Noun-modifying clause constructions in Ainu [PDF]

Citation preview

Noun-Modifying Clause Constructions in Ainu Anna Bugaeva, National Institute for Japanese Language and Linguistics (NINJAL) 1. Introduction There are languages like Japanese which, unlike English and many other European languages, make no syntactic distinction between relative clauses and noun-complement constructions (Matsumoto 1997). It has been claimed that such languages have a single general noun-modifying clause construction (‘GNMCC’) simply attaching a modifying clause to a head noun with no explicit expression of the relation between the two, forming a noun phrase which lacks extraction (and thus, necessarily, constraints on extraction) (Comrie 1998: 76). According to Comrie (1996: 1079), the ‘Japanese-type’, or the ‘general noun-modifying clause construction type’ is widespread in Asia (hence also referred to as the ‘Asian type’) encompassing “at least Ainu, Japanese, and Korean; Chinese, and probably some (but not all) other Sino-Tibetan languages; the Dravidian languages; some Turkic languages (but definitely not all – Turkish itself is excluded...), Thai and Khmer…” (Comrie 1996: 1077). Indeed, Ainu shares with Japanese a good deal of properties which are believed to correlate with the GNMCC. Just like Japanese, Ainu is a language with low referential density, the modifying clause is in a form that can be used independently without having the head noun and an extended array of semantic and grammatical relations can be represented by the head noun. However, unlike Japanese, Ainu is extremely strict with the argument structure.1 Moreover, Ainu makes little use of nominalization, see §7, which normally helps, at least partially, to reduce argument structure. In this paper, based on the data from Southern Hokkaido Ainu, the best documented group of dialects which used to be spoken in the proximity to mainland Japanese, I will describe a range of possible grammatical and semantic relations between the modifying clause and noun. I will argue that the interpretation of noun-modifying constructions in Ainu (here: Southern Hokkaido) depends largely on grammatical relations, much less on pragmatics; thus Ainu lacks the prototypical GNMCC. Ainu formally distinguishes between relative clauses (1) and noun-complement clause constructions (2), (3). The latter include sentential complements of a noun (2) and fact-S (or “the story that…”) noun-complements (3) both requiring a possessive suffix on the head noun, just like Turkish (Comrie 1998). Consider the following examples from Comrie (1998: 54-5) with my reinterpretation of morphemic boundaries and glosses in (2), cf. the nominal possessive construction in (14). (1) taan aynu koyki yuk that man hunt deer ‘the deer that that man hunts’ (2) a. kamuy ek bear come 1



hum-i perception



The distinction between arguments and adjuncts is clear cut because only the latter are marked with case postpositions. It is not possible to add or introduce arguments without additional verbal morphology, which is the distinctive property of many polysynthetic languages including Ainu, cf. the Morphological Visibility Condition in Baker (1996: 17). This property also largely facilitates zero anaphora in Ainu: topical arguments can easily be omitted because they are indexed on the verb.



bear come sound-POSS ‘the perception that the bear came’ b. kamuy apkas ru-w-e bear walk track bear walk track-EP-POSS ‘the track of the bear’s walking (i.e. that the bear left while walking)’ (3) eraratki patek e=ki asur-uhu a=nu kor an=an fornication only 2SG.A=do rumor-POSS IND.A=hear and exist.SG=IND.S ‘I constantly heard rumors that you have been only sleeping around.’ (NV137) (‘NV’ indicates the source from which the example is cited. The same applies to other examples, see the list of Sources in the end.) Towards the end of the paper in §7 and §8, I will discuss a number of peripheral (in terms of frequency and overall weight) noun-modifying strategies in Ainu which exhibit more features of the prototypical GNMCC. 2. Basic properties of Ainu Genetically, Ainu is considered an isolate. It used to be spoken in Hokkaido (the northern island of Japan), Sakhalin and Kurile Islands. It is now on the verge of extinction. Ainu shows considerable dialectal variation. Its dialects are primarily divided into the Sakhalin, Hokkaido and Kurile groups. The Hokkaido dialects can be roughly divided into Northeastern (Northern, Eastern, and Central) and Southwestern (Southern and Southwestern) groups, which are further subdivided into local sub-dialectal forms (for details see Hattori 1964: 18). A couple of native speakers remain for two of the southern dialects: Saru and Chitose. The data used in the present paper are mainly taken from the earlier documentation of these dialects. Ainu has no written tradition, and all the data is cited from the oral language. Ainu is agglutinating, polysynthetic and incorporating. It employs more prefixation than suffixation, which is unusual for this area of the world. It is predominantly head-marking. It has postpositions, but no prepositions. The basic constituent order is SV/AOV. Arguments in Ainu (either nouns or pronouns) are not marked for case. Grammatical relations are distinguished by (i) the relative position of A and O in the clause, and also (ii) obligatory verbal cross-referencing, which employs mainly prefixes and proclitics, with the exception of two enclitics. Adjuncts are marked by postpositions. All modifiers are prepositive. There is no separate word class of adjectives. The concepts that are expressed by adjectives in other languages are expressed by intransitive verbs in Ainu. Ainu has mixed alignment: in verbal cross-referencing, the first person singular has nominativeaccusative alignment, the second and third persons neutral alignment, and the first person plural exclusive and indefinite tripartite alignment, see Table 1. The indefinite form (‘IND’) has four functions: (i) the indefinite person proper (=the impersonal), (ii) the first person plural inclusive, (iii) the second person singular/plural honorific, and (iv) logophoric (person of the protagonist). The logophoric use is common in folktales because they have the structure of reported discourse with the whole story being a quote.2 For convenience, the indefinite form with the logophoric function is 2



Most examples in this paper are from folktales, hence a high frequency in the use of the indefinite person.



translated as ‘I’, although it is glossed as ‘IND’, see (10). All pronouns show neutral alignment: A=S=O. Pronominal arguments are usually omitted (pro-dropping) since there is obligatory crossreferencing for 1st and 2nd persons on the predicate. As Table 1 shows, the 3rd person crossreferencing markers are phonologically zero; they are not shown in examples. There is a tendency to omit topical argument NPs (zero anaphora). Table 1. Person-number marking in the Saru dialect of Ainu (Southwestern group) A/S/O pronouns káni3 cóka eani ecioká sinuma oka aoká



person-number 1SG 1PL.EXC 2SG 2PL 3SG 3PL IND



A markers ku= ci= e= eci= Ø= Ø= a=



S markers ku= =as e= eci= Ø= Ø= =an



O markers en= un= e= eci= Ø= Ø= i=



The opposition of transitive and intransitive verbs is clear-cut, and so is the opposition of transitive verbs and nouns. However, all intransitive verbs can, in principal, function as nouns without any change in their stem morphology, e.g. uwepeker i. ‘to tell a folktale’, ii. ‘a folktale’. In the verbal morphology, there is no distinction between finite vs. non-finite forms, i.e. Ainu lacks any kind of special subordinate morphology. A number of aspectual, modal, and evidential meanings may be expressed with particles and auxiliaries; there are no pure tense markers. Ainu has an extensive voice system, viz. applicatives e-, ko-, o- and causatives -e/-te/-re, -ka, -ke for increasing valency and reflexives yay-, reciprocals u-, antipassives i-, and anticausatives si-, -ke for decreasing valency, and also noun incorporation (of O, So and non-agentive A). Some verbs employ different stems for singular and plural. In the case of intransitive verbs, plurality refers to the number of S referents while, in the case of transitive verbs, plurality refers to the number of P or patientive A referents. 2.1. Basic clause types in Ainu The basic distinction is between intransitive and (bivalent) transitive clauses: intransitive predicates are indexed for S (4), while transitive predicates for A and O (5); there is no case-marking for S/A/O. (4) [cóka]S aynu or un inkar=as 1PL.EXC human place ALL look=1PL.EXC.S ‘We (I and he/she/they) looked to the place of humans.’ (OI) (5) a. [cóka]A 1PL.EXC 3



[aynu human



kotan]O village



ci=nukar 1PL.EXC.A=see



rusuy DESID



In most cases, the accent is assigned automatically: it falls on the first syllable if the first syllable is closed, as in kír.pu (CVC.CV) ‘fat’, and on the second one if the first syllable is open, viz. sa.pá.ha (CV.CV.CV) ‘his/her head’. However, there are some exceptions; in this paper, only the irregular accent will be indicated.



