Kabardian causatives, reflexives, and case marking domains/Kabardinski kauzativi, povratne konstrukcije i domene pripisivanja padeza.
O. Introduction (1)0.1. Kabardian (East Circassian) is an ergative head-marking language spoken in the Caucasus (NW Caucasian, Adyghe- Kabardian subgroup; ca. 500.000 speakers). It is spoken mostly in the Kabardino-Balkar Republic of the Russian Federation. There are also many speakers in the diaspora (Turkey, Jordan, Israel, USA, and elsewhere).
0.2. Kabardian is written in a modified Cyrillic script. The principles of translitteration used in this paper can be found in my "Short Grammar of Kabardian" (Matasovic 2009a).
0.3. The outline of this paper is as follows: in Section 1, I present the basic typological properties of the causative construction in Kabardian, using Dixon's (2000) typology of causatives. Section 2 discusses the case marking in Kabardian causative construction; it is argued that arguments of causativized verbs preserve the same cases they are assigned by the underlying non-causatives, in accordance with the "Dependent-first" strategy of case assignment (Matasovic 2009). In Section 3, I address the problem of the juncture/nextus type of Kabardian causatives. The key concepts of juncture and nexus are adopted from Role and Reference Grammar (RRG, see VVLP, Van Valin 2005), and will not be explained here for reasons of space. A number of arguments is used to show that we are dealing with nuclear coordination. An examination of reflexivized causatives in Kabardian, and the comparison of similar structures in Spanish, English, and Jakaltek, shows that languages can differ in their choice of the argument which serves as the binder of the reflexive in a nuclear juncture: in some languages this is the highest ranking macrorole of the causative verb (e.g. in Jakaltek), but in others this can be the highest ranking macrorole of the underlying base verb, which is the case in Kabardian. Finally, in Section 4 I discuss a number of theoretical issues relevant to RRG, especially the problem of the domain of case assignment. The apparent problem that case-marked independent RPs in Kabardian are outside the core, which is supposed to be the domain of case assignment, is resolved by positing the clause as the universal case assignment domain for all languages. The typological differences between languages such as English (where only one RP in the clause can receive the Nominative case) and Icelandic (where RPs in different co-ordinated cores can be marked for the Nominative) boil down to the contrast between "Head-first" case assignment (as in English) and "Dependent-first" case assignment (as in Icelandic and Kabardian).
1. Basic typological properties of Kabardian causatives
1.1. The causative is formed with the prefix -ga-. (2) It is regularly the last prefix in the prefix chain, occurring immediately before the verbal root: (3)
(1) w[??]-q'-ya-z-ga-h-a-s 2SG.U-DIR.-3SG.-1SG.A-CAUS.-carry-PRET.-AFF. "I made him carry you"
The only prefix that can occur between the causative prefix and the root is the factitive prefix -w[??]-. It is used to derive verbs from nominal and adjectival roots, e.g. w[??]f'[??]c'[??]n "blacken, make something black" from f'[??]c'a "black". Such derived verbs can be freely causativized (gaw[??]f'[??]c'[??]n "make someone blacken something"). If factitives are regarded as a type of causatives, as some linguists suggest, this construction may be viewed as a kind of double causative (on which see below).
1.2. Causatives can be formed from both transitive and intransitive verbs:
(2) k'wan "go": ga-k'wan "send"
(3) h[??]n "carry": ga-h[??]n "make someone carry"
Note that in many languages the formation of causatives is limited to intransitive bases only (Nedjalkov & Sil'nickij 1973: 7-8, Song 1996, Dixon 2000). Such languages, however, seem to be generally rarer in Eurasia than in other macro- areas. In the Caucasus, most languages seem to freely form causatives from both transitive and intransitive bases (Hewitt 2004, Klimov 1986). A notable exception to this claim is Lezgian (Haspelmath 1993), in which causatives are formed from intransitives only. In some Caucasian languages, including, e.g. Abkhaz, causatives can be derived from both intransitive and transitive verbs, but not from ditransitive verbs (e.g. the verb "to give").
1.3. In Kabardian, causatives can be formed from ditransitive verbs, but these often sound "unnatural", fabricated. The following example is elicited (I have not been able to find a causative of a ditransitive verb in my text corpus):
(4) sa wa ab[??]-xa-m s[??]-ra-w-z-ga-t-a-s I you(sg.) he-pl.-erg. 1sg.-3pl.-2sg.-1sg.-caus.-give-pret.-aff. "I made you give me to them"
1.4. "Double causatives" (causatives of causatives) are possible, as, e.g., in Turkish:
(5) sawp[??]-r (q'a-)v-a-s "the soup was boiling" soup-NOM (DIR.)-boil-PRET.-AFF. (6) s'ala-m sawp[??]-r q'-y[??]-ga-v-a-s "the boy was cooking soup" boy-ERG soup-ABS DIR.-3SG.-CAUS.-boil-PRET.-AFF. (7) sala-m y[??] ana-m sawp[??]-r y[??]-r-y[??]-ga-ga-v-a-s boy-ERG his mother-ERG soup-ABS 3SG.-3SG.-3SG.-CAUS.-CAUS.- boil-PRET.-AFF. "The boy made his mother cook the soup"
1.5. Double causatives can also be formed from transitive verbs; it has already been noted by Dixon (2000: 61) that such causatives are rare, especially in texts (see also Kulikov 1993); this is indeed also the case in Kabardian. The following example is elicited from a native speaker--I was unable to find a similar example in my text corpus:
(8) sala-m ana-m l'[??]z-am py[??]smaw-r boy-ERG mother-ABS old.man-ERG letter-ABS y[??]-r-y[??]-ga-ga-tx-a-s 3SG.-3SG.-3SG.-CAUS.-CAUS.-write-PRET.-AFF. The boy made mother make the old man write the letter"
1.6. Causativized verbs are freely combined with other valence-changing affixes, e.g. with applicatives and reflexives. Kabardian has two applicative prefixes, -d[??]- for comitatives (in (9)), and -xw[??]- for benefactives (in (10)): (4)
(9) psasa-m l'[??]-m f[??]z-xa-m girl-ERG man-ERG woman-PL.-ERG z-a-d[??]-r-y[??]-ga-s-a-s REFL.-3PL.-APPL.-3SG.-3SG.-CAUS.-carry-pret.-aff. "The girl made the man carry her (together) with the women" (Kumaxov & Vamling 2006: 57) (10) s-xw[??]-ya-b-ga-s-a-s 1SG.-APPL.-3SG.-2SG.-CAUS.-lead-PRET.-AFF. "You made him carry it for me"
The example (10) can only be interpreted as the causative of applicative, i.e. the sentence means "You did something, and this action caused him to carry it for me" rather than *"In my interests, you did something, and this action caused him to carry it" (applicative of causative).
