CC BY
19
А.Б. Летучий
Школа лингвистики,
Национальный исследовательский университет «Высшая школа экономики», 105066, Москва, Россия
Согласование
с сентенциальными актантами в адыгейском языке1
В работе изучается необычное свойство конструкций с сентенциальными актантами в адыгейском языке (для анализа взят материал бжедугского диалекта). Оно состоит в том, что сентенциальные актанты могут контролировать числовое согласование глагола, в то время как во многих языках такое согласование маргинально (английский) или невозможно вообще (русский).
Мы показываем, что возможность согласования в целом зависит от двух факторов:
1) синтаксической позиции сентенциального актанта;
2) его формального типа (формы глагола в придаточном предложении). Однако согласование с сентенциальными актантами также сильно варьирует
от одного информанта к другому.
Информанты, не допускающие согласование с сентенциальными актантами вообще или в определенном контексте, могут вместо этого использовать конструкции, напоминающие дистантное согласование. Эти примеры рассматриваются в последней части статьи.
Ключевые слова: сентенциальные актанты, согласование, адыгейский язык, синтаксическая позиция, иерархия доступности именных групп, полисинтетизм.
1 Работа над статьей была бы невозможна без носителей адыгейского языка, работавших со мной в ходе экспедиции 2013 г. Кроме того, я исключительно благодарен коллегам, принимавших участие в адыгейских экспедициях НИУ ВШЭ и РГГУ (в особенности П.М. Аркадьеву, К.А. Ершовой и Я.Г. Тестельцу). Наконец, мне хотелось бы поблагодарить двух анонимных рецензентов за критику, вопросы и замечания.
A.B. Letuchiy
School of Linguistics,
National Research University «Higher School of Economics», Moscow, 105066, Russia
Agreement with complement clauses in Adyghe1
In my paper, I will analyze a special feature of clausal complements in the Bzhedug dialect of Adyghe, a polysynthetic language of the West Caucasian family. While in many languages, clausal complements cannot trigger verbal agreement, in Adyghe, the matrix verb can bear plural agreement with a clausal complements, along with the default singular agreement. If there is a coordinate structure including several clausal complements, the agreement slot they correspond to can contain a plural marker. However, this is not obligatorily the case. I will discuss the conditions of this unusual agreement pattern.
I will show that the possibility of agreement depends on at least two syntactic parameters: namely, the syntactic position of the clausal complement and the morphological verb form which is used in the complement clause. Non-standard agreement is also subject to a significant inter-speaker variation. Sometimes speakers who do not allow agreement with clausal complements admit instead long distance agreement.
The most unexpected thing is that even if the complement clauses are marked with a (typically) non-argument suffix, they can control agreement. This points to the fact that neither the canonical view of the pronominal argument hypothesis, nor classical approaches to agreement, represented in many works on European languages is plausible for Adyghe. The data of this language must be accounted for in a perspective that regards verbal personal markers and verbal arguments as types of items, which are not isomorphic to each other and which both play role in the agreement marking.
Key words: clausalcomplements, agreement, Adyghe, syntacticposition, acces-sibilityhierarchy, polysynthesis.
1 I am extremely grateful to the Adyghe native speakers who worked as language experts with participants of the field trip in 2013. I also thank my colleagues who took part in the expeditions (especially Peter Arkadiev, Ksenia Ershova and Yakov Testelets). Last but not least, I am thankful to the two anonymous reviewers for their critiques, questions and remarks.
1. Introduction
In the world's languages, verbs can, along with noun phrase (NP) arguments (e.g., I know this story), have clausal complements (I know that Peter is here). While nominal arguments represent the canonical argument type, clausal complements can be more or less like nominal arguments or like sentences (see [Noonan, 2007] for a comprehensive classification of CCs).
On the one hand, clausal complements across languages often demonstrate a number of nominal properties. For instance, some of CCs can become subjects of passive constructions (1)1, or control some types of floating quantifiers, e.g., samo po sebe 'by itself' (2).
Russian:
(1) Mne by-l-o objavle-n-o
I. DAT2 be-PST-SG.N announce-PARTCP.PASS.PST
cto poezdk-a ne sosto-it-sja.
that trip-sG.NOM not occur-PRs .3 sg-refl
'It was announced to me that the trip was canceled.'
(2) Sam-o po sebe prijatn-o cto on priexa-l-0. self-N.sG. on own.DAT pleasant-PRED that he.NOM come-PST-sG.M 'The fact that he came is pleasant by itself.'
On the other hand, in most languages clausal complements do not possess the whole set of typically nominal features. For instance, [Davies, Dubinsky, 2009; Delahunty, 1983] and others show that in English, clausal 'subjects' in fact do not show many typical subject properties.
Similarly, in Russian, clausal complements behave differently from NPs in many diagnostic contexts, including nominalization. Nominal arguments of verbs tend to be retained under nominalization of the verb form. The same is not necessarily true for clausal complements. [Letuchiy, 2012; Knjazev, 2014] show that in many cases, as in (4), they cannot be retained:
(3) Vyjasni-l-o-s' cto vinovat sofer-0. turn.out-PST-sG.N-REFL that guilty.M.sG driver-sG.NOM 'It turned out that the driver was guilty.'
1 Note that the subject status of the clausal complement in (1) has itself to be checked, but this issue is beyond the scope of the present article.
2 Abbreviations: ABS - absolutive agreement, ADV - adverbial marker, AP - antipassive, AUX - auxiliary, BEN - benefactive, CAUS - causative, COM - comitative, COND - conditional, COORD - coordination marker, DIR - directional marker, ERG - agreement with A' of transitive verbs, INF - infinitive, INS - instrumental, IO - indirect object agreement, LNK - linking element, LOC - locative, MSD - masdar, OBL - oblique case, OBLIQUE - default applicative marker, PARTCP - participle, PRS - present tense, POSS - possessive marker, PST - past tense, RE - refactive, REFL - reflexive, VN - deverbal noun.
(4) *vyjasneni-e cto vinovat sofer-0. turn.out.VN-SG.NOM that guilty .M.SG driver-SG.NOM Intended: 'Turning out that the driver was guilty.'