‘We (I and he/she/they) wanted to see an Ainu village.’ (OI) b. [unarpe]A [(cóka)]O un=nukar aunt 1PL.EXC 1PL.EXC.O=see ‘(His/her) aunt saw us (me and him/her).’ (constructed example) Theoretically, on transitive predicates, A cross-referencing markers precede O markers, but in practice one can find few examples where both are overtly expressed, as in (6a), because there is no overt marking for the third person, as in (5). Moreover, in no variety of Ainu, is the combination of first person singular/plural A and second person singular/plural O ever marked as *ku=e= (1SG.A=2SG.O=), *ku=eci= (1SG.A=2PL.O=), *ci=e= (1PL.EXC.A=2SG.O=), and *ci=eci= (1PL.EXC.A=2PL.O=). For instance, in Saru (6b), we find instead the prefix eci= (1SG/PL.A+2SG/PL.O=), which has the key function of marking 2PL for A, S, or O. (6) a. eci=en=hotuyekar yak pirka p 2PL.A=1SG.O=call if be.good but ‘You (PL) may have called out to me.’ (T1 36) b. eci=ekanok kusu ne 1SG.A+2SG.O=go.to.meet intention COP ‘I’ll come to meet you.’ (AB 124)



na FIN



Three-place transitive verbs employ the same A and O series of cross-referencing affixes as twoplace transitive verbs (5), (6) but are differentiated from the latter by the number of unmarked objects, which are, more or less, similar in terms of their encoding and behavioral properties, see §4.1.5. (7)



Anna ninup káni en=kor-e. (name) sewing 1SG 1SG.O=have-CAUS ‘Anna gave me some sewing.’ (OI)



A special clause type is represented by the copula verb ne ‘be, become’ which is a kind of transitive verb. The copula is cross-referenced only for the transitive subject A, see ci= in (8a), cf. A column in Table 1, but not for the S or the O (= copula complement) (8b). (8) a. tu-n ci=ne wa arki=as. two-people. CLF 1PL.EXC.A=COP and come.PL=1PL.EXC.S ‘There are two of us and we came.’ lit. ‘we are two…’ (T1 70) b. cóka ne. 1PL.EXC COP ‘[It] is us.’ (T1 23) As mentioned, the use of case postpositions is restricted to adjuncts: locative ta, allative un (directed Motion), dative eun, ablative wa, instrumental ani, comitative tura-(no), degree/allative pak-(no) ‘as far as, till/until’, traversal peka (‘over’), and mutative ne (‘as’).



Locative, allative, and ablative postpositions cannot generally attach to common nouns directly: a special relational noun4 - neutral or(-o) ‘place (of)’ (9) or specific sam(-a) ‘near (to)’ (10) etc. - is added between noun and postposition. 5 Only a few nouns (e.g. kotan ‘village’ (102a), kim ‘mountains’, pis ‘seashore’ (11), rep ‘open sea’, ya ‘shore’, put ‘mouth of a river’) and geographical names do not require a relational noun.6 (9)



somo cise or un hosip-pa=an pe ne na. NEG house place ALL return-PL=IND.S NMZ COP FIN ‘(If you are out playing and hear unfamiliar voices,) do not come home.’ (K7908051UP.023)



(10) hopuni=an h_ine7 ape sam ta ek=an h_ine mono stand.up.SG=IND.S and fire near LOC come.SG=IND.S and silently a=an. sit.SG=IND.S ‘I stood up, came up to the fire and sat down.’ (K8106233UP.053) (11) pis ta san ruwe ne h_ine, seashore ALL go.down.SG INF.EV COP and ‘He went down to the seashore and...’ (K7708241UP.100) There is a group of case postpositions which clearly originate in verbs and are still in an early stage of grammaticalization as they can also occur without nouns, e.g. eun ‘to/toward (sb)’, ani ‘by/with’, pak-(no) ‘as far as, till/until’, and tura-(no) ‘together with’. In (12), aeyayramekotep ‘utensils’ before ani ‘with’ (< ani ‘hold sth’) is anaphorically omitted as if ani were still a verb and aeyayramekotep ‘utensils’ its direct object (lit. ‘holding (utensils)’). However, if ani were really a verb here, we would expect its plural form anpa instead and a coordinating conjunction (cf. h_ine ‘and’ in (10)), but none are present, meaning that synchronically it has intermediate status between verb and postposition. (12) nep ka aeyayramekotep oka yak-ne ani a=supa something even utensil exist.PL if-COP INST IND.A=cook.PL ‘If we had the utensils we would boil our catch with them.’ (K7803231UP.012) Emphasizing the ability of these postpositions for independent use, Tamura (2000: 96) regards them as postpositional adverbs distinct from the above-mentioned locative/allative/ablative postpositional particles. She also mentions that some postpositional adverbs can take object marking, e.g. pak-(no) ‘as far as, until’ and tura-(no) ‘together with’, some would be expected to take it, but this does not 4



A class of words non-existent in European languages and translatable into English by adverbs. The choice of the respective citation/possessive forms of relational nouns (e.g. or ‘place’ or or-o ‘the place of’) is to some extent determined by the animacy properties of their preceding nouns: animates require the possessive forms and inanimates most commonly co-occur with the citation forms (less commonly with the possessive forms). 6 In Tamura (1984: 39-40), nouns which require relational nouns in locative/spatial phrases are described as inherently ‘thing’ nouns and nouns which do not are described as inherently ‘place’ nouns. 7 The symbol “_” indicates the occurrence of morphophonological alternations (sandhi), for the details of alternations see Bugaeva (2012: 468). 5



show up because other conditions are not met: they are unlikely to co-occur with human objects, e.g. ani ‘by/with’, and some cannot, e.g. eun ‘to/toward (sb)’. Moreover, a few postpositions allow even the form with a cliticized coordinating conjunction no ‘and’. However, none of them allow the use of the corresponding plural forms which are required with plural ‘objects’ in the case of transitive verbs. As shown, postpositional adverbs exhibit varying degrees of grammaticalization from verbs, yet none of them can be regarded as verbs synchronically. 2.2. Noun phrase structure A noun can be the head of an NP with preposed modifiers: a determiner (numeral determiner or nominal/anaphoric demonstrative), an adverb, a noun, a possessor NP, a relative clause or a nouncomplement clause. Unlike Japanese, Ainu has distinct nominal attributive and possessive constructions. In the attributive construction, dependent and head nouns are merely juxtaposed. (13) a.



sísam uwepeker Japanese old.tale ‘an old Japanese folk story’



b.



kamuy rus bear fur ‘bear skin’ (T1 187)



In the possesssive construction, the possessee takes the so-called possessive form with possessive suffixes -hV or -V(hV)8 which indicates the bound status of the form in question and is marked with one of the A prefixes for the person and number of the possessor (3-rd person is zero), so pronominal (and even nominal) possessor NPs are commonly omitted (14a-c). This is parallel to the zero anaphora phenomenon in verb phrases. (14) a.



c.



ku=sapa-ha 1SG.A=head-POSS ‘my head’ e=haw-e(he) 2.SG.A=voice-POSS ‘your voice’



b.



d.



ci=setur-u(hu) 1PL.EXC.A=back-POSS ‘our backs’ kamuy rus-i(hi) bear fur-POSS ‘the bear skin’



Unlike Sakhalin Ainu, not all nouns in Southern Hokkaido have the respective possessive forms. Those that lack them employ citation forms to express possessive relations, cf. (45) and (46). The possessive construction is used in Southern Hokkaido mainly for inalienable possession, including relations with head nouns referring to body parts/excretion, strength, feelings, some kinterms and also part-whole relations. Tamura (1988: 27-28) suggests that it is necessary to use a possessive form when the possessor is definite, thus (14d) could have been “the skin of the bear which I killed yesterday” while (13b) is just “a bear skin”. Her approach was criticized in Satō (1997: 150-53), who provided an example of the possessive form occurring with an indefinite possessor and attempted to explain it in terms of topicalization of the possessor; but Satō’s proposal is controversial. Indeed, a lot is still unclear about the use of possessive forms of nouns. I think that 8



Possessive suffixes -hV or -V(hV) often copy a root-final vowel with epenthetic /h/ being inserted, i.e. -ha/hu/-ho/-he/-hi for vowel-final (14a) and -a(ha)/-u(hu)/-o(ho)/-e(he)/-i(hi) for consonant-final roots. However, roots ending with w/y always trigger -ehe (14c) and a few other irregular nouns take a default suffix -i(hi) (14d) originating in the bound noun hi ‘place/time/thing’, cf. § 5.1. The difference between “short” (e.g. ci=setur-u) and “long” (ci=setur-uhu) possessive forms is unclear.



Tamura’s explanation is basically correct because definite possessors are most likely to trigger the possessive construction; however, the crucial point is not whether the possessor is definite, but whether contextually the relation is viewed as possessive (14d) or just attributive (13b). Alienable possession in Southern Hokkaido is encoded by the relative clause-based periphrastic construction with kor ‘have sth/sb’ as the predicate and the possessor as the subject; the possessee is left unmarked. (15) [ku=kor] seta 1SG.A=have dog ‘my dog’ (lit. ‘the dog (that) I have’) (OI) 3. Noun-modifying clause constructions in Ainu: relative clauses vs. noun-complement clauses As mentioned, the relative clause construction in Ainu, though similar to the noun-complement construction, is not identical. The relative clause construction (16) is built on the model of the nominal attributive construction, cf. (13), while the nominal complement construction (17) is built on that of the nominal possessive construction, cf. (14). In the latter (17), a possessive suffix on the head noun cross-references the whole clause, which relates to the head noun as its complement viz. Ø=haw-e (3SG.A=voice-POSS) ‘the voice of him/her/it (=the whole clause)’. (16) [a=roski a] inaw opitta… hácir wa okay IND.S=stand.PL PERF.SG inaw.prayer.sticks all fall.down and exist.PL ‘All the inaw willow prayer sticks which I had erected (on the previous day at the hunting cabin) fell down.’ (AB 187-8) (17) [kamuy-utar nuwap kor okay] haw-e god-PL groan and exist.PL voice-POSS ‘I heard the voices of (fish) gods’ groaning.’ lit. ‘I heard the voice (that) gods groaned.’ (AB 339)



a=nu IND.A=hear



Importantly, in both cases we are dealing with a genuine clause without any signs of nominalization either on the predicate or elsewhere; note the use of perfect (16), analytical progressive (17), admirative (18a,b), and evidential markers (18b). (18) a. [nuynak wa an aan] menoko-po hide and exist.SG ADM woman-DIM ‘A young woman who appeared to be hiding in the mountains…’ (TS2 12) b. [i=pa siri ne aan] haw-e a=nu IND.O=find VIS.EV COP ADM voice-POSS IND.A=hear ‘(It was hunger in my village.) I heard the story (that people who were starving kept walking and finally) found me.’ (TS2 74) lit. ‘... I heard the story (that) it appeared (that) it was the sight (that) they found me.’ Relative clauses may also be marked for modality with dubitative (19) and intentional (20) markers, which I could not attest in noun-complement clauses.



