, the boy escaped. (e.g. helped him by unlatching the window). (65) direct or indirect (non-curative) ru].a ]ynyK kur.a-s
girl(ERG) child(NOM) become_dirty.IPF-INF

q'.u-ne.
do.PF-PFT

I. The girl made the child dirty. (on purpose, as e.g. being angry with the child) II. The girl let the child get dirty. (she didn't want to, but was diverted and did not prevent him from falling in the mud) In sections 2.2 through 2.4 we considered several highly irregular causative patterns for verbs (compound causatives, lexical causatives, labiles) as well as more regular but still not fully productive causatives based on adjectives, nouns and statives (compound `do'causatives), which we call non-productive causatives. For every non-productive causative there exists a parallel periphrastic causative (for causative compounds of adjectives, statives and nouns it is formed on the corresponding `become'-compound). The two causatives contrast semantically.

27


(66) compound causative vs. periphrastic causative: noun a. gada.ji Dinaba kun-ar degi]-q'.u-ne.
boy(ERG) secretly clothes-PL(NOM) change-do.PF-PFT

The boy has secretly changed the clothes. (e.g. he hid away some clothes and planted some other clothes instead). b. BakW at'us.u-na, kun-ar degi]-x.a-s q'.u-ne
light(NOM) put_out.PF-CVB clothes-PL(NOM) change-become.IPF-INF do.PF-PFT

wun.
you(ERG)

Incidentally turning off the lights, you caused the clothes to be changed. (e.g. the guests put on wrong coats because it was dark) (67) compound causative vs. periphrastic causative: adjective a. [u pijan-di qaj.i-[i, ]ynyK-ar Din-aq'
brother(NOM) drunken-ADV come.PF-COND child-PL(NOM) hidden-do.IMP

gi-q-as.
that-POST-ELAT

If the brother will be boozed when comes home, hide the children from him. b. [u pijan-di qaj.i-[i, ]ynyK-ar.i-w Dinu-x.a-s
brother(NOM) drunken-ADV come.PF-COND child-PL-APUD

q'-e.

hidden-become.IPF-INF do-IMP

If the brother will be boozed when comes home, make the children hide (e.g. tell them to go away to another room). (68) compound causative vs. periphrastic causative: dynamic verb a. gi zun reQ.y-n jaBan.i-l Ruza-r-q'.u-ne.
that(ERG) I(NOM) road-GEN middle-SUPER stand-CMP-do.PF-PFT

He stopped me (right) in the middle of the street . b. gi zun Bu seBeT.i raR.una-l
that(ERG) I(NOM) two hour(ERG) sun-SUPER

Ruz.a-s
stand.IPF-INF

q'.u-ne.
do.PF-PFT

Because of him I had to stand in the sun for two hours. (69) compound causative vs. periphrastic causative: dynamic verb a. ru].a kun-ar q'e]a-r-q'.u-ne
girl(ERG) clothes-PL(NOM) soak(Intr)-CMP-do.PF-PFT

The girl let soak the clothes. (e.g. by putting them in a bowl full of water to wash them later). b. ru].a lak-ar q'e].a-s q'.u-ne.
girl(ERG) foot-PL(NOM) soak.IPF-INF do.PF-PFT

The girl's feet got wet. (e.g. she walked carelessly on the river stones) (70) compound causative vs. periphrastic causative: dynamic verb a. gada.ji dad.a-s pul Dik'a-r-q'.u-ne.
son(ERG) father-DAT money(NOM) find-CMP-do.PF-PFT

The son managed to find money for his father. b. gada.ji dad.a-s pul Dik'.a-s
son(ERG) father-DAT money(NOM) find.IPF-INF

q'.u-ne.
do.PF-PFT

The son helped his father to find the money. (e.g. by helping him to get in touch with rich people) (71) transitive labile vs. periphrastic causative of the intransitive labile a. zun BakW at'us.u-ne.
I(ERG) light(NOM) put_out.PF-PFT

I turned off the lights. b. sin-ar.i-k
wire-PL-SUB/CONT

kerH.a-j
touch.IPF-CVB

wun

BakW

at'us.a-s

q'.u-ne

you(ERG) light(NOM)

put_out.IPF-INF do.PF-PFT

Because you meddled with the wires, the light is gone.

28


(72) lexical causative vs. periphrastic causative of the non-causative correlate a. dad.a zun waXTuna keRk.i-ne.
father(ERG) I(NOM) on.time wake_up(tr).PF-PFT

Father woke me up on time. b. dad.a-n haraj-ar.i
father-GEN shout-PL(ERG)

]ynyK
child(NOM)

ket'.a-s
wake_up(Intr).IPF-INF

q'.u-ne.
do.PF-PFT

Because of his father's yelling the child woke up. (73) lexical causative vs. periphrastic causative of the non-causative correlate a. baw.a alurq'.u ]ynyK Rahad.u-ne.
mother(ERG) fall.PF child(NOM) put_upright.PF-PFT