1.7. Causatives can be formed from all verbal bases, irrespective of their Aktionsart.
1.8. The causative construction does not imply that the causer has control over the caused act. Inanimate causers are possible as well as the animate ones:
(11) d[??]ga-m m[??]l-[??]r y[??]-ga-v-a-s sun-ERG ice-ABS 3SG.-CAUS.-melt-PRET.-AFF. "The Sun melted the ice"
1.9. Almost universally, with animate causees, the caused action is intentional. Kabardian has a verbal prefix used to indicate unintentional action, ?as'a-, but this prefix seems to be incompatible with the causative. However, I found one instance of possible "involuntary causation" in my corpus:
(12) bzaxwy[??]-pl'[??]-r zady[??]l-ry[??] c[??]c' c'[??]k'w[??]-r bee-keeper-ABS together.rise-and he.goat small-ABS d[??]gw[??]z[??]-m q'[??]-?as'a-ga-xw-a-s wolf-ERG DIR.-INVOL.3PL.-CAUS.-put-PRET.-AFF.
"The four bee-keepers rose together and made the wolf (unintentionally) drop the little goat" (an example from the folk-tale about Zabagy Qazanoqo).
Note that the prefix -?asa- modifies the action of the original actor (the wolf), which is the derived causee, rather than the action of the derived actor (the four bee-keepers). This is an indication that the causee preserves some properties of the original actor under causativization.
1.10 Kabardian causatives generally imply that the action caused by the causer is actually performed by the causee. This means that it is impossible to say (without contradicting oneself), e.g. "The boy made the girl go, but she did not go". Such a sentence would be possible in some languages, e.g. in Korean (Park 1993, Song 1996: 13), in which there is no implication that the caused action is actually performed.
1.11. The Kabardian causatives can, with some hesitation, be characterized as direct and coercive (Shibatani 1975, Bishop 1992, Dixon 2000): the causative implies a straightforward means of bringing about the effect event, and it may imply the resistance on the part of the causee. However, the implication is not necessary. The causative form can also mean that the causer asked the causee to perform the caused action.
(13) l'[??]z[??]-m s'ala-m pxa-r y[??]-r-y[??]-ga-q'w[??]t-a-s old.man-ERG boy-ERG tree-ABS 3SG.-3SG.-3SG.-CAUS.-break-PRET.-AFF. "The old man made the boy cut the tree" / "The old man had the boy cut the tree"
The preceding sentence can mean both that the old man ordered the boy to cut the tree, but also that he asked him to do so. The correct interpretation must be inferred from the context.
2. Case marking in causative constructions
2.1. Kabardian has two grammatical cases: Ergative/oblique (glossed ERG) and Absolutive (sometimes also called "Nominative", here glossed ABS). Absolutive is the case of the lowest ranking macrorole argument, while Ergative is the case of the other core arguments (it can also mark some adjuncts). Indefinite NPs do not receive case marking. For a RRG account of case marking in Kabardian, see Matasovic 2008.
2.2. The case of the arguments in a causative construction is not determined by the argument structure of the causativized verb, which is always transitive, but by the verb from which the causative verb is derived.
2.2.1. If the base is an intransitive monovalent verb (14), its single argument becomes the causee of the derived causative, and receives the Absolutive (15). This is, of course, the same case it received in the underived, non-causative construction:
(14) s'ala-r gw[??]bgwa-m ma-k'wa boy-ABS field-ERG 3SG.PRES.-go "The boy goes into the field" (15) ana-m s'ala-r gw[??]bgwa-m y[??]-ga-k'wa mother-ERG boy-ABS field-ERG 3SG-CAUS.-go "The mother sends the boy into the field"
This case is unproblematic, since s'ala "boy" is the lowest ranking macro-role of both the base verb and the derived causative verb.
2.2.2. Again, if the base is an intransitive bivalent verb (e.g. dzan "read"), its single macrorole argument (its undergoer) remains in the Absolutive in the derived causative construction:
(16) yagadzak'wa-m yadzak'wa-r w[??]sa-m q'-r[??]-y-ga-dz-a-s teacher-ERG student-ABS poem-ERG DIR.-3SG.-3SG.-CAUS.-read-PRET.-AFF. "The teacher made the student read the poem"
Here the case marking pattern is unexpected, since w[??]sa "poem" is the lowest ranking macrorole of the derived transitive causative verb, and would therefore receive the Absolutive case, if the case were assigned by the causativized verb. However, it remains in the Ergative, since w[??]sa is the non-MR core argument of the base intransitive verb dzan "read", and non-MR core arguments, like Actors, receive the Ergative case.
2.2.3. Finally, if the base is a transitive verb (e.g. h[??]n "carry"), its lowest ranking macrorole (its Undergoer) again remains in the Absolutive:
(17) l'[??]z[??]-m s'ala-m x[??]dzabz[??]-r y[??]-r-y[??]-ga-h-a-s old.man-ERG boy-ERG girl-ABS 3SG.-3SG.-3.SG.-CAUS.-carry-PRET.-AFF. "The old man made the boy carry the girl" (18) l'[??]z[??]-m s'ala-m pxa-r y[??]-r-y[??]-ga-q'w[??]t-a-s old.man-ERG boy-ERG tree-ABS 3SG.-3SG.-3SG.-CAUS.-cut-PRET.-AFF. "The old man made the boy cut the tree"
Here the lowest ranking macrorole of the derived causative verb is in the Absolutive, since it is also the lowest- ranking macrorole of the base transitive verb.