(19) [ek nankor] kur come.SG probably man ‘a man who will probably come’ (=one’s fiancée) (T3 64) (20) [e=kor kusu ne] kur or ta e=arpa 2SG.A=have intention COP man place LOC 2SG.S=go.SG ‘If you go to the place of the man you are going to marry…’ (KI 206)



wa, and



There are few general restrictions on what can occur in the modifying clause of relative and noun-complement clauses. Naturally, they cannot include sentence-final particles na, wa, and so expressing illocutionary force (cf. (6b) and (9)), just like Japanese which, in similar syntactic contexts, does not allow sentence-final particles yo, zo, na etc. The use of imperative mood (plain or polite with particle yan) and hortative construction (marked with particle ro) is also ruled out. Syntactically, unlike main clauses with zero anaphora, relative clauses do not allow for the use of a “gapped” NP, but this fact alone is not sufficient to postulate extraction which is usually taken as a key property distinguishing relative clauses from noun-complement clauses. The question requires further consideration. In fact, relative and noun-complement clauses are distinct not only in terms of the presence/absence of possessive marking, but also in terms of their respective head nouns (Okuda 2001 (1989) 248). Relative clauses may involve all kinds of common nouns and a few bound nouns (see §4), but noun complement clauses allow only a few nouns with very specific semantics (see §5), half of them are even considerably grammaticalized (see §6). The point that head nouns of noun complement clauses in Japanese are more semantically specific than common nouns has been made in Teramura (1969: 68-73) and Matsumoto (1988: 173).



4. Relative clauses With the exception of alienable possessor and standard of comparison proper, all the positions on Keenan and Comrie’s (1977) accessibility hierarchy can be relativized on. The formation of all relative clauses involves the NP gap strategy or, to be precise, the zero anaphora strategy, as it is labeled in Givón (2012: 21). However, what we have in Ainu only superficially resembles Japanese. Unlike Japanese, Ainu involves the gapping of the relativized NP only while all its “accompanying” material from the main clause (if present there) is left intact in the relative clause, which helps to track the relation between the head noun and the clause. The clause constituent of the relative clause construction is indeed always the same as the corresponding main clause with zero anaphora which is possible in both verb and noun (possessive) phrases (cf. §2.1. and §2.2). Based on the presence/absence and nature of the gapped NP tracking material in the relative clause which is retained from the hypothetical main clause, it is possible to distinguish the following cases: [1] Relativization of arguments9(§4.1): the absence of retention10. Correlates with the lack of case marking for A/S/O and zero cross-referencing on the verb for the third person in the main clause, cf. (4)-(7); [2] Relativization of non-arguments (§4.2): the presence of retention. 9



Also including the second object9 (typically, a Recipient) of three-argument verbs. Recall that in Ainu, in the case three-argument verbs, T and R show neutral alignment (7), i.e. there is no such thing as “indirect object”. 10 In principle, it is possible to regard transitivizing affixes (applicatives/causatives) on the verb as the gapped NP tracking material.



Correlates with the lack of cross-referencing for non-arguments on the verb in the main clause. (a) Relativization of instrumental/comitative/dative adjuncts (§4.2.1.1) and the standard of comparison equivalent (=degree adjunct) (§4.2.1.2): a postposition retention strategy. Correlates with the ability of the respective postpositions to occur independently in the main clause, cf. (12); (b) Relativization of possessor (§4.2.2): a possessive suffix (on the possessee) retention strategy. Correlates with the head-marking structure of the possessive construction in the main clause, cf. (14); (c) Relativization of locative/allative adjuncts (§4.2.3): the relational noun or-o ‘the place-POSS’ and a postposition retention strategy. Correlates with the inability of the respective postpositions to occur with most nouns without the relational noun or-(o) ‘the place(-POSS)’11 in the main clause, cf. (9). Overall, I suggest that the structure of relative clauses in Ainu may be best viewed as a manifestation of the head-marking polysynthetic character of this language which generally presupposes the traceability of grammatical relations from the head (verbal or nominal). As mentioned, relative clauses may involve as their heads not only common nouns, but also a few so-called bound nouns, and the latter are, in fact, much more common as relative clause heads. Bound nouns are peculiar in that they cannot occur on their own and must be modified with a determiner, noun or relative clause, e.g. tan kur ‘this man/person’, but not *kur ‘a man/person’. Their semantics is associated with a generic category: kur ‘person’, utar ‘people’, us(i)-ke-(he) ‘place-POSS-(POSS))’, pe/p ‘thing/person (pejorative)’, hike ‘the one/side’ (similar to Japanese hō), and hi/i ‘place/time/thing’. All of them are in the process of grammaticalization into suffixes; the latter three have already developed additional usages as derivational suffixes in lexical nominalizations (see §6.1) and as subordinating conjunctions (6a). A similar class of bound nouns is attested in Hup and a few other northwest Amazonian languages where it is also extensively used in the formation of relative clauses (Epps 2012: 197). 4.1. Relativization of arguments In terms of frequency, O-relativization and S-relativization are much more common than Arelativization in Ainu. 4.1.1. Subject of intransitive verb (S) There is no syntactic distinction between relativizing agentive (21) and patientive (22) intransitive subjects; the latter also include the subjects of stative verbs (“adjectives”) (22b). Note the use of a bound noun as head in (21). (21) [cise soy pak-no arki] utar anak-ne house outside till-ADV come.PL men TOP-COP a=ahun-ke yak pirka wa IND.A=enter-CAUS if be.good FIN ‘As to the men who came up to the doorway, I may let them in.’ (AB 95) (22) a. [onne=an kaspa be(come).old=IND.S overdo.AUX 11



ka somo ki no ray] katkemat even NEG do ADV die woman



While in the main clause there is a fluctuation between the use of the citation form or ‘place’ and possessive form or-o ‘the place of’ (see footnote 5), it is always the possessive form that is retained in the relative clause.



a=ne IND.A=COP ‘I am leaving the world at a young age, (so that is why I am telling you this.)’ lit. ‘I am a woman (who) is dying without having become too old…’(K8010301UP.175) b. [pirka] menoko a=etun be.good/beautiful woman IND.A=get.as.a.bride ‘I got a beautiful woman as a bride.’ (AB 94) 4.1.2. Subject of two-place transitive verb (A) Examples of this kind are less common in texts; most of them involve bound nouns as heads (23), (24c). Along with inherent transitives, as in (23), there are many derived transitives which are derived from intransitives through causative and applicative derivations, cf. (24). In the case of causatives, it is the causer (A) that is relativized (24a). (23) rapok-ke [kem sawot] pe [tumi sawot] between-POSS famine flee thing/person war flee pe payeoka. thing/person pass.by.PL ‘During that time, people (who) were fleeing famine, and people (who) were fleeing war passed by.’ (K8109193.UP.128) (24) a. [nea i=siknu-re] okkaypo a=tura wa ek=an. that IND.O=be.alive-CAUS man-DIM IND.A=bring and come.SG=IND.S ‘I have brought that youngster (who) revived me.’ (K8106233UP.246) b. [usa usa mosir e-punkine] kamuy opitta, various various land APPL-be.a.guard god all u-koramkor ruwe ne. REC-consult.with INF.EV COP ‘Gods (that) watch over various lands consulted with each other.’ (TS3 32) c. nen ka [[kim un] aca someone even/also mountains belong.to uncle ko-y-puni] kur ka isam na. APPL.to-APASS-offer man even/also not.exist FIN ‘There is no one to serve the old man of the mountain, (so serve yourself and eat.)’ (K8109193.UP.038) lit. ‘There is no person, no man (who) will serve the old man (who) belongs to the mountains.’ 4.1.3. Subject of three-place transitive verb (A) No examples have been attested so far. 4.1.4. Object of two-place transitive (O) Here as well, there are not only inherent two-place transitives (25) but also many derived verbs, i.e. causatives (26a) and applicatives (26b). In the case of applicatives, a newly added applicative object gets relativized.



