Relative Clauses in Japanese Sign Language ENDO, Eita A Thesis Submitted in Partial Fulfilment of the Requirements for the Degree of Master of Arts in Linguistics The Chinese University of Hong Kong June 2015 1 ABSTRACT Japanese Sign Language (JSL) is an indigenous sign language that the Deaf people in Japan use every day. JSL has some types of constructions that occur frequently, which some researchers preliminarily coined them as relative clauses (RCs). To date, only a few researchers have attempted to examine such clauses. To gain further understanding about them, an elicited production task was designed in which Deaf JSL signers were invited to produce several sentences after viewing some pictures, which were designed to induce them to sign RCs. The second task involved the participants to produce monologues. They were given three familiar topics such as „What was your most enjoyable experience of travelling?‟, „What is your most impressive experience at the deaf school?‟, and „What topic/activities are you most interested in now?‟ All the data were recorded and transcribed by ELAN, and subsequently analyzed with a focus on the distribution of nonmanual (NM) features. On the basis of the findings, I have confirmed the existence of two types of RCs: prenominal and internally headed, which the previous research has proposed. The findings also showed that NMs, such as movement of the head and eye blinks, play a role in marking the relative clauses. 2 Contents Abstract ...................................................................................................................................... 2 1. INTRODUCTION ................................................................................................................. 4 1.1. Relative clauses: defining properties? ............................................................................ 5 1.2. Typology of relative clauses ......................................................................................... 14 1.2.1. Relative clauses are subordinate clauses................................................................ 15 1.2.2. Nominalizer and nominalization ............................................................................ 18 1.2.3. Embedded or adjunction ........................................................................................ 21 1.2.4. Externally headed: prenominal or postnominal ..................................................... 22 1.2.5. Internally headed relative clauses .......................................................................... 25 1.2.6. Correlatives ............................................................................................................ 25 1.3. Relative clauses in sign languages: previous studies .................................................... 27 1.4. Relative clauses in Japanese Sign Language: some previous analysis ......................... 30 1.5. Interim conclusion ........................................................................................................ 32 2. THE CURRENT STUDY .................................................................................................... 33 2.1. Methodology ................................................................................................................. 33 2.1.1 Participants and data collecting conditions ............................................................. 33 2.1.2 Types of data ........................................................................................................... 34 2.1.3. Transcription of the data ........................................................................................ 37 3. RESULTS ............................................................................................................................ 37 3.1. Prenominal relative clauses........................................................................................... 39 3.2. Internally headed ........................................................................................................... 41 4. DISCUSSION ...................................................................................................................... 44 5. CONCLUSION .................................................................................................................... 48 Appendix: abbreviations .......................................................................................................... 49 BIBLIOGRAPHY .................................................................................................................... 51 3 1. INTRODUCTION Relative clauses (RCs)1 attract much attention in diverse fields. For example, Fabb (1990) syntactically investigated two types of RCs in English: restrictive RCs and nonrestrictive RCs. He argued that the differences between the two types are explainable by the fact that the former is a modifier and the latter is not. In addition, he claims that a restrictive RC is a „part of the same sentence as its antecedent N‟ (i.e. the head noun) but the nonrestrictive RCs is not. RCs also catch interests of the researchers into the language acquisition. They carried out their investigation into how children become capable of establishing RCs through examining the comprehension and production of RCs by young speakers. For example, the early literature on the acquisition of RCs attributed the children‟s error in understanding RCs to a hypothesis that children do not have „the hierarchical structure that is characteristic of adult syntactic representations‟ and „children assign “flat” phrasal structure‟ (Crain & LilloMartin 1999: 389) to RCs. According to this hypothesis, children analyze a sentence including an RC (1), not as a sentence (2a) (adults‟ analysis), but as a sentence (2b). (1) The dog pushed the sheep that jumped over the fence. (2) (a) [The dog pushed [the sheep jumped over the fence]. (b) [The dog pushed the sheep][… jumped over the fence]. (Crain & Lillo-Martin 1999: 392) The analysis as in (2b) is frequently referred to as the conjoined-clause analysis, because the analyzed sentence looks like a conjoined sentence „The dog pushed the sheep and (the dog) jumped over the fence‟. However, the research that was conducted afterwards suggested the difference of the RC interpretation between children and adults should be due to elicitation of RCs. In other words, young children at the age of three or four can produce sentences that include a RC „if conditions (to produce RCs) are met‟ (Crain & Lillo-Martin 1999: 397) as in (3). (3) Jabba, please come over to point to the one that‟s asleep. (3;5) (Crain & Lillo-Martin 1999: 397) 1 In this paper, square brackets ([ ]) are placed around the RCs in the original sentence and/or its glosses. E.g., He loved the woman [who killed him]. 4 Sentences which have RCs are generally complex. Nevertheless, they are found in human languages because they have some functions. One of them is a discourse function. Relevant to it, Fox and Thompson (1990), in their study of the grammar of RCs in English conversations, pointed out that information-flow factors decide how to combine the head NP (including the determiner) and the relative pronoun in the discourse. One of such factors is grounding as in (4). (4) (talking about upkeep on houses) But uh – the original price of it, eh – you can‟t even (inaud.) the original price, just that little screen porch alone is five hundred dollars, the air condish- the uh heater thing [we put in] I think was a hundred uh five six hundred dollars, (Fox and Thompson 1990: 300) In (4): the RC grounds the head NP „the heater thing‟ and makes it clear that the heater (the air conditioner) is not the old one but the new one that „we‟ (the speaker and his/her coworker(s)) put in. Thus, RCs make it clear what the speaker is talking about. Thus, RCs attract linguists in various fields. But what kind of properties define RCs? 1.1. Relative clauses: defining properties? There have been various definitions of RCs such as (5). (5) A construction consisting of a (possibly empty) nominal and subordinate clause which semantically modifies the nominal. The nominal is the head, and the subordinate element the R[elative] C[lause]. (Lehmann 2003: 460) For example, Downing (1978) observed ten syntactic properties that may be found among a variety of RC constructions in his survey (see (6)). (6) Syntactic properties of relative clause constructions (Downing 1978: 377) (a) A RC contains a finite verb. (b) The verb of a RC assumes a distinctive nonfinite form. 5 (c) A RC contains a pronoun coreferential with a noun that immediately precedes (or follows) the RC. (d) No nominal in the RC is coreferential with a preceding (or following) noun. (e) A RC together with a nominal expression forms a noun phrase (NP) constituent. (f) A RC is the sole constituent of a NP. (g) A RC is not a constituent of a NP. (h) A RC begins (or ends) with a distinctive marker. (i) A RC contains a marker that is linked by cooccuance with a nominal marker outside the clause. (j) The internal structure of a RC is indistinguishable from that of (some) nonrelative clauses. Since some of them are contradictory to each other (e.g., (6a) and (6b); (6f) and (6g)), one given type of RCs cannot have those contradictory properties at the same time. For example, in an example of English such as „Mary knows the man who killed (*kill) the boss last week‟, the verb kill must be finite, since verbs must be finite in the RCs. If the verb is nonfinite, it is ungrammatical. Then, how do these properties appear in the natural languages? Before its discussion, I mention RC types. Typologically, RCs may be categorized2 as follows: (7). (7) Types of relative clauses in this paper (a) i. Subordination -- embedded Externally headed 1. Prenominal3 2. Postnominal ii. 2 Internally headed4 De Vries (2002) divided RCs into six types: prenominal relatives, prenominal participial relatives, postnominal relatives, postnominal participial relatives, circumnominal relatives, and correlatives. There is little documentation of participials among signed languages, I would leave these two types of RCs aside in my current discussion. 3 Although both prenominal and postnominal are called adnominal (Lehmann 1986), the difference between adnominal RCs and other types of RCs needs further research. 4 Internally headed RCs are also called circumnominal relative clauses. 6 (b) Subordination -- adjunction (c) Subordination -- correlative5 Then, let us discuss some examples, each of which includes one of RC types in (7), from natural languages, in order to observe which property in (6) is involved. The first example from Walbiri shows the property „A RC contains a finite verb‟ (6a): (8). (8) njuntulu-ḷu kutja-Ø -npa wawiri [you-ERG kangaroo spear-PAST] SR-AUX-SBJ.2 ŋula lapi-ṇa pura-mi DEM FUT-SBJ pantu-ṇu, ŋatyulu-ḷu. cook-PRES I-ERG „The kangaroo that you speared, I will cook.‟ (Lehmann 1986: 6 citing from Hale 1976: 79) In (8), inside the RC „njuntulu-ḷu kutja-Ø -npa wawiri pantu-ṇu‟ (that you speared), we can find the finite verb „pantu-ṇu’ (speared), which is marked by the past suffix -ṇu. The type of this RC is called a correlative, which is preposed (placed to the left peripheral edge), with its head noun „wawiri‟ (kangaroo) inside6, and the anaphoric relationship is established between the noun „wawiri‟ in the RC and the demonstrative „ŋula‟in the main clause. Next, in Quechua, we can find the second property „The verb of a RC assumes a distinctive nonfinite form.‟ (6b) such as (9). (9) [[hamu-q] runa-ta] come-AGT man-ACC riqai-ni know-I „I know the man who comes.‟ (Muysken 1994: 2813) 5 Correlatives can be analyzed as internally headed RCs as well as adjunct RCs, but in this paper, it is classified into one type, because they are „very common‟ among adjunct RCs (Lehmann 1986). 6 Literally, correlatives are internally headed RCs (IHRCs), but they are so different from other IHRCs that I classified correlatives into one group. 