According to Dixon (2000: 49) Kabardian would belong to a small group of languages in which the causee in a causative derived from a transitive verb retains its A-marking (marking of agents of transitive verbs, or Actors, in RRG terms). As a similar case he adduces an isolate, Trumai (Brasil), in which both the causer and the causee take the ergative marking in a causative construction. Note, however, that the Kabardian ergative is not just the case of the Actor, but also the case of the non-macrorole core argument, and in the causative construction the causee has precisely the status of non-macrorole core argument.
2.3. Moreover, the situation in Kabardian should be seen from a broader typological perspective. In Kabardian, as in many languages, the single macrorole argument of an intransitive verb becomes the Undergoer of the causative verb; this follows from two facts:
1. it is marked by the Absolutive case;
2. it is cross-referenced by the Undergoer form of the person prefix:
(19) w[??]-z-aw-ga-txa 2SG.U-1SG.A-PRES.-write "I make you write" (20) w[??]-z-aw-h 2SG.U-1SG.A-PRES.-carry "I carry you"
In causativized transitives, the causee becomes the non-MR core argument, as in French, Turkish, and many other languages; this also follows from two facts:
1. it is marked by the Ergative case, as, e.g., the recipient arguments of the verbs of giving;
2. it is indexed by the non-Undergoer form of the person prefix:
(21) 0-wa-z-ga-s'-a-s 3SG.U-2SG.-1SG.-do-PRET.-AFF. "I made you do it" (22) 0-wa-s-t-a-s 3SG.U-2SG.-1SG.A-give-PRET.AFF. "I gave it to you"
Note that the form of the 2nd person prefix is different for Undergoers (w[??]-) and for non-MR core arguments (w-).
2.4. There are many other languages in which subjects retain some subject properties as causees in causatives (Kozinsky & Polinsky 1993). (5) For example, Japanese subjects preserve their ability to bind the reflexive pronoun zibun (Comrie 1985: 336); this is also the case in Kabardian (see 2.3 above):
(23) Taroo ga Hanako ni zibun no hon o watasi-ta T. NOM H. DAT self GEN book ACC hand-PAST. "Taroo handed Hanako his/*her book" (24) Taroo ga Hanako ni zibun no huku o ki-sase-ta T. NOM H. DAT self GEN clothes ACC wear-CAUSE-PAST "Taroo made Hanako put on his/her clothes"
In Qiang (Tibeto-Burman) the original transitive subject preserves its ergative case marking in the causative construction (Dixon 2000: 49):
(25) qa the:-wu p[??]its[??]-e-ze z[??]-p[??]:-dza 1SG. 3SG.-AGENT cup-one-CLASSIFIER DIR.-buy-CAUS.1SG. "I made him/her buy a cup"
The pronominal suffix to the verb (-dza) cross-references the causer in (25); however, the causer NP never takes the agentive/subject marker -wu. It is the causee (the original Subject) that takes -wu. Still, the causer is cross- referenced on the verb as subjects normally are.
2.5. However, in Kabardian, the intransitive "subjects" also remain "subjects". What is special about Kabardian is that in causatives derived from intransitive verbs, the case marking of arguments remains the same as in the non-derived intransitives. If we assume that the causative prefix is the head, and the base verb the dependent element in a complex causative construction, we clearly have what I have termed the "Dependent-first" (DF) pattern of case assignment (Matasovic 2009). Since case marking is a consequence of the macrorole status of the arguments in the causative construction, we can also say that the exceptional status of M-intransitivity (6) of intransitive bivalent verbs (such as dzan "to read" in (16)) is preserved under causativization.
There are other languages which allow the DF pattern of case marking in causative constructions. For example, in Korean we find the DF pattern of case assignment in causatives formed from both transitive and intransitive verbs (Sohn 1999: 377):
(26) Minho ka [Mia ka/lul wus-key] hay-ss-e.yo M. NOM M. NOM/ACC laugh-to do-PAST-POLITE "Minho made Mia laugh" (27) Minho ka Mia lul /*ka wus-ky-ess-e.yo M. NOM M. ACC/*NOM laugh-CAUS.-PAST-POLITE "Minho made Mia laugh"
The example (28) is from Park (1993: 38):
(28) nay-ka Chelswu-ka say os-ul ip-key ha-ess-ta I -NOM Chelswu-NOM new clothes-ACC wear-COM do-PAST-DEC. 'I made Chelswu wear the new clothes'
It is important to note that this pattern of case marking in Korean is possible only in the analytic causative construction, as in (26) and (28) (Sohn's "long-form" or periphrastic causatives), but not in the synthetic causative, as in (27) (Park's "lexical causatives", cf. Park 1993), which is comparable to Kabardian causatives. Park (1993: 38ff.) argues that the periphrastic causatives involve core coordination, while the lexical causatives represent a nuclear juncture. Does this suggest that the causative construction in Kabardian is also a nuclear juncture, since it not only allows, but requires the DF-pattern of case marking? This is the question to which we turn in the next paragraph.
3. The Juncture and Nexus type of Kabardian causatives
3.1. Previous work on causatives in RRG (Park 1993, Toratani 2002:136ff., Paris 1999), has shown that languages can have several types of causative constructions, involving different types of nexus and different levels of juncture. The interclausal relations hierarchy proposed by RRG (e.g. Van Valin 2005: 206-208) predicts that the closer the connection between the two events, the stronger the syntactic relation between the syntactic elements depicting these events. Since the causative relation of the Kabardian type (direct causation) is at the very top of the hierarchy, we would expect the juncture to be at the nucleus level. And indeed, there are some indications that this is the case. For one thing, nothing except the factitive prefix can intervene between the causative marker and the verbal root, implying the iconic closeness of their connection.