(25) a. [na u-oyak ta oka a=hok] okkay-po utar yet REC-another.place LOC exist.PL IND.A=buy/hire man-DIM PL ka po-kor pa wa, even/also child-have PL and ‘The youngsters we had hired at each place were also blessed with children.’ (K8010291UP.492) b. [toani un ku=nukar] cise hemanta an? there ALL 1SG.A=see house what exist.SG ‘What is that house I see over there?’ (KS #3347) (26) a. makan… [a=ray-ke] p nea a=uni un sometimes IND.A=die-CAUS thing/person that IND.A=house.POSS ALL a=se IND.A=carry ‘Sometimes I took my catch (lit. ‘the things I killed’) to my home.’ (K8010291UP.353) b. rapok-e ta [Okikirmuy e-ak] [pon] ay ek between-POSS LOC Okikurmuy with.APPL-shoot be.little arrow come.SG ‘While a small arrow (which) Okikirmuy shot reached (the goal)’ (CY 58) Ainu applicatives are derived with the prefixes e-, ko-, and o-, which can only generally be tagged as instrumental, dative, and locative. Though all three markers increase valency, their specific semantic function may vary depending on interaction between the semantics of the prefix and verb, see Bugaeva (2010); for the availability of different functions of the applicative markers with different verb classes see Bugaeva (forthcoming). Applicatives typically offer an alternative way of introducing peripheral arguments/adjuncts as direct objects, which makes them immediately available for relativization with the strategy that does not require retention of the gapped NP tracking material, see relativization of instrumental applicative object introduced with the prefix e-in (26b); recall that relativization of adjuncts requires a “heavy” postposition stranding strategy. To paraphrase, a lot of oblique relations in Ainu can (but do not necessarily have to) be relativized as object arguments because of the availability of applicative derivations. Here are some examples illustrating the relativization of applicative objects of applicatives in e-, ko-, and o- with just a few representative semantic roles. (27) a. katkemat [a=e-hotke] usi kar housewife IND.A=at.APPL-lie.down place make wa i=kor-e and IND.O=have-CAUS ‘The housewife prepared a sleeping place (lit. ‘place to lie at’) for me.’ (AB 400) b. hemanta [a=e-oripak] kamuy why IND.S=APPL-show.respect god ene sik-i poro ruwe an? like.this eye-POSS be.big INF.NR exist.SG ‘Why does the god (whom) we respect (=the Owl god) have such big eyes?’ (AB 335) (28) a. sekor [a=ko-rewsi] Q IND.A=with.APPL-stay.overnight



nispa hawean ruwe rich.man say.SG INF.EV



ne COP



‘Said the rich man I stayed with.’ (N 174) b. [[pet etok un] kur, [kotan kor_] nispa river front belong.to person village have rich.man ko-asur-u-as] kamuy ne awa because.of.APPL-rumor-POSS-stand.SG god COP and ‘It is a god (here: hound) (who) the man from the source of the river, the village chief (lit. ‘the rich man (who) has/rules the village), became famous because of.’ (N 8) (29) araysir merayke wa [o-oman] mosir ka dead.person be.cold and to.APPL-go.SG land even/also ko-erampewtek APPL-not.understand ‘The dead person got cold and didn’t understand where he should go.’ lit. ‘did not understand the land to go to.’ (SK 43) Note that some Ainu applicatives, viz. e- applicatives with the roles of Theme/Content/Stimulus, as in (27b), are obligatory because they lack corresponding non-applicative paraphrases with case postpositions (see the end of §2.1). 4.1.5. Object of three-place transitive (O) As mentioned, three-place transitives have two unmarked objects which are similar in terms of their encoding and behavioral properties, see (7). In principle, both objects can be indexed on the verb (one at a time) but in practice it is usually the Recipient object, as in (31a), that is indexed because Recipients are typically animate and therefore may have SAP referents, while Theme/Instrument/Location objects are typically inanimate, so they automatically fall under the third person with no overt marking. In fact, I have attested only one example of Theme object indexing (‘I have just shown myself (like this) to the gods’) (Bugaeva 2011: 245) and, in the case of a competition for the slot, i.e. when there are two animate SAP objects, it is always Recipient that receives indexing. With only one exception (viz. o ‘put sth (PL) on sth’), all three-place verbs in Ainu are derived verbs, i.e. causatives (7) or applicatives (30), (31a). Even the verb kor-e ‘give’ is the causative of ‘have’ (31b). Both base and derived objects can be relativized, cf. (30a), (31b) and (30b), (31a). And there are no restrictions on the sematic role of a relativized object: it can be either Recipient/Source (30) or Theme (31). (30) a. [kampi a=e-pakasnu] hekattar paper/letter IND.A=about.APPL-teach.to children ‘Children who are taught to read and write.’ (KS #1237) b. [sintoko patci ka a=ko-uk] utar anak-ne lacquer-ware bowl even IND.A=from.APPL-take people TOP-COP nep ka pirka pirka nukar somo ki yak-ka somehow even be.good be.good look.at NEG do if-even ‘The people who were robbed of lacquer-ware and bowls, didn’t look carefully at (the things) at all…’ (AB 411) (31) a. [kunneywa morning



an exist.SG



kor, when



ne, that



a=i=y-e-pakasnu] IND.A=IND.O=EP-about.APPL-teach.to



usi-ke un sittemraypa=an wa arpa=an place-POSS ALL grope.one’s.way=IND.S and go.SG=IND.S ‘(Early) in the morning, I went groping to the place I was taught about.’ (TS2 6) b. nep [ku=kor-e] p ka isam. something 1SG.A=have-CAUS thing/person even/also not.exist ‘(Today Anna visited me and brought me many various things but) I have nothing to give to her.’ (OI) lit. ‘The thing (that) I give (to her) does not exist at all.’ 4.2. Relativization of non-arguments Relativization of non-arguments, i.e. adjuncts (=obliques) and possessors, also employs the gap strategy like that for arguments. However, to make the grammatical relations of the gapped NP traceable, (a) a case postposition-retention strategy must be used for instrumental, comitative and dative adjuncts (§4.2.1.1) and also for the degree adjunct whose relativization is regarded here as an equivalent of the standard of comparison relativization (§4.2.1.2), (b) a possessive suffix (on the possessee)-retention strategy must be used for the possessor, and finally (c) the relational noun or-o ‘the place-POSS’ and a case postposition retention strategy must be used for locative and allative adjuncts (§4.2.3). All these strategies perfectly facilitate the recovery of the case relation involved but, for some reason, examples with strategies (a) and (c), are few in texts. Instead, we find many applicative examples, in which oblique relations are relativized without the retention of the gapped NP tracking material, cf. relativization of object in §4.1.4 and §4.1.5. 4.2.1. Relativization of adjuncts 4.2.1.1. Instrumental, comitative, and dative adjuncts Only those postpositions which can occur independently (i.e. without associated nouns) in main clauses (i.e. zero anaphora use, see (12)) can be stranded in relativization, viz. intrumental ani (32), comitative tura-(no) (33), dative eun (34), and degree/allative pak-(no) (§4.2.1.2). The ability of several Ainu case postpositions to occur independently (and also that of some particles that originate in verbs) was first noticed in Chiri (2001 (1953) 19) where he contrasts it with Japanese case postpositions. (32) [ani OCA12 a=ku] CAWAN sine-p NAGASI corpok ta INST tea IND.A=drink bowl one-thing sink under LOC a=nuyna IND.A=hide ‘I hid a bowl for drinking tea under the sink.’ lit. ‘a bowl (which) I drink tea with’ (K8108012UP.007) (33) [tura-no ku=yupo kamuy cotca] COM-ADV 1SG.POSS=elder.brother.POSS bear/god shoot ‘the uncle with (whom) my elder brother shot the bear.’ (AB 96)



12



Recent loan-words from Japanese are shown in capitals.



acapo uncle



(34)



[eun a=nu13] usi ka isam DAT IND.A=hear/ask place even/also not.exist ‘There was no place (where) I (could) ask about (it).’ lit. ‘there was no place towards (where) I could ask.’ (AB 390)



4.2.1.2. Standard of comparison equivalent Essentially, standard of comparison cannot be relativized but there is a conventionalized relative clause based strategy for forming superlatives which is regarded here as a closest possible equivalent of the standard of comparison relativization 14 : lit. ‘He was a person/rich man [till (whom) there was no (such) person/rich man] (i.e. ‘a person/rich man with no equals’).’ In such a phrase, the main clause is encoded with the equative copular construction, e.g. ‘My father is a rich man…’ in (35a), in which the copular complement contains a relative clause modifier (‘a rich man i [till __i there was no rich man j]’), cf. (8). The relative clause always has the same intransitive predicate isam ‘not exist’. The most commonly used relative clause subject NPs are general (bound) nouns kur ‘man’ (36), (38) and pe ‘thing/person’ (37) or a few common nouns, e.g. katkemat ‘fine woman’ (39), (40), and nispa ‘rich/wise man’ (35), which include referents’ quality characteristics as part of their meaning. The head noun of the relative clause (see bold-faced nispa ‘rich man’ in (35)) semantically corresponds to the standard of comparison and has an underlying function of degree adjunct, which is specified in the relative clause by the retention of the postposition pak-(no) ‘as far as, till/until’. (35) a=ona [SOREKUSU15 IND.A=father.POSS especially pak-no nispa isam] nispa ne. till-ADV rich.man not.exist rich.man COP ‘Father was really a rich man, with no peers.’ lit. ‘My father was a rich man till (whom) there was really no rich man.’ (‘a rich man i [till __i there was no rich man j]’) (K8106233UP.002) (36) [pak ison kur isam] kur_ ne. till be.a.successful.hunter man not.exist man COP ‘(He) was a master of hunting, with no equals.’ lit. ‘(He) was a man till (whom) there was no man (who) is a successful hunter.’ (K8106233UP.003) The head noun and relative clause subject are often encoded with the same NP, e.g. nispa ‘rich man’ in (35), but the subject may also contain a restrictive modifier or a relative clause if its semantics is not sufficiently specified, cf. ison kur lit. ‘a man (who) is a successful hunter’ in (36) and sermaka yupke] p ‘a person (whose) guardian god is strong’ in (37).