Mother helped the child (who fell) to get back to his feet. b. baw.a ]ynyK-ar.i-w Dil.i-l-as Raj].a-s
mother(ERG) child-PL-APUD earth-SUP-ELAT stand_up.IPF-INF

q'.u-ne.
do.PF-PFT

Mother made the children get up from the floor. (e.g. by starting washing the floor). The contrast between non-productive and productive periphrastic causatives is clearly related to the so-called direct vs. indirect (or contact vs. distant) causation distinction, as predicted already in (Nedyalkov, Silnickij 1969: 32). (In the case of (71) it is worth noting that in another Nakh-Daghestanian language, Godoberi direct vs. indirect contrast is observed between transitive use of labiles and morphological causative of intransitive labile (Kibrik 1996), and the obvious reason for that is that morphological causativization in Godoberi, unlike Agul, is a very productive pattern.) This opposition is considered to be basic for the semantic typology of causative constructions; contrast between direct and indirect causation is probably a universal distinction of the human language (Shibatani, Pardeshi 2001). At the same time, the categories of direct vs. indirect causation are examples of construct categories, i.e. they have no immediate semantic interpretations but represent clusters of features that may, in principle, vary from language to language. We need to investigate what exactly this opposition means in Agul, considering the examples in more details. In (66a) the Causer (the boy) is intentional and directly manipulates with the Causee/Patient (clothes); in (66b) the result is not necessarily anticipated by the Causer and he does not physically deals with the Patient. In (67a) the Causer deals with the Causees (Patients) in a more direct way, probably bringing them to another room, leading them by hand or pushing them by force; in (67b) the Causer merely asks or urges them to move to the other room. In (68a) the Causer is present in the situation, he stops the Causee by barring his way, starting to talk to him or grabbing his hand; in (68b) the Causer is not necessarily present and makes the Causee remain outside by for instance not leaving him the key to the door. In (69a) the Causer puts the clothes into the water intentionally by a direct manipulation, as e.g. intending to wash them later; in (69a) the negative effect on the Causer makes it clear she simply did not prevent her feet from getting wet, probably by not being cautious enough. In (70a) the Causer tries and finds the money and gives it to his father; in (70b) he makes it possible for his father to collect the money, without dealing with the money directly. In (71a) the Causer turns light off intentionally and in a regular, default way, using the switch; while in (71b) he achieves the same result unintentionally, as a side effect of his manipulations with the wiring. In (72a) the Causer wakes up the Causee, again, intentionally, by direct manipulation or address, in a regular way children are woken up; in (72b) he does so inadvertently, his yelling being caused by some other reasons. Finally, in (73a) the Causer puts a child upright or helps it to stand up from the floor or picks the baby who fell down intentionally, by taking it in hands or by hand; in (73b) mother probably did not have making

29


the children stand up as her main goal, but her activities made it impossible for the children to stay seated on the floor. Let us now generalize these distinctions. (a) Intentionality. The causation is always intentional with non-productive causatives, while not so with periphrastic causatives: it may be intentional (67, 70), unintentional and not envisaged (66, 73) and even clearly undesirable for the Causer (69, 72); or allow both interpretations (68, 71). (b) Causer's control after change of state. Periphrastic causatives often describe situations where the Causer, after the change of state took place, does not control the situation and can not (easily) revert to the previous state; non-productive causatives describe those situations where the Causer's control is preserved after the change of state: cf. (68), (71). (c) Manipulation. Non-productive causation suggests a physical interaction, direct contact with the Causee, while periphrastic causation tends to designate non-contact situations, as creating intermediate situations for which the caused situation is but an effect. Typically for instance, with non-productive causatives the Causer changes the state of the Causee, while with periphrastic causatives he or she changes the world so that this leads to an autonomous change of the state of the Causee. (d) Default way of achieving the result. If it is an intentional periphrastic causation, nonproductive causation is often a more regular, direct and easy way (default way) to achieve the needed result, cf. (71a and b). Note that, although speech causation is not a physical manipulation, this interpretation may be available for non-productive causatives, as in (67a, 68a, 72a); most likely because speech interaction is the default way of interpersonal causation. (e) Single (vs. multiple) event. Indirect causation in Agul clearly correlates with the multiple event model of indirect causation discussed in (Shibatani, Pardeshi 2001); unlike non-productive causatives that tend to consider the causing and caused situations as a single event, periphrastic causatives are perceived as a combination of two events. With periphrastic causatives the cause event often needs to be specified, while with non-productive causatives the cause is the default way to get the effect and is indivisible from the caused situation. Linguistically, this is reflected in explicit description of the cause by a separate clause as in (66b), (71b) or action nominal as in (72b), hardly possible for non-productive causatives. The distinctions are summed up in Table. Table 9. Non-productive Non-productive causatives Intentional Causer's control preserved after the change of state Physical interaction with the Causee Default way of causing the change of state or process vs. periphrastic causatives Periphrastic causatives Unintentional, undesirable No control after the change of state No physical interaction with the Causee Non-default, specific way of causing the change of state or process Single event, no need to make Two events, often necessary to the cause explicit make the cause explicit