7 In (9), in the RC „hamu-q‟, the verb „hamu’ (come) is non-finite, although suffixed by the agentive marker „-q’, since it is not changed in form by tense, person, or number, while the verb of the matrix clause7 „riqai-ni’ (I know) is finite with the person marker „-ni’ (I). This RC can be called an externally headed prenominal RC, since the head noun „runa‟ (man) is outside the RC which precedes its head noun. The third property „A RC contains a pronoun coreferential with a noun that immediately precedes (or follows) the RC.‟ (6c) can be observed in Modern Greek such as (10). (10) o anthropos pu (ton) skotosa the man [SR him killed :1.SG] „the man that I killed‟ (Lehmann 1986 : 7) In (10), the pronoun „ton‟ (him) co-refers to a man with the head noun „anthropos‟ (man) immediately preceding the RC „pu (ton) skotosa‟ (that I killed). The type of this RC is called a postnominal RC, because the RC follows the head noun. The fourth property (i.e. 6d), that „No nominal in the RC is coreferential with a preceding (or following) noun‟, can be identified in Japanese such as (11). (11) Satoo-sensei-ga [gakusei-ga kaita] ronbun-o Prof.Sato-NOM yondeiru. student-NOM wrote article-ACC is-reading „Prof. Sato is reading the article that the student wrote.‟ (Tsujimura 2007: 301) 7 I use the distinction between the term „main clause‟ and „matrix clause‟ by Velupillai (2012: 315): „A matrix clause is a main (or superordinate) clause minus its subordinate clause part.‟ While a main clause is always independent, a matrix clause does not always be independent. For example, in the sentence Mary told John that she saw him, the matrix clause Mary told John is not independent without a subordinate clause that she saw him. 8 In (11), the RC „gakusei-ga kaita’8 (the student wrote) is referred to as a prenominal RC, since it precedes its head noun „ronbun‟ (article). We cannot find any nominal inside the RC that is coreferential with a following head noun. Since the head noun is outside the RC, the clause is an externally headed RC (EHRC). We can also find the fifth property in (11), which is „An RC together with a nominal expression forms a noun phrase (NP) constituent.‟ (6e). That is, the RC plus the nominal forms an NP „gakusei-ga kaita ronbun ‟, so that the NP can be attached with an accusative marker „-o’. The sixth property „A RC is the sole constituent of a NP.‟ (6f) is represented in Japanese such as (12). (12) John-wa [RC [ringo-ga John-TOP sara-no ue-ni at-ta] -no] -o toriage-ta apple-NOM plate-GEN upper.area-at be-PAST -NMR -ACC pick.up-PAST „John picked up an apple, which was on a plate.‟ (Hendery 2012: 23 citing from Fuji 2010: 47) In (12), since the RC „ringo-ga sara-no ue-ni at-ta-no‟ (An apple was on a plate) has the head noun „ringo‟ inside itself, it is an internally headed relative clause (IHRC). It is also nominalized by „no‟, thus becoming the object, the internal argument of the matrix verb „toriage‟ (pick up). In addition, the seventh property, „The internal structure of a RC is indistinguishable from that of (some) nonrelative clauses.‟ (i.e. 6j), can be observed in this example. That is, this type of RCs has the same structure inside as that of a Japanese simple sentence „ringo-ga sara-no ue-ni at-ta‟ (An apple was on a plate). The eighth property is „A RC is not a constituent of an NP.‟ (6g), and it can be observed in Hindi such as (13). (13) [Jo larkaa mere paas rhtaa hai], vah meraa chotaa bjaaii hai REL boy me near living is he my small brother is „The boy who lives near me is my small brother.‟ (Alexiadou et al. 2000: 21) 8 The verb „kak-‟ (write) and the past suffix -ta turns phonologically into „kai-ta‟ (wrote). 9 In (13), the RC „Jo larkaa mere paas rhtaa hai‟ (which boy lives near me) is a correlative, since it is preposed and the anaphoric relationship between the RC and the main clause is marked by the relative marker „Jo‟ in the correlative and the pronoun „vah‟ in the main clause. Also, the correlative does not constitute an NP argument of any given verb. It is followed by the pronoun „vah‟ (he). However, the RC has a link with the matrix sentence in that the NP „jo larkaa‟ co-refers to the same boy with „vah‟. The ninth property, that is „A RC contains a marker that is linked by cooccuance with a nominal marker outside the clause‟ (i.e. 6i), can also illustrated by (13). In this sentence, the RC contains a relative marker „jo‟ (which) that co-refers with a nominal marker „vah‟ (he), that is a pronoun in this case, outside the clause. The last property, „A RC begins (or ends) with a distinctive marker.‟ (i.e. 6h), is observed in one dialect of Ojibwe (Downing 1978). Compare a simple declarative sentence (14a) and the same clause embedded as an RC (14b). (14) (a) Ikwe ogī wābamān ininiwan bijīnāgō woman did see man yesterday „The woman saw the man yesterday.‟ (b) Ikwe gā wābamān ininiwan bijīnāgō woman [did:REL see man yesterday] „the woman who saw the man yesterday‟ (Downing 1978: 386) In (14b), the RC „gā wābamān ininiwan bijīnāgō‟ (who saw the man yesterday) is a postnominal RC, because the head noun „Ikwe’ is followed by the RC. The only difference is the ogī/gā between these two examples. Both include the identical past marker „-gī’. However, while it appears as „ogī‟9 in (14a), it occurs „gā‟ with the vowel change in (14b). According to Downing (1978: 386), in this dialect, this vowel change is „generalized as a present and past tense relative marker‟. Thus, different RCs show different RC properties proposed by Downing (1978). Lehmann (1986), in his study on the typology of RCs, proposed that a relative construction is formed by a „bundle of operations‟: (15). 9 Downing (1978) did not mention the origin of „o‟. 10 (15) A bundle of operations forms a relative construction (Lehmann 1986: 4) Subordination – nominalization10, Anaphora – empty place formation, Attribution/head formation. The first operation means how much an RC is nominalized as a subordinate clause. For example, if a RC is preposed or postposed, Lehmann (1986) analyzed it is not embedded and least nominalized such as (16) and (17) (the data come from Walbiri and Ancient Greek, respectively cited by Lehmann (1986)). (16) njuntulu-ḷu kutja-Ø -npa wawiri [you-ERG kangaroo spear-PAST] SR-AUX-SBJ.2 ŋula lapi-ṇa pura-mi DEM FUT-SBJ pantu-ṇu, ŋatyulu-ḷu. cook-PRES I-ERG „The kangaroo that you speared, I will cook.‟ (=(8)) (17) autíka d‟ égnō oulēn, at.once however recognized scar:ACC.SG.F tēn poté min sŭs ēlase [DEM:ACC.SG.F once him boar:NOM.SG stroke:3.SG] „At once she recognized the scar which once a boar had struck him.‟ (Lehmann 1986: 6 citing from Od. 19, 392f) In (16), the RC „njuntulu-ḷu kutja-Ø -npa wawiri pantu-ṇu‟ (the kangaroo that you speared) is preposed, while in (17), the RC „tēn poté min sŭs ēlase‟ (which once a boar had struck him) is postposed11. Both the RCs do not show any marking of nominalization. On the other hand, if the RC is externally headed (prenominal or postnominal), it is nominalized to a large degree (Lehmann 1986). For example, Turkish prenominal RCs are analyzed as nominalized: (18). 10 Nominalization will be discussed more in section 1.2.2. 11 According to Lehmann (1986: 3), both preposed and postposed RCs are not a constituent of the main clause, not inside (embedded in) it, and always at its margin. Their original positions were not discussed. 11 (18) iç-in-den ev cık-tıg-ım-ız [interior-POSS.3-ABL leave-NR-POSS.1-PL] house „house which we left‟ (Lehmann 1986: 6 citing from Andrews 1975: 54) In (18), the RC „iç-in-den cık-tıg-ım-ız‟ (which we left) is prenominal, because it is followed by its head noun „ev‟ (house). We can find the RC is nominalized by the nominalizer „-tıg‟. Furthermore, the possessive suffix „-ım‟ and the plural suffix „-ız‟ at the end of the RC indicate the subject of the RC and its number and link the RC to the head noun. As for postnominal RCs, we can find an example in Yaqui such as (19). (19) lni-meɁe hu-me haam-im in DEM-PL [POSS.1 3.PL-about speak-REAL-REL] DET-PL woman-PL ame-t noka-k-aɁu „Those are the woman that I spoke about.‟ (Lehmann 1986: 7 citing from Lindenfeld 1973) In (19), since the RC „in ame-t noka-k-aɁu’ (that I spoke about) follows its head noun „haamim‟ (women), it is postnominal. Inside the RC, we cannot find such an obvious nominalizer as (18), while the possessive „in‟ (our) represents the subject of the RC in the similar way to (18). Let us return to the bundle in (15). The second operation (anaphora – empty place formation) accounts for how to relate the RC to the head. Languages around the world employ various strategies for that. For example, the most common way is to leave an empty space in the RC such as (20). (20) The book [you are consulting] is up-to-date. (Lehmann 2003: 460; partly modified) In (20), the transitive verb consulting within the RC „you are consulting‟ lacks the object NP inside the clause, but the RC finds the object NP „The book‟, that is the head noun, outside. Thus, the anaphoric relationship is established between the head and the RC. The next common strategy is to use an anaphoric pronoun to refer to the head. Persian provides us with examples such as (21) and (22): 12 (21) Mard-i [ke u man-IND SR va zan-aš diruz āmad-and] mi-rav-ad. he and wife-POSS.3.SG yesterday came-3.PL IMPF-leave-3.SG (lit.) „The man that he and his wife came yesterday is leaving.‟ (Lehmann 2003: 460) (22) Kārxāne-yi ke dar ān kār mi-kard-am baste šode factory-IND [SR in closed become is D3 work IMPF-did-1.SG] ast. „The factory in which I used to work has been closed.‟ (Lehmann 1986: 3 citing from Amin-Madani & Lutz 1972: 175) In (21), the pronoun „u’ (he) in the RC „ke u va zan-aš diruz āmad-and’ (that he and his wife came yesterday) refers to the head NP „mard-i’ (the man). In (22), on the other hand, the demonstrative element „ān‟ in the RC „ke dar ān kār mi-kard-am‟ (in it I used to work) refers to the head NP „Kārxāne-yi‟ (the factory). In both examples, the anaphoric elements agree with the head NPs in person12. Thus, we can observe the connection between the head and the RC. The last operation of the bundle in (15) is attribution/head formation. It is to combine the head with the RC together. In fact, it is a matter of the position the head takes: inside (internally headed) or outside (externally headed), before (prenominal) or after (postnominal) -- the head. The positions of the head will be discussed later. Among the studies on RCs, De Vries (2002) redefined RCs as follows: (23) Defining properties of relative constructions: (De Vries 2002: 14) (a) A relative clause is subordinated. (b) A relative clause is connected to the surrounding materials by a pivot constituent. While (23a) follows the previous researchers‟ point of view, (23b) is very different from the previous analyses mentioned above. In order to understand this definition, compare the following examples: (24)–(26). 12 Lehmann (1986) did not mention whether the demonstrative element „ān‟ is singular or plural. Therefore, it is unclear whether the anaphoric elements in the RC agree with the head nouns in number only through (21) and (22). 13 (24) The mouse [that I caught yesterday] was hungry. (De Vries 2002: 15) (25) The mouse [that I caught yesterday] (26) The mouse was hungry. The example (24) is a sentence embedding the RC „that I caught yesterday.‟ We can analyze the sentence in this way. The NP in (25) (the head NP plus the RC) and the matrix sentence in (26) (the subject NP plus the predicate) are connected together using the head NP „the mouse‟ as a pivot constituent. As a result, the RC is connected to the surrounding materials (was hungry). I illustrate this idea with Figure 1. Figure 1 „The mouse that I caught yesterday was hungry.‟ In sum, an RC is a subordinate clause to be related to the matrix clause through the head NP. At the same time, the RC modifies the head NP. 1.2. Typology of relative clauses So far, we have taken a brief look at the various definitions of RCs as well as their characteristic properties. Technically, all clauses that fit the definitions are considered as RCs. As mentioned above, RCs may be typologically categorized as follows: (27). (27) Types of relative clauses in this paper (=(7)) (a) Subordination -- embedded i. Externally headed 1. Prenominal 2. Postnominal 14 ii. Internally headed (b) Subordination -- adjunction (c) Subordination -- correlative In the next section, I will discuss why RCs are subordinate clauses. 