13



The verb nu ‘hear/ask sth’ is a two-place transitive which can take only Theme as its object. The Source role can be expressed with the applicative prefix ko- (ko-nu ‘hear/ask sth from sb) or postpositions wa (ABL) and eun (DAT). 14 The possibility of a standard of comparison equivalent analysis of this construction was pointed out to me by Peter Sells (p.c.). 15 lexical nominalizations As mentioned, some relative clause heads are bound, so they tend to cliticize and eventually turn into derivational suffixes which occur in a great number of lexical nominalizations. (77) sar-un-kur Saru(place.name)-attach.to-person ‘a person from Saru’ (T1 123) (glosses are mine) (78) a. a-e-p IND.A-eat-thing/person ‘food’ lit. ‘things (that) people eat’ b. yar-pe-oma-p worn.out-thing/person-be.placed.at.SG-thing/person ‘baby’ lit. ‘a person placed in worn-out cloth’ (T1 221) (glosses are mine) c. ci-ronnu-p 1PL.EXC-kill.PL- thing/person ‘fox’ lit. ‘things (that) we kill’ (T 59) (79) ape-o-y < ape-o-hi fire-attach.to-thing/place/time ‘fireplace’ lit. ‘a place (which) the fire attaches to’ (T1 222) (glosses are mine)



29



Or-o-wa-no (place-POSS-ABL-ADV) ‘from that time/place on’ should not be mistaken for or-o ‘the place of’, to which it is etymologically related. Or-o-wa-no is an adverb (not a relational noun); it is topicalized and is not a constituent within the relative clause.



6.2. Noun-complement clause construction with perception nouns as head > ‘mermaid construction’ Ainu is remarkable for its rather rich evidential system consisting of four markers. Moreover, Ainu is typologically unusual in that all four markers originate in nouns, whereas it has been claimed that “evidentials often come from grammaticalized verbs” and “rarely come from nouns” (Aikhenvald 2004: 271, 284). In fact, the evidential markers in question are the above-mentioned grammaticalized perception nouns as the heads, cf. Table 2, which will be repeated here for convenience. (80) inferential ru-w-e [trace-EP-POSS] (< ‘the trace of’) (81); reportative haw-e [voice-POSS] (< ‘the voice of’) (82); visual sir-i [sight-POSS] (< ‘the sight of’) (83); non-visual sensory hum-i [sound-POSS] (< ‘the sound of’) (84) This section is concerned only with cases where these four perception-evidential nouns are followed by the equational copula ne ‘be(come) sth/sb’. Such a construction is commonly regarded as biclausal (Tamura 1988; Okuda 2001 (1989); Satō 2008a): the copula is analyzed as matrix clause predicate, the evidential noun as copula complement (the copula subject is left unexpressed, since it is non-referential) and the clause preceding the evidential noun as a clausal nominalization functioning as modifier of the evidential noun, with the formal structure of a noun-complement clause [ [NP] N-POSS]NP. In Bugaeva (2013: 671), I show that both monoclausal and biclausal properties co-exist, i.e. the biclausal evidential construction is on the verge of turning into a monoclausal complex predicate construction consisting of verb + auxiliary ruwe/hawe/siri/humi-ne (evidential noun + copula), hence no bracketing in (81)-(84). Therefore I prefer to consider this construction separately from the less grammaticalized O-complement clause construction, cf. §5.2. (81) a. mosirpaunsar or wa topattumi ek ruwe ne. (place.name) place ABL night.raider come.SG INF.EV COP ‘It is a fact that (lit. ‘it is the trace of’) the night raiders came from Mosirpaunsar.’ (N1 130) b. tane e=poro ruwe ne. already 2SG.S=be.big INF.EV COP ‘It is a fact that (lit. ‘it is the trace of’) you are already big.’ (K7708242UP.223) (82) a=kor nispa ek hawe ne. IND.A=have wise.man come.SG REP.EV COP ‘It is said that (lit. ‘it is the voice of’) our wise man has come.’ (K8010281.UP.059) (83) húci ek kor an siri ne. grandmother come.SG and exist.SG VIS.EV COP ‘It looks like (lit. ‘it is the sight of’) grandmother is coming.’ (Tamura 1988: 77) (84) poy-seta en=sam ta ek humi be.little-dog 1SG.O=near LOC come.SG NONVIS.EV ‘It feels like (lit. ‘it is the sound of’) a puppy came up to me.’ (ST 14)



ne… COP



This construction has a structural equivalent in Japanese 30 , mostly with modal/aspectual functions, which has been recently referred to as the ‘mermaid construction’ because it is a conglomerate of an erstwhile adnominal clause ending with a verb and an erstwhile head noun (Tsunoda 2013a). The same grammaticalization path is attested in the development of structurally similar modal/aspectual constructions in a few Tibeto-Burman languages (DeLancey 2011: 243). 6.3. Noun-complement clause construction with relational nouns as head > subordinate temporal clause construction This is a commonly attested grammaticalization path (Heine & Kuteva 2007: 254; Ōhori 2012: 637, 642). Examples below include hi ‘place/time/thing’ > ‘when’, oka-ke ‘behind (of)’ > ‘after’, etok-o ‘front (of)’ > ‘before’, rapok-i ‘interval (of)’ > ‘while’, but the list is not exhaustive. (85) [rek a rek a kor an] hi ta patek, sing ITR sing ITR and exist.SG place/time/thing LOC only a=antehoku ipe ka somo ki no IND.A=husband.POSS eat even/also NEG do and ‘Only when (the cuckoo) was singing and singing, my husband didn’t eat.’ lit. ‘Only at the time (when) (the cuckoo) was singing and singing…’ (AB 140) (86) [a=ona-ha ka i-hoppa wa isam] IND.A=father-POSS even/also APASS-leave and not.exist ‘After my father also left this world…’ (K8010311UP.156)



oka-ke ta, behind/after-POSS LOC



(87) SOREKUSU [a=wen-pakasnu] etok-o ta, especially IND.A=be.bad-punish front-POSS LOC ‘(The mad cat had told us where it had put mother’s body) before it was severely punished.’ (K8010311UP.175) It is not uncommon that case postpositions in such constructions are omitted, as in (88). This is probably indicative of a more advanced stage in the grammaticalization of relational nouns into subordinating conjunctions in these constructions. (88) [ape okari u-ko-terke rok u-ko-terke fire around REC-to.APPL-jump PERF.PL REC-to.APPL-jump kor oka] rapok-i, menoko suke and exist.PL interval/between-POSS woman cook.SG ‘While the dogs grappled around the fire, the woman cooked.’ 30



rok PERF.PL



Cf. According to Tsunoda (2013b: 105-107, 130-131, p.c.), in the case of the ‘mermaid construction’ of Japanese, the subject marked with the topic marker =wa (e.g. Hanako=wa) has all properties of the subject that Tsunoda (2009: 214-215) proposes for Japanese in general, i.e. Hanako=wa is the subject of the entire ‘mermaid construction’. Hanako=wa Nagoya=ni ik-u yotee=da. Hanako=TOP Nagoya=DAT/LOC go-NPST plan=COP.NPST ‘Hanako plans to go to Nagoya.’ lit. ‘Hanako is a plan [such that she] goes/will go to Nagoya.’ (Tsunoda 2013b: 69)



lit. ‘At the interval (that) the dogs grappled around the fire…’ (N 407) There are even more advanced cases of grammaticalization when an erstwhile relational noun occurs as conjunction in the sentence initial position (without a preceding clause), cf. rapok-ke (ta) ‘during, while, then’ in (23), (26a), and (106). This kind of development is possible because the original relational noun bears the possessive marking cross-referencing the presence of a possessor (here, the presence of a clause), so the possessor NP itself is subject to zero anaphora, cf. the structure of possessive NP in (14). Thus, the formation of noun-complement clauses in Ainu, just like that of relative clauses (cf. §4), may also be accounted for by the head-marking polysynthetic character of Ainu which presupposes the traceability of grammatical relations from the head (here, a nominal head) and predilection for zero anaphora. 7. Noun+noun compounds In this section, I will consider a structure which is superficially similar to the noun modifying construction and GNMCC in particular since it looks as if an instrumental or locative adjunct is relativized without the retention of the respective postpositions, which is normally not possible in Ainu, cf. §4.2.1.1. and §4.2.3. Yet, a closer look reveals that the modifying element in it is not clausal but is rather a nominalization. The existence of this structure was first pointed out in (Kirikae 2001 (1984)) who suggests distinguishing it from (clausal) “modification structures” and terms it “pseudo-modification structure”. (89) [repa] go.out.to.sea.to.fish ‘a fishing boat’



cip boat



(90) *[yuk san] nay deer descend river intended meaning: ‘a river (where) deer come out’ (Chiri 1974 (1956) 380) According to Kirikae (2001 (1984): 208), this kind of structure is possible only if there is no overtly expressed subject, cf. Chiri’s (1974 (1956): 380) famous example of “broken Ainu” registered in the speech of those speakers of Ainu who were more fluent in Japanese (90). The predicate may be either an inherent intransitive verb, as in (89), or a derived intransitive verb with the antipassive (indefinite object) prefix i-, as in (91), or a transitive verb with an overtly expressed object (92). (91) [i-ma] san APASS-grill shelf ‘a fish-grilling shelf’ (92) [SAKE kar] sintoko sake.wine make vessel ‘a large vessel used in the sake making process’ lit. ‘a sake-making vessel’ In all these cases, the modifying element of the pseudo-modification structure is regarded as “nominalized, so this structure is assigned to the noun+noun construction which is in turn a type of compound nouns.” (Kirikae 2001 (1984): 217)