30


But even this cluster approach is not enough to explain some distinctions where nonproductive causatives are far from being prototypical direct causations. (74) direct vs. indirect causation: interpreting the opposition a. wun guni kysuQa-r-q'.u-ne.
you(ERG) bread(NOM) go_stale-CMP-do.PF-PFT

The bread went stale because of you. (e.g. you were supposed to clear the table but forgot to take away and cover the bread). b. wun guni kysuQ.a-s q'.u-ne.
you(ERG) bread(NOM) go_stale.IPF-INF do.PF-PFT

Your bread went stale. (e.g. you did not finish it). The process of staling can only be initiated by an external Causer, but can not be controlled by him or her; it is very rarely intentional or desirable; no direct manipulation with bread can be done to achieve this result faster than in a natural course of events. If leaving bread uncovered is to be considered as default way of staling, then both (a) and (b) follow it. And it is not obvious that the event structure in (a) is simpler than it is in (b). Thus, none of the typical parameters discussed above allows to distinguish between direct and indirect causation in (74). Both seem to be instances of non-curative causation25. And yet, this contrast uses the same formal opposition. Native speaker distinguishes the causatives in (74a and b) by suggesting the Causer is respectively more vs. less responsible for what's happening; probably, in (74a) the Causer was responsible for removing the bread to keep it from staling, while in (74b) he/she simply did not finish his/her bread. It seems that the distinction in (74 a and b) is another manifestation of direct vs. indirect causation, adapted for the situational semantics of the verb. Similarly, no direct interpretation is available for the non-productive causative in (75). (75) direct vs. indirect causation: interpreting the opposition a. haraj, zun le k'Walas hat.u-ne Xi
Oh I(ERG) this from.heart send.PF-PFT PTCL

b.

My god, I forgot all about that! ]ynyK.i le dy]y]
child(ERG) this sorrow(NOM)

za-s
I-DAT

k'Walas
from.heart

go.IPF-INF

q'.u-ne
do.PF-PFT

Because of the child I forgot my troubles. (e.g. by being kind to me, or because the child fell ill and this problem took my mind away from my own troubles) c. za-s le k'Walas ].u-ne
I-DAT that(NOM) from.heart go.PF-PFT

I forgot about that. The periphrastic causative in (b) is clearly an indirect causation. The question is, in what sense (a) is a more direct causation. With lexical causative of `forget', it is always the person who forgets who is conceptualized as the Causer, the Causer is coreferential to the Experiencer, as in (a). The Causer here has a more direct access to his own memory than an external Causee in (b), and this seems to be the way in which the category of direct causation is realized with this verb. (The question arises then why the causative is used in (a), in the first place. The difference from (c) is that in (a) the speaker assumes responsibility for forgetting, he/she admits he/she was supposed and could have remembered about some important issue; unlike (c), he/she assumes some level of control over his memory.)
25

Note that the more prototypically direct interpretation is also available for (74a), as if someone would take some bread and leave it to stale intentionally, as to get some crumbles to feed birds. Important, however, is that it is not the only interpretation.

31


Thus, the distinction between direct vs. indirect causation strongly depends on the situation designated by the verb. Situations, even those that seem to be close to each other conceptually, may be compatible with prototypical direct causation to very different extent. Cf. for instance `fall asleep' and `wake up'. It is much harder to make someone fall asleep if he/she does not want to sleep than to wake up someone who doesn't want to wake up. That is why direct causation with `wake up' means that you shake someone or yell at him/her (i.e. direct manipulation), while direct causation with `fall asleep' may mean lulling the child or singing to it, i.e. creating environment that would induce sleep ­ something which would be typically indirect causation with other verbs. In other words, the non-productive causative chooses the more (or the most) direct causation, while periphrastic causative is left with some kind of less direct causation than that chosen by the non-productive causative. However, it must be kept in mind that these distinctions are very fine and are best seen when confronting the two examples, well thought of and thoroughly considered, which of course does not happen in real speech. It is not very clear how the opposition of direct vs. indirect causation is actually realized. And of course, if a verb has no non-productive causative altogether, as in (65), periphrastic causative covers both what is coded by a nonproductive and periphrastic causative with other verbs. Some periphrastic causatives further distinguish two patterns of case marking for the Causee, discussed in Section 2.1.2: original (ergative for transitive Causees and nominative for intransitive Causees) vs. apudessive, or, more rarely, apudelative marking. Although the relative plausibility of the original vs. apud marking seems to be different for different verbs and different Causees, there are minimal contexts where both markings are available. (76) original vs. apudessive Causee marking a. gi ]ynyK-ar.i waK.a-n jaK
that(ERG) child-PL(ERG) pig-GEN


q'.u-ne.
do.PF-PFT

meat(NOM) eat.IPF-INF

He let children eat pork. (e.g. he forgot that they are Muslims, or he neglected the dietary restrictions). b. gi ]ynyK-ar.i-w waK.a-n jaK that(ERG) child-PL-APUD pig-GEN meat(NOM) eat.IPF-INF do.PF-PFT