The present paper focuses on one property of clausal complements in Adyghe: namely, their ability to control number agreement in the verb form. I will show that this property is not equally demonstrated by all clausal complements, but rather depends on their syntactic position and formal type.
The article is structured as follows. In Section 2, I show that clausal complements usually do not control agreement, though some exceptions exist. In Section 3 I sketch the relevant grammatical features of Adyghe. Section 4, central for the paper, focuses on agreement with complement clauses and the relevant factors, which facilitate or block it. In Section 5, constructions with long distance agreement are analyzed, which are sometimes used by native speakers as an alternative to agreement with CCs.
2. Absence of plural agreement
in constructions with clausal complements
Control of agreement is one of the nominal features which are not characteristic of clausal complements. As a rule, a coordinate construction including two or more CCs cannot trigger plural predicative agreement (see, for instance, [Quer, 2008] for details and example (6) from Russian). In the Russian example (6) below, the plural variant is ungrammatical -the singular form must be chosen. By contrast, (5), where two nominal arguments are coordinated, either plural or singular is possible, but plural sounds definitely better:
(5) Mne nravj-at-sja/ nrav-it-sja Petj-a i
I. dat like-PRs .3 pl -refl/ like-PRs.3sG-REFL Petja-sG.NOM and
Vasj-a.
Vasja-sG.NOM
'I like Petja and Vasja.'
(6) Mne nrav-it-sja/ *nravj-at-sja plava-t' i nyrja-t'. I.dat like-PRs.3sG-REFL/ like-prs.3pl-ref swim-iNF and dive-iNF 'I like swimming and diving.'
This can occur for various reasons, but in general, it is explained by the fact is that clausal complements are syntactically different from canonical (NP) arguments. For instance, CCs lack nominal categories, such as (in Russian) nominal number and case.1 It seems that their absence
1 This is why nominalizations in Russian are able to control agreement, though this option is restricted and in some contexts even impossible.
prevents the predicate from bearing plural agreement. On the other hand, the internal structure of CCs is also different from that of NPs: Davies and Dubinsky (2009) demonstrate the relation between the fact that CCs and NPs differ in the category of the head (N vs. V) and the syntactic organization and (im)possibility of agreement.1
However, it turns out that the absence of agreement with sentential arguments is not universal. It has been noted that some languages, such as English, have constructions where plural clausal subjects control agreement.
(7) That he'll resign and that he'll stay in office seem at this point equally possible. ([McCloskey, 1991, p. 564], cited by [Davies, Dubinsky, 2009, p. 123])
Usually these constructions are infrequent, compared to standard (non-agreeing) constructions. However, the very possibility of agreement shows that clausal complements (in languages like these) are closer to NPs than it may seem.
In my paper, I will show that in Adyghe (a language of West Caucasian group) CCs can control plural agreement, though its possibility is conditioned by a number of factors.
To analyze agreement with clausal complements, I will first list the positions that can theoretically trigger verbal agreement. Adyghe has four of them: A (Agent / active argument of transitive verbs), P (Patient / non-active argument of transitive verbs), IO (indirect objects of transitive and intransitive verbs - one verb form can have several IOs and contain several IO cross-reference markers) and S (Subject / the most privileged argument of intransitive verbs). Though the question of syntactic position of a sentential argument is much more complicated than for NPs, I will consider CCs which correspond to NPs as occupying one of these four positions.
3. General grammatical features of Adyghe
Before starting the analysis of clausal complements, relevant grammatical features of Adyghe have to be sketched. I will mention only two of them: polysynthetic verb forms and multiple agreement and the ergative alignment.
1 In the Russian cases like (6), the different syntactic organization of clausal complements and NPs seems to be more important than the fact that CCs lack the category of case. This is obvious from the fact that clausal complements, marked with the pronoun to + the complementizer cto cannot control verbal agreement either, though their syntactic head - the pronoun to - is marked for case.
3.1. Polysynthetic verb forms
Adyghe is a polysynthetic language in terms of [Jelinek, 1984; Baker, 1996; Fortescue, 1994, 2007] and others. This means that many relations which cross-linguistically tend to be expressed in syntax (= with autonomous constituents), in Adyghe are moved to morphology and are coded inside the verb form.
The Adyghe verb form is organized in slots. In other words, most morphological markers, at least in the prefixal part of the verb form, have a fixed position with regard to each other and to the verbal root, and their mutual order does not correspond to the order of derivation or the relative scope (see, though, [Lander, 2012] for exceptions). For the reader's convenience, I provide here a complete list of slots showing in general words which grammatical meanings are encoded in each prefixal or suffixal slot.
The following example from Temirgoy Adyghe, taken from Lander (2012), shows that the verb form includes aspectual, temporal, modal, locative markers, valency change operators and agreement markers.
Temirgoy Adyghe:
(8) Sa-qa-t-de-p-fa-r-a-Ka-je-s'ta-K.
1SG.ABS-DIR-1PL.IO-COM-2SG.IO-BEN-3SG.IO-3PL.A-CAUS-read:AP-AUX-PST 'They were making me read it to you together with us.' [Lander, 2012]
The verb form in (8) contains an absolutive 1SG cross-reference marker ss- (slot -9 of the a directional prefix qa-; an oblique 1PL IO cross-reference marker t- introduced by the comitative applicative marker de-; an oblique 2SG IO cross-reference marker p- introduced by the benefactive applicative marker fa-; an oblique 3SG IO cross-reference marker r-; an oblique agent 3PL cross-reference marker a-; the causative marker Ka-; the root je- 'read'; and a composed imperfective suffix -s'ts-K).
The structure of the noun is much simpler, as it is in most polysynthetic languages (see [Jelinek, 1986; Baker, 1995; Fortescue, 1994, 2007]). Typically, a noun includes case markers (though they can be omitted under the non-specific reading).
However, necessary to say, the base distinction in Adyghe is not between nouns and verbs, but rather between predicate and argument positions. If a word which we normally regard as a noun (e.g., 'teacher') occupies the predicate position ('He is a teacher'), it can take almost all markers which are usually referred to as 'verbal' in grammars (see, for example, [Rogava, Kerasheva, 1976]).