3.2. We may first try a test used by Toratani (2002) in her RRG analysis of Japanese causatives. Temporal modifiers in Kabardian necessarily modify the causative structure as a whole, rather than modifying individual verbal cores (time adverbials are core modifiers in the Layered Structure of the Clause). This means that it is impossible to say things like "Today, the girl had the boy read the poem" in the sense that the girl said to the boy today to read the poem at a later stage (perhaps tomorrow). (7)
(29) psasa-m f[??]z[??]-r s'ala-m y[??]-r-y[??]-ga-w-a-s d[??]gwasa girl-ERG woman-ABS boy-ERG 3SG.-3SG.-3SG.-CAUS.-hit-PRET.-AFF. yesterday "Yesterday, the girl made the woman hit (at) the boy"
The preceding sentence cannot be taken to mean that the girl asked the woman yesterday to hit the boy at some other time; it is implied that both the causing and the caused act were performed yesterday. This could mean that the sentence contains one core rather than two distinct cores, which implies that the juncture level is the nucleus.
3.3. However, there are problems with this analysis. Whereas it may be true that the possibility of modifying two causal events by different temporal modifiers implies that we are dealing with core juncture (two distinct cores being independently modified), the converse does not hold: the obligatory scope of a temporal modifier over the whole causative construction may be the result of different factors, e.g. the culturally specific way in which causation is conceptualized in a particular language. Languages vary greatly with respect to the scope of time modifiers (Bohnemeyer et al. 2007), but it appears that generally individual cores have the "Macro-Event property" (MEP, the property of being necessarily within the scope of temporal modifiers). However, core junctures appear to be of two types (J. Bohnemeyer and R. Van Valin, p.c.). In core subordination, both cores often behave as a single core (i. e. the construction has the MEP), while in core coordination they do not (i. e. the construction lacks the MEP: individual subevents expressed by the core juncture can be in the scope of different temporal modifiers). In languages such as Ewe and Thai (Bohnemeyer et alii, to appear: 14-15), periphrastic causatives formed with light verbs have the MEP, i.e. the two parts of the causal event cannot be modified by two different time adverbials (with scope over only one subevent), although the causative constructions in question look like prototypical core junctures. According to the same authors (ibid, p. 20), Japanese is typologically peculiar in requiring the use of multiple Macro-Event expressions to encode the causal relation between the initial cause and the final change of state.
Thus, even if the construction involves two individual cores, it is still possible for the default interpretation of core adjuncts to be one in which they modify both cores simultaneously, rather than one of the two fused cores. One important piece of evidence in deciding the question of juncture level comes from reflexivization.
3.4. If a causative of a transitive verb is reflexivized, the reflexive prefix z- is bound by the causer, rather than by the causee (cp. also (9) above):
(30a) psasa-m s'ala-m z-r-y[??]-ga-w[??]c'-a-s girl-ERG boy-ERG REFL.-3SG.-3SG.-CAUS.-kill-PRET.-AFF. "The girl made the boy kill her" (lit. "kill herself") (30b) s'ala-m psasa-r z-r-y[??]-ga-w-a-s boy-ERG girl-ABS REFL.-3SG.-3SG.-CAUS.-hit-PRET.-AFF. "The boy made the girl hit (at) him" (lit. "at himself") (30c) s'ala-m psasa-r gwadza-m z-r-y[??]-ga-pi-a-s boy-ERG girl-ABS mirror-ERG REFL.-3sG.-3sG.-CAUS.-look at-back-PRET. -AFF. "The boy made the girl look at him in the mirror" (lit. "at himself")
In languages in which causatives are core junctures, e.g. in Spanish (Paris 1999: 56) and English, the reflexive is bound by the causee:
(31) Juan lo hizo a Pedro peinar-se J. him-ACC made to P.i comb-REFL.i "Juan made Pedro comb himself" (i.e. comb Pedro).
(32a) *John made Mary hit himself
(32b) John made Mary hit herself
In (32a), reflexivization cannot operate across the core boundary, but within the embedded core it is perfectly possible (in 32b).
In nuclear causative junctures, e.g. in Jakaltek French, and German, causers can bind reflexives (VVLP 607): (8)
(33) X-0-w-a' maka-'hin-ba t-aw-et (Jakaltek) PAST-3ABS.-1SG.ERG.-make hit-INF.-1SG.ERG-self. AUG.-2SG.ERG.-to "I made you hit me" (lit. "I made you hit myself")
3.5. However, the preceding arguments from reflexivization are not quite conclusive. If the suffix -z- "back" is added to the reflexivized causative verb, the causee becomes the binder of the reflexive, as in (34a-c):
(34a) s'ala-m psasa-m z-r-ya-ga-w[??]'[??]-z-a-s boy-ERG girl-ERG REFL.-3SG.-3SG.-CAUS.-kill-back-PRET.-AFF. "The boy made the girl kill herself" (34b) s'ala-m psasa-r z-r-y[??]-ga-wa-z-a-s boy-ERG girl-ABS REFL.-3SG.-3SG.-CAUS.-hit-back-PRET.-AFF. "The boy made the girl hit (at) herself" (34c) s'ala-m psasa-r gw[??]dza-m z-r-ya-ga-pl[??]-z-a-s boy-ERG girl-ABS mirror-ERG REFL.-3SG.-3SG.-CAUS.-look at-back-PRET.- AFF. "The boy made the girl look at herself in the mirror"
We see that it does not matter whether the base verb is transitive (as in 34a) or an intransitive bivalent verb (as in 34b-c): (9) the causee binds the reflexive in all of the examples. The difference with the examples (30a-c) lies in the fact that the binder of the reflexive is now not the actor of the causative verb, but rather a different argument. When the base verb is transitive, as in (34a), the binder is the non-macrorole core argument of the causative verb, as well as the Actor of the base verb. When the base verb is intransitive, as in (34b-c), the binder is the Undergoer of the causative verb, (10) as well as the single macrorole (the Actor) of the base verb. Note that there are independent arguments showing that in Kabardian the single macrorole of intransitive verbs can bind reflexives, as in the case of the intransitive bivalent verb wan "to hit (at)" in (35):
(35) s'ala c'[??]k'w[??]-r za-wa-z-a-s boy little-ABS REFL.-hit-back-PRET.-AFF. "The little boy hit (at) himself"
The suffix -z- appears to be obligatory with reflexivized intransitive bivalent verbs, which have only one macrorole. This shows that its primary function is to indicate that the binder of the reflexive is not the default (the highest ranking macrorole of the two macroroles of a transitive verb). In causatives, the suffix will consequently indicate that the binder is not the causer (the highest ranking macrorole of the derived verb), but the other macrorole argument, i.e. the causee. This is fully consistent with the other "Dependent-first" syntactic patterns in Kabardian, e.g. the "dependent-first" pattern of case assignment, which is also found in the Kabardian causative construction. What we observe in Kabardian reflexivized causatives with the suffix -z- could aptly be called "Dependent-first reflexive binder selection".