He made children eat pork. (e.g. although, being Muslims, they didn't want to) (77) a. Qun]i
neighbour(ERG)

Xurur.i
dog:Pl(ERG)

ze
my

Kel
lamb(NOM)

fac.a-s
catch.IPF-INF

q'.u-ne.
do.PF-PFT

Due to my neighbor's negligence the dogs attacked the lamb. (e.g. he carelessly left them off lead, or did not close the gate) b. Qun]i Xurur.i-w ze Qel fac.a-s
neighbour(ERG) dog:Pl-APUD my lamb(NOM) catch.IPF-INF

q'.u-ne.
do.PF-PFT

The neighbor sicced the dogs on the lamb. (78) a. dad.a ru] raR.una-k
sun-SUB/CONT

Rut'.a-s
stand.IPF-INF

q'.u-ne.
do.PF-PFT

father(ERG) girl(NOM)

I. Dad let the girl stand in the sun. II. Because of Dad the girl had to stand in the sun. (e.g. he forgot to leave her the keys)

32


b.

dad.a

ru].a-w

raR.una-k
sun-SUB/CONT

Rut'.a-s
stand.IPF-INF

q'.u-ne.
do.PF-PFT

father(ERG) girl-APUD

Dad made the girl stand in the sun. (e.g. as a punishment) (79) a. ru].a
girl(ERG)

gada.ji
boy(ERG)

k'arab
bone(NOM)

alhat.a-s

q'.u-ne

swallow.IPF-INF do.PF-PFT

Because of the girl the boy swallowed a bone. (e.g. she did not took bones out of the fish and he got one of them stuck in his throat) b. ru].a gada.ji-w k'arab alhat.a-s q'.u-ne
girl(ERG) boy-APUD bone(NOM) swallow.IPF-INF do.PF-PFT

The girl helped the boy swallow the bone. (e.g. she gave him some dried bread to help him swallow the bone stuck in his throat) The semantic distinctions between (a)'s and (b)'s are close to those considered above in that the Causers in (a)'s at least tend to be less intentional than in (b)'s. However, this contrast is seemingly more specific than between direct and indirect causation; we suggest that it focuses specifically on the Causer's control, which is higher in (a)'s and lower in (b)'s. The subtlety of contrast is aggravated by the fact that transitive and intransitive Causees behave differently. For intransitive Causees, original marking is an unmarked option associated with a regular level of control, it does not add anything in particular to the causative meaning. Apud intransitive Causee, on the contrary, does not control the situation the way he or she normally does. This marking is strongly associated with coercive causation; the Causee is forced to do what he/she does by the Causer, against his/her own will. In (78b), it is clear that the girl would prefer to avoid standing in the sun and run away, but has to do that because she is punished. In (78a) standing in the sun is much more volitional ­ she could have chosen to go to a friend, but she preferred to stay near home waiting for her father to get inside as soon as possible. For transitive Causees apud is unmarked, while the use of ergative puts some focus on an increase in control. In (76a) it is absolutely excluded that the Causer forced the children eat pork against their will, it was their wish. It is most likely an instance of non-curative causation. Note that more control on behalf of the Causee means here less control and probably unintentionality on behalf of the Causer, so that the whole causative situation is very close to a non-curative indirect causation. In (76b) unmarked case assignment is used. The coercive reading is still quite probable, but not obligatory as with intransitive apud Causee; it could as well mean the Causer let or allowed the children eat pork. Out of context, the type of causation is not very clear here; this is a situation where a specifying question is possible (He forced them to eat it, or what?). The example in (77) is very similar. In (a), again, the Causees (dogs) act on their own account, the Causee just being too careless, while in (b) the Causee orders the dogs to attack, sics them on the lamb, which is also the default way of causing dogs. The whole opposition is thus, again, similar to that of indirect vs. direct causation. The example (79) is different. It is true that the Causer, again, acts non-curatively. But, unlike other cases, the ergative Causee in (a) acts unintentionally, his control is reduced. This apparently contradicts our claim. However, cf. (a) and (b). In (a) it is he who swallows a fish bone all by himself, while in (b) the control has further decreased; he was unable, even intentionally, to swallow the fish bone that stuck in his throat and needed help that was provided by the girl, so this is an instance of assistive causation.

33


In general, the marked, original marking for the transitive Causee gives the semantic effect that is indeed close to indirect causation. This is in conformity with our claim that apud vs. ergative marking deals with less or more control on behalf of the transitive Causee, respectively. Indeed, giving more control to the Causee the speaker naturally takes this control away from the Causer, especially making him or her unintentional, which naturally lead to indirectness, especially non-curative effects. There is probably another factor also corroborating the indirect reading of the ergative marking. Transitive verbs do not have non-productive causatives, so the only variance in causative construction formation available to them is the marking of the Causee, which is then likely to be used as conveying the indirect vs. direct causation opposition, the fundamental semantic contrast in causatives. However, this factor alone would be unable to explain the data. It is important that the first approach (decrease vs. increase of the Causee's control) explains why the marking is used this way and not vice versa (by analogy with control properties of the intransitive Causee). Second, it seems important that apud marking to some extent (though not obligatory) implies coercion, which is by no means a necessary correlate of direct causation. And third, as was already mentioned, there is an apparent correlation between apud marking and control, apudelative being used to mark involuntary Agent. The model of the semantic contrast connected to the Causee marking is shown in Table 10. White and gray areas indicate original and apud marking, respectively. Table 10. Semantic Interpretation of the Causee Marking Options Marked control increased
Intransitive Causee Transitive Causee