Table 1
Structure of the Adyghe verbal complex
Prefixes
Argument structure zone Pre-stem elements
Absolutive Directional prefix Prefix referring to the temporal argument Applicatives Indirect object Agent 'Dynamic' prefix
Optative Negation
-9 -8 -7 -6 -5 -4 -3 -2
Prefixes Root Suffixes
Stem ' Endings'1
Causative Root Directional, inceptive, antipassive Prepositional operators Plural 'Dynamic' suffix Negation, illocutionary force, epistemic modality
-1 0 +1 +2 +3 +4 +5
1 The difference between suffixes proper and "endings" is not important for the purposes of this article; see [Smeets, 1984, pp. 206, 282-287].
-N
Fl3biK03HaHi/ie
3.2. Ergative alignment
Adyghe is an ergative language: both the system of cross-reference markers and nominal case affixes follow the ergative pattern (see [Kumakhov et al., 1996] for details, and [Letuchiy, 2012] for the analysis of ergativity in the Adyghe valency change patterns). The case ending of A of transitive verbs is -m (see c'alem 'boy' in (9), while both P of transitive verbs (psaser 'girl' in (9)) and S of intransitive verbs (psaser 'girl' in (10) and c'alem 'boy' in (11)) are marked with -r.
However, Adyghe cannot be regarded as a canonical ergative language: the ergative case has a broader range of functions than in many other ergative languages. The suffix -m (ergative) is used not only for transitive subjects (agents), as in (9), but also for indirect objects, certain adverbials, and adnominal possessors. In example (11), it marks an indirect object ('girl') of the intransitive verb bewan 'kiss'.1 This is why, throughout the article, we term what has been traditionally called 'ergative' with another term 'oblique' and gloss it as OBL.
(9) C'ale-m psase-r 0-3-XeEW3-K.
boy-OBL girl-ABS 3SG.ABS-3SG.A-see-PST
'The boy saw the girl.' (transitive verb)
(10) Psase-r 0-ma-kWe.
girl-ABS 3SG.ABS-DYN-go
'The girl goes.' (intransitive verb)
(11) C'ale-r psase-m 0-0-je-bews-K.
boy-ABS girl-OBL 3SG.ABS-3SG.IO-OBLIQUE-kiss-PST
'The boy (intransitive subject, absolutive) kissed the girl
(IO, oblique).'
3.3. Complementation in Adyghe
The system of complementation in West Caucasian is described by [Serdobol'skaja, Motloxov, 2009] for the Temirgoy dialect of Adyghe and [Ershova, 2012] for Besleney Kabardian, a Kabardian dialect which is in many respects intermediate between Adyghe and Kabardian. However, the issue of agreement is not thoroughly described in any of these studies. In my paper, I will use examples from the Bzhedug dialect of Adyghe. This
1 Note that the classes of transitive and intransitive verbs in Adyghe are not always isomorphic from these classes in European languages. Importantly, these two classes can mainly be distinguished on the morphological (cross-reference) basis. In (11), it is crucial that the patient (the 'indirect object') of 'kiss' is not marked in the same way as the 'direct object' of 'see' in (9), as well as verbs with high semantic transitivity such as 'kill'.
dialect is in many respects different from Standard Adyghe, which is based on the Temirgoy dialect. The study is based on data elicited from during the field trip organized by the Higher School of Economics and Russian State University for Humanities in 2014.
In Bzhedug Adyghe, the following strategies of complementation have been analyzed and are used throughout the article (they correspond ones listed by Ershova in her description, the most detailed for West Caucasian languages, though some syntactic properties can differ between the two language varieties):
Factive complementation prefix zere-.
Potential / future / masdar marker -n.
Potential / future / masdar marker -n + suffix -ew.
Potential / future / masdar marker -n + suffix -je.
Conditional marker -me.
The first strategy differs from others in that it has a rather narrow specialization: zere- is mainly used with factive verbs like sen 'know'.
I do not consider the distribution of the three other strategies. In any case, I will concentrate on number marking on emotional complement-taking predicates, such as 'be afraid', 'be good'. Some of them are factive (such as 'be surprised' or 'be good'), while others are non-factive (like 'be afraid'). The precise distribution of these strategies is beyond the scope of the present paper.
The main function of the conditional marker -me is to mark adverbial (more specifically, conditional) clauses. However, it is equally used in complement clauses, where it marks that the whole situation is unreal, as in (12):
(12) deKn3 s-j3-nebj'eKw3-xe-r qa-kwe-xe-me.
good 1SG-POSS-friend-PL-ABS DIR-go-PL-COND
'It would be good if my friends came.'
4. Agreement with complement clauses
In what follows, I describe possibilities of agreement with complement clauses. In Adyghe, the verb agrees with its A and P (if it is transitive), S (if it is intransitive) and every indirect / oblique object in the oblique case it has. Notably, base verbs cannot have an indirect object. All indirect objects, marked with the oblique (traditionally, ergative) case are introduced by a special prefix. There are many morphological devices that add an indirect object to the verb valency (I call all of them applicatives): here belong, for instance, locative prefixes, benefactive and malefactive, inadvertitive (accidental action) marker -Pec'e, and so on, as well as the non-specialized applicative, marked with je-. Even those indirect objects which are tightly
integrated into the verbal semantics must be introduced by an applicative: here belong, for instance, IO of verbs of giving like jetan 'give'. Many verbs of physical contact, such as jebewan 'kiss', which are transitive in standard average European languages, are intransitive in Adyghe. Their second argument is marked with the oblique case: thus, it is also an indirect object and is introduced by one of applicative devices (in this case, by the non-specialized applicative je-).