Thus, Kabardian differs from languages such as French and Jakaltek not in the juncture type of the causative construction, but in that rules governing reflexivization may apply to the highest ranking macrorole argument of the base verb, rather than of the derived causative verb. In both French and Jakaltek the binder of the reflexive must be the highest ranking argument of the derived causative verb, which is the causer. In Kabardian, both the causer and the causee are possible binders of the reflexive, since the causee is the highest ranking argument of the base verb. In English and Spanish, on the other hand, the binder of the reflexive cannot be the causer, since the bound element (the reflexive) does not belong to the same core as the intended binder. Therefore, in (31) above, only the causee can bind the reflexive clitic se, while the causer cannot: Juan lo hizo a Pedro, peinar-se. "Juan made Pedro comb himself" is fine, while *Juani lo hizo a Pedro peinar-se. "Juan made Pedro comb him (i.e. Juan)" is predictably impossible.
Therefore, we reach the following conclusion: if a language forms the causative construction as a core juncture, then the causee must be the binder of the reflexive in such a construction, as in the English example (32b) and in the Spanish example (31). The converse, however, does not hold: if the causer is the binder of the reflexive in a causative construction, it does not follow that the causative is of the core juncture type: it is still possible that we are dealing with nuclear juncture. Now, in a language with causatives formed as nuclear junctures, the question arises which argument is selected as the binder of the reflexive. It can be the highest ranking argument of the derived causative (the causer), as in French (11) and Jakaltek, or in Kabardian examples (9) and (30). On the other hand, the argument selected as the binder can also, in principle, be the highest ranking argument of the base verb. This is the case in Kabardian examples (34a-c), in which the marked selection of the binder is indicated by the suffix -z-. Of course, a language can have two different types of causative construction, say one which represents a core juncture, and another which is a nuclear juncture. Indeed, the two types co-occur in Japanese (Toratani 2002), Korean (Park 1993), and Spanish (Paris 1999).
3.6. It remains to determine the nexus type of Kabardian causatives. Modal operators, which take the core as their scope, by default modify just the first nucleus in the causative construction:
(36) s-xw-ya-ga-tx[??]-r-q'[??]m 1SG.-POT.-3SG.-CAUS.-write-PRES.-NEG. "I cannot make him write it"
The preceding example cannot mean *"I do not make him be able to write it". This is fully expected in nuclear juncture, since both nuclei have to be in the single core.
3.7. The aspectual adverb q'ana sama'awa "completely" appears to be able to modify just one event in the causative construction (for a similar test in Japanese see Toratani 2002):
(37) l'aza-m s'ala-m pxa-r ya-r-ya-ga-q'wat-a-s old.man-ERG boy-ERG tree-ABS 3sG.-3sG.-3sG.-CAUs.-cut-PRET.-AFF. q'ana s[??]-ma-'a-w[??] rest DIR.-NEG.-have-GER. "The old man made the boy cut the tree completely"
According to my informant (37) can also mean that the old man forced the boy to cut the tree, but that the tree was not cut completely. The other meaning is also possible: the old man made the boy cut the tree, and the tree is completely cut. Although this would have to be checked in more examples, including those found in the corpus, it appears that the nuclear modifier does not necessarily modify both nuclei in the juncture.
3.8. Since it seems likely that the operators do not necessarily have scope over both nuclei (see above), we can determine the juncture/nexus type of Kabardian causatives as nuclear coordination (FIGURE 1). (12)
[FIGURE 1 OMITTED]
4. The domain of case assignment in Kabardian causative construction
4.1. Kabardian is a Head-Marking (HM) language (Nichols 1992). In a HM language, personal affixes are core arguments (Van Valin 2005), and lexical Referential Phrases (RPs) are in the "extra-core position" (or ECP, Van Valin, p. c.); I believe this is the same structural position as the Pre-core/Post-core slot, but elements occurring in it are not necessarily focal (i.e. their status in the focus structure projection is not specified). (13) Pre-core slot and Post- core slot are just Extra Core Positions with special features imposed to them by the focus structure projection. In HM languages the ECPs do not necessarily interact with focus structure, which is why there can be more than one of them (for two or more RPs). On the other hand, I believe there is a universal prohibition against two instantiations of a lexical RP within a single clause, which explains why RPs do not normally occur in ECPs in dependent-marking languages: the only possible occurrence, in pragmatically unmarked contexts, of lexical RPs in those languages is in the argument positions within the core of the clause. (14) Since they have to occur in the core, they obviously cannot occur again in the ECP, because this would violate the restriction against more than one instantiation of lexical RPs in a single clause. (15)
This restriction may appear to be ad hoc, since it was introduced to account for a theoretical claim that needs to be justified in the first place, but I believe it is actually independently motivated. The prohibition against two or more instantiations of lexical RPs per clause also helps us understand why all languages have reflexives, either as independent pronouns (as in English or Croatian), or as bound markers on verbs (e.g. in Kabardian and in Bantu languages such as Xhosa). (16) There is, to my knowledge, no language in which the equivalent of English John saw himself would be expressed as *John saw John, because that would represent a violation of the restriction against more than one instantiation of a lexical RP per clause.