1.2.1. Relative clauses are subordinate clauses As mentioned above, RCs are subordinate clauses. Then, what is a subordinate clause? A subordinate clause can only occur as part of a larger clause, so it cannot stand alone. A subordinate may be a complement of a matrix verb. Compare the two sentences: (28). (28) (a) Angela reported [the result]. (b) Angela reported [that the team had won]. (Vincent 1994: 4391) In both examples, the verb report is used. This predicate takes two arguments: the source/agent (where the report comes from or who delivers the report) and the theme (what is reported). How many and what kind of arguments are required depends on each predicate13. The verb report takes an NP (e.g., the result in (28a)) or a CP (e.g., „that the team had won‟ in (28b). In other words, the clause „the team had won‟ is a subordinate clause within the sentence, with the complementizer „that‟, theoretically the head of CP, preceding the clause. In many languages, subordinate clauses have the following properties (Vincent 1994): (29). (29) Properties of subordinate clauses Subordinating conjunctions introducing a subordinate clause 13 This may also be an interesting topic for researchers. For example, in Japanese, the adjective „takai‟ (high) has multiple meanings referring to such as position, level, price (meaning „expensive‟), and sound (in high frequency), as well as in the idiom „me-ga takai‟ (lit. the eye is high) (to have a sharp eye (e.g., for antiques)). In JSL, EXPENSIVE is frequently used for price but also for sense organs such as eye, nose, tongue, ear, and skin: „YOU NOSE EXPENSIVE!‟ (lit. „Your nose is expensive!‟ meaning „Your nose is very sensitive!‟ 15 Different word order from that of the main clause Special verb form Nonfinite forms such as infinitives In English, for instance, subordinating conjunctions like the complementizer that in (28a) and adverbial conjunctions (e.g., when, because) are used, as in (30) and (31). (30) [When the jar was full], he turned the water off. (31) I couldn‟t feel angry [because I liked him too much]. (Sinclair (Ed.) 2012: 515–516) As for the different word order, in German, the finite verbs always occur in the second position in the main clause but in the final position of the subordinate clause, as in (32a) and (32b). (32) (a) Joachim liebt Joachim love:3.SG.IND.PRES das Mädchen. DA:NEU.SG.ACC girl „Joachim loves the girl.‟ (b) Ich glaube, I think:1.SG.INDIC.PRES daß Joachim das Mädchen liebt. CMR Joachim D:NEU.SG.ACC girl love:3.SG.INDIC.PRES „I think that Joachim loves the girl.‟ (Vincent 1994: 4392) Note the position of the finite verb liebt. In (32a), the simple declarative sentence, the finite verb liebt occurs in the second position. On the other hand, in (32b), the complex sentence, the finite verb liebt occurs at the end of the subordinate clause „daß Joachim das Mädchen liebt‟, while the finite verb glaube still occurs in the second position of the main clause. Next, let us look at how verbs appear in a special form when they are used in the subordinate clauses in Italian. Before looking at its subordinate clauses, I will mention its declarative sentence as in (33). 16 (33) Giorgio non capisce. Giorgio NEG understand:3.SG.PRES.IND „Giorgio does not understand.‟ (Vincent 1994: 4392) The verb capire „understand‟ is used in the indicative form capisce, when it is used in the declarative sentence. However, the verb is in the subjunctive form capisca, when it is used in the subordinate clauses as in (34) and (35). (34) benché Giorgio non capisca although Giorgio NEG understand:3.SG.PRES.SUBJ „although Giorgio does not understand‟ (35) temo che Giorgio non capisca fear:1.SG.PRES.IND CMR Giorgio NEG understand:3.SG.PRES.SUBJ „I fear Giorgio does not understand.‟ (Vincent 1994: 4392) Example (34) is adverbial, and it is not a complement clause, but an adjunct clause. On the other hand, example (35) is the complement clause to the verb „temo‟ (I fear). Lastly, verb nonfinite forms such as infinitives can be observed in subordinate clauses such as (36). (36) Credo [Caesarem trans flumen transisse] believe.1.SG [Caesar-ACC across river cross-PERF.INF] „I believe Caesar has crossed the river.‟ (Vincent 1994: 4394) In (36), we can see a subordinate clause (complement clause), and it has the subject Caesarem in the accusative and the verb transisse in the infinitive form. Having gone discussed the properties of subordination, we need to ask in what way an RC is considered as a subordinate clause to the matrix clause? Lehmann (1986) pointed out two processes: embedding and adjoining (adjunction). Before entering the discussion of the processes, I will first discuss two important terms related to embedding: nominalizer and nominalization. 17 1.2.2. Nominalizer and nominalization Brown & Miller (2013) defined the term nominal as a „constituent that functions like a noun phrase‟, and nominalization as „the process by which clauses or verb phrases are converted into constituents functioning like noun phrases‟. But how does it actually occur? According to the web site of Oxford English Dictionary (accessed 1/4/2015), the term nominalizer is defined as „an affix which may be added to a word or phrase so as to convert it into a nominal‟. For example, there are two nominalizer suffixes in English. One is the verb suffix -er (hunt/hunter), while the other is the adjective suffix -ness (white/whiteness). However, among languages around the world, there is a variety of nominalization, and we will see some examples. According to Muysken (1994), nominalized elements are noun phrases which have something verbal. He mentioned two types of nominalization. One type is to turn a verb phrase into a noun phrase (e.g., in English, to apply for a job application for a job). The other is to change a clause into a noun phrase with the same marker as that is used for verbnoun changing (e.g., in English, „I apply for a job.‟ „my application for a job‟). However, cross-linguistically, there are cases in between, so it is not easy to draw the line between them. First, let us look at nominalization morphology. If nominalization happens at the morphological level, it is through the process of derivation. Among languages, the two main types of morphological nominalization are agentive (e.g., hunt/hunter in English) and action nominals (e.g., sell/sale in English) (Muysken 1994). In Quechua, some suffixes nominalize verbs in different ways: (37)–(39). (37) suwa-q stealing-AGT „someone in the act of stealing; thief‟ (38) puqlla-na playing-INST „playing (some time); playing, toy‟ (39) puñu-na sleeping-LOC „sleeping (some time); bed‟ (Muysken 1994: 2812) 18 In (37), the agentive suffix -q converts the verb „suwa‟(stealing) into a noun „thief‟, while in (38), the instrumental suffix -na nominalizes the verb „puqulla‟ (playing), and the noun means playing (n.) or a toy. Similarly in (39), the locative -na changes the verb „puñu‟ (sleeping) into a noun meaning „bed‟. In many languages, an agentive nominal can modify another nominal. If a verb is nominalized by an agentive marker and modifies another nominal, we can interpret the agentive nominal as an RC and the modified nominal as its head noun (Muysken 1994). Let us see an example from Quechua: (40). (40) [[hamu-q] runa-ta] come-AGT man-ACC riqai-ni know-I „I know the man who comes.‟ (=(9)) In this example, the agentive marker -q attaches to the verb „hamu‟ (come) in the process of nominalization. The nominalized verb (nominal) modifies the following nominal „runa‟ (man). As a result, the whole phrase „hamu-q runa-ta‟ (the man who comes) is an NP, and it is an object of the verb „riqai-ni‟ (I know). How deeply the clause is nominalized varies. Compare these English examples: (41) and (42). (41) They had a strong disagreement about capital punishment. (42) John‟s disagreeing so strongly about capital punishment has not made things any easier. (Muysken 1994: 2814) Both sentences show nominalization of the verb disagree, but the ways they are nominalized are different. In (41), the verb is nominalized with a suffix -ment, while in (42), the verb is nominalized with the gerund suffix -ing. This difference affects how each nominal is modified. For the former, the adjective strong can be used. On the other hand, the adverb phrase „so strongly‟ is used for the latter. Next, Turkish shows a different but similar phenomenon as in (43). (43) [Ahmet-in ev-i yap-tığ-ın-ı] bil-i-yor-um Ahmet-GEN house-ACC build-NMR-3-ACC know-PRES-3-1 19 „I know that Ahmet has built the house.‟ (Muysken 1994: 2814) In (43), since the verb yap „build‟ is nominalized by the nominalizer -tığ, it is assigned the accusative marker -ı by the verb „bil-i-yor-um‟ (I know). However, the nominal „yap-tığ-ın-ı‟ (that s/he has built) also assigns the accusative case „-ı‟ to its object „ev‟ (house). On the contrary, the nominal „yap-tığ-ın-ı‟ takes the genitive-marked noun „Ahmet-in‟ (Ahmet‟s) as its subject. Then, what is the relation between nominalization and RCs? Lehmann (1986) proposes that the positions of the RC (prenominal or postnominal) have something to do with how much the clause is nominalized. If the RC is prenominal, the clause is well nominalized and the subject of the RC is attached with a genitive (possessive) affix, as shown by the Japanese and Turkish shows us examples: (44) and (45), respectively. (44) Kore-wa [ano hito-no D1-TOP D3 kai-ta] hon desu. person-GEN write-PAST book COP „This is the book which that man wrote.‟ (Lehmann 1986:7 citing from Kuno 1973) (45) [Orhan-ın gör-düğ-ü] adam cık-tı Orhan-GEN see-NMR-POSS.3 man leave-PAST „The man Orphan saw left.‟ (Lehmann 1986: 3 citing from Andrew 1975: 152) In (44), the verb „kai-‟ (to write) is affixed by the past marker, not by the nominalizer. However, by being preceded by the genitive-marked nominal „hito-no‟ (that man), the verb works as if it was a nominal. In (45), the clause is nominalized by the nominalizer „-düğ‟, and its subject „Orhan‟ attached by the possessive affix „-ın‟. In addition, the RC has the thirdperson possessive marker „-ü‟ at its end to link the RC to the head noun „adam‟ that the RC modifies. In sign languages, nominalization is represented in different ways. In American Sign Language (ASL), for example, reduplication is frequently used to change a verb into a noun (Valli et al. 2005). The language also has an agentive affixes -ᴇʀ to change a verb into a person who does it as an occupation as in ᴛᴇᴀᴄʜᴇʀ (<ᴛᴇᴀᴄʜ + -ᴇʀ). Australian Sign Language (Auslan) has the similar nominalization strategy. Reduplication, in many cases, differentiates 20 between a verb (no reduplication) and a noun (one reduplication), depending on the context, such as OPEN-DOOR vs. DOOR (Johnston & Schembri 2007: 27). So far, we have looked at various ways for nominalization. Then, I will mention the strongly nominalized RCs (embedded RCs) and others (adjunct RCs). 1.2.3. Embedded or adjunction First, let us see an example of embedded RCs in Turkish: (46). (46) Orhan-ın gör-düğ-ü [[Orphan-GEN see-NMR-POSS.3] adam cık-tı man] leave-PAST „The man Orphan saw left.‟ (= (45)) In (46), the verb „gör‟ (see) is nominalized by a nominalizer „-düğ‟. The genitive suffix „-ın‟ indicates that the noun „Orhan‟ is the subject of the verbal noun „gör-düğ‟ and links „Orhan‟ with the verbal noun „gör-düğ‟. The possessive marker „-ü‟, whose person (third-person) agrees with that of the subject of the verbal noun, is suffixed to the verbal noun and connects the verbal noun to the following noun „adam‟. As a result, the whole chunk „Orhan-ın gördüğ-ü adam‟ becomes a NP. Next, an adjunct RC is also a subordinate clause. The language Bambara shows us one of its examples: (47)14. (47) [n ye I COMPL ty m n ye], man REL be fn f re. saw D3 IMPF cloth:DEF sell „The man I saw (, he) sells the cloth.