All the four positions triggering agreement (S, IO, P, A) are available for a complement clause. For instance, the causative verb Kes'anen 'frighten' has a complement clause in the A position (it is clear from the fact that the causative always introduces a causer in the A position). With the verb s'es'hanen 'fear, be afraid' the complement clause occupies the IO position. The verb sen 'know' requires a complement clause in the P position. Finally, predicates like deKw3 '(it is) good' have an S complement clause. Thus, examples under analysis will include complement clauses in all of these positions. Note, though, that my judgments on the syntactic position of the complement clause are based on the analogy with an NP argument of the same verb, which is not always reliable. Thus, a claim like 'The verb P has a complement clause in the position X' means roughly 'The verb P can have an NP argument in the position X, and we suppose the complement clause with the same semantic role to be situated in the same position X'.
However, it should be noted that the absolutive number agreement in the third person is sometimes 'omitted' even with NP arguments (see [Kumakhov et al., 1996]): with a plural NP in the absolutive (= S or P) position, singular absolutive agreement is found. This is why the data of oblique-marked (A and IO) arguments are more reliable in this respect and will be used throughout the article. The 3PL marker in the A and IO positions is a-, while 3SG is marked in the A position and remains unmarked in the IO position.
Like in many other languages, agreement in Adyghe is in number and person. Since complement clauses are always 3rd person by definition, the only domain where any variation can be found in the domain of complementation is number.1 Thus, the main type of examples includes two coordinated complement clauses (e.g., 'After the accident, I am afraid to run and to ride my bike'). Coordinate phrases of this sort always ttrigger plural agreement if NPs are coordinated:
1 I do not consider here the opposition between complement clauses and adjunct clauses which do not control agreement at all. In my paper, all examples seem to contain complement clauses, syntactically parallel to argument NPs, which can control agreement with the same verb heads.
(13) C'ale-m-re psase-m-re hal3Kn3-r q-a-s'ef-3K. boy-OBL-COORD girl-OBL-COORD bread-ABS DIR-3PL.ERG-buy-PST
'The boy and the girl bought some bread.'
Example (13) shows the possibility of agreement with sentential arguments:
(13a) se mas3ne-r s-f3-n-ew-j3 parasw3t3-m
I car-ABS 1 SG-lead-MSD-ADV-LNK parachute-OBL
s3-q-je-pc'ex3-n-ew-j3 s-a-s'e-s'h3ne/
1SG.ABS-DIR-DAT-jump-MSD-ADV-LNK 1SG.ABS-3PL.IO-LOC-fear/ s3-s'e-s'h3ne. 1SG.ABS-LOC-fear
'I am afraid of driving a car and jumping with a parachute.'
In the first variant of (13a), the plural cross-reference prefix a- in the indirect object slot shows the plurality of situations: 'jump with parachute' and 'drive a car'. In the second variant, the same slot is not filled (= filled by the 3SG zero marker).
It should be noted that the variant without plural agreement is the default one for Adyghe constructions with clausal complement. The absence of agreement rarely leads to ungrammatically. By contrast, the (un)accep-tability of examples with agreement seem to be regulated by two factors: (1) the verb form used in the clausal complement and (2) the syntactic position of the clausal complement. Both factors are addressed in the following sections.
4.1. Verb form in the clausal complement
The choice of the verb form in the embedded clause is relevant for agreement with CCs. For instance, in (14), the combination of the masdar marker -n with the adverbial suffix -ew makes either plural or singular agreement possible. In contrast, in (15), the native speaker judges the singular marking as impossible or awkward, because the instrumental case is chosen for CC marking.
(14) se mas3ne-r s-f3-n-ew-j3 parasw3t3-m I car-ABS 1 SG-lead-MSD-ADV-LNK parachute-OBL
s3-q-je-pc'ex3-n-ew-j3 s-a-s'e-s'h3ne/
1SG.ABS-DIR-DAT-jump-MSD-ADV-LNK 1SG.ABS-3PL.IO-LOC-fear/ s3-s'e-s'h3ne 1SG.ABS-LOC-fear
'I am afraid of driving a car and jumping with parachute.'
(15) se masane-r s-fa-n-j'-ja paraswata-m
I car-ABS 1 SG-lead-MSD-INS-LNK parachute-OBL
sa-q-je-pc'exa-n-j'-ja s-a-s'e-s'hane/
1SG.ABS-DIR-DAT-jump-MSD-INS-LNK 1SG.ABS-3PL.IO-LOC-fear/ *sa-s'e-s'hane *1SG.ABS-LOC-fear
'I am afraid of driving a car and jumping with parachute.'
Both variants of marking of the clausal complement in examples (14) and (15) are non-canonical in that the dependent verb is in a peripheral (non-argument) verb form (see the following remark on this type of cases). Canonically, since the matrix verb contains an applicative marker s'e-, introducing the stimulus of emotion, the stimulus participant must become an IO and be marked with the oblique case. This is the case in example (16), the most canonical one. Necessary to say, in these examples, plural agreement is almost always judged as acceptable (evaluation of the variant without agreement varies from one native speaker to another).
(16) se masane-r s-fa-na-m-ja paraswata-m
I car-ABS 1SG-lead-MSD-OBL-LNK parachute-OBL
sa-q-je-pc'exa-na-m-ja s-a-s'e-s'hane/
1SG.ABS-DIR-DAT-jump-MSD-OBL-LNK 1SG.ABS-3PL.IO-LOC-fear/
sa-s'e-s'hane 1SG.ABS-LOC-fear
'I am afraid of driving a car and jumping with parachute.'
A remark on peripheral marking of clausal complements
A remark should be made here on clausal complements. Serdobol'skaya and Motlokhov (2009) note that the marking of CCs in Adyghe can differ from NP marking with the same predicate (see also [Letuchiy, 2014; Schwabe, 2014] for the same phenomenon in German and other European languages).
Temirgoy dialect:
(17) sa-tje-fe-n-c'e/ sa-tje-fa-n-ew s-e-s'ane.
1SG.ABS-LOC-fall-POT-INS/ 1 SG-LOC-fall- POT-ADV 1SG.ABS-DYN-be.afraid
'I am afraid that I will fall.'
