1

The UG Code of Hentai Kambun1 Edith Aldridge, University of Washington

eca1@u.washington.edu edithaldridge@gmail.com

DRAFT: June 2012 1. Introduction: This paper presents an analysis of word order patterns in the 8th century hentai kambun (変体漢

文) text Kojiki (古事記). Hentai kambun was one of three ways to write Japanese in the eighth century. Since the beginning of the transmission of the Chinese writing system to the Japanese islands in approximately the 5th century, Chinese characters were used to record Japanese according to either their phonetic or semantic value. When employed purely for their phonetic value, Chinese characters were used to represent Japanese syllables, one character per syllable, as in the following poem from the 8th century anthology Manyoshu. (1) 佐欲比賣能 故何 比列 布利斯 夜麻 (MYS 868) Sayopimye=no kwo=ga pire puri-si yama2 Sayohime=Gen child=Gen scarf wave-Past.Adn hill ‘the hill where the girl Sayohime waved her scarf’ Chinese characters could also be employed for their semantic value. In the following, for example, the middle Chinese wh-word 何物 ‘what’ is used to represent the Old Japanese wh-word nani ‘what’, and the Chinese character 人 for ‘person’ is used for the Old Japanese pito with the same meaning. The same is true for the other content words. The main verb kazasu is particularly interesting, since it is represented in writing as a compound with two Chinese characters, one meaning ‘stick in’ and the other meaning ‘head’. The interrogative and focus particles ka and mo, respectively, are represented indirectly but still illustrate the principle at hand. The Chinese character means ‘duck’ (type of bird). The Old Japanese word for ‘duck’ is kamwo. This character is co-opted to represent the sequence of particles indirectly through the homonym ‘duck’. Another grammatical particle is the genitive marker on the subject 之 used to represent the Old Japanese genitive particle no. Finally, the Old Japanese modal suffix –mu is represented by the Chinese modal 将. As it would in Chinese, the modal precedes the verb stem. Grammatical morphemes used to represent Old Japanese inflectional forms appeared in pre-verbal position, where they are expected to appear in Chinese. Word-internal order thus adheres to the Chinese pattern. Word order among independent lexical items, on the other hand, follows Old Japanese order. (2) is particularly interesting in this regard, since it is a wh-question. Wh-phrases in Old Japanese always preceded genitive subjects. This word order, however, is not attested in either Classical or Middle Chinese.

1 This paper was presented in the Department of Linguistics, University of Washington on December 3, 2010. I would like to thank Barbara Citko, Zev Handel, Julia Herschensohn, and Karen Zagona, for their questions and comments. 2 Readings are taken from the 1962 edition of the Nihon Koten Bungaku Taikei 『日本古典文学大系』(Iwanami Press). Romanizations are based on Old Japanese reconstructions in Frellesvig & Whitman (2008).

2

(2) 何物 鴨 御狩 人之 折而 将挿頭 Nani=wo ka-mwo mi-kari =no pito=no wori-te kazasa-mu? what=Acc Q-Foc Hon-hike=Gen person=Gen break-Conj stick.head-Mod ‘What should the hikers break off and wear in their hair?’ (MYS 1974) In the third type of Old Japanese writing, hentai kambun (HTKB), Chinese characters were employed for both their lexical semantic value. Surface word order also strongly resembles Chinese. Specifically, verbs and adpositions generally appeared before their complements, contra Old Japanese word order, which (like modern Japanese) was OV and postpositional. (3a) shows a verb and its complement DP, (3b) an adposition and its complement DP, and (3c) an embedding verb and its complement CP. When these are read, the Japanese head-final word order is restored by switching the order of head and complement. Japanese lexical items are also inserted for Chinese characters, and the phrases are read in Old Japanese. (3) a. 採 其 地 之 青菜 (Kojiki, Nintoku) pick [Dem place Gen vegetable] => [soko=no awona]=wo tumu there=Gen vegetable=Acc pick ‘pick the vegetables of that place’ b. 自 其 地 (Kojiki, Jimmu) from [Dem place] => soko ywori there from ‘from there’ c. 吾 欲 取 其 猪。 (Kojiki, Ohojin) I [VP want [CP ... [VP take Dem boar]]] => Ware [VP [CP ... [VP sono winosisi=wo tora]-mu to] omopu]. I Dem boar=Acc take-Mod C want ‘I want to get that boar.’ The translation process is not, however, always so straighforward as this. There are many types of word order which are not attested in either Chinese of the time or otherwise in Old Japanese texts. (4) shows one such type of word order. In ditransitive clauses in which both of the objects are overtly expressed, the verb appears between its two internal arguments. (4) 名 賜 曙立 王 (Kojiki, Suinin) name give Aketatu prince => [VP na=wo [V’ [Aketatu=no Opokimi]=ni tamapi-te name=Acc Aketatu=Gen prince=Dat give-Conj ‘(He) gave a name to Prince Aketatu, and….’ At least since Motoori (1798), there has been an awareness in the field that hentai kambun texts were intended to be read in Japanese. However, the precise characterization of the texts themselves has remained elusive. The general assumption seems to be that hentai kambun texts are some sort of hybrid, containing elements of both Chinese and Japanese (Miller 1967;