Satō (2008b) elaborates on this suggesting that the modifying element is a kind of verbal noun converted from the corresponding intransitive verb, recall that intransitive verbs may function as nouns without any change in the morphology. He reinterprets the cases of a transitive verb with overtly expressed object as O-incorporation, which first results in the intransitivization of the verb and then in its conversion into a verbal noun: sake-kar ‘sake-making’, cf. (92), see also (93). (93) [i-sapa-kik] ni IND.O-head-beat tree/stick ‘a wooden stick used to strike the head of freshly caught salmon’ lit. ‘a stick for striking the thing’s head’ (cf. [i-sapa kik] ni in (KH 204) The incorporation explanation finely accounts for the absence of the subject NP (S or A) in the respective collocations (see (90)): incorporating subject, although not impossible in Ainu, is very rare. Some “pseudo-modification structures” resemble noun-complement clauses, but the possessive marking on the head noun is missing and the subject in the modifying part is omitted, therefore they are also regarded as noun+noun compounds in which the first noun is converted from an intransitive verb often containing O-incorporation. (94) [upas-ru-re] cup snow-melt-CAUS month ‘the month of snow-melting (=April)’ (cf. [upas ru-re] cup in Kirikae (2001 (1984): 210)) The fact that in noun+noun compounds O-incorporation really takes place and then gets converted into a noun can be proven by a number of syntactic tests. For instance, the incorporation does not allow the insertion of a restrictive particle ka ‘even/also’, cf. (95a) and (95b), and verbal noun converts cannot take adverbial modifiers, cf. (96a) and (96b). (95) a. [wakka-ku] ontaro water-drink barrel ‘a water tank’ lit. ‘a barrel (for) water-drinking’ (T 471) b. *[wakka ka ku] ontaro water even/also drink barrel intended meaning: ‘a barrel (for) (that is) also water-drinking’ (ST2 60) (96) a. [husko ni-tuy-pa] ru ka an ru-w-e a=nukar be.old tree-cut-PL trace even/also exist.SG trace-EP-POSS IND.A=see ‘I saw that there was the trace of a tree that had been cut before.’ lit. ‘the trace of old tree-cutting.’ (AB 254) b. *[husko-no ni-tuy-pa] ru ka an ru-w-e a=nukar be.old-ADV tree-cut-PL trace even/also exist.SG trace-EP-POSS IND.A=see intended meaning: ‘I saw that there was the trace of a tree that had been cut before.’ lit. ‘the trace (such as) a tree is cut before/previously’ (my constructed example), cf. (103) Kirikae (2001 (1984): 206) provides only one example with an overtly expressed subject (97), which he could not account for. I suggest that this example may be assigned to a “noun+noun+noun



structure”, the first two nouns are verbal nouns derived by conversion from intransitive verbs. (97) [komni] [ran] cup oak.tree fall month ‘the month of oak tree’s (acorn) falling’ cf. ‘the month (when) the oak tree (its acorns) falls (=October)’ (KH 206) The overall number of noun+noun compounds of “pseudo-modification structure” is rather low. They are mainly restricted to tools, cultural objects and names of some plants/animals. 8. Emerging GNMCC As mentioned in §5, noun-complement clauses require possessive marking on the head noun. However, occasionally, in actual texts, we can find examples without the possessive marking, which I suggest regarding as emerging GNMCCs. (98) ora-no [a=ko-caranke] haw ne noyne haw-as kor sir-an. then-ADV IND.A=to.APPL-argue voice COP seems voice-stand.SG and sight-exist.SG ‘I could hear someone getting into an argument (with my husband).’ lit. ‘it sounded as if it was a voice of someone arguing towards (my husband).’ (K7908032UP) (99) ape ka pirka-no a=úna pekor yay-nu=an fire even/also be.good-ADV IND.A=cover.a.fire.with.ash as.if REFL-hear=IND.S a p, [ape a] ru an. PERF.SG but fire sit.SG trace exist.SG ‘I thought I had buried the embers, but there is a trace (such as) a fire was lit.’(K7908032UP) (100)



[pirka aynu a=ne] hum as. be.beautiful human IND.A=COP sound stand.SG ‘I felt like an admirable human being.’ lit. ‘(There) stood a sound (such as) I am an admirable human being.’ (K7803232UP.071)



(101)



ora-no [korkoni tane-po hetuk-pa] rapok ne then-ADV butterbur now/already/finally-DIM spring.up-PL interval/between COP ‘It was the season (when) butterbur had finally sprouted.’ (K7803233UP.112)



The overall number of such examples (98)-(101) is very low (less than 3%)31 compared to those in which the head noun bears the possessive suffix; the conditions of their use are not all clear. According to Satō (2008b: 172), there is a tendency that the head noun is likely to occur in its possessive form (vs. citation form) when it refers to what has already been introduced in discourse or is mutually understood by both the speaker and hearer, otherwise the head noun occurs in its citation form. Generally, this seems to be true but yet there are many examples which require an extra explanation. Consider the following passage discussing ‘watching the sight of worshipping water chestnuts’ in which the word ‘sight’ occurs as a head noun three times: first, in its citation form sir as a newly introduced participant (102a), then in this possessive form sir-i as an already 31



The estimation is made based on Tamura (1991, 1992, 1993).



introduced participant (102b), and finally, quite unexpectedly, as sir again (102c); the latter use cannot be accounted for with Satō’s rule. (102) a. ranma poro-n-no okay pe ne wa kusu, now be.many-EP-ADV exist.PL NMZ COP and because tono utar, [ne i-nomi=an] sir nukar rusuy-pa kusu, lord PL this APASS-pray=IND.S sight see want-PL because toop, tono kotan wa-no, yap wa, be.far lord village ABL-ADV descend.PL and b. ne, [pekanpe a=nomi] sir-i, nukar-pa. this water.chestnut IND.A=worshipsight-POSS see-PL c. horip-pa ne ya upopo ne ya, kí-pa kor dance-PL COP Q sing.in.a.circle COP Q do-PL and [kamuy-nomi-pa] sir a=nukar wa, a=e-rayap hi, god-worship-PL sight IND.A=see and IND.A=APPL-admire NMZ tane ka, kanpi or ta ka, a=nuye ka ki. now even/also paper place LOC even/also IND.A=write even/also do ‘(Since even now there are many (water chestnuts)), the Japanese, in order to see the sight of this worshiping, come from remote Japanese areas and watch the sight of water chestnut worshiping. Now, it is even written in the books that the sight of dancing, singing in a circle and worshiping the gods is being watched and admired.’ (TS1 28) I suggest that the occurrence of the head noun sir ‘sight’ in its citation form in (102c) is triggered by the presence of a modifier encoded with a nominalization kamuy-nomi-pa ‘godsworshiping’, cf. §7. Recall that noun+noun compounds do not require the use of the possessive suffix on the second noun since their relationship is viewed as purely attributive and does not imply possession. In principle, most examples with the head noun in the citation form which have a modifier containing an intransitive predicate, e.g. (99), can be assigned to the “(noun)+ noun+noun structure”, cf. i. and ii. in (103), and also (97). (103)



oraun [hinak un hekaci cis] haw a=nu then where ALL boy cry voice IND.A=hear ‘Then I (could) hear a child crying somewhere.’ lit. i. ‘the voice (such as) a child is crying somewhere’; ii. ‘the voice of a child cry somewhere’ (K7908051UP)



Examples involving transitive predicates which are indexed for the person and number of the subject, as in (98) and (100), cannot be accounted for in this manner. I suggest that there is another general tendency: the head noun is used in its citation form in order to distinguish it from strongly grammaticalized perception and relational nouns as head which always occur in the possessive form. Consider the following example in which ru ‘trace’ is used in its lexical meaning and ru-w-e ‘the trace of’ in its grammatical meaning – as an inferential evidential marker ‘it is a fact (that)’, cf. also (96a). (104)



[asin-no be.new-ADV



ni tree



a=tuy-pa] IND.A=cut-PL



ru trace



ka even/also



oka, exist.PL



[hoski-no ni a=tuy-pa] ru ka oka ru-w-e be.old-ADV tree IND.A=cut-PL trace even/also exist.PL trace-EP-POSS>INF.EV an h_i kusu, exist.SG NMZ because ‘It is a fact (that) there were the traces of freshly cut trees (and) there were the traces of trees that had been cut before, so…’ (K7908032UP) In noun-complement clauses, the use of the head noun in its citation form may be regarded as an innovative process with the GNMCC as an output. 9. Island violation Relativizing from inside the relative clause, i.e. violating island (or subjacency) constraints, is not possible in Ainu. However, in the case of noun-complement clauses, island violation is possible. Example (105) involves a standard noun-complement clause with the head noun in its possessive form, while example (106) involves the so-called “emerging GNMCC”, i.e. a noun-complement clause with the head noun in its citation form. (105) [[ene SOREKUSU tura-no oka=an] hum-i wen] like.this especially COM-ADV exist.PL=IND.S sound/feeling-POSS be.bad pe SOREKUSU nep a=ronnu pa ruwe ene an. thing/person especially somehow IND.A=kill.PL PL INF.EV like.this exist.PL ‘(Even if just from now, we would be grateful if we could live quietly in peace. We have heard about you. We have heard about the wise man.) The people we didn’t want as neighbors were killed in that manner.’ lit. ‘Peoplei [(such as) especially the sound [(such as) to be with __i ] is bad] were killed in this manner.’(K8109171UP.224-5) (106) rapok-ke nea okkaypo [[ani kik kuni, between-POSS that man INST beat in.order kisma] hum pirka-no an, an] cikuni tuy-e grab sound/feeling be.good-ADV exist.SG exist.SG tree cut-CAUS ‘Then that young man cut a tree (such that) the feeling is good of grabbing (it) [i.e. which fits his hand] in order to beat with (it).’ lit. ‘Then that young man cut a treei [(such that) the feeling [(that) he grabs __i in order to beat with __i] is good].’ (K7708242UP.199-200) Crosslinguistically, noun-complement clause islands are much easier to violate than relative clause islands, cf. island violation in the case of English fact-S construction with a non-lexical head (that), e.g. The hat which I believed that Otto was wearing__ is red (Ross 1967: 70). The examples of noun-complement island violation in Ainu are remarkable in that they involve a bit more lexical heads, i.e. pe ‘thing/person’ in (105) and hum ‘sound/feeling’ in (106), cf. the impossibility of island violation in English fact-S construction with a lexical head (claim): *The hat that I believed the claim that Otto was wearing __ is red (Ross 1967: 70). 10. Concluding remarks In this paper, I have argued that the relative clause construction in Ainu (Southern Hokkaido), though similar to the noun-complement construction, is not identical. The relative clause construction (16) is built on the model of the nominal attributive construction, cf. (13), while the