(18) tje-fe-na-m sa-s'-e-s'ane. LOC-fall-POT-OBL 1SG-LOC-DYN-be.afraid
'I am afraid of falling.' [Serdobol'skaya, Motlokhov 2009: 524]
With the verb s'anen 'be afraid', two variants are registered for Temirgoy: either the verb has an applicative marker, and the clausal complement
is marked with the oblique case, as in (18), or the verb has no applicative marker and the CC is marked by one of peripheral cases, which do not control cross-reference prefixes, as in (17). The situation is not the same with NP arguments of the same verb: if s'3nen 'be afraid' has a stimulus, expressed by an NP, this stimulus has to be introduced by the locative applicative marker s'-, as in (19). No structure with an NP argument, parallel to (17), seem to exist:1
(19) ha-m s3-s'-e-s'3ne. dog-OBL 1SG-LOC-DYN-be.afraid 'I am afraid of the dog.'
In the Bzhedug dialect, a third variant is also possible with a complement clause: the verb contains an applicative marker, yet a peripheral (adverbial) case marker -ew is chosen (see example (20)):
(20) se s3-tje-fe-n-ew s3-s'-e-s'h3ne.
I 1SG.ABS-LOC-fall-POT-ADV 1SG.ABS-LOC-DYN-be.afraid
'I am afraid of falling.'
Normally, at least in the NP domain, phrases marked by adverbial and instrumental cases do not control agreement. This gives additional support to the claim made by [Serdobol'skaya, Motlokhov, 2009] that the formal marking of clausal complements in Adyghe does not always correspond to their syntactic status. Even if a CC is marked with a non-argument marker, such as -ew, it can behave as a core argument (A, S, IO or DO) as soon as agreement is considered. However, this account does not explain what controls the agreement in examples like (20): the fact that the complement clause has some argument properties does not allow it to trigger agreement markers, which is only possible for absolutive-marked and oblique-marked argument clauses.
Examples like (20) can be given different accounts. One of possible ways is to suppose that applicativization always generates, along with the initial variant of the NP or CC, marked with a non-argument device, a second instance which is marked for the oblique case. Any of the the two instances can further be eliminated. If the peripheral instance is eliminated, this leads to a standard structure like (18) - however, elimination of the oblique-marked argument instance is also possible and leads to a non-standard structure in (20).
1 It should be said, though, that with some verbs having an NP argument, parallel structures with and without an applicative may exist (though this seems to be an exception, rather than a rule). This is why I do not cite here an example, parallel to (16), as ungrammatical.
Another variant of analysis does not presuppose elimination of a copy of CC. We can just adopt the pronominal argument hypothesis, proposed by Jelinek (1984) and Baker (1996). It claims that in polysynthetic languages, cross-reference markers, and not NPs really function as arguments of the verb. Argumental NPs (or clause-like constituents, if we discuss CCs) rather have adjunct properties - in this sense, the choice of the marking for CCs is not entirely motivated by the verb form. However, I will show that the canonical pronominal argument analysis does not account for the differences between the formal types of complement clauses in their (in)ability to control verbal agreement.
The possibility of agreement sheds some light on the possible analysis of constructions with non-canonical marking. On the one hand, the radical variant of the pronominal argument hypothesis seems to be implausible -otherwise, agreement (im)possibility would not depend on the verb form in the clausal complement, but this is the case, as examples (14)-(16) show. On the other hand, the 'copy' hypothesis could in principle be accepted: the logical form contains a canonically-marked argument, which can afterwards be eliminated, the peripherally-marked argument being retained in the surface structure. However, this point of view is not irreproachable either: again, if it is the core-marked copy, which triggers the plural agreement, it remains unexplained why the possibility of agreement depends on the verb form in the peripherally-marked copy.
It seems most plausible to adopt a view, intermediate between the radical pronominal hypothesis and the treatment of agreement, adopted for SAE languages, where autonomous NP arguments seem to be real arguments of the verb. The Adyghe data in examples (14) through (16) show that agreement markers do not correspond ideally to the argument marking, but are not autonomous from it either.
Let us consider the nature of the two devices, opposed in (15) and (14) (the markers -¿'e and -ew) in more detail. The former bears many functions in the nominal domain: it marks many types of participants, especially with instrumental and locative roles. The latter mainly functions in the verbal (predicate) domain, where it marks dependent clauses and secondary predications, e.g., depictives.
It seems that this functional difference correlates well with the fact that -¿'e-complements control agreement more readily than -ew -complements. If it is true that -¿'e is regarded by native speakers as a nominal marker, it can exist in the same functional domain as the oblique case marker -m. As I mentioned before, the structure where clausal complements are marked with -m, as in (16), is canonical for Adyghe and usually agreement is possible there. The fact that -¿'e is a nominal marker, parallel to -m, may
allow the former to adopt the functional properties of the latter, for instance, to trigger verbal agreement. This is why plural agreement is possible (and sometimes preferred) by native speakers. Note that the precise mechanism in which -j'e adopts the properties of -m is not clear. It may be that the structure with -j'e and a verb with applicative in (15) results from the fact that applicative was applied, and then the instance of the complement, marked with the canonical oblique case, was eliminated. However, this mechanism needs to be elaborated in more detail.
The marker -ew is systematically less close to nominal (argument) markers, such as -m. This is why -ew-complements have less chance to adopt the properties of the canonical argument marker, represented in (16).1
4.2. The syntactic position of the clausal complement
The second factor of (im)possibility of agreement is the syntactic position of the clausal complement. For some speakers, the variant with plural agreement is impossible if the clausal complement occupies the position of A of a transitive verb (e.g., a causer, introduced by the causative marker), but possible if the CC occupies an indirect object position (I do not consider here the P position, because the absolutive agreement is not always obligatory even with an NP argument).2
This type of opposition is illustrated by examples (21) vs. (22). Although in both contexts there are two situations ('when I drive a car' and 'when I jump with a parachute') which could potentially trigger plural agreement, actual agreement is only possible in (22), where the 3PL plural marker a- occupies the indirect object position, which is introduced by the locative applicative marker s'e-, marking the stimulus of emotion in this context. In (21), where the stimulus occupies the A position, introduced by the causative Ke-, the plural agreement is impossible.
Note that the verb form in the sentential argument is the same in (21) and (22).