3

Minegishi 1980, 1986; Feng 1995; Uchida 1995; Rabinovitch 1996). Others treat hentai kambun texts as primarily Japanese (Tsukishima 1963; Nishimiya 1993; Yamaguchi 1995; Sema 1999). However, given the existence (or even preponderance) of word order types like (4) which cannot be characterized directly as Japanese or as Chinese, some account of these patterns in and of themselves does appear to be warranted. Unfortuantely, such systematic analysis of the word order in these texts are few and far between. Nakagawa (1995) and Aldridge (2001) are the only syntactic accounts of word order patterns that I am aware of. In this paper, I propose a simple encoding process from Japanese to the word order found in the Kojiki text. The first is the reordering process referred to above. Head-final order in Japanese VPs and PPs is switched to Chinese head-initial order. When the text is read, this process is reversed. What is important is that setting of head-complement order is a process which must be applied in spelling out word order in the derivation of an utterance in natural human language. Therefore, I am proposing that the author of the Kojiki text is able to call on this same process of Universal Grammar in the composition of a text. The second step similarly borrows from generative grammatical processes. Old Japanese had a process of object shift which moved a definite, specific, or focused constituent from VP to a position in the edge of vP above the base position of the subject. I propose that a similar raising process was employed in encoding Old Japanese in the Kojiki, which accounts for a collection of word order types which otherwise seem to violate basic principles of constituency. I show, however, that even these cases can be given a straightforward account in light of the object shift process which was present in Old Japanese. 2. Unexpected Word Order Patterns In this section, I examine some word order types which are unexpected in either Classical or Middle Chinese. I argue that these are not simply mistakes in attempting to write Chinese. Rather, I show that they are structurally principled and systematic and directly reflect Old Japanese surface word order. 2.1. Systematic Deviation 1: Indefinite subjects First note that verbs follow subjects in the text, even when they are indefinite, as in (5b). (5) a. 於 頭 者、 大 雷 居 (Kojiki, Izanagi) on head Top great thunder.god be => Kasira=ni=pa opo ikaduti wo-ri head=Dat=Top great thunder.god be-Conj “On her head was the great thunder god, ….” b. 於是 到 気多之前 時、 裸 兔 伏 也。 (Kojiki, Okuninusi) then arrive Ketanosaki when naked rabbit lie Decl => Kokoni, ketanosaki=ni itari-si toki then Ketanosaki=Dat arrive-Past when akapada=no usagi puseri-ki. naked=Gen rabbit lie-Past “When (they) arrived at Ketanosaki, a naked rabbit was lying (there).”

4

This is true also for existential constructions. (6) a. 火打 有 其 裏。 (Kojiki, Keiko) flint exist Dem inside => Pwiuti sono uti=ni ari-ki. flint Dem inside=Dat exist-Past ‘There was a flint inside it.’ b. 於是 須佐之男命 以為 thus Susano’o think 人 有 其 河 上 (Kojiki, Amaterasu) person exist Dem river up Kokoni Susanowonomikoto pito sono kapa kami=ni ari thus Susano’o person Dem river up exist と 以為ひて to omopi-te C think-Conj “Thus, Susanowo thought there was someone upstream.” This is highly unexpected from the perspective of Chinese. Locative or temporal expressions appear in preverbal position, as in the Middle Chinese examples in (7). However, the DP whose existence is being asserted appears postverbally and never in subject position. (7) a. 村 中 有 好 美 水。 (Baiyujing 4.2) Cun zhong you hao mei shui. village middle exist good beautiful water ‘There was very good water in the village.’ b. 昔 有 愚 人 養育 七 子。 (Baiyujing 7) Xi you yu ren yangyu qi zi. past exist stupid person raise seven son ‘In the past, there was a stupid person who raised seven sons.’ 2.2. Systematic Deviation 2: Passive agents Another anomalous type of word order is the position of agents in passive constructions. These always precede the verb and do not necessarily accompany an adposition. (8) a. 汝 者 我 見欺。 (Kojiki, Okuninushi) you Top me Pass.deceive => Na=pa ware=ni azamuka-e-tu. you=Top me=Dat deceive-Pass-Past ‘You have been deceived by me.’ b. 其 身 皮 悉 風 見吹折。 (Kojiki, Okuninushi) Dem body skin all wind Pass.blow.tear