‟ (Lehmann 1986: 2 citing from Bird 1968: 43) In (47), there are two clauses. One is „n e t m n ye‟ (I saw which man), which is an RC. The other is „ be f n f re’, which is the matrix clause. The RC is an adjunct to the left side of the matrix clause, and they form a whole sentence. They are related to each other, because the anaphoric relationship is established between the relative pronoun „m n’ and the demonstrative „ ‟ (he). 14 The RC can also be analyzed as a correlative. See 1.2.6. 21 On the other hand, an RC may be an adjunct to the right side of the matrix clause such as (48) (from Ancient Greek). (48) autíka d‟ égnō oulēn, at.once however recognized scar:ACC.SG.F tēn poté min sŭs ēlase [DEM:ACC.SG.F once him boar:NOM.SG stroke:3.SG] „At once she recognized the scar which once a boar had struck him.‟ (=(17)) In (48), the RC „tēn poté min sŭs ēlase‟ (which once a boar had struck him)‟ is an adjunct to the right side of the matrix clause. We can observe the anaphoric relationship between the head noun „oulēn‟ and the demonstrative „tēn‟ because they agree in number (singular) and gender (feminine). 1.2.4. Externally headed: prenominal or postnominal If an RC has a head, in other words, if an RC is headed, one needs to decide whether the head is outside the RC or not. If the head is outside an RC, the clause is called an externally headed RC (EHRC). EHRCs are further divided into two types, according to whether an RC occurs before or after the head: prenominal and postnominal, respectively15. First, let us see an example of prenominal RCs from Japanese: (49). (49) Satoo-sensei-ga [gakusei-ga Prof.Sato-NOM kaita] ronbun-o yondeiru. student-NOM wrote article-ACC is-reading „Prof. Sato is reading the article that the student wrote.‟ (=(11)) In (49), the RC „gakusei-ga kaita‟ (that the student wrote) precedes and modifies the head noun „ronbun-o‟ (the article). The verb „kak-‟ (write) and the past suffix „-ta‟ turns phonologically into „kai-ta‟ (wrote). The clause lacks any anaphoric pronouns such as relative pronouns, and this lack of an anaphor is one of the properties that most prenominal RCs have (Lehmann 1986). 15 Prenominal and postnominal RCs are also referred to as RelN and NRel, respectively (Velupillai 2012). 22 Next, let us see an example of postnominal RCs from Modern Greek: (50). (50) o onthropos [pu (ton) skotosa] the man SR him killed:1.SG „the man that I killed‟ (Lehmann 1986: 7) In (50), the RC „pu (ton) skotosa‟ (that I killed) comes after the head „onthropos‟ (man). The subordinator „pu‟ (that) comes at the beginning of the RC. The pronoun „ton‟ (him) is anaphoric to the head, but it is optional. In fact, most of postnominal RCs have zero anaphora (i.e. the pronoun can be omitted) (Lehmann 1986). Typological studies have shown interesting distribution of RCs. For example, most languages have EHRCs (Velupillai 2012). According to Dryer (2013a), 720 out of 824 languages around the world have EHRCs (prenominal: 141; postnominal: 579). In addition, VO languages tend to have postnominal RCs (Velupillai 2012). According to Dryer (2013b), out of 829 languages around the world, 416 languages with the VO word order have postnominal RCs, while only five have the VO order and prenominal RCs. On the contrary, OV languages have no such strong correlation (out of 245 OV languages in the same data; prenominal: 132; postnominal: 113). These figures are illustrated as below Table 1 and Table 2. Num. EHRC 720 Prenominal Postnominal Others Total 141 579 104 824 Table 1 Externally headed relative clauses (Dryer 2013a) 23 Num. VO 421 Prenominal Postinomial 5 416 OV 245 Prenominal Postinomial 132 113 Others Total 163 829 Table 2 VO/OV languages and postnominal relative clauses (Dryer 2013b) In other words, RCs in most languages are externally headed. If a language has a basic VO order, its RCs are postnominal in most cases. However, other researchers observe the distribution of RCs differently. According to Malmkjær (1991: 283), for instance, „Some language universals may be explained by reference to the processing demands placed on language users by, for instance, memory constraints and by the relative ease or difficulty involved in processing certain structures in comprehension and production‟. Therefore, languages generally avoid centre embedding, and it is more difficult to process centreembedded RCs than to process left-peripheral RCs or right-peripheral RCs. Hence, Malmkjær (1991: 283) argued that most SOV languages have prenominal RCs; that is, if an RC is postnominal, the RC would be embedded and interrupt the process of the utterance. For example, let us look at an example from Japanese: (51). (51) Satoo-sensei-ga [gakusei-ga Prof.Sato-NOM kaita] ronbun-o yondeiru. student-NOM wrote article-ACC is-reading „Prof. Sato is reading the article that the student wrote.‟ (=(11)) If the RC of (51) „gakusei-ga kaita‟ were postnominal such as „*Satoo-sensei-ga ronbun-o [gakusei-ga kaita] yondeiru‟, the addressee‟s process of the sentence would be interrupted by the RC, and the addressee would be confused about who is the subject of the verb „yondeiru‟ at the end of the sentence. Thus, in order to confirm the more precise distribution of RCs, further research needs to be conducted on the RC distribution of much larger number of languages around the world. 24 1.2.5. Internally headed relative clauses For internally-headed RCs, the head is inside the RC. Internally headed RCs (IHRCs) are also called circumnominal RCs (Lehmann 1986). They are found in some languages such as Japanese: (52). (52) John-wa [RC [ringo-ga John-TOP sara-no ue-ni at-ta] -no] -o toriage-ta apple-NOM plate-GEN upper.area-at be-PAST -NMR -ACC pick.up-PAST „John picked up an apple, which was on a plate.‟ (=(12)) In (52), the head „ringo-ga‟ (apple) is case-marked as nominative, not by the verb of the matrix sentence („toriage-ta‟ (picked up)), but by the verb of the RC („at-ta’ (be-PAST)). Therefore, the head is inside the RC, hence the clause „ringo-ga sara-no ue-ni atta-no‟ (which an apple was on a plate) is an IHRC. The whole RC is nominalized by the nominalizing particle „-no‟ (Hendery 2012: 23) and becomes an argument of the main verb „toriage-ta‟ due to the accusative marker „-o‟. 1.2.6. Correlatives The last type of RCs is correlatives. According to Lipták (2009: 1), „in a correlative relativization strategy a left-peripheral relative clause is linked to a (possibly phonetically unrealized) nominal correlate in the clause that follows the relative clause‟. In addition, their construction has relative and/or demonstrative pronouns in the RC and the main clause to „mark the anaphoric relationship‟ (Lehmann 1986) between the head and the personal or demonstrative pronoun in the main clause16. A typical correlative construction is illustrated as (53). (53) [correlative clause … relative phrase …][main clause … correlate …] (Lipták 2009: 2) What does a correlative construction look like? We will look at some example from natural languages. 16 The phrase of the personal or demonstrative pronoun is especially called the correlate (De Vries 2002). 25 First, the language Bambara shows us one of its examples: (54). (54) [n ye I COMPL ty m n ye], man REL be fn f re. saw D3 IMPF cloth:DEF sell „The man I saw (, he) sells the cloth.‟ (=(47)) In (54), the RC „n e t m n ye‟ (I saw which man) is located on the left periphery of the sentence. There is the head noun „t ‟ (man) inside the RC. The anaphoric relationship is established between the relative pronoun „m n’ in the RC and the demonstrative „ ‟ (he) in the matrix clause. Next, let us see an example from Hindi: (55). (55) [Jo larkaa mere paas rhtaa hai], vah meraa chotaa bjaaii hai REL boy me near living is he my small brother is „The boy who lives near me is my small brother.‟ (=(13)) In (55), the head „larkaa’ (boy) is inside the correlative „Jo larkaa mere paas rhtaa hai‟ (which boy lives near me), and the clause is left-adjoined to the matrix clause „vah meraa chotaa bjaaii hai‟ (he is my brother). The anaphoric relationship is maintained between the relative pronoun „jo‟ (which) next to the head and the pronoun „vah‟ (he) in the main clause. Interestingly, in the same language, the head of correlatives may occur twice: in the correlative and the main clause again as in (56). (56) [jo laRkii khaRii REL girl hai], vo laRkii lambii hai standing is that girl tall is „The girl who is standing is tall‟ (Lipták 2009: 3) In (56), the head noun „laRkii’ (girl) appears twice: inside the correlative „jo laRkii khaRii hai‟ (which girl is standing) and in the main clause „vo laRkii lambii hai‟ (that girl is tall). Like (55), the anaphoric relationship is built by the relative pronoun „jo‟ (which) and the demonstrative „vo‟ (that) in the main clause. Velupillai (2012) mentions a unique type of RC. It is a paratactic relative clause. It is like a simple sentence, but it is unmarked. Also, it is only very loosely connected with the 26 main clause. The language Amele in Papua New Guinea has an example of that kind of RCs: (57). (57) mel mala heje on [boy chicken illicit take.3SG.SUBJ.REM.PAST] ((mel) eu) busali boy nu-i-a that run.away go-3SG.SUBJ-TOD.PAST17 „The boy [that stole the chicken] ran away.‟ (Comrie & Kuteva 2013) This type of RCs, unlike the correlatives, has no relative pronouns inside, while like correlatives, there is a demonstrative „eu‟ in the matrix clause. However, (57) is analyzed not as two independent clauses, but as one sentence, because of intonation (Velupillai 2012: 327). To sum, we can divide various RCs into two large groups: embedding and adjunction. Embedded RCs are further divided into three types. If an RC precedes the head, it is a prenominal RC and strongly nominalized. If an RC follows the head, it is a postnominal RC and not nominalized so much as prenominal ones. If a RC has the head inside, it is an internally headed relative clause (IHRC). If an RC is not a constituent with the head but an adjunct to the matrix clause, it is an adjunct RC. There is one more type. If a RC has the head inside and the matrix clause has a determiner or something to corefer with the head, it is a correlative. Thus, there is a wide range of RC types in languages around the world. Signed languages, as natural languages, are also expected to display such diversity. What do we know about the signed language in Japan so far? Before discussing this issue, I will summarize the findings based on previous studies. 1.3. Relative clauses in sign languages: previous studies It is very difficult to identify the basic word order of a sign language since the order is quite free. However, according to Leeson & Saeed (2012), ASL and Brazilian Sign Language (LSB) have a basic SVO order, while a basic SOV order is recognized in Italian Sign Language (LIS) and German Sign Language (DGS). JSL has a basic SOV order (Yonekawa 17 TOD.PAST=today‟s past tense 27 1984; Ichida 2005c; Oka & Akahori 2011; Kimura & Ichida 2014). In this section, we will take a brief look at what RC types have been documented in sign languages. If based on word order typology and RC distribution illustrated in Table 2, we would expect to find postnominal RCs in SVO sign languages. First, let us see the RCs of SVO sign languages. Nonmanual (NM) markers play a role in most of them. In ASL, an RC is marked by NM features such as „raised eyebrows, a backward tilt of the head, and contraction of the muscles that raise both the cheeks and the upper lip‟ (Liddell 2003: 54) as in (58). r (58) RECENTLY DOG CHASE CAT COME HOME „The dog which recently chased the cat came home.‟ (Liddell 1980: 136) In (58), the relative NMs or marker co-occurs with the whole RC (marked by the line above), which contains the head, DOG. Therefore, this type of RC is interpreted as an IHRC. On the other hand, Liddell (1980) reported ASL also has EHRCs such as (59). r (59) 1ASK3 GIVE1 DOG [[URSULA KICK] THATc] „I asked him/her to give me the dog that Ursula kicked.