(21) *mame a-Ke-s'h3ne masjane s-fa-me
mummy 3pL.A-CAUS-fear car 1 SG.ERG-lead-COND
paraswat-j'-ja sa-q-je-pc'exa-me.
parachute-INS -LNK 1SG.ABS-DIR-DAT-jump-COND
Intended: 'It frightens my mother if I drive a car and if I jump with a parachute.'
1 Note, however, that some native speakers allow two variants (with or without plural agreement) both for ew-marked and for j'-marked clausal complements. This means that the vague notion of systematic similarity between -ew vs. -j'e and the oblique marker -m, proposed above, can be interpreted differently by different native speakers.
2 Although I did not carry out a corpus-based work, it seems that this factor is relevant for a greater number of speakers than the first one.
(22) mame a-s'e-s'h3ne maijane s-fa-me
mummy 3pL.io-LOC-fear car 1sg .ERG-lead-COND
parasw3t-¿'-ja sa-q-je-pc'exa-me.
parachute-INS -LNK 1SG.ABS -DIR-DAT-jump-COND
'My mother is afraid if I drive a car and if jump with a parachute.'
This fact seems to contradict the hierarchy of syntactic roles (accessibility hierarchy), proposed by Keenan and Comrie (1977). In general, the higher the syntactic position of the argument is, the more it is possible that the argument controls agreement. For instance, agreement with S and A is more frequent across languages than agreement with IO and DO. However, in Adyghe, agreement with a clausal complement in a lower position (IO) is allowed by the speakers easier than agreement with a CC in the A position.
The possible explanation is related to the prototypical type of the argument in the A vs. IO position. The position of A is semantically related to the most active participant: this is why it is typically filled by an agentive or agentlike argument. This makes it possible for the language system to match the syntactic hierarchy, where the A position is the highest one, with the semantic hierarchy, where the Agent is the participant that is more responsible for the situation than any other participant. Non-canonical arguments, such as clausal complements, are not easily accepted in this position. Complement clauses, which virtually correspond to the A position, behave as non-canonical subjects in that they cannot control agreement.
By contrast, the position of IO is not so restrictive, because it is not expected to contain an active participant. This is why non-canonical arguments are easily accepted as IOs and can trigger plural agreement.
A question arises why clausal complements are, nevertheless, possible in the A position, though they do not control plural agreement. A possible answer is that the restriction on the filler of the A position acts on the morphological level (see [Baker, 1996], with the hypothesis that many typically syntactic restrictions work on the morphological level in polysynthetic languages). This does not mean that clauses are in general prohibited to be in the A position.
Note that the distinction between A and IO only has to do with the syntactic status of the arguments. It is not related to the formal marking of the clausal complement, since the complements in the A and IO positions bear identical case-marking: in (21) and (22) they are both marked with the conditional suffix -me. The same is true for NP arguments, which, as mentioned above, are marked with the ergative / oblique marker -m, both in the A and IO position.
The situation we observe in the domain of agreement is reminiscent of the one in the reflexivization domain. Here also, the syntactic hierarchy
(A is ranked higher than IO), is in fact semantically motivated. If a transitive A and a DO are bound by the reflexive relation, the reflexive marker is situated in the DO slot and controlled from the A slot:
(23) za-s-ja-Ke-XeK^a-K. (Temirgoy) REFL.ABS-1SG.IO-3SG.ERG-CAUS-see-PST
'He showed himself to me.' (lit. 'He made me see himself').
In contrast, under the DO = IO type of reflexivity, the reflexive marker is controlled from the DO slot:
(24) Kwanje-m-c'e s.ja-ne sa-ze-rja-Ke-pXa-z'a-K. (Temirgoy) mirror-OBL-INS 1SG.POSS-mother 1SG.ABS-REFL.IO-3SG.ERG-CAUS-look-RE-PST The mother made me look at myself in the mirror.
(Causative > Reflexive)
This means that the syntactic hierarchy, valid for the reflexive criterion, looks like A > DO > IO. At the same time, the difference between (23), where the reflexive marker is controlled from the A slot and (24), with the control from the DO slot can well be accounted in semantic terms: the reflexive marker is always controlled by the most agentive argument and occupies the slot of the least agentive one.
For agreement, as shown above, the explanation can also be semantic. If what really matters in Adyghe is cross-reference, we can claim that the A position of the causative is prototypically filled by an agent or, at least, a name of a physical object. The fact that a clausal complement itself is possible with causative verbs, is not really surprising in this case, because we suppose that free NPs and dependent clauses do not have the whole set of argument properties. By contrast, the IO position is not expected to contain an agent or a physical object. Note that it has been shown in many studies (see, for instance, [Buring, 2005, p. 16; Jackendoff, 1972]) that processes, which are virtually purely syntactic, often allow or even require a semantic (or partly semantic) explanation, for instance, in terms of thematic hierarchy [Jackendoff, 1972] or the animacy feature [Silverstein, 1976].1
The problem of the proposed analysis, based on the correspondence of semantics and syntax is that Adyghe causatives do not necessary have an agentive subject. However, in examples like (25), the A of transitive verbs is coded by an NP in the oblique case, not a clausal complement.
1 I thank an anonymous reviewer for reminding me an analogous case, discussed in Bresnan and Mchombo's (1987) article on Chichewa. However, it seems that the analogy is not ideal here. While in Chichewa, different argument positions use different techniques of combining morphological and syntactic argument expression, in the case I discuss for Adyghe, the difference between A and IO is valid for agreement with complement clauses, while in the constructions with NP arguments, no asymmetry between the same argument positions is found.
(25) ruslane ja-tetez zaKe-m ja-Ke-Xa-K. (Temirgoy) Ruslan 3SG.POSS-grandfather Old.age-OBL 3SG.ERG-CAUS-die-PST 'Ruslan's grandfather died of old age.' (lit. 'Old age killed Ruslan's grandfather').
I propose that the semantic restrictions on the A position only become relevant when the structure contains a non-canonical (e.g., clausal) A-argument. If the position is occupied by an NP, as in (25), the syntactic validity of the structure prevails over the semantic restrictions.