5

=> Sono mwi=no kapa kotogotoni kaze=ni puki-saka-e-ki. Dem body=Gen skin all wind=Dat blow-tear-Pass-Past ‘All the skin on his body was torn by the blowing of the wind.’ Classical and Early Middle Chinese also formed passives with the morpheme 見 jian. But in Chinese, the agent always follows verb and was contained within a PP. (9) 吾 長 見 笑 於 大方 之 家。 (Zhuangzi, Qiushui) Wu chang jian xiao yu dafang zhi jia. I Mod Pass laugh by enlightened Gen person ‘I would have been laughed at by an enlightened person.’ 2.3. Systematic Deviation 3: Causees Causative constructions in the Kojiki are somewhat more varied but still provide evidence for the structural basis of word order in the text. (10) shows a Middle Chinese causative marked with 令 ling. Chinese causative constructions are biclausal, with ling in the position for the matrix verb. The causee follows ling. (10) 云何 能 令 是 麥 茂 好？ (Baiyujing 82) Yunhe neng ling shi mai mao hao? how can make this wheat dense rich “How can you make this wheat dense and rich?” In contrast to this, when the embedded clause is transitive, the causee precedes the verb in the Kojiki. The causative morpheme in Old Japanese (as it does in modern Japanese) appeared as a suffix on the verb. In hentai kambun, the Chinese character ling appears in immediate preverbal position. This is analogous to the preverbal position of the modal in (2). What is relevant to the current discussion is that the complex verb follows the causee. (11) a. 於 髪長 比売 令握 大御 酒 柏 P Kaminaga princess Caus.take Hon wine oak(.leaf) 賜 其 太子。 (Kojiki, Ojin) bestow Dem prince => Kaminaga-pimye=ni opomi-sake=no kasipa=wo twora-sime-te, Kaminaga-princess=Dat Hon-wine=Gen oak.leaf=Acc take-Caus-Conj sono mikwo=ni tamapi-ki. Dem prince=Dat give-Past ‘(He) had Princess Kaminaga take the oak leaf of wine, and then (he) gave her to the

prince.’ b. 其 御子 令拝 其 大 神 宮 Dem prince Caus.pray Dem great god shrine 将 遣 之 時 (Kojiki, Keiko) Mod send Gen time

6

=> sono mikwo=wo site sono Opo kamwi=no miya=wo Dem prince=Acc do Dem great god=Gen shrine=Acc worogama-sime-ni tukapasa-mu-to-se-si toki pray-Caus-Purp send-Mod-C-do-Past.Adnom time ‘When (he) was about to send the prince to pray at the shrine of the great god....’ Interestingly, the verb follows the causee in (11). This is also true in (12), in which there is no overt DP following the verb. In (12a), the embedded verb is unergative. In (12b) it is transitive, but the object is not overtly expressed. (12) a. 思金神 令思 而 (Kojiki, Amaterasu) Omoikanenokami Caus.think Conj => Omopikanenokamwi=ni omopa-sime-te O=Dat think-Caus-Conj ‘(They) had Omoikanenokami think, and….’ b. 百 官 令拝。 (Kojiki, Richu) all official Caus.bow => Mwomwo=no tukasa=wo si-te worogama-sime-tamapu=ni all=Gen official=Acc do-Conj bow-Caus-Hon=Conj ‘(He) made all the officals bow (to them).’ The correct generalization, I suggest, is that when the causee is agentive, i.e. is base merged in the specifier of a vP projection, then it precedes the causative verb. Unsurprisingly, then, an internal argument causee follows the verb. (13) a. 令為舞 其 嬢女。 (Kojiki, Oryaku) Caus.do.dance Dem maiden => Sono wotomye=ni mapi-se-sime-tamapi-ki. Dem maiden=Dat dance-do-Caus-Hon-Past ‘(He) made the girl dance.’ b. 令副 誰 人 者 吉？ (Kojiki, Suinin) Caus.accompany who person Cond fortuitous => [Tare bito=wo sope-sime-ba] yokye-mu? who person=Acc accompany-Caus-Cond good-Mod ‘If we assign whom to him, will it be fortuitous?’ This asymmetry is reminiscent of well-known contrast in modern Japanese causatives in which the causee when the embedded clause is transitive must appear in dative case (14a), while the causee when the embedded clause is intransitive can be either dative (14b) or accusative (14c). But when the causee is dative, it can receive an agentive (or at least volitional) interpretation, while it cannot be agentive when it has accusative case. (14) a. Taroo=ga kodomo=ni hon=o yoma-se-ta. Taro=Nom child=Dat book read-Caus-Past ‘Taro made/let the child read a book.’