‟ (Liddell 1980: 162) In (59), since the relative NMs do not co-occur with the head noun „DOG‟, the RC „URSULA KICK‟ is interpreted as an EHRC. Also, it is a postnominal RC because it follows the head noun. HKSL, according to Sze (2000), has a basic SVO order and has a similar NM marker. Tang & Lau (2012) proposes only brow raise as the relative marker as in (60). (60) br HEY! IX3 LIKE [IXi MALE EAT CHIP IXi] „Hey! She likes the man who is eating chips.‟ (Tang & Lau 2012: 359) 28 In (60), brow raise scopes over the whole RC „IXi MALE EAT CHIP IXi‟ (the man who is eating chips) including the head noun „MALE’. Therefore, the RC is an IHRC. In addition, the pointing IXi occurs both at the beginning and the end of the RC. Thus, the distribution of RCs in SVO sign languages does not seem to match that of spoken languages illustrated in Table 2. IHRCs may be more frequent in sign languages than in spoken languages. Further research on RCs in SVO sign languages will confirm that. So far, we have seen what types of RC markers are used in the sign languages that have a basic SVO word order. The next question is whether SOV sign languages have any different ways to sign RCs? For example, LIS is a SOV language, but the analysis of RCs has been controversial. On the one hand, Cecchetto et al. (2006) termed RCs as „PROREL constructions‟, where there consists of a PROREL clause (correlative) with a demonstrative marker which they call PROREL at the end of the clause as in (61). br (61) [BOY iCALL PRORELi] LEAVE DONE „A boy that called left.‟ (Cecchetto et al. 2006: 953) In (61), the RC „BOY iCALL PRORELi‟ (a boy that called) is left-adjoined, because LIS usually „avoid center-embedded structures‟ (Cecchetto et al. 2006: 973). For example, the translation of sentence (62), which includes a complement clause, would not be (63) (center-embedded), but (64) or (65) (staying peripheral). (62) Gianni says that Maria left. (63) *GIANNI MARIA LEAVE DONE SAY (64) MARIA LEAVE DONE GIANNI SAY (65) GIANNI SAY MARIA LEAVE DONE (Cecchetto et al. 2006: 973) Cecchetto et al. (2006) hypothesized that the avoidance of center embedding of clauses is due to lower memory span of deaf signers than that of hearing people. As for NMs, the brow raise scopes over the verb and the special sign PROREL. There are no anaphoric elements such a demonstrative in the matrix clause. According to Cecchetto et al. (2006), RCs in LIS are correlative constructions illustrated as (66). 29 (66) [PROREL CLAUSE … NPi … PRORELi][MAIN CLAUSE … PRONOUNi …] (Cecchetto et al. 2006: 952)) In fact, the construction in (66) matches a typical correlative construction illustrated in (53). Cecchetto et al. (2006) attributed the lack of anaphoric elements in the main clause to prodrop, because LIS is a pro-drop language. Thus, Cecchetto et al. (2006) concluded that LIS has a correlative with the demonstrative marker PROREL occurring at the end of the clause. On the other hand, Branchini & Donati (2009) proposed to term an RC as a „PE clause‟. It is internally headed and the RC is an embedded clause, with a PE marker (determiner), which consists of the unvoiced bilabial stop and downward shaking pointing, that appears at the end of the clause as in (67). (67) rel [TODAY MANi PIE BRING PEi] YESTERDAY (IXi) DANCE „The man that brought the pie today danced yesterday.‟ (Branchini & Donati 2009: 164) In (67), the RC „TODAY MANi PIE BRING PEi‟ (the man that brought the pie today) is embedded in the main clause, and there is the marker PE at the end of the clause. NMs glossed „rel‟ are made up of „raised eyebrow, tension of eyes and upper cheeks‟ (Branchini & Donati 2009: 163). The NMs scope over the whole RC. Thus, Cecchetto et al. (2006) and Branchini & Donati (2009) agreed that a pointing occurs at the end of the RC, but they disagreed over two points. One is how the relative NMs scope over the RC. The other is how the RC is related to the main clause. 1.4. Relative clauses in Japanese Sign Language: some previous analysis Before discussing the present study on RCs in JSL, we will take a look at what the previous study has to say about this construction. JSL is an indigenous sign language used by about 45000 Deaf people in Japan (Kanda 2009). To date, only a few papers have reported on RCs in JSL; however, many properties of RCs have not been thoroughly discussed. In a study by Yonekawa (1984), it was claimed that JSL has prenominal EHRCs18 as in (68). 18 Yonekawa (1984) did not refer to whether JSL has INRCs. 30 rel (68) I [PRICE HIGH] TOKYO GO „I go to Tokyo where prices are high.‟ (Yonekawa 1984: 211) In (68), a subordinate clause „PRICE HIGH‟, which is an RC, plus the head noun ‘TOKYO’ fills the argument position of the verb „GO‟ of the main clause. As for NMs, the special marker, glossed „rel‟ in (68), which consist of slight backward head tilts, brow raise, and lip pout, scope over from the beginning of the sentence to the end of the head noun. In addition, a short interval occurs between the head noun before the next sign. Furthermore, Yonekawa (1984) argued that JSL has postnominal RCs. It seems rare for a language to have both postnominal and prenominal RCs, because out of 172 languages around the world, only eleven, such as Finnish and Tagalog, have both (De Vries 2002). Let us see an example of JSL: (69). (69) I TOKYO [PRICE HIGH] GO (= (68)) (Yonekawa 1984: 213) In (69), according to Yonekawa (1984), the noun TOKYO is the head, and that the RC „PRICE HIGH‟ follows it. However, there are the constraints with this type of RCs, which is that not every noun can be the head of a postnominal RC. Non-concrete nouns (e.g., TIME, PLACE, MEANS, REASON) as in (70), locative nouns (e.g., SIDE, FRONT, UPPER-PART), and nouns stimulating senses (e.g., SMELL, SOUND) as in (71) can never be followed by RCs. (70) [SHE MARRIAGE REJECT] REASON KNOW LIKE „I want to know the reason she rejected the marriage.‟ (71) [FISH GRILL] SMELL HATE „I hate the smell when fish is grilled.‟ (Yonekawa 1984: 214) However, without NMs in the description of these examples, we cannot exclude other analyses that deny the existence of postnominal RCs. For example, (69) can be analyzed as „I 31 [TOKYO PRICE HIGH] GO‟; that is, a sentence including an IHRC meaning the same as (69). We can also analyze (70) as two conjoined sentences that mean: „She rejected the marriage, and I want to know its reason‟. Similarly, we can also interpret (71) as the complex sentence of an adverbial clause and a matrix clause meaning: „Whenever fishes are grilled, I hate the smell.‟ In order to confirm the postnominal RCs in JSL, further research needs to be conducted. Besides EHRCs, JSL has one more RC type: internally headed relative clauses (IHRCs) (h=hold of the head, shn=small head nod) as in (72). h (72) TANAKA (shn) [SUZUKI BENTO MAKE] EAT „Mr. Tanaka eats the bento lunch which Mr. Suzuki makes.‟ (Kimura & Ichida 2014: 37) In (72), we can observe two NMs: hold of the head (h) and small head nod (shn). Hold of the head is that the head is held at a specific position for a very short moment (Ichida 2005c: 93). According to Ichida (2014), the head movement, or the change of the head position, indicates phrase or syllable boundaries. Therefore, since the clause „SUZUKI BENTO MAKE‟ is sandwiched by the two head movements (h and shn), it is not analyzed as an independent clause but a subordinate clause embedded in the main clause. Recall that the basic word order of JSL is SOV. Example (72) yields an SOV order, so the whole clause „SUZUKI BENTO MAKE‟ must be a nominal (NP) to take the argument of the verb „EAT‟, representing something to eat. Hence, the noun „BENTO‟ inside the clause can be interpreted as the head. Therefore, the clause can be interpreted as an IHRC. To sum up, previous literature has made the following claims: JSL has EHRCs. They are prenominal and postnominal RCs, but the latter needs further research to be confirmed. Besides, JSL has IHRCs. Based on my knowledge, I think these observations are correct and postnominal RCs are non-restrictive or appositive RCs. Further research need to be done for the confirmation. 1.5. Interim conclusion So far, we have taken a glance at what the previous research has discussed concerning RCs. In Section 1.1, we have mainly discussed RC properties proposed by Downing (1978). The verb inside the RC may be finite or non-finite. It differs how RCs are related to the head noun, according to language to language. Relative markers differ as well. 32 In Section 1.2, we have observed what several types of RCs actually look like, by watching the natural data from various languages. Some RCs are nominalized to the various degree, connected to the head noun and embedded to become an NP with the head. Or, they are adjuncts to the matrix clause. The head noun may be located inside the RC (internally headed relative clause: IHRC). Or, outside (externally headed relative clause: EHRC). If it is an EHRC, is it before the RC or after (postnominal RC and prenominal RC, respectively)? Correlatives have the head noun and are preposed to the left edge of the matrix clause. Then, it matters how is anaphoric relationship established between the head noun and the relative element in the matrix clause. In Section 1.3, we have taken a brief look at what has been researched with regard to sign languages. The distribution of IHRCs may be different in sign languages from that in spoken counterparts. NMs, one of unique features signed languages have, play a role in the relative construction of sign languages, but their ways are varied from language to language. However, there are still unknown things about sign RCs such as controversial RCs in LIS. In Section 1.4, previous literature has reported three types of RCs: prenominal, postnominal, and internally headed RCs. However, postnominal RCs leave more controversy than the rest. Still, further research must be performed for the investigation into RCs in JSL. 2. THE CURRENT STUDY So far, we have got a general view on the diversity of RCs that both spoken and signed languages represent. This paper will focus on JSL‟s RCs for the following purposes: (a) to confirm the RC types (prenominal, postnominal, and IHRC) proposed by the previous studies, (b) to verify the available NM adopted and use of NM features, and (c) to find out the grammatical status of a relative marker, if any. 2.1. Methodology 2.1.1 Participants and data collecting conditions In order to obtain spontaneous signing data, two strategies were adopted. I aimed to recruit native signers as participants and followed specific elicitation procedures during the recording sessions. First, I narrowed down the participants to the native adult signers that were born deaf, acquired JSL at home and/or at school, and are profoundly deaf. In order to minimalize the 33 influence of spoken Japanese, hearing CODAs (children of deaf adults) were excluded as participant, because their productions might be influenced by knowledge of spoken Japanese. Secondly, I made sure that no hearing people were in the recording room. Based on my experience, I found out it is very difficult to see the native deaf signers produce natural JSL when there is a hearing person within their sight. In fact, they do code-switch among JSL, signed Japanese, and spoken Japanese. Therefore, except for before and after data recording, hearing people were refrained from entering the recording room in order to make the JSL data as genuine as possible. In this current study, the data were produced by two native signers of JSL. 2.1.2 Types of data 2.1.2.1 Picture elicitation task RCs are difficult to elicit just passive sentences (Fischer, S.: PC). It was highly expected that the participants would not use RCs but produce other types of structures such as conjoined sentences. Sign linguistic researchers have struggled with this difficulty and used various methodologies. For example, Liddel (1980: 139) asked native ASL signers to translate English RCs into ASL, not on the spot but after a few days. Cecchetto et al. (2006: 946) first asked native deaf signers to translate Italian RCs into LIS, and then sorted out the LIS RCs from the data and made sure that the clauses meant the same as the original spoken Italian RCs by asking other native signers to judge the adequacy of the translation. Branchini & Donati (2009: 158) asked the LIS informants to translate the RCs of spoken Italian „through the aid of a LIS interpreter who presented the signers with a relevant context‟ and sometimes asking the signers to sign in the most natural LIS. Li (2013) asked a Deaf signer to be an interviewer, and the interviewer told a story to a native Deaf signer (i.e. participant). Then, the interviewer asked the participant multiple-choice questions, which would elicit RCs, and obtained answers including RCs. In this study, for the constraints of time and human resources and in order to eliminate the influence of spoken Japanese, I did not utilize these translation methods. In addition, it was very problematic to predict which types of RCs would occur. A subject is more easily relativized than other syntactic functions such as an object, according to Lehmann (1986). Therefore, in this study, as a preliminary exploration, an elicitation task was designed to elicit only SS RCs (e.g., in English, „A man who is holding a book is intelligent.