5. Long distance agreement
Some native speakers which do not allow clausal complements to control agreement use a structure like in (26) and (28) instead. Here the matrix predicate agrees with the A of the embedded clause.
(26) deKwa(-x) c'ale-me pc'ashe-mashe-xe-r good(-PL) boy-OBL.PL fruit-PL-ABS
bew zer-a-ska-re-r.
many REL.FCT-3SG.ERG-eat-DYN-ABS
'It is good that our boys eat much fruit.'
(27) deja-(*x) ma c'ale-r zere-sna-Ke-r bad-PL this boy-ABS REL.FCT-get.Wet-PST acja za-zer-ja-wapcepa-z'a-Ke-r.
and REFL.ABS-REL.FCT-3SG.ERG-get.dirty-RE-PST-ABS
'It's bad that the boy got wet and dirty.'
(28) deja-x ma c'ale-xe-m za-zer-a-wapcepa-z'a-Ke-r
bad-PL this boy-PL-OBL REFL.ABS-REL.FCT-3PL.ERG-get.dirty-RE-PST-ABS
acja za-zer-a-Ke-sna-z'a-Ke-r.
and REFL.ABS-REL.FCT-3PL.ERG-get.wet-RE-PST-ABS
'It's bad that the boys got wet and dirty.'
In (26), it is evident that the optional plural marking on the adjective deKwa 'good' is not motivated by the plurality of clausal complements (the sentence contains only one CC 'they eat much fruit'). It seems that the plural marker -x denotes the plurality of the agent in the embedded clause ('boys'). In other words, the structure is like (29):
(29) 'They are good that the boys eat much fruit'.
In (28), the number of the matrix predicate might potentially agree with the two clausal complements ('that they got wet' and 'that they got dirty'). However, the native speaker who provided this data does not allow agreement with clausal complements in this context, as shown by the ungrammati-
cality of (27). Thus, the number on the matrix predicate - the adjective deja(x) 'it's bad' - is triggered by the agent ('boys') of the embedded clause.
Not surprisingly, in constructions of this sort the embedded predicate ('eat' in (26) and 'get dirty and get wet' in (28)) also carries plural agreement, which it gets from its own argument ('boys').
It seems plausible to analyze contexts of this sort as long-distance agreement. If this analysis is correct, then we should consider that the agent of the embedded clause controls agreement in the main clause.
The main syntactic position where the LDA is observed is the subject position of adjectives having a clausal subject ('It is good that'). However, sometimes the same type of phenomena is observed with 'verbs proper'.
(30) Sweldata-r a-s'a-s'hana-K a-wa?e-n-j'e.
soldier-ABS 3PL.IO-LOC-be.afraid-PST 3PL.ERG-wound-MSD-INS 'The soldier was afraid that he would be wounded (lit. 'was afraid of them that they would wound him)'.
While in (26) and (28), we apparently observe a long distance agreement, this is not so obvious in (30). In this example, it can be interpreted as a structure like 'The soldier was afraid of them that they wound him', e.g., the verb has a nominal argument 'they', while the clausal complement is added afterwards to denote the ground of the fear. It seems that the same analysis is more problematic in examples like (26) and (28): the reading, supposed for (30), would look like 'Boys are good that they eat much fruit' for (26). Under this reading, the boys themselves would be evaluated as good, which is not the case, according to the native speaker's judgment.
[Serdobol'skaya, Letuchiy, 2005] show that 'mixed' structures of this sort, where a clausal complement is added to a full structure with an NP argument (e.g., 'I got angry with Peter that he was late') are not rare in Russian -a language where no real raising or long distance agreement is found. For instance, this kind of pattern may contain an afterthought: the speaker chooses the nominal pattern ('I am afraid of them'), and then adds a clausal complement which makes the ground of the emotion clearer. However, the afterthought analysis needs to be confirmed by empirical tests which show that the speaker re-analyzes the syntactic structure in the process of the speech. Currently, no empirical data of this sort are available for Adyghe.
6. Conclusions
In this paper I have considered an unusual phenomenon observed in Adyghe: agreement with clausal complements. The existence of this sort of agreement proves that Adyghe CCs have a subset of nominal properties. In many other languages clausal complements never control agreement.
This fact agrees well with another feature of Adyghe. It has been shown (see, for instance, [Lander, 2012]) that Adyghe does not show a strict distinction between nouns and verbs as parts of speech. It is more plausible to speak of argument positions and the predicate position, and both the former and the latter can be occupied by words of different parts of speech (as they are traditionally defined). Not surprisingly, the distinction between nominal arguments and sentential ones is also not as strict as, for instance, in Standard Average European languages.
At the same time, we cannot say that the two types of arguments - nominal and sentential ones - behave similarly. While agreement with argument NPs is always possible, though sometimes optional, the same is not true for clausal complements. We have shown two parameters to be relevant for the (im) possibility of agreement: (1) formal type of the clausal complement and (2) syntactic position of the clausal complement.
Concerning parameter (1), it is surprising that even clausal complements with peripheral (non-argument) marking can trigger agreement. However, inside the system of peripheral markers, there is a difference in the (in)ability to trigger agreement. As noted before, CCs marked with the instrumental -je marker are more plausible candidates for plural agreement than those marked with the secondary predication marker -ew. I proposed that this is due to the fact that the instrumental marker - is regarded by native speakers as a marker of the nominal paradigm, parallel to the argument marker -m of the oblique case. This is why complements, marked with the instrumental suffix, easier adopt some features of argument NPs than complements marked with the converb marker -ew, which is a marker of dependent clauses which has not much to do with argument marking.
The very fact that the presence / absence of 'nuclear' (oblique or absolutive) marking of the clausal complement is not always crucial to the possibility of agreement is very interesting This supports the hypothesis proposed in [Serdobol'skaya, Motlokhov, 2009] that the formal marking of the verb in a clausal complement does not fully define the syntactic position of the complement.
Parameter (2) says that that the verb agrees with clausal complements in the position of A of transitive verbs (e.g., the causer of causative verbs) less readily than with CCs in the IO position. I claimed that this fact, seemingly conflicting with the hierarchy of syntactic positions, results from restrictions on the filling of the A position, which are not valid for the lower IO position.