4.2. Kabardian allows multiple Absolutive RPs in a clause (in constructions with core coordination):
(38) a-r tx[??]l-[??]r y[??]-tx[??]-nw[??] xwayz-a-s he-ABS book-ABS 3SG.-write-INF. begin-PRET.-AFF. "He began to write the book" (Kumaxov & Vamling 1998: 201)
If presence of more than one Absolutive RP in a clause is taken as indication of the domain of case assignment, then it would appear that in Kabardian individual cores are domains of case assignment, rather than clauses (as in English, which does not allow more than one Nominative RP per clause). In the preceding example, the intransitive verb xwayzan "begin" assigns the Absolutive to its only macrorole, the pronoun a "he" (the argument it shares with the linked verb), and the verb tx[??]n "write" assigns the Absolutive to its lowest ranking macrorole (tx[??]l "book"); however, since the linked verb tx[??]n is transitive, the sentence (38) is also possible with the shared argument (a) in the Ergative (with a slightly different word order), since it is the other core argument of the linked verb:
(39) a-b[??] tx[??]l-[??]r y[??]-tx[??]-nw[??]xwayz-a-s he-ERG book-ABS 3SG.-write-INF. begin-PRET.-AFF. "He began to write the book"
This is again the "Dependent-first" (DF) pattern of case assignment (Matasovic 2009), which appears to be an areal feature in the Caucasus (Matasovic 2007).
As we saw above, we find the same DF pattern in Kabardian causative constructions. It is easiest to assume that, in causative constructions, the head nucleus (the causative "root") assigns case to its arguments after the dependent nucleus (the lexical verbal "root") had already assigned case to its arguments. But now an interesting problem comes to mind. Since the lexical RPs receiving case are outside the core, in the "extra-core position", they are outside the domain of case assignment. How can this be?
4.3. In the RRG linking algorithm, case is assigned to RPs before they receive their place in a constructional template (Van Valin 2005: 225-6). This means, e.g., that Wh-words can be assigned case if they are in a precore-slot (i.e. outside the core but within the clause) in a language in which cores are domains of case assignment (e.g. in Icelandic). The same solution might apply to Kabardian: the case is assigned to lexical RPs in the linking algorithm before the constructional template is applied. But still, I feel a little uneasy about this, since in Kabardian, the lexical RPs will be outside their "case assignment domain" in every single constructional template that exists in the language, which is awkward.
4.4. Therefore, here is a different proposal: let us assume that the clause is the domain of case assignment in all languages, in the sense that case-marked RPs are always within the clause, but that individual cores assign case to their arguments. (17) The differences between languages like English, on the one side, and Icelandic and Kabardian, on the other, can be reduced to the difference between the Head-first and Dependent-first patterns of case assignment. Here is how.
4.4.1. Consider the following English example:
(40) SheNOM believed himACC to have hit herACC
In the English sentence above, the syntactic structure is core coordination. There are two cores, the matrix (believe) and the linked core (have hit). Since English always applies the Head-first pattern of case assignment, the case is first assigned to the arguments of believe, and consequently, the single macrorole of believe (she) receives the nominative, and the other core argument (him) the accusative case. Since him is the argument that the matrix core shares with the linked core, its case is already assigned when the linked verb have hit takes its turn at case assignment. Consequently, it assigns the accusative case to its remaining argument, and the whole clause ends up with only one argument in the nominative. If the order of case assignment had been reversed, we would have *She believed she to have hit her, but this would be the consequence of the Dependent-first pattern of case assignment, rather than of the domain of case assignment.
4.4.2. Unlike English, Icelandic allows the Dependent-first case assignment in at least some constructions; the following example is taken from Van Valin 2005: 259:
(41) Jon-0 tel-ur mer haf-a alltaf pott J.-SG.NOM believe-3sG.PRES. 1SG.DAT have-INF. always think.PAST.PART. Olaf-ur leidinleg-ur O.-SG.NOM boring-SG.NOM "John believes me to have always considered Olaf boring"
In this example there are two nominative RPs, one in each core. This is because case is first assigned to the arguments of the linked core in which Olaf-ur is the highest ranking macrorole marked by the nominative. In the second step, case is assigned to the arguments of the matrix verb, so that the highest ranking macrorole of the verb telja "believe" can be assigned the second nominative in the clause (Jon). (18) It is important to note that a language can have the HF pattern of case assignment in some constructions, and the DF pattern (which is cross-linguistically much rarer) in others. In Enga (VVLP 580), the HF pattern in purposive constructions is the default, but the DF pattern is also possible:
(42) baa-(me) mend doko pyd-la pe-ly-d-mo 3SG.-(ERG) pig DEF. kill-INF. go-PRES.-3SG.-DECLARATIVE "He is going (somewhere) to kill the pig"
In (42) the ergative marking on the shared argument is preferred, but not necessary; it is assigned by the linked verb in accordance with the DF pattern of case assignment. Note that in Enga it would be awkward to explain the possibility of variable case marking in (42) by involving a difference in case-marking domains, rather than by the difference between the HF and DF strategies. (19)
4.5. If this analysis is accepted, the typological contrast between languages differing in the domain of case assignment is reduced to the difference in the order of case assignment, i.e. to the difference between "Head-first" and "Dependent-first" patterns of case assignment. As argued elsewhere (Matasovic 2007, 2008, 2009), this difference is independently motivated in syntactic theory.
References
Aissen, J. "Topic and focus in Mayan", Language 68/1992: 43-80.
Bishop, N. "A Typology of Causatives, Pragmatically Speaking", in: S. J. J. Hwang & W. R. Mer-rifield (eds.), Language in Context: Essays for Robert E. Longacre, SIL & The University of Texas at Arlington, 1992: 295-304.