‟) from the participants. 34 In order to elicit sentences including SS RCs, I modified the Preference Task used by Utzeri (2007), in which hearing children at the age of six to eleven were encouraged to produce Italian RCs. For example, in order to elicit subject RCs, the experimenter showed a picture (Figure 2) and asked a participant a question (73) aimed to elicit the target answer (74). Figure 2 (73) Elicitation of a subject relative: Ci sono due bambine. Una bambina sta leggendo una storia, l‟altra bambina sta ascoltando una storia. Quale bambina preferiresti essere? Inzia con “(Preferirei essere) la bambina che… „There are two children. One child is reading a story, the other child is listening to a story. Which child would you rather be? Start with “(I would rather be) the child that…‟ (74) Target answer subject RC: Il bambino che sta leggendo una storia The child that is reading a story (Utzeri 2007: 289) In the present task, I used 24 picture sets as stimuli. Out of 24 sets, 21 consist of two pictures, and three consist of four pictures. An example of the stimuli may be found in Figure 3. 35 Figure 3 (75) Question: Who is more intelligent? (76) Target answer: A man who is holding a book is intelligent. In most sets, each picture represents a person doing one or two things at a time. The elicitation was conducted between the Deaf interviewer and the participant. The interviewer was also a fluent deaf signer. First, the interviewer showed one set of pictures (e.g., Figure 3) to the participant, and after confirming the participant that he/she understood the contents of the stimuli, the interviewer hid the picture set. Then, the interviewer asked a question (e.g., „Who is more intelligent?‟ (75)), and the participant was invited to answer the question (e.g., „A man who is holding a book is intelligent.‟ (76)). A total of 24 stimuli were used in the elicitation. I expected the replies signed by the participant to include an RC. It is important to note that hiding the stimulus before asking the questions was crucial in order to prevent the participant from answering the question only by pointing directly at the picture. Also, the questions were not which questions, which Utzeri (2007) used, but open questions using who, in order to prevent obtaining the answers without a head. The data collected were transcribed by ELAN19 (e.g., Figure 4), the language annotation software, with four parameters: manual signs, extra manual signs, head movement, eye brow movement, and eye opening (eye blinks). 19 For more details, see the website (https://tla.mpi.nl/tools/tla-tools/elan/). 36 Figure 4 The reason why there are two parameters for manual sign is the simultaneity shared by sign languages. That is, sign languages have two articulators (both hands) and can express two different signs at the same time. Thus, I collected and analyzed the data produced by one of the participants in order to obtain the prototypes of JSL‟s RCs. Through using 48 sets of elicitation pictures (24 sets per participant) and I obtained eight examples to be RCs. 2.1.2.2. Monologues I also examined other data produced by the participants in order to capture more examples that can be analyzed as RCs. Some of the data were produced by another set of picture stimuli, which was originally designed for word order research. Other data involved monologues, in which I gave the Deaf participants three topics (questions) such as „What is your most enjoyable experience of travelling?‟, „What is your most impressive experience at the deaf school?‟, and „What topic/activities are you most interested in now?‟ 2.1.3. Transcription of the data As for the transcription, I glossed the targeted RCs including manual and NM elements of the signing produced by the participants. After the clauses that I suspected to be RCs were confirmed by the two Japanese Deaf signers. I asked them whether the clauses were not simple sentences but one which modifies a nominal (EHRCs), or one which can be analyzed as a nominal (IHRCs). Thus, all the RCs (13 tokens) in this study have been confirmed. 3. RESULTS From the data, two types of RCs have been confirmed: prenominal and internally-headed RCs. Postnominal RCs (Yonekawa 1984) are not been identified in the present data. Before discussing the results, I will present the usual declarative sentence in JSL. 37 As mentioned above, the basic word order is SOV in JSL. However, it is difficult to identify a sentence in JSL because the language, unlike spoken Japanese, does not have any obvious sentence final markers (Kato 2013). Therefore, in this study, I defined a point as a sentence boundary where I found long pauses or longer sign duration. I sometimes consulted Deaf signers about the point when it was difficult to confirm the boundaries. Therefore, I analyze the following line of signs as one sentence although it seems to be a long predicate: (77)20. (77) (Describing the man on the right side of Figure 3) „A man is holding a book, bored, and smart.‟ In (77), there are a head nod (hn) and an eye blink (eb) before „BOOK‟. By analyzing other declarative sentences, I found out that a head nod and an eye blink tend to occur between the subject and predicate in a declarative sentence. Therefore, we can interpret that the subject is „MAN‟ and the predicate is „BOOK …SMART VERY-EAS(ILY) IX-1‟. Inside the predicate, we can observe a role shift, which is illustrated by the square brackets. During the role shift, „IX-1‟ represents not the signer but the man. The long brow raise (br) that co-occurs with some part of the „CL:holding-a-book‟ does not have a syntactic role. It just depicts the man‟s smartlooking appearance. The predicate consists of several phrases that are divided by eye blinks, but the classifier construction on the non-dominant hand (CL: holding-a-book), kept in space till the end, suggests that the predicate can be analyzed as forming a larger and complete constituent. At the end, a head nod and an eye blink occur, but I usually observed only an eye blink at the end of other sentences. In sum, a declarative sentence is illustrated as in (78). 20 „Gloss 1‟is for the main transcription of the manual signs, while „Gloss 2‟ is used when separating two hands is necessary. 38 (78) a declarative sentence in JSL hn eb Subject eb [Predicate …] Keeping (78) in mind, let us look at the first type of RCs: prenominal RCs. 3.1. Prenominal relative clauses The first type of RCs in JSL to mention is prenominal RCs. Among 13 tokens of RCs, I have recognized nine as prenominal RCs. They precede and modify the head as in (79). (79) „The man who is sitting and eating noodles (is angry).‟21 In (79), the participant answered to the question „Who is angry?‟ with an NP, not a simple sentence (subject and predicate). There are two „MAN‟ in the NP, and they can be taken as appositive. Between them, there is the RC „SIT EAT-NOODLES‟. The question is: which „MAN‟ is the head noun of the RC. Since we can observe an eye blink between the first „MAN‟ and the RC, we can analyze the „MAN‟ and the RC as different constituents. On the contrary, we cannot find any constituent-cutting NMs between the RC and the second „MAN‟. Therefore, we interpret that the RC and „MAN‟ form a constituent (NP). Also, there are two head nods. The first nod that is associated with „SIT‟ is gestural; that is, when you sit down, your head sometimes moves like a nod. Hence, the nod is not syntactic. The second nod can be interpreted as the marker of the end of the NP including the RC. Eye blinks occur at three points. However, they do not occur at the end of RC (between „EAT-NOODLES‟ and „MAN‟) or sentence (IX3), although eye blinks normally are observed at the beginning of the sentences 21 „the man sitting and eating noodles‟ would be more natural, but for the sake of comparison between the original and translation, I used the RC for the translation. 39 and the right edge of a prosodic phrase (i.e. end of a phrase or syntactically a constituent) in sign languages (Tang, G.: PC). Further research need to be done to find out the functions that eye blinks in JSL have. As for the linkage between the RC and its head noun, why are not there any connecting elements (manual or NM) to link the RC and the head? I assume that the word order signals the modification relationship between the RC and head, because the order VVS („SIT EAT-NOODLES MAN) is different from a basic SOV order; this unusualness hints that the utterance is not a simple sentence. Are prenominal RCs accompanied with by a pointing sign? In four out of nine prenominal RCs produced by the Deaf participant in this research, a pointing occurs immediately after the head noun. In (79), for example, IXdet occurs immediately after the head noun „MAN‟. I think that this kind of pointings is a simple determiner, rather than a relative determiner, because it is common that prenominal RC languages lack a relative determiner (28 out of 38; De Vries (2002)). Eye blink sometimes occurs before and after the RC as in (80). (80) „The man on the right side, smart, holding a book, looks smart‟ In(80), after the second „TO-THE-RIG(HT)‟, the head nod and eye blink represent the end of the constituent representing the subject. After that, the NMs (the back head tilt (htb) and brow raise (br)) are interpreted as depicting the attitudes of the man, hence they are not syntactic markers. The rightward head tilt (htr) occurs twice, and both seem to be lexical (TO-THERIGHT and SMART), rather than syntactic markers. The head movement such as head nod and backward head tilt that occur at the end of the NP here are interpreted as prosodic cues for making the RC and the head noun. The part of the RC, which ranges from „CL:holding-abook‟ to the end of two NMs (the back head tilt and brow raise), represents the role shift (the 40 smart-looking man), and the ending of the role shift matches the end of the RC. Thus, I suppose the end of the role shift may also indicate the end of the RC. Thus, prenominal RCs can be illustrated as in (81). (81) Prenominal relative clauses in JSL (eb) (eb) hn [NP [RC … V] Nhead] In the next section, we will look at one more RC type confirmed in this study: IHRCs. 3.2. Internally headed Some RCs have the head inside and are embedded in the main clause as the NP. This type of RCs is commonly referred to as an internally headed relative clause (IHRC). Among 13 tokens of RCs, I have recognized four as IHRCs. They typically no change of the word order; however, in JSL, these IHRCs are not marked with clause-cutting NMs. However, JSL‟s IHRCs are frequently followed by a pointing sign and different NMs according to its position in the main clause. For example, when it is the subject, the clause itself is not marked by any prominent NMs but followed by the pointing with the head nod and eye blink as in (82). (82) „The man with a stethoscope hanging from his neck is licking a lollipop.