Unmarked regular control nominative

Marked control decreased apud
(coercive)

ergative
(indirect, especially non-curative)

apud
(coercive, but not necessarily)

Thus, we argue that the semantic contrasts between different Causee case assignments, on the one hand, and between non-productive vs. periphrastic causatives, on the other, are not identical. The difference between the two choices is highlighted when we compare verbs for which all three choices of causativization are available ­ non-productive causative, periphrastic causative with original marking of the Causee and periphrastic `do'-causative with an apud marked Causee. These contrasts are only available for intransitive verbs, because transitive verbs form no non-productive causatives. Cf. (80), (81): (80) triple contrasts a. dad.a ru]
father(ERG) girl(NOM)

raR.una-k
sun-SUB/CONT

Rut'.a-r-q'.u-ne
stand.IPF-CMP-do.PF-PFT

I. Father put the girl to the sun. (e.g. to get her dry and warm after she fell in water) II. Father told the girl to stay outside in the sun while he himself went inside. b. dad.a ru] raR.una-k Rut'.a-s q'.u-ne
father(ERG) girl(NOM) sun-SUB/CONT stand.IPF-INF do.PF-PFT

Because of the father, the daughter had to stay in the sun. (e.g. he locked the door and forgot to leave her the keys)

34


c.

dad.a
father(ERG)

ru].a-w
girl-APUD

raR.una-k
sun-SUB/CONT

Rut'.a-s
stand.IPF-INF

q'.u-ne
do.PF-PFT

Father made the girl stand in the sun. (as a punishment, or sent her to work when the sun was hot) (81) a. dad.a
father(ERG)

gada
son(NOM)

dukan.i-s
shop-DAT

hat.u-ne
send.PF-PFT

Father sent the boy to shop. b. dad.a gada
father(ERG) son(NOM)

dukan.i-s
shop-DAT

go.IPF-INF

q'.u-ne
do.PF-PFT

I. Father permitted the boy to go to shop. II. Because of the father the boy had to go to shop. (e.g. the father would not let him drink, so that the son had to go to the shop himself) c. dad.a gada.ji-w dukan.i-s father(ERG) son-APUD shop-DAT go.IPF-INF do.PF-PFT

Father ordered the boy to go to shop. In (80a), the father could have put his small daughter under the sun to make her warm by a direct manipulation, or ordered her to stay out of the house. In (81a), all we know is that the son is told to go to the shop; this is a natural interpretation of a direct causation in combination with the situation of going to the shop. In both cases the agentivity of the Causee is not in question in any way. In (b)'s, the Causee is much more free in his/her choices, the Causer being unintentional or at least not directly interested in the result; these are typical instances of indirect causation. Finally, in (c)'s, the focus is on the loss of the control by the Causee, and the causation is very clearly coercive. Table 11. Semantic Interpretation of Formal Contrasts with Different Classes of Predicates Productive (periphrastic) Original Apud P-intransitive A-intransitive Transitive
indirect unmarked increased Causee control (indirect?) non-curative, permissive decreased Causee control coercive unmarked (direct?) coercive?

NonProductive
direct

4 Overview and Daghestanian perspective
This section first summarizes the data presented in the paper, and then proceeds to a very brief characterization of Agul causatives against the typological background of causativization in Daghestanian in general. We have considered various aspects of causative formation in Agul, a Lezgic language of Nakh-Daghestanian, North Caucasian, focussing on formal properties of causative constructions and semantic contrasts between different causatives of the same verb (when several causatives are available). There is only one fully productive model available for all verbs except few statives. This is periphrastic `do'-causative, a combination of the infinitive of the lexical verb with the verb (a)q'as `do' (Section 2.1). There are also several less productive models, including numerous causative `do'-compounds (combining an adjectival, nominative, stative or verbal stem and the verb (a)q'as `do') considered in Section 2.2, over 35