The phenomenon of agreement with complement clauses is typologically rare. The fact that it is observed in Adyghe is illustrative in two respects. First, it shows once more that the distinction between nouns and verbs (NPs and VPs) is not as sharp in Adyghe as, for instance, in SAE languages). Second,
it shows that this distinction, nevertheless, exists, which is evident from the fact that agreement with complement clauses is much more restricted in Adyghe than agreement with NP arguments.
References
Baker, 1996 - Baker M. The polysynthesis parameter. Oxford, 1996.
Bresnan, Mchombo, 1987 - Bresnan J., Mchombo S. Topic, Pronoun, and Agreement in Chichewa. Language. 1987. Vol. 63. № 4. Pp. 741-782.
Buring, 2005 - Buring D. Binding theory. Cambridge, 2005.
Davies, Dubinsky, 2009 - Davies W.D., Dubinsky S. On the Existence (and Distribution) of Sentential Subjects. University of South Carolina. Faculty publications, 2009.
Delahunty, 1983 - Delahunty G. But sentential subjects do exist. Linguistic Analysis. 1983. Vol. 12. Pp. 379-398.
Ershova, 2012 - Ершова К.А. Конструкции с сентенциальными актантами в бесленеевском диалекте кабардино-черкесского языка: Дипломная работа. РГГУ, 2012. [Ershova K.A. Konstrukcii s sentencial'nymi aktantami v besleneevskom dialekte kabardino-cerkesskogo jazyka [Constructions with complement clauses in Besleney Kabardian]. BA thesis. Moscow, 2012.]
Fortescue, 1994 - Fortescue M. Polysynthetic morphology. Encyclopedia of Language and Linguistics. R.E. Asher et al. (eds). Oxford, 1994. Рр. 2601-2602.
Fortescue, 2007 - Fortescue M. The typological position and theoretical status of polysynthesis. Linguistic Typology. J. Rijkhoff (ed.). Erhus, 2007. Рр. 1-27.
Jackendoff, 1972 - Jackendoff R. Semantics in Generative Grammar. Cambridge, Mass., 1972.
Knyazev, 2014 - Knyazev M.Yu. Structural licensing of sentential complements: Evidence from Russian noun-complement constructions. Bucharest working papers in linguistics. Vol. XVI. № 2. Cornilescu A., Avram L. (eds.). Bucharest, 2014.
Kumakhov et al., 1996 - Kumakhov M., Vamling K., Kumakhova Z. Ergative case in the Circassian languages. Working Papers 45, Lund University, Dept. of Linguistics. 1996. Рр. 93-111.
Lander, 2012 - Ландер Ю.А. Релятивизация в полисинтетическом языке: адыгейские относительные конструкции в типологической перспективе: Дис. ... канд. филол. наук. М., 2012. [Lander Yu.A. Reljativizacija v polisinte-ticeskom jazyke: adygejskie otnositel'nye konstrukcii v tipoloiceskoi perspective [Relativization in a polysynthetic language: Adyghe relative constructions in the typological perspective]. PhD dis. Moscow, 2012.]
Letuchij, 2012 - Летучий А.Б. О некоторых свойствах сентенциальных актантов в русском языке // Вопросы языкознания. 2012. № 5. С. 57-87. [Letuchij A.B. On some properties of Russian complement clauses. Voprosy jazykoznanija. 2012. Vol. 5. Pp. 57-87.]
Letuchiy, 2012 - Letuchiy A.B. Ergativity in the Adyghe system of valency-changing operations. Ergativity and voice. J. Authier, K. Haude (eds.). Berlin, 2012. Рр. 323-353.
Letuchiy, 2014 - Letuchiy A.B. Esli-clauses: arguments or adjuncts? Factive or non-factive? Workshop on (Non)factivity in complementation. DGFS conference, Marburg. 2014.
McCloskey, 1991 - McCloskey J. There, it, and agreement. Linguistic Inquiry. 1991. Vol. 22. Pp. 563-567.
Noonan, 2007 - Noonan M. Complementation. Language Typology and Syntactic Description 2: Complex Constructions, 2nd ed. Shopen T. (ed.). Cambridge, 2007. Pp. 52-150.
Quer, 2008 - Quer J. Argument clauses and nominal features. Lingue e linguaggio.
2008. Vol. VII. Issue 1. Pp. 93-110.
Schwabe, 2014 - Schwabe K. Conditional clauses with argument features. Workshop on (Non)factivity in complementation. DGFS conference, Marburg. 2014.
Serdobol'skaya, Letuchiy, 2005 - Serdobol'skaya N.V., A.B. Letuchiy. Object or situation: the factors of expressing the arguments of mental verbs in Russian. East West Encounter: Second international conference on Meaning - Text theory. Ju.D. Apresjan, L.L. Iomdin (eds.). Moscow, 2005.
Serdobol'skaya, Motlokhov, 2009 - Сердобольская Н.В., Мотлохов А.В. Семантика конструкций с сентенциальными актантами в адыгейском языке // Аспекты полисинтетизма / Под ред. Я.Г. Тестельца и др. М., 2009. С. 498-543. [Serdobol'skaya N.V., Motlokhov A.V. Semantics of constructions with complement clauses in Adyghe. Aspekty polisintetizma. Testelets Ya.G. et al. (eds.). Moscow,
2009. Pp. 498-543].
Silverstein, 1976 - Silverstein M. Hierarchy of features and ergativity. Grammatical Categories in Australian Languages. R.M.W. Dixon (ed.). Canberra, 1976. Рр. 112-171.
Статья поступила в редакцию 10.12.2016 The article was received on 10.12.2016
Летучий Александр Борисович - кандидат филологических наук; доцент Школы лингвистики, Национальный исследовательский университет «Высшая школа экономики»
Letuchiy Alexander В. - PhD in Philology; Assistant Professor of School of Linguistics, National Research University «Higher School of Economics»
E-mail: alexander.letuchiy@gmail.com