‟ In (82), since the RC „MANa DOCTOR CL-STETHOSCOPE-HANGING-FROM-THE-NECK‟ is not marked by any phrase dividing NMs, it can be interpreted as a phrase. After the RC is followed by a pointing IX-3, a head nod and eye blink occur. This pointing does not seem obligatory because another example lacks it as in (83). (83) „The woman who is sitting in a chair is eating a chocolate bar.‟ 41 In (83), the RC „WOMAN SIT WOMAN SIT‟ (a sitting woman) is in the subject position, not marked by any NMs. When this clause ends, we observe a head nod and an eye blink. It is difficult to analyze the pair of hn and eb as markers for RC because the set also appears in the same position (cf. (78)) of the declarative sentence. These may not be syntactic marker for RC but they are prosodic marker for clause boundary (Tang, G: PC). In (83), if we analyze the chunk of signs (the RC plus the pointing) as the subject NP, the chunk is followed by the eye blink and head nod, then, the sentence has the same structure as a declarative sentence such as (77). Next, the IHRC comes as the object. First, let us look at (84). (84) „The boy succeeded in catching the frog inside the jar.‟ In (84), the clause „FROG STAY INSIDE‟ is interpreted as one phrase, because it is not marked inside by any phrase-dividing NMs such as a head nod. In the main clause, there is no NP to be assigned the argument (object) by the verb „CATCH‟ except the chunk. Therefore, the chunk is an NP. Inside it, there is only one noun to be the head of the NP: „FROG‟. Besides, since the NP (clause) has its head FROG inside, so the clause can be analyzed as an IHRC. Example (84) cannot be analyzed as two conjoined sentences, because the FROG cannot be the subject of „CATCH‟ as well. After the RC, there is no pointing sign, probably because the pointing probably fused with the former sign INSIDE, because INSIDE consists of the pointing sign of the dominant hand and the big C-handshaped non-dominant hand. In order to mark the end of the IHRC, a role shift and an agreement verb were used as in (85). (85) 42 „I see various footprints mixed on the beach.‟ In (85), the clause, which is illustrated between the brackets, can be interpreted as a preposed clause to the front, because the subject „IX-(1)‟ comes after it. Besides, the agreement verb „(IX-1)SEE(IX-3p)‟ needs to agree with an argument in the preceding discourse (i.e. FOOTPRINTS). There are no NMs inside the clause to make it a simple sentence, and the clause includes the argument of the verb. Therefore, the clause can be analyzed as an IHRC. As for NMs, inside the RC there are two instances of brow raise. The first one is used to form the CL construction „FOOTPRINTS CL:two-feet-step‟, while the other is to attract the addressee‟s attention, or to focus the footprints. The chunk „FOOTPRINTS CL:two-feet-step EX(IST) CL:twofeet-step‟ is analyzed as the NP („FOOTPRINTS CL:two-feet-step‟ and „CL:two-feet-step‟ are appositive connected with the verb EXIST), and it is also the subject of the RC. Furrowed brow (bf) is strongly connected to the sign „MIX‟. Head tilt forward (htf) often co-occur with CL:two-feet-step because feet are physically lower than arms. The end of the RC is very obvious for two strategies. One is role shift: the signer changed the neutral description of footsteps on the beach to the role „I saw the footsteps (i.e. „(IX-1)SEE(IX-3p)‟)‟ at the end of the RC. The other strategy is the agreement verb „(IX-1)SEE(IX-3p)’. The verb is directed to the footsteps established through a locus in space when signing the classifier constructions. This directing verb is almost like a pointing sign, directing the location where the argument is so that the referent can be identified through agreement. In other words, instead of using a determiner or a demonstrative, native deaf signers of JSL can make use of verb agreement to „relativize‟ a head noun, so to speak. To sum up, I may formulate the IHRCs in JSL: (86). (86) Internally headed relative clauses in JSL (eb) (eb) (hn) [NP [IHRC … Nhead …] (IXhead)] In a similar manner as prenominal RCs, the eye blink frequently occurs before and after the RCs, and the head movement such as the head nod and backward head tilt occur at the end of the NPs consisting of the RCs and the pointing sign. However, the pointing sign following the RC is not obligatory. It may be a relative determiner. Further research is needed to reveal how it occurs and to confirm what it is. 43 To sum up, through this research, two types of RCs have been confirmed in JSL: prenominal and internally headed RCs. In many cases, at the beginning and ending of the RC, eye blink occur, which enables the identification of a syntactic constituent which I would claim to be an RC. Also, the pointing sign that occurs after the head noun (prenominal) or the RCs (postnominal) may be a relative determiner, which needs to be confirmed by further research. These manual signs or NM features are still yet to be investigated to further consolidate our understanding of RCs in JSL. 4. DISCUSSION In this study, I have identified 13 RCs in the data of 103‟29 (picture elicitation task: 35‟58”; monologues: 66‟31”). Although the number of identified RCs was small, I have managed to categorize them into the two types of RCs: prenominal (nine tokens) and internally headed (four tokens). Postnominal RCs are not attested in my data. Since the size of data is small, this does not exclude the possibility of postnominal RCs in JSL. Therefore, further investigation on RCs is needed to confirm postnominal RCs. As for the relative strategy in JSL, it has been revealed that both types share three features. First, the eye blink frequently occurs immediately before and after the RCs. It seems that the blink is omitted or merged to other eye blinks when the RCs is located at the beginning of the sentence, followed by the predicate, or at the end of the sentence. Secondly, the head movement, mostly head nod, appears at the end of the NP including the RC. This is the feature ones previous studies such as Ichida (2005c) pointed out. Based on my knowledge about JSL, I think other prosodic features, such as sign duration, intervals between the signs, and the speed of signs, mark the RCs. Lastly, role shift may also mark the RC. In the process of conducting the present study, I have encountered several problems that are to be solved through further research in the future. First, more descriptive research on RCs is needed. In this research, the data recorded and investigated for the search of RCs in JSL amount to over 100 minutes (elicitation task: 35‟58; monologues: 66‟31). However, I could identify only 13 RCs (elicitation task: 8; monologues: 5), which I could categorize into two types – prenominal RCs and IHRCs. There are two reasons for such a small number. First, RCs may rarely occur by default. In fact, RCs are rare in spoken language in its oral mode, but surface more frequently in the written mode. In the future, more precise descriptions about JSL and quantitative analysis are required. The other reason is the 44 elicitation task. It is true that the tasks were designed to elicit only the SS RCs, but still they did not work well. As a result, I could obtain only a small number of RCs. With this number, it is difficult to generalize the features of the RCs in JSL. In the coming research, it is necessary to design elicitation materials that are sensitive to the production of RC. Secondly, subordinate clauses in JSL must be explored. It seems to me that they have been researched to some degree through the study of the conjunctive functions of head nod, but there appears to be little research into coordination and subordination of JSL. The new knowledge on subordinate clauses in JSL will help investigate the differences between embedding and adjunction in JSL, which the current study has not examined. Thirdly, we can investigate more on eye blinks (i.e. eye aperture) and word lengthening. In this study, I have annotated head movement, brow raise, and eye blink as the data transcription. Head movement, especially head nod22, seems to have been researched more than other NMs. We will be able to research NMs in more details, owing to the development of video recording technology. Fourthly, the JSL types of RCs, prenominal and internally headed, remind me of how spoken Japanese, which also has these two types of RCs. First, let us look at one type: prenominal RCs. As noun modifiers such as adjectives (87), adjectival nouns (88), and APs of the noun and the Genitive Case particle (-no) (89), RCs come before the head NP as in (90). (87) Taroo-ga omosiroi hon-o kaita. [adjective] Taro-NOM interesting book-ACC wrote „Taro wrote an interesting book.‟ (88) Ziroo-ga kirei-na hana-o Ziro-NOM pretty Satiko-ni okutta. [adjectival noun] flower-ACC Sachiko-DAT sent „Ziro sent pretty flowers to Sachiko.‟ (89) Hanako-ga tomodati-no uti-o Hanako-NOM friend-GEN katta. [noun + -no] house-ACC bought. „Hanako bought her friend‟s house.‟ 22 Head nod is said to be one of the most unique features in JSL. In fact, the reduction of head nod has been emphasized during my leaning of International Sign. However, in order to prove the uniqueness of head nod in JSL, more scientific research may be necessary. 45 (90) Satoo-sensei-ga [gakusei-ga kaita] ronbun-o yondeiru. [RC] Prof. Sato-NOM student-NOM wrote article-ACC is-reading „Prof. Sato is reading the article that the student wrote.‟ (=(11)) (Tsujimura 2007: 301)23 What is common with these examples is that the modifiers (omosiroi, kirei-na, tomodati-no, and gakusei-ga kaita) precede their heads (hon-o, hana-o, uti-o,and ronbun-o, respectively). According to Siewierska (1994: 4996), „Non-rigidly verb-final SOV languages (SOV languages which allow for the placing of some modifiers to the right of the verb) tend to have some nominal modifiers following the noun, while rigidly verb-final SOV languages generally serialize all the modifiers prenominally‟. Therefore, although it needs to be investigated whether JSL is a rigidly verb-final SOV language or not, it is no wonder that JSL and Japanese share prenominal RCs, because they share a basic SOV word order. As for the verb form, in (90), the RC has the verb „kaita‟ (wrote), which is in the same form as in a simple sentence. Compare these two sentences. (91) Satoo-sensei-ga [gakusei-ga kai-ta] ronbun-o yondeiru. [RC] Prof. Sato-NOM student-NOM write-PAST article-ACC is-reading „Prof. Sato is reading the article that the student wrote.‟ (=(11)) (92) Gakusei-ga ronbun-o kai-ta. [simple sentence] student-NOM article-ACC write-PAST „The student wrote the article.‟ In both examples, the verb „kak-‟ (write) appears in the same form, kai-ta „wrote‟, in the RC (92), as well as in the simple sentence (92). Also, there are no relative elements in the prenominal RCs in Japanese. Next, the other RC type Japanese has is internally headed RCs (Kuroda 1992; Tsujimura 2007). They resemble simple sentences, but we can find the head inside them as in (93). (93) 23 Taroo-wa [ringo-ga sara-no ue-ni atta]-no-o totte, Prenominal RCs are also called sentence modifiers (Tsujimura 2007). 46 Taroo-TOP apple-NOM plate-GEN top-at existed-one-ACC took poketto-ni ireta. pocket -to put-in „Taro picked up an apple which was on a plate and put it in a pocket.