thirty labiles (Section 2.3) and few lexical causatives (Section 2.4). All verbs that have nonproductive causative correlates also form periphrastic causatives; the contrast between a nonproductive causative and periphrastic `do'-causative clearly lies in the domain of direct vs. indirect causation (Section 3). In periphrastic causatives, some of the Causees (animate Causees with A-intransitive and transitive verbs) may be marked either as they were marked in the original, non-causative construction (i.e. by nominative for intransitive Causees or ergative for transitive Causees) or by apudessive (or, more rarely, apudelative). Apud marking is more readily available to human and transitive Causees and is completely ungrammatical with inanimate or Pintransitive Causees and Experiencers (Sections 2.1.2 and 2.1.5). The contrast between original and apud marking is strongly intertwined with indirect vs. direct causation opposition; in addition, the default options are different in transitive vs. intransitive clauses. Still, it seems that the primary factor behind the choice of the case is the degree of the Causee's control; it is higher with original marking and lower with apudessive marking, the latter often resulting in coercive causative semantics (Section 3). `Do'-causativization is probably the most widespread causativization mechanism, crosslinguistically. The most regular and productive causative pattern of Agul, periphrastic causativization, is also based on the use of (a)q'as `do'. The interesting point about `do'causativization in Agul is that the same `do'-pattern is also present in another model, that of `do'-compounds. It seems that grammaticalization of `do' as a means of causativization occurred twice in Agul, apparently at different time. The two `do'-causatives have different morphosyntactic status. Periphrastic causatives are intermediate between two clauses and monoclausal construction (Section 2.1.4); `do'-compounds are clearly monoclausal constructions, though, speaking in terms of morphological autonomy of lexical stems, they are rather `loose' words (Section 2.2.5). Periphrastic causatives are treated differently in different Daghestanian languages; cf., on the one hand, (Haspelmath 1993a: 358) who, although quoting some other points of view on Lezgian `do'-causatives, assigns them biclausal structure, or similar solution argued for in (Kibrik et al 1982) for Tabassaran `let'causatives, and, on the other hand, (Ljutikova 2001: 384-6) who argues that Bagvalal `let'causatives are analytical rather than biclausal, or (Kibrik et al 1977: I, 98-107) view of lexical verb plus `do' combinations in Archi as complex verbs. The derivational pattern with pairs of inchoative ~ causative verbs produced from the same stem (2.2), so widespread in Agul, occurs elsewhere in Daghestanian. These are numerous as- `do' / kes- `become' complex verbs in Archi (Kibrik et al 1977: I, 98-107). Another, even more interesting parallel is with Tsez, where a structurally similar pattern of inchoative ~ causative derivation involves affixes which are not in any clear way connected to `become' and `do' meanings; similarly to Agul, in Tsez this derivation is typical of nonverbal stems (Comrie 2000: 366). A more distant analogy is -li or -di vs. -e: patterns in Bagvalal (Ljutikova 2001: 394). There is a widely known morphosyntactic problem of marking transitive Causees. On the one hand, there are two pretendents to A-marking, the Causer and the Causee; on the other, there is a strong cross-linguistic tendency not to use agentive (or any other core argument) marking twice in the same clause. In this conflict, the Causer is a universally preferred role in the languages of the world, so that the Causee has to come up with some other, construction-specific (constructional) marking. In most general terms, the choice is usually between the recipientive (dative) and some locative form. Agul, as well as most other

36


Daghestanian languages, opts for the latter. The choice of a specific locative form, however, varies greatly across Dagestan ­ compare apud (essive and elative) marking in Agul with cont-essive in Bagvalal (Ljutikova 2001), super-essive in Godoberi (Kibrik 1996), poss-essive in Tsez (Comrie 2000), or in-lative (illative) in Icari Dargwa (Sumbatova & Mutalov 2003). Apud marking of the Causee in Agul, though showing a tendency to grammaticalize in this conflict-resolving function (i.e. as `transitive Causee'-specific marking), is also available for intransitive Causees. This marking, at least at the present stage of its evolution, has a semantic function which, we argue, has to deal with the Causee's control. And it is exactly this function which probably presents the most typologically controversial fact about causativization in Agul. Indeed, we suggest that apud vs. original (nominative for intransitive, ergative for transitive verbs) marking correlates with the degree of control the Causee exerts over the causative situation. Correlation between Causee marking and degree of the Causee's control is a very well known phenomenon, but in Agul it works in the direction opposite to what seems to be universal. (Comrie 1981) discusses the Causee's demotion into oblique as a means of increasing rather than decreasing his or her control over the causative situation. However, we believe that this conflict is superficial. Indeed, Comrie's hierarchy is supposed to work primarily for intransitive Causees: nominative < dative < instrumental (ordered from the left to the right with increasing control). The reason behind this hierarchy is the principle of role marking of the Causee. Nominative marks a patientive role, dative marks a beneficiary/recipientive role, and instrumental marks causer/agentive role. Movement from nominative through dative to instrumental goes together with increase in agentivity and thus correlates with increase in control. In Agul, changing from the default nominative marking to the marked apud option as a means to decrease control is also role-based, even though the direction is opposite, because, most probably, this use of apud is associated with the role of involuntary Agent, a most nonagentive, control-lacking human role. It is probably not a coincidence that in Tsez the marking of (transitive) Causees is identical to that of involuntary Agent marking (Comrie 2000: 367). With transitive verbs, changing from the default apud marking to the marked ergative option increases agentivity and control because ergative is obviously the most agentive marking available in the language. Interestingly, North Tabassaran, a close relative of Agul, displays important differences from the causative profile of the former. An analysis of agreement patterns makes (Kibrik et al 1982) consider the causative `let'-constructions as biclausal (sentential complement) structures rather than causative auxiliary constructions. They also show that in North Tabassaran (Dyubek dialect), just as in Agul, there is a competition between the original (nominative/ergative) and constructional (dative) marking of the Causee. Formally, dative marking here is an analogue of Agul apud marking, but there is a controversy in semantic interpretation. (Kibrik et al 1982) suggest that the Tabassaran original vs. dative marking opposition conveys direct vs. indirect causation contrast, which in Agul is rather conveyed by non-productive vs. productive causative formation. Of course, the difference between increasing the Causee's control and changing from direct to indirect causation is vague in Agul, especially under causativization of transitive clauses. Note also that the meaning of coercive causation to which apud marking often amounts in Agul, especially for intransitive Causees, in Bagvalal is reported to be associated with productive periphrastic vs. less productive morphological causative opposition (which also conveys direct vs. indirect causation contrast).