Bohnemeyer, J. et alii "Principles of event segmentation in language: The case of motion events", Language 83, 3/2007: 495-532.
Bohnemeyer, J. et alii "The macro-event property: The segmentation of causal chains", to appear in: J. Bohnemeyer and E. Pederson (eds.), Event representation in language and cognition. Cambridge: Cambridge University Press.
Colarusso, J. A Grammar of the Kabardian Language, University of Calgary Press, Calgary 1992.
Comrie, B. "Causatives and Universal Grammar", Transactions of the Philological Society 1974: 1-31.
Comrie, B. "The syntax of causative constructions: cross-language similarities and divergences", in: Shibatani, M. (ed.), Syntax and Semantics, Vol VI: The grammar of causative constructions, Academic Press, NY 1976: 261-312.
Comrie, B. "Causative verb formation and other verb-deriving morphology", in: T. Shopen (ed.) Language Typology and Syntactic Description, III, C.U.P., Cambridge 1985: 309-348.
Comrie, B. & Polinsky, M. (eds.) Causatives and transitivity (Studies in Language Companion Series, 23), Benjamins, Amsterdam 1993.
Dixon, R. M. W. "A Typology of Causatives: Form, Syntax, and Animacy", in: R. M. W. Dixon and A. Aikhenvald (eds.), Changing Valency, C.U.P., Cambridge 2000: 30-83. Haspelmath, M. A Grammar of Lezgian, De Gruyter, Berlin 1993.
Hewitt, G. Introduction to the Study of the Languages of the Caucasus, Lincom, Munich 2004. Klimov, G. A. Vvedenie v kavkazskoe jazykoznanie, Nauka, Moskva 1986.
Kozinsky, I. & Polinsky, M. "Causee and patient in the causative of transitive: Coding conflict or doubling of grammatical relations?" in: Comrie & Polinsky (eds.) 1993: 121-154.
Kulikov, L. I. "The 'second causative': a typological sketch", in: Comrie & Polinsky (eds.) 1993: 121-154.
Kumaxov, M. A. Sravnitel'no-istoriceskaja grammatika adygskix (cerkesskix) jazykov, Nauka, Moskva 1989.
Kumaxov, M. A. (ed.) Kabardino-cerkesskij jazyk (I-II), Izdatel'skij centr El'-Fa, Nal'cik 2006. Kumaxov, M. A. & K. Vamling, Dopolnitel'nye konstrukcii v kabardinskom jazyke, The Lund University Press, Lund 1998. Kumaxov, M. A. & K. Vamling, Ergativnost' v cerkesskix jazykax, Malmo University, Malmo 2006.
Matasovic, R. "The Dependent First Syntactic Patterns in Kabardian and in other Languages of the Caucasus", paper presented at the Conference on the Languages of the Caucasus in the Max-Planck Institute, Leipzig, in December 2007; available from the site https://www.eva.mpg.de/lingua/conference/07-CaucasusConference/pdf/talk/ Matasovic_talk.pdf
Matasovic, R. "Transitivity in Kabardian", in: R. D. Van Valin Jr. (ed.), Investigations of the Syntax-Semantics- Pragmatics Interface, John Benjamins, Amsterdam 2008: 59-74.
Matasovic, R. "A new typology of control constructions within the framework of Role and Reference Grammar", in: Studies in Role and Reference Grammar, Lilian Guerrero, Sergio Ibanez-Cerda, Valeria A. Belloro (editors), Mexico City: Universidad Nacional Autonoma de Mexico, 2009: 305-318.
Matasovic, R. A Short Grammar of Kabardian (pdf): https://www.ffzg.hr/srmatasov 2009a.
Minger, D. An Analysis of Grammatical Relations and Case Marking in Icelandic, MA Thesis, UC Davis 2002, downloadable from the RRG website.
Nedjalkov, V. P. & Sil'nickij, G. G. "The typology of morphological and lexical causatives", in: Kiefer, F. (ed.) Trends in Soviet theoretical linguistics, Reidel, Dordrecht 1973: 1-32.
Nichols, J. Linguistic Diversity in Space and Time, The University of Chicago Press, Chicago 1992.
Paris, L. A. The Spanish Causative Construction 'hacer-infinitive'. A Role and Reference Grammar Description, MA Thesis, Buffalo 1999, downloadable from the RRG website.
Park, K. Korean Causatives in Role and Reference Grammar, MA thesis, Buffalo 1993, downloadable from the RRG website.
Peterson, D. A. Applicative Constructions, O.U.P, Oxford 2007.
Shibatani, M. A linguistic study of causative constructions, Indiana University Press, Bloomington 1975.
Sohn, H-M. The Korean Language, CUP, Cambridge 1999.
Song, J.J. Causatives and causation: A universal typological perspective, Longman, London 1996.
Toratani, K. The morphoyntax and the logical structures of compound verbs in Japanese, Ph.D. (Buffalo 2002), downloadable from the RRG website.
Van Valin, R. Exploring the Syntax-Semantics Interface, CUP, Cambridge 2005.
VVLP = Van Valin, R. & LaPolla, R. Syntax. Structure, Meaning, and Function, CUP, Cambridge 1997.
Wierzbicka, A. "Why 'kill' does not mean 'cause to die': The semantics of action sentences", Foundations of Language 13/1975: 491-528.
Ranko Matasovic
Faculty of Humanities and Social Sciences, Zagreb [email protected]
(1) A version of this paper was presented at the International Conference on Role and Reference Grammar in Berkeley, in August 2009. I would like to thank to Robert D. Van Valin Jr. and Johanna Nichols for their comments and suggestions. I am also indebted to my friend Lemma Maremukova for checking my Kabardian data.
(2) This prefix, as well as the causative formation itself, are inherited from Proto-Northwest Caucasian. The Abkhaz cognate, the causative prefix -ra- is also found in the position immediately preceding the verbal root (Hewitt 2004: 125).