‟ (Kuroda 1992: 147) In (93), the noun „ringo‟ (an apple) has two roles in the sentence. First, it is the object or internal argument of the verbs „totte‟ (took) and „ireta‟ (put in), hence marked by the accusative „-o‟. The other role of „ringo‟ is that it is the head of the RC. In addition, the head is inside the clause, marked with the nominative particle „-ga‟. Thus, this is an IHRC, not an EHRC. Furthermore, the clause is embedded with the particle „-no‟. This particle can be interpreted as a „nominalizing complementizer‟ (Kuroda 1992), or more precisely, „relative additional nominalizing suffix‟ (De Vries 2002). IHRCs have no relative pronoun, either. Thus, spoken Japanese has two types of RCs: prenominal RCs and internally headed RCs. Both have no relative pronouns. These features are shared with JSL, but is this merely a coincidence? I think language contact may be one of the reasons for the similarities between JSL and Japanese. According to Hendery (2012), language contact could induce one language (stronger) to change another (weaker). One of its examples is Singapore English. Singapore English is changing through the influence of surrounding major languages such as English, Malay, and the three Chinese varieties: Mandarin, Hokkien, and Cantonese (Alsagoff & Lick 1998: 128–129). (94) That boy pinch my mother one very naughty „That boy who pinched my mother is very naughty‟ (Hendery 2012: 219 citing from Alsagoff & Lick 1998: 129, ex. 3) (95) The fruit they grow one very sweet „The fruit that they grow is very sweet‟ (Hendery 2012: 219 citing from Alsagoff & Lick 1998: 134, ex. 31) 47 The word one in the two examples is the relative marker, and it occurs at the end of the RC „on analogy with Chinese de‟ (Hendery 2012: 220)24. According to Hendery (2012), multilingual speakers create „a new strategy in one language on analogy with a construction in the other‟ (p. 235). In fact, most Deaf people around the world cannot help being bilingual in order to survive surrounded by the hearing majority. In Japan, since the first deaf school was established in 1878, deaf education had been conducted with the manual method; that is, using sign language, for approximately half a century. Then, after the oral method had been introduced from Europe in the early twentieth century, spoken Japanese has been dominant at deaf schools (Yonekawa 1998). Also, deaf people have been surrounded by written Japanese. These circumstances may have influenced the grammar of JSL, which includes relative constructions. However, it is extremely hard to prove the existence of this influence, because it would never be possible without the data produced by the illiterate Deaf signers, if any. Or, the comparison of the elicited RCs between the elder generation and younger generation could provide some clues regarding the Japanese influence on JSL. 5. CONCLUSION Research in to RCs in JSL will enhance the training of translation or interpretation in JSL. The structural complexity, such as subordinate clauses, of the target language sometimes causes difficulties for interpreters. The more the structure of JSL is examined and revealed in detail, the more effectively the interpreters perform. Therefore, I highly expect this research to benefit the Deaf community in Japan as well as other people involved in the community. This study, for the data number constraint, has left much unknown about RCs in JSL. More researches are expected to figure out what other strategies to mark the end of IHRCs. 24 Hendery (2012: 220) footnoted: „Mandarin de, Cantonese ge and Hokkien ge or e all function the same way (Alsagoff & Lick 1998: 134), but I will use de as shorthand to encompass any of these here‟. 48 APPENDIX: ABBREVIATIONS 2h two hands ew eye widening (nonmanual) ABL ablative FUT future ACC accusative GEN genitive AGT agentive h hold of head (nonmanual) ASL American Sign Language hb backward head shift (nonmanual) Auslan Australian Sign Language hf forward head shift (nonmanual) AUX auxiliary hn head nod (nonmanual) bf furrowed brow (nonmanual) hs head shake (nonmanual) br brow raise (nonmanual) htb head tilted backward (nonmanual) CMR complementizer htf head tilted forward (nonmanual) COMPL hts head tilted sideward completive CONN connective IMPF imperfect(ive) COP INAN inanimate copula D1/2/3 demonstrative element of 1/2/3. ps. deixis. DA definite article DAT dative DEF definite DEM demonstrative DGS German Sign Language (Deutsche Gebärdensprache) IND indicative INST instrumental IX indexing (pronoun) JSL Japanese Sign Language (Nihon Shuwa) LIS Italian Sign Language (Lingua Italiana dei Segni) lit. literally locative dh dominant hand LOC eb eye blink (nonmanual) ec eye closing (nonmanual) LSB Brasilian Sign Language (Língua de Sinais Brasileira) ERG ergative es eye slit (nonmanual) NEU neuter NM nonmanual 49 NMR nominalizer REM remote NOM nominative SBJ subject (personal affix slot) O SG singular PART participle sh subordinate hand (non-dominant) POSS possessive shn small head nod (nonmanual) PP SR subordinator object past participle PRES present SUBJ subjunctive PRET preterite tense TOD.PAST r(el) relative nonmanual TOP RC relative clause today‟s past tense topic REAL realis REL relative element 50 BIBLIOGRAPHY Alexiadou, A., Law, P., Meinunger, A. & Wilder, C. (2000). Introduction. In A. Alexiadou, P. Law, A. Meinunger & C. Wilder (Eds.), The Syntax of relative clauses. Amsterdam: John Benjamins Pub. Co. 1–51. Alsagoff, L., & Lick, H. C. (1998). The relative clause in colloquial Singapore English. World Englishes 17(2). 127–138. Amin-Madani, S. & Lutz, D. (1972). Persische Grammatik. Heidelberg: J. Groos. Andrews, A. D. I. (1975). Studies in the syntax of relative and comparative clauses. Bird, C. (1968). Relative clauses in Bambara. Journal of West African Languages 5(1). 35–48. Branchini, C. & Donati, C. (2009). Relatively Different: Italian Sign Language Relative Clauses in a Typological Perspective. Correlatives Cross-Linguistically. 157–191. Brown, E. K. & Miller, J. (2013). The Cambridge dictionary of linguistics. Cambridge; New York: Cambridge University Press. Cecchetto, C., Geraci, C. & Zucchi, S. (2006). Strategies of Relativization in Italian Sign Language. Natural Language & Linguistic Theory 24(4). 945–975. Comrie, B. & Kuteva, T. (2013). Relativization on subjects. The World Atlas of Language Structures Online. [Available at http://wals.info/chapter/122 (2015-06-18)] Crain, S., & Lillo-Martin, D. (1999). An introduction to linguistic theory and language acquisition. Malden: Blackwell Publishers. Downing, B. T. (1978). Some universals of relative clause structure. Universals of Human Language 4. 375–418. Dryer, M. S. (2013a). Order of Relative Clause and Noun. The World Atlas of Language Structures Online. [Available at http://wals.info/chapter/90 (2015-05-22)] Dryer, M. S. (2013b). Relationship between the Order of Object and Verb and the Order of Relative Clause and Noun. The World Atlas of Language Structures Online. [Available at http://wals.info/chapter/96 (2015-5-22)] Fabb, N. (1990). The difference between English restrictive and nonrestrictive relative clauses. Journal of Linguistics, 26(1), 57-78. Fox, B. A.;Thompson,S. A. (1990). A Discourse Explanation of the Grammar of Relative Clauses in English Conversation. Language 66(2). 297–316. Fuji, M. (2010). On internally headed relative clauses in Japanese and Navajo. Journal of the Tokyo University of Marine Science and Technology, 6. 47–58. Hale, K. (1976). The adjoined relative clause in Australia. Grammatical Categories in Australian Languages 784105. 78–105. 51 Hendery, R. (2012). Relative clauses in time and space: a case study in the methods of diachronic typology. Amsterdam; Philadelphia: John Benjamins Pub. Co. Ichida, Y. (1998). Nihonshuwa-no bunpoo: tokushuu shuwa-no sekai–gengo-tosite deau-toki. [Grammar of Japanese Sign Language: world of sign language–when you meet it as a language]. Gengo 27. 44–51. Ichida, Y. (2005c). Shuwa-no gengogaku (10) bunkoozoo-to atama-no ugoki–Nihon Shuwano bunpoo (6) gojun, hobun, kankeisetu. [Linguistics of Japanese Sign Language (10) sentence construction and head movement--grammar of Japanese Sign Language (6) word order, complement, and relative clauses]. Gengo 34. 91–99. Ichida, Y. (2005a). Shuwa-no gengogaku (5) jikan kuukan-to te-no undoo--Nihon Shuwa-no bunpoo (1) aspect etc. [Linguistics of Japanese Sign Language (5) time and space and hand movement--grammar of Japanese Sign Language (1) aspect etc]. Gengo 34. 92– 99. Ichida, Y. (2005b). Shuwa-no gengogaku (6) kuukan-no bunpoo--Nihonshuwa-no bunpoo (2) daimeesi-to doosi-no icchi. [Linguistics of Japanese Sign Language (6) grammar of space--grammar of Japanese Sign Language (2) pronouns and verb agreement]. Gengo 34. 90–98. Johnston, T., & Schembri, A. (2007). Australian sign language: an introduction to sign language linguistics. Cambridge; New York: Cambridge University Press. Kanda, K. (2009). Shuwa no omosirosa. [Interestingness of sign language] Gekkan Gengo, 8, 8–15. Kato, N. (2013), Shuwa-ni okeru gengoshigen-no kenkyuudookoo. [Research trend towards language resource in sign language]. NHK STRL R&D 139. 10–19. Kimura, H., & Ichida, Y. (2014). Hajimete-no shuwa shoho-kara yasashiku manaberu shuwa-no hon. [First sign language: sign language book to learn easily from the rudiments with]. Tokyo: Seikatsushoin. Kuno, S. (1973). The structure of the Japanese language. Cambridge, Mass.: MIT Press. Kuroda, S. (1992). Japanese syntax and semantics: Collected papers. Dordrecht; Boston: Kluwer Academic Publishers. Leeson, L., & Saeed, J. (2012). Word order. In R. Pfau, M. Steinbach & B. Woll (Eds.), Sign language: an international handbook. Berlin; Boston: De Gruyter Mouton. 245-265. Lehmann, C. (1986). On the Typology of Relative Clauses. Linguistics 24(4), 1–13. Lehmann, C. (2003). Relative clauses. In J. Frawley, & W. Frawley (Eds.), International encyclopedia of linguistics. Oxford; New York: Oxford University Press. 460–462. 52 Li, J. (2013). Relative constructions in Hong Kong Sign Language. (Unpublished MA). Chinese University of Hong Kong, Liddell, S. K. (2003). Grammar, gesture, and meaning in American Sign Language. Cambridge; New York: Cambridge University Press. Liddell, S. K. (1980). American Sign Language syntax. The Hague; New York: Mouton. Lindenfeld, J. (1973). Yaqui Syntax (University of California Publications in Linguistics 76). Lipt k, A. K. ( ). The landscape of correlatives: An empirical and analytical survey. In A. K. Lipt k ( d.), Correlatives cross-linguistically. Amsterdam; Philadelphia: John Benjamins Pub. Co. 1–46. Malmkjær, K. (1991). Language universals. In Malmkær, K. (ed.). The linguistics encyclopedia. London; New York: Routledge. 277–284. Muysken, P. (1994). Nominalizations. In R. E. Asher, & J. M. Y. Simpson (Eds.), The Encyclopedia of language and linguistics. Oxford; New York: Pergamon Press. 2811– 2815. Oka, N., & Akahori, H. (2011). In Bilingual Bicultural Education Center for Deaf Children (Ed.), Bunpoo-ga kiso-kara wakaru Nihonshuwa-no sikumi. [Strucure of Japanese Sign Language to understand from the basic about its grammar]. Tokyo: Taishukan Shoten. Siewierska, A. (1994).Word order and linearization. In R. E. Asher, & J. M. Y. Simpson (Eds.), The Encyclopedia of language and linguistics. Oxford; New York: Pergamon Press. 4993–4999. Sinclair, J. (Ed.). (2012). Collins COBUILD English Usage (3rd ed.). Glasgow: HarperCollins Publishers. Sze, F. Y. B. (2000). Word order of Hong Kong Sign Language. Cross-Linguistic Perspectives in Sign Language Research: Selected Papers from TISLR. 163–192. Tang, G., & Lau, P. (2012). Coordination and subordination. In R. Pfau, M. Steinbach & B. Woll (Eds.), Sign language: an international handbook. Berlin; Boston: De Gruyter Mouton. 340–365. Tsujimura, N. (2007). An introduction to Japanese linguistics (2nd Ed.). Malden, MA: Blackwell Pub. Utzeri, I. (2007). The production and acquisition of subject and object relative clauses in Italian: a comparative experimental study. Nanzan Linguistics 2. 283–313. 53 Valli, C., Lucas, C., & Mulrooney, K. J. (2005). Linguisitcs of American Sign Language: an introduction (4th Ed.). Washington, D.C.: Clerc Books. Velupillai, V. (2012). An introduction to linguistic typology. Amsterdam; Philadelphia: John Benjamins Pub. Co. Vincent, N. (1994). Subordination and complementation. In Asher, R.E. & Simpson, J.M.Y. (Ed.), The Encyclopedia of language and linguistics. Oxford; New York: Pergamon Press. 4391–4396. Vries, M. d. (2002). The syntax of relativization. Utrecht: LOT. Yonekawa, A. (1984). Shuwagengo-no kijututekikenkyuu. [Descriptive study of sign language]. Tokyo: Meiji Shoin. Yonekawa, A. (1998). Tokushu shuwa-no sekai: korekara shuwa-o manabu hito-no tameni. [Special features sign language: for those who begin to learn sign language from now on]. Gengo 27. 20–25. (This is the end of this thesis.) 54