37


Even for a Daghestanian language, Agul is rich in P-labiles; it seems that the number of labiles is reported to be higher only in Tabassaran. There are no strong indications that labiles split into clearly distinct classes of primarily transitive and primarily intransitive labiles (as Kibrik 1996 argues for Godoberi); the situation is more similar to that of Bagvalal labiles (Ljutikova 2001: 381-3). Only some verbs can reasonably be argued to prefer transitive or intransitive usage, basing first of all on an analysis of their lexical meaning. Considering some situations that require an Agent but can still be labile in Agul, such as `kill' or `give birth', we extend the invariant of lability from situations that may be both agentless and agentive (as e.g. Haspelmath 1993b argues) to, more generally, situations where the Agent may either be present in the real-world situation but out of focus of linguistic conceptualization. Typical of Daghestanian languages is special treatment of Experiencers. Indeed very few experiential verbs use transitive alignment, most Experiencers are marked locatively or by a dative, and in some languages even by a dedicated case marker (affective). This tendency is the reason why studies of causativization in Daghestanian pay special attention to causativization of experiential verbs (Kibrik 1996; Comrie 2000: 368; Ljutikova 2001: 3878). In Agul, from the point of view of causativization experiential verbs group together with intransitives in at least two ways: some of them form non-productive causatives and they do not allow apud marking of the Causee (the latter property groups them more specifically with patientive intransitives). Non-productive causativization of experiential verbs is further interesting in that it produces most or all of Agul semantically irregular causatives, recurring, at least in some of them, to the mechanism of Experiencer-to-Agent promotion (2.5). Competition of non-productive vs. productive causatives of the same verb is typologically widespread. (Nedjalkov & Silnickij 1969) suggest that such competition, when it occurs in a language, typically amounts to direct vs. indirect causation contrast, and this holds perfectly for Agul. Note, however, that an analysis of various interpretations of this contrast shows very clearly that the notions of direct and indirect causation are construct typological categories rather than cognitive primitives and include many parameters; in Agul, the most prominent parameters are default way of causation, intentionality and multiple event model; cf. (Shibatani & Pardeshi 2001). The list of glosses:
ADV APUD COND CVB COP DAT ELAT ERG GEN IMP IN INF INTER IPF LAT NEG NMLZ NOM adverbial localization: near the landmark conditional converb copular verb dative (case) elative (orientation) ergative (case) genitive (case) imperative localization: inside hollow landmark infinitive localization: inside homogeneous (compact) landmark imperfective (verbal stem category) lative (orientation) negative nominalizer used in adjectives and participles nominative (case)

38


OPT optative ordinal numeral ORD PF perfective (verbal stem category) PFT perfect (tense marker) plural PL POST localization: behind the landmark prohibitive PROH PRS present (tense marker) purposive converb PURP PTCL emphatic particle RE refactive (verbal prefix) reflexive (pronoun) REFL Q general question marker SUB/CONT localization: under or at the landmark CMP "empty" derivational suffix used in compounds; see Section 2.2.5 localization: on the landmark SUP Nominative has no overt marking and is thus enclosed in brackets (e.g. girl(NOM)). The oblique stem marker is not glossed but delimited by a dot (all other case markers are added to the oblique stem). Ergative is formally identical to the oblique stem and is glossed following the same principle as glossing nominatives, e.g. father(ERG).