(3) Throughout the paper I will use the following grammatical glosses: A = Actor, ABS = absolutive, ACC = accusative, aff. = affirmative, appl. = applicative, DAT = dative, dec. = declarative, def. = definite, dir. = directional (directional prefix), ERG = ergative, fut. = future, GEN = genitive, ger. = gerund, inf.= infinitive, invol. = involuntative, caus. = causative, neg. = negation, NOM = nominative, pl. = plural, refl. = reflexive, pres. = present, pret. = preterite, sg. = singular, U = Undergoer.
(4) The term "applicative" is here used in the sense of Peterson 2007, so that includes both comitatives and benefactives. The traditional term for benefactives in Russian grammars of Kabardian is "version", by analogy with the objective version in Kartvelian languages.
(5) Needless to say, the traditional concept of "subject" does not have any theoretical status in typologically- oriented syntactic theories such as RRG. A more precise formulation (in the RRG terminology) would be that the causee retains some properties of the highest macrorole argument of the underived verb.
(6) A verb is macrorole-intransitive (or M-intransitive) if it only has one macrorole in its logical structure. The M- transitivity of a verb is not the same thing as its valence, since semantically bivalent verbs can be M-intransitive, and trivalent verbs such as send, or give are M-transitive (Van Valin 2005: 64).
(7) See Wierzbicka 1975 for an early discussion of the scope of temporal modifiers in complex action sentences.
(8) In Georgian, reflexive causatives additionally have permissive meaning (Nedjalkov & Silnic'kij 1973: 13), e.g. Svil-ma amxanag-s tavi a-cem-in-a (son-ERG comrade-DAT himself caus-beat-caus-3sg) 'The son allowed his comrade to beat him (lit. 'himself)'.
(9) Note the different case marking, showing that the base verbs in (34b-c) are intransitive: the only macrorole of the base verb in (34b-c) receives the absolutive (psasa-r), while in (34a) the other macrorole (the non-undergoer) of the base verb is in the ergative (psasa-m). The rules of case marking in Kabardian were established in Matasovic 2008.
(10) We have seen in 2.1. above that the single macrorole argument of an intransitive verb becomes the Undergoer of the causative verb. Note that Undergoers cannot bind reflexives in any language.
(11) Cf. the French equivalent of the Spanish example (31), Jean s'est fait peigner par Pierre, where the causer (Jean) is the binder of the reflexive se. Unlike in Spanish, the French construction is a nuclear juncture (Van Valin 2005: 235f.).
(12) Note, however, that we have shown in 3.2. and 3.3. that Kabardian causatives have the MEP ("Macro-Event Property"), which means that temporal adverbs cannot modify individual nuclei in the juncture separately. According to Bohnemeyer and Van Valin (p. c.) this is expected in nuclear coordination.
(13) Wh-questions in languages such as English are fine examples of structures with a pre-core slot. Of course, wh- words occurring in such questions (e.g. in English Who did John see in the library yesterday) are focal (see Van Valin 2005: 6-7).
(14) In pragmatically marked contexts, lexical RPs can occur in the ECPs, but then they do not co-occur within the core. This is the case with the focalized RPs in the pre-core slot in English, as in the example Bean soup I can't stand (Van Valin 2005: 5). Note that it would be ungrammatical to repeat the extracted RP within the core (*Bean soup I can't stand bean soup), since this would violate the prohibition against two instantiations of a lexical RP in a single clause. The same holds in those HM languages, like the Mayan language Tzotzil (Van Valin 2005: 6, Aissen 1992), which have the pragmatically marked pre-core slots for focal elements besides the pragmatically unmarked ECPs., in which their RPs normally occur.
(15) This, in short, is the reason why there are no languages in which it would be possible to say the equivalent of English *John John saw Mary (with two instantiations of the lexical RP John), while a HM language such as Kabardian apparently allows the same individual to be referred to twice in the same clause (by a personal prefix on the verb and by the independent RP in the ECP).
(16) There are languages (e.g. Old English) in which the form of the reflexive pronoun does not differ from the form of the demonstrative pronoun, but this does not contradict our generalization.
(17) Of course, as an anonymous reviewer points out to me, case-marked RPs can also occur in the "left-detached position" (LDP), which is outside the clause, as in Croatian Ivan, on je jucer dosao (Ivan-NOM he(NOM) aux.3sg yesterday come(participle.sg.masc.)) "Ivan, he came yesterday". However, case-marked nominals in the LDP are not assigned case by any verb, but presumably receive case by agreement with co-referent case-marked nominals within the clause. In any case, this hypothesis should be investigated more thoroughly.
(18) The linked verb pykja (pple pott) also takes the dative case of its non-Macrorole core argument (the experiencer mer) shared with the matrix verb (Minger 2002: 38).
(19) In principle, one could perhaps argue that the ergative marking in (42) implies that only one Absolutive RP per clause is allowed, hence that the clause is the case marking domain in Enga. If the ergative on baa- is omitted, on the other hand, then one might say that the individual cores are domains of clause assignment, each of them permitting one Absolutive (unmarked) RP. But then it is difficult to see why the Ergative is disfavored when the linked verb is "to get" rather than "to kill" (VVLP 580): this clearly shows that it is the logical structure of the linked verb that is crucial to the issue of case assignment, i.e. that we are dealing with the DF pattern of case assignment, and that the clause is the domain in Enga. As Van Valin justly observes (VVLP 581), if the core were the domain, it would be impossible for the linked verb to affect the case assignment of the matrix verb in examples such as (41).
 Printer friendly
 Cite/link
 Email
 Feedback
| |
| Author: | Matasovic, Ranko |
|---|---|
| Publication: | Suvremena Lingvistika |
| Date: | Jan 1, 2010 |
| Words: | 7818 |
| Previous Article: | Slavic-style aspect in the Caucasus/Slavenski aspekt u kavkaskoj regiji. |
| Next Article: | Vladimir Anic, Nalicje kalupa / Sabrani spisi. |