7

b. Taroo=ga kodomo=ni ika-se-ta. Taro=Nom child=Dat go-Caus-Past ‘Taro made/let the child go.’ c. Taroo=ga kodomo=o ika-se-ta. Taro=Nom child=Dat go-Caus-Past ‘Taro made the child go.’ 2.4. Systematic Deviation 4: Wh-phrases Finally, I return to the position of wh-phrases. There are examples in the Kojiki in which a wh-phrase can precede the subject. (15) a. 何 由 汝 泣 伏？ (Kojiki, Okuninushi) what from you cry lie.down => Nani=no ywosi=ni=ka na=ga naki-puseru? what=Gen reason=Dat=Foc 2s=Gen cry-lie.down ‘Why are you lying here crying?’ b. 何 汝 兄 久 不 参出。 (Kojiki, Keiko) what you brother long.time not Hum.come.out => Nanisikamwo imasi=no irose=pa pisasiku mawi-de-zaru? why you=Gen brother=Top long.time Hum.come-out-Neg ‘Why has your brother not come out in a long time?’ According to Nomura’s (1993) generalization, Old Japanese wh-phrases were required to precede genitive subjects. (16) Nomura (1993) XPTop … YPQ … DPGen … VAdnom Classical and Middle Chinese wh-phrases, on the other hand, had to follow the subject. (17) a. 汝 何 故 哭？ (Baiyujing 11) Ru [he gu] ku? you what reason cry ‘Why are you crying?’ b. 汝 何以 自 毀？ (Baiyujing 96) Ru heyi zi hui? you why self mutilate ‘Why did you mutilate yourself?’ 2.5. Section Summary In this section, I have considered several types of word order in the Kojiki which deviate from Classical and Middle Chinese. However, the deviations are not the result of random errors in attempting to write Chinese. Rather, these word order patterns are systematic and conform to Old

8

Japanese word order patterns. I suggest here that these word orders are sensitive to phrase structure. The account I propose is as follows. (18) Account, Part 1 HTKB word order is derived from surface Japanese by switching head and complement

order within VP and PP. Each of the word order types considered so far in this paper can be accounted for by assuming that the author employs (18) as his word order encoding process in creating the text from surface Japanese sentences. This is straightforward in the case of a verb and its complement direct object. The verb is reordered from its Japanese position following the object to a position immediately to the left of the object. (19) [VP [V’ [soko=no awona]=wo tumu]] there=Gen vegetable=Acc pick => 採 其 地 之 青菜 pick [Dem place Gen vegetable] ‘pick the vegetables of that place’ For the existential construction, there are two possible accounts. On one, we could assume that nominative subjects are always required to move to [Spec, TP]. Therefore, reordering in the VP will not affect the position of the subject with respect to the verb. The other possibility is that the indefinite subject remains in its base position. However, this position will be a specifier position. Consequently, reordering of the verb and its complement will likewise not affect the position of the subject. (20) [TP Pwiuti ... [VP tSubj [V’ [sono uti=ni] ari-ki]] flint Dem inside=Dat exist-Past => 火打 有 其 裏。 flint exist Dem inside ‘There was a flint inside it.’ In the case of a passive construction, regardless of whether the agent is treated as an argument in [Spec, vP] or as an adjunct in the verbal domain, it will not be located in the verb’s complement position. Reordering will consequently not change its position with respect to the verb. (21) Na=pa [vP ware=ni [VP tTop azamuka-e-tu]] you=Top me=Dat deceive-Pass-Past => 汝 者 我 見欺。 you Top me Pass.deceive ‘You have been deceived by me.’ The asymmetry between the transitive/unergative and unaccusative causative construction also receives a straightforward account. In the first case, the causee is located in a specifier of vP, so reordering within the VP will have no effect on its position in the text. For the unaccusative case, the causee is an internal argument, merged as complement to the verb.