Bibliography: Comrie, B. 1981. Language universals and linguistic typology. Chicago: University of Chicago Press. Comrie, B. 2000. Valency-changing derivations in Tsez. In: R.M.W. Dixon & A. Aikhenvald (eds), Changing Valency. Case studies in transitivity, 30-83. Cambridge: Cambridge University Press. Dixon, R.M.W. 2000. A typology of causatives: form, syntax and meaning. In: R.M.W. Dixon & A. Aikhenvald (eds), Changing Valency. Case studies in transitivity, 30-83. Cambridge: Cambridge University Press. Dobrushina, N.R. & Merdanova, S.R. 2004. Ustupitelnyje konstrukcii v agulskom jazyke. In: V.S. Khrakovski (ed) Tipologija usutpitelnykh konstrukcij, 523-541. St. Petersburg: Nauka. Ganenkov, D.S. & Merdanova, S.R. 2002. Semantika prostranstvennyx form sushestvitelnyx v agulskom jazyke. In: V.A. Plungian (ed) Issledovanija po teorii grammatiki. Vyp. 2: Grammatikalizacija prostranstvennyx znachenij v jazykax mira, 127-149. Moscow. Ganenkov, D.S., Maisak, T. A. & Merdanova S.R. In print. Prostranstvennyje i neprostranstvennyje znachenija lokalizacii APUD `vozle orientira' v agulskom jazyke. In: Gora jazykov... i jeshe odin. K 100-letiju E.A. Bokareva. Moscow. Ganenkov, D.S., Maisak, T.A. & Merdanova, S.R. In print. Non-canonical Agent marking in Agul. In: H. Hoop & P. de Swart (eds) Differential Subject Marking. Haspelmath, M. 1993a. A grammar of Lezgian. Berlin ­ New York: Mouton de Gruyter. Haspelmath, M. 1993b. More on typology of inchoative/causative verb alternations. In: B. Comrie & M. Polinsky (eds) Casuatives and transitivity. Amsterdam ­ Philadelphia: Benjamins. Hopper, P.J. & Thompson, S. 1980. Transitivity in grammar and discourse. Language 56 (2), 251-299. Kibrik, A. A. 1996. Transitivity in lexicon and grammar. In: A.E. Kibrik (ed) Godoberi. Muenchen: Lincom Europa. 39


Kibrik, A. E., Kodzasov, S.V., Olovjannikova, I.P. & Samedov, Dzh.S. 1977. Opyt strukturnogo opisanija archinskogo jazyka. Volumes I through IV. Moscow: MSU. Kibrik, A.E., Plungian, V.A. & Rakhilina, E. V. 1982. Kauzativnaja konstrukcija v tabasaranskom jazyke. In: Tabasaranskije etjudy. Moscow: MSU. Klimov, G.A. & Alexeev M.E. 1980. Tipologija kavkazskikh jazykov. Moscow: Nauka. Kulikov, L. 1993. The "second causative": a typological sketch. In: B. Comrie & M. Polinsky (eds) Causatives and transitivity, 121-154. Amsterdam/Philadelphia: Benjamins. Ljutikova, E.A. 2001. Perekhodnost' i diateznyje preobrazovanija. In. A. E. Kibrik (ed) Bagvalinskij jazyk. Grammatika, texty, slovari, 377-394. Moscow: Nasledie. Ljutikova, E.A. 2002. Kauzativy, dekauzativy i labilnyje glagoly v dagestanskikh jazykakh: opyt tipologii v predelakh jazykovoj gruppy. (Paper read at the XIth workshop of the Societas Caucasologica Europaea, Moscow, June 20-22, 2002) Magometov, A.A. 1970. Agulskij jazyk: Issledovanija i texty. Tbilisi: Mecniereba. Maisak, T.A. & Merdanova, S.R. 2002. Sistema prostranstvennyx preverbov v agulskom jazyke. In: V.A. Plungian (ed) Issledovanija po teorii grammatiki. Vyp. 2: Grammatikalizacija prostranstvennyx znachenij v jazykax mira, 251-294. Moscow: Russkie slovari. Maisak, T.A. & Merdanova S.R. 2004. «Proverjatelnaja forma» v agulskom jazyke: struktura, semantika i gypoteza o proiskhozhdenii. In: Yu.A. Lander, V.A. Plungian & A.Yu. Urmanchieva (eds) Issledovanija po teorii grammatiki. Vyp. 3: Irrealis i irrealnost', 430-455. Moscow: Gnozis. Nedjalkov, V.P. & Silnickij, G.G. 1969. Tipologija morfologicheskogo i leksicheskogo kauzativa. In: A.A. Kholodovich (ed) Tipologija kauzativnyx konstrukcij. Morfologicheskij kauzativ, 20-50. Leningrad: Nauka. Noonan, M. 1985. Complementation. In T. Shopen (ed) Language typology and syntactic description. Vol 2, 42-140. Cambridge: Cambridge University Press. Shaumyan, R.M. 1941. Grammaticheskij ocherk agulskogo jazyka (s tekstami i slovarem). Moscow/Leningrad. Shibatani, M. & Pardeshi, P. 2001. The Causative Continuum. In: M. Shibatani (ed) The Grammar of Causation and Interpersonal Manipulation. 85-126. Amsterdam/Philadelphia: Benjamins. Sheikhov, E.M. 1980. O formalnom vyrazhenii perekhodnosti glagola v lezginskom jazyke. In: Glagol v jazykakh Dagestana. Makhachkala. Sulejmanov, N.D. 1993. Sravnitelno-istoricheskoje issledovanije dialektov agulskogo jazyka. Makhachkala. Sumbatova, N.R. & Mutalov, R.O. 2003. A Grammar of Icari Dargwa. Muenchen: Lincom. Tarlanov, Z.K. 1994. Aguly: ix jazyk i istorija. Petrozavodsk.

40