noun-complement construction (17) is built on that of the nominal possessive construction, cf. (14). In the latter (17), a possessive suffix refers to the content of the whole (noun-modifying) clause, which relates to the head noun as its complement. Relative and noun-complement clauses behave differently with regard to island violation constraints: only the latter allow to violating them (§9). In the case of the noun-complement clause construction, the possessive marking may be regarded as explicit expression of the relation between the head noun and modifying clause, hence this is not a GNMCC (§5). The formation of all relative clauses involves the NP gap or zero anaphora strategy. However, only the relativization of arguments does not require the retention of the gapped NP tracking material. Relativization on non-arguments requires (a) a case postposition retention strategy for instrumental, comitative and dative adjuncts (§4.2.1.1) and also for the standard of comparison equivalent (=degree adjunct) (§4.2.1.2), which is typologically unusual, (b) a possessive suffix retention strategy for the possessor (§4.2.2), and (c) the relational noun or-o ‘the place-POSS’ and a case postposition retention strategy for locative and allative adjuncts (§4.2.3). All these retention strategies contain explicit expression of the relation between the head noun and modifying clause and thus deviate from the prototypical GNMCC. Importantly, in the case of relative and noun-complement clause constructions, except for a few conventionalized lexical nominalizations, which are mainly restricted to tools, cultural objects and names of some plants/animals (§7), we are dealing with a genuine clause without any signs of nominalization either on the predicate or elsewhere (see §3). In this respect Ainu is similar to Japanese, but is different from such Eurasian languages as Korean, Chinese, some Turkic and Uralic languages. As suggested in §4 and §6.3, the structure of both relative and noun-complement clauses in Ainu may be best accounted for by its head-marking polysynthetic character which generally presupposes the traceability of grammatical relations from the head (verbal or nominal) and predilection for zero anaphora in both verb and noun (possessive) phrases. I have shown that noun-modifying clause constructions are diachronically unstable, since, in many cases, they are in the process of grammaticalization into other construction types. Thus, in relative clauses, bound nouns as heads often grammaticalize into suffixes deriving lexical nominalizations (§6.1). In noun-complement clauses, perception nouns as heads followed by the copula ne tend to grammaticalize into auxiliaries with evidential and aspectual meanings and the construction in question turns into a mono-clausal complex predicate construction (the so-called ‘mermaid construction’) (§6.2). In noun-complement clauses, relational nouns as heads tend to grammaticalize into subordinating temporal conjunctions and the construction in question turns into an adverbial subordinate clause construction (§6.3). And finally, in §8, I have pointed to the existence of a small number of noun-complement clauses without the possessive marking on the head noun. I suggest that in such cases the head noun is used in its citation form in order to dissociate it from the above-mentioned strongly grammaticalized perception and relational nouns as heads which always occur in a possessive form. This may be regarded as an innovative process resulting in the emergence of the GNMCC. It might be interesting to find out whether a similar development is attested in other Ainu dialects and other languages of the world. * Parts of this paper were presented at the 1st meeting of the PFIH -- NMC project headed by Yoshiko Matsumoto (Stanford University on July 28, 2009) and 1st meeting of the project “Typological and Historical/Comparative research on the Languages of the Japanese Archipelago and its Environs” headed by John Whitman (NINJAL, July 6, 2013). I very am grateful to participants of these meeting for their questions



and comments. I would also particularly like to thank Bernard Comrie, Yoshiko Matsumoto and my colleagues at NINJAL, i.e. John Whitman, Iku Nakasaki, Heiko Narrog, Prashant Pardeshi, and Peter Hook, for most valuable inspiring comments on the earlier version of this paper. None of them are responsible for any errors. My deepest heartfelt thanks go to the late Vladimir Nedjalkov (1928–2009) who was my mentor, and to the late Ainu language speaker Mrs. Ito Oda (1908–2000) and other speakers of Ainu who I was blessed to work with. The present study was partially supported by the Japanese Ministry of Education, Science and Culture, International Scientific Grant-in-Aid Research Program (grant #22720169).



Abbreviations 1/2/3 = 1st/2nd/3rd person, Ø = zero-marked 3rd person, = = inflectional boundary in the morphemic line, - = derivational boundary in the morphemic line, _ = morphophonological alternations in the text line, A = transitive subject, ABL = ablative, ACAUS = anticausative, ADM = admirative, ADV = adverbial, ALL =allative, APASS = antipassive, APPL = applicative, AUX = auxiliary verb, CAUS = causative, CLF = classifier, COM = comitative, COMP = complementizer, COP = copula, DAT = dative, DESID = desiderative, DIM = diminutive, EP = epenthetic consonant, EXC = exclusive, FIN = final particle, INC = inclusive, IND = indefinite, INF.EV = inferential evidential, INST = instrumental, ITR = iterative, KY = kamuy yukar ‘songs of gods’, LOC = locative, NEG = negation, NMZ = nominalizer, NONVIS.EV = nonvisual evidential, NPST = nonpast, O = object, PERF = perfect, PL = plural, POSS = possessive, Q = question marker, QUOT = quotation marker, REC = reciprocal, REFL = reflexive, REP.EV = reportive evidential, S = intransitive subject, sb = somebody, SG = singular, SOC = sociative, sth = something, TOP = topic, UP = uwepeker ‘prosaic folktales’, vi = verb intransitive, VIS.EV = visual evidential, vt = verb transitive.



Sources Bugaeva, Anna (2004) Grammar and folklore texts of the Chitose Dialect of Ainu (Idiolect of Ito Oda). ELPR Publication Series A–045. Suita: Osaka Gakuin University. CY Chiri, Yukie (1978) Ainu shin’yōshū [Ainu songs of gods]. Tokyo: Iwanami Bunko. K Nakagawa, Hiroshi & Bugaeva, Anna (2012) Folktales of the Saru dialect of Ainu by Mrs. Kimi Kimura (1900-1988) [deposit 1]. ELDP, SOAS, University of London. http://elar.soas.ac.uk/projects/0107 KH Kirikae, Hideo (2001 (1984)). KI Kubodera, Itsuhiko (1977) Ainu jojishi shin’yō seiden no kenkyū [The study of Ainu heroic epics and songs of gods]. Tokyo: Iwanami shoten. KS Bugaeva, Anna, Shiho Endō, and Shirō Akasegawa (2015). A Topical Dictionary of Conversational Ainu (English HP) NINJAL. http://ainutopic.ninjal.ac.jp/en/ N Nakagawa, Hiroshi (1995) Ainugo chitose hōgen jiten [A dictionary of the Chitose dialect of Ainu]. Tokyo: Sōfūkan. N1 Nakagawa, Hiroshi (2002) Ainu kōshōbungei tekisuto 3. Yūrashia gengo bunka ronshū 5. Chiba: Chiba University, 111–143. NV Nevskij, Nikolaj (1972) Ajnskij folklor [Ainu folklore]. Moskva: Nauka. OI Oda, Ito (1908-2000), informant of the Chitose dialect of Ainu; the author’s fieldnotes. SK Sunazawa, Kura (1983) Watashi no ichidai no omoide Ku sukup opuspe [My life story]. Sapporo: Miyama Shobō. ST Satō, Tomomi (2011) Ainugo chitose hōgen ni okeru suyryō no jodōshi nankor no imi ni AB



ST1 ST2 ST3



T T1 T3 TS1 TS2 TS3 TS5 TS6



tsuite [A semantic analysis of a modal auxiliary verb nankor in the Chitose dialect of Ainu]. Hokkaido-ritsu ainu minzoku bunka kenkyū sentā kenkyū kiyō 17. Sapporo: Hokkaido-ritsu ainu minzoku bunka kenkyū sentā, 1-18. Satō, Tomomi (2008a). Satō, Tomomi (2008b). Satō, Tomomi (1998) Ningen no kigen o monogataru Penanpe to Pananpe no mukashibanashi [A story of Penanpe to Pananpe who told about the origins of humans]. In: Watanabe, Hitoshi; Ōshima, Minoru; Kirikae, Hideo & Satō, Tomomi (eds.), Ainu minzoku bunkazai chōsa hōkokusho 18. Sapporo: Hokkaidō kyōiku iinkai, 120-130. Tamura, Suzuko (1996). Tamura, Suzuko (2000 (1988)). Fukuda (Tamura), Suzuko (2001 (1960)) Ainugo saru hōgen-no jodōshi [Auxiliary verbs of the Saru dialect of Ainu]. In: Minzokugaku kenkyū, 24-4, Seibundō Shinkōsha, 67-78. Tamura, Suzuko. 1984. Ainugo shiryō (1) [Ainu audio materials (1)]. Tokyo: Waseda daigaku gogakukyōiku kenkyūjo. Tamura, Suzuko (1985) Ainugo shiryō (2) [Ainu audio materials (2)]. Tokyo: Waseda daigaku gogakukyōiku kenkyūjo. Tamura, Suzuko (1986) Ainugo shiryō (3) [Ainu audio materials (3)]. Tokyo: Waseda daigaku gogakukyōiku kenkyūjo. Tamura, Suzuko (1988) Ainugo shiryō (5) [Ainu audio materials (5)]. Tokyo: Waseda daigaku gogakukyōiku kenkyūjo. Tamura, Suzuko (1989) Ainugo shiryō (6) [Ainu audio materials (6)]. Tokyo: Waseda daigaku gogakukyōiku kenkyūjo.



References Aikhenvald, Alexandra (2004) Evidentiality. Oxford: OUP. Baker, Mark (1996) The polysynthetic parameter. Oxford: OUP. Bugaeva, Anna (2008) Reported discourse and logophoricity in Southern Hokkaido dialects of Ainu. Gengo Kenkyū 133, 31–75. Bugaeva, Anna (2010) Ainu applicatives in typological perspective. Studies in Language 34(4). 749-801. Bugaeva, Anna (2011) Ditransitive constructions in Ainu, Sprachtypologie und Universalienforschung (STUF) 64 (3). 237-255. Bugaeva, Anna (2012) Southern Hokkaido Ainu. In: Tranter, Nicolas (ed.) The languages of Japan and Korea. London: Routledge, 461-509. Bugaeva, Anna (2013) Mermaid constructions in Ainu. In: Tsunoda, Tasaku (ed.) Adnominal clauses and the ‘mermaid construction’: grammaticalization of nouns. (NINJAL Collaborative Research Project Reports 13-01.) Tachikawa, Japan: National Institute for Japanese Language and Linguistics, 667-675. Bugaeva, Anna (2015) Valency classes in Ainu. In: Malchukov, Andrej & Comrie, Bernard (eds.) Valency classes in the world’s languages, Berlin, New York: De Gruyter Mouton, 807-854. Chiri, Mashiho (2001 (1953)) Ainugo no joshi [Auxiliaries in Ainu]. In: Kindaichi hakase koki kinen gengo minzoku roshū. Sanseidō. Reprinted in: Ainugo-kō, Bunpō I, v. 4, 1-20. Tokyo: Yumani Shobō. Chiri, Mashiho (1974 (1956) ) Ainugo nyūmon: toku-ni chimei kenkyuūsha no tame ni. Tokyo. Reprinted in: in Chiri Mashiho chosakushū, v. 4, Tokyo: Heibonsha.



Comrie, Bernard (1996) The unity of noun modifying clauses in Asian languages. Pan-Asiatic Linguistics: Proceedings of the Fourth International Symposium on Languages and Linguistics, January 8-10, Volume 3, 1077-1088. Salaya, Thailand: Institute of Language, and Culture for Rural Development, Mahidol University at Salaya. Comrie, Bernard (1998) Rethinking the typology of relative clauses. Language Design 1: 59–86. DeLancey, Scott (2011) Finite structures from clausal nominalizations in Tibeto-Burman. In: Foong Ha Yap, Karen Grunow Harsta, and Janick Wrona (eds.) Nominalizations in Asian languages (TSL 96), Amsterdam, Philadelphia: John Benjamins, 343-359. Epps, Patience (2012) Between headed and headless relatives: evidence from Amazonia. In: Comrie, Bernard & Estrada-Fernández, Zarina (eds) (eds) Relative clauses in languages of the Americas: A typological overview (TSL 102). Amsterdam, Philadelphia: John Benjamins, 191212. Givón, Talmy (2012) Toward a diachronic typology of relative clause. In: Comrie, Bernard & Estrada-Fernández, Zarina (eds) Relative clauses in languages of the Americas: A typological overview (TSL 102). Amsterdam, Philadelphia: John Benjamins, 3-25. Hattori, Shirō (ed.) (1964) An Ainu dialect dictionary. Tokyo: Iwanami Shoten. Heine, Bernd & Kuteva, Tania (2007) The Genesis of grammar. A reconstruction. Oxford: OUP. Keenan, Edward L. & Comrie, Bernard (1977) Noun phrase accessibility and universal grammar. Linguistic Inquiry Vol.8, No.1: 63-99. Kirikae, Hideo (2001 (1984)) Ainugo no meishiku no kōzō to gōsei meishi [The structure of noun phrases and compound nouns in Ainu]. Gengo kenkyū 86: 105-21. Reprinted in: Ainugo-kō, Bunpō II, v. 5, 201-217. Tokyo: Yumani Shobō. Matsumoto, Yoshiko (1988) Semantics and pragmatics of noun-modifying constructions in Japanese. Proceedings of the 14th Annual Meeting of the Berkeley Linguistics Society, 166-175. Matsumoto, Yoshiko (1997) Noun-modifying constructions in Japanese: A frame semantic approach (Studies in Language Companion Series 35). Amsterdam: John Benjamins. Matsumoto, Yoshiko (this volume) General noun modifying constructions in Japanese. In: Matsumoto, Yoshiko, Comrie, Bernard & Sells, Peter. Noun-Modifying Clause Constructions in Languages of Eurasia: Reshaping theoretical and geographical boundaries. Amsterdam, Philadelphia: John Benjamins. Ōhori, Toshio (2012) The grammaticalization of subordination. In: Narrog, Heiko & Heine, Bernd (eds) The Oxford handbook of grammaticalization. OUP, 633-642. Okuda, Osami (2001 (1989)) Nihongo to ainugo-ni okeru rentai shūshoku to bun-no meishika [Noun-modifying and clausal nominalization in Japanese and Ainu]. In: Waseda daigaku gogaku kyōiku kenkyūjo kiyō 39. Reprinted in: Ainugo-kō, Bunpō II, v. 5, 243-256. Tokyo: Yumani Shobō. Ross, John (1967) Constraints on variables in syntax. MIT PhD dissertation. Satō, Tomomi (1997) Possessive expressions in Ainu. In: Hayashi, Tōru & Bhaskararao, Peri (eds.) Studies in Possessive Expressions, 143-160. Tokyo: Tokyo University of Foreign Studies. Satō, Tomomi (2008a) Ainugo bunpō no kiso [The basics of Ainu grammar]. Tokyo: Daigakushorin. Satō, Tomomi (2008b) Ainugo chitose hōgen ni okeru gōsei meishi [Compound noun word formation in the Chitose dialect of Ainu]. Journal of the Center for Northen Humanities, 1:55-68. Tamura Suzuko (1984) Aynu Itak. Aynugo nyūmon. Kaisetu [The Ainu language. An introduction into Ainu. Commentary], (Goken kyōzai sensho 32). Tokyo: Waseda daigaku gogaku kyōiku kenkyūjo.



Tamura, Suzuko (2000 (1988)) Ainugo [The Ainu language]. In: Kōji, Takashi; Kōno, Rokurō & Chino, Eiichi (eds.) Gengogaku daijiten [Encyclopedia of linguistics] Tokyo: Sanseidō. English translation: The Ainu language, (ICHEL Linguistic Studies 29). Tokyo: Sanseidō. Tamura, Suzuko (1991, 1992, 1993) Ainugo onsei shiryō (1-6) Goi: zenyōrei-tsuki [Ainu Audio Materials (1-6) Lexicon: All with examples]. v. 1-3, Tokyo: Waseda daigaku gogaku kyōiku kenkyūjo. Tamura, Suzuko (1996). Ainugo Saru hōgen jiten [A dictionary of the Saru dialect of Ainu]. Tokyo: Sōfūkan. Teramura, Hideo (1969) The syntax of noun modification in Japanese. The Journal-Newsletter of the Association of Teachers of Japanese. Volume 6, No. 1: 64-74. Teramura, Hideo (1992) Teramura Hideo Ronbunshu I: Nihongo Bunpoo Hen (Papers by Hideo Teramura I : Japanese Grammar). Tokyo: Kurosio Shuppan. Tsunoda, Tasaku (2009) Sekai no gengo to nihongo, kaiteiban [The world’s languages and Japanese, rev. ed.]. Tokyo: Kurosio. Tsunoda, Tasaku (2013a) Mermaid construction: an introduction and summary. In: Tsunoda, Tasaku (ed) Adnominal Clauses and the ‘mermaid construction’: grammaticalization of nouns. (NINJAL Collaborative Research Project Reports 13-01.) Tachikawa, Japan: National Institute for Japanese Language and Linguistics, 15-66. Tsunoda, Tasaku (2013b) Mermaid construction in modern Japanese. In: Tsunoda, Tasaku (ed) Adnominal Clauses and the ‘mermaid construction’: grammaticalization of nouns. (NINJAL Collaborative Research Project Reports 13-01.) Tachikawa, Japan: National Institute for Japanese Language and Linguistics, 67-150.