9

(22) a. [vP Omopikanenokamwi=ni [VP omopa-sime-te]] O=Dat think-Caus-Conj => 思金神 令思而 Omoikanenokami Caus.think.Conj ‘(They) made Omoikanenokami think, and….’ b. [VP [V’ Sono wotomye=ni mapi-se-sime-tamapi-ki]] Dem maiden=Dat dance-do-Caus-Hon-Past ‘(He) made the girl dance.’ => 令為舞 其 嬢女。 (Kojiki, Oryaku) Caus.do.dance Dem maiden Finally, the position of wh-phrases falls out naturally, since these will be located in a position well above the VP and therefore will be unaffected by reordering of the verb. (23) Nani=no ywoshi=ni=ka [vP na=ga [VP naki-puseru]]? what=Gen reason=Dat=Foc 2s=Gen cry-lie.down => 何 由 汝 泣伏？ what from you cry-lie.down ‘Why are you lying here crying?’ 3. Further evidence for Head-Complement order switch: Ditransitives In this section, I discuss additional evidence for the analysis proposed in section 2. As seen briefly in (4) above, ditransitive clauses in the Kojiki deviate from the expected word order in Chinese. In Chinese, ditransitive clauses are divided into the following two types. Chinese has a dative construction, in which the direct object is a DP and the indirect object a PP. Chinese also has a double object construction, in which both objects are packaged as DPs. (24) shows examples of each type. (24) 天子 能 薦 人 於 天， Tianzi neng [jian ren yu tian] ruler can recommend person to heaven 不 能 使 天 與 之 天下。 (Mencius 9) bu neng shi tian [yu zhi tianxia] not can make heaven give 3.Obj world ‘The ruler can recommend someone to heaven, but (he) cannot make heaven give him the

world.’ It is possible for the direct object to precede the verb, but this in turn is preceded by the object marker yi. This is the precursor of the modern Chinese disposal construction using ba3. The archaic Chinese equivalent was yi. I gloss it below as ‘object marker’ for convenience4.

3 For discussion of the modern Mandarin disposal construction the reader is referred to Li and Thompson (1981), Cheng (1988), Sybesma (1999), Bender (2000), among many others.

10

(25) 天子 不 能 以 天下 與 人。 (Mencius 9) Tianzi bu neng yi tianxia [yu ren]. ruler not can OM world give person ‘The ruler cannot give the world to someone.’ In middle Chinese, PPs could also appear preverbally. (26) 於 樓上 得 一 磨 石。 (Baiyujing 18) Yu loushang [de yi mo shi] P upstairs obtain one sharpening stone “He got a sharpening stone upstairs.” This order is attested in the Kojiki, as in (27) DAT V ACC (27) 於 其 隼人 賜 大臣 位。 (Kojiki, Richu) to Dem Hayabito give minister rank => Sono Hayabito=ni opomayetukimi=no kurawi=wo tamapi. Dem Hayabito=Dat minister=Gen rank=Acc give ‘(He) gave the Hayabito the rank of minister.’ What is unexpected from the standpoint of Chinese is the type seen in (28), in which the direct object appears unmarked before the verb. ACC V DAT (28) 多 禄 給 其 老 女 (Kojiki, Oryaku) many thing give Dem old woman => Amatano mono=wo sono omina=ni tamapi-te many thing=Acc Dem old.woman=Dat give-Conj ‘He gave many things to the old woman, and….’ 3.1. Information Structure One question which must be addressed at this point is whether the appeared preverbally in Classical and Middle Chinese. But they preceded the subject in surface order and did not appear between the subject and VP5.

4 Note that glossing yi as an object marker is intended as an endorsement of any particular analysis of yi. Precise analysis of the disposal construction is immaterial to the discussion at hand which merely offers the empirical observation that nontopicalized objects could appear preverbally in this construction. 5 Topicalized objects were resumed by pronouns in argument position in the VP, a fact which is tangential to the discussion at hand.

11

(29) 子路，人告之以有過。 (4th C. BCE; Mencius, Gongsun Chou 1) Zilu, ren gao zhi yi you guo. Zilu person tell 3.ACC that have error ‘Zilu, someone told him he made a mistake.’ In addition to the word order, information structure also makes it unlikely that (28) is a topic construction. A topicalized constituent generally occurs with a topic marker 者. Note that the topicalized object refers back to the preceding sentence. (30) a. 生 子 水蛭 子。 (Kojiki, Izanagi) bear child leech child => Ume-ru kwo piru kwo. bear-Adn child leech child ‘The child which (she) bore was a leech child.’ b. 此 子 者 入 葦 船 而 流 去。 this child Top put reed boat Conj float discard => Kono kwo=pa asi.bune=ni ire-te nagasi-ute-ki. this child=Top reed.boat=Dat put-Conj float-discard-Past ‘This child, (they) put into a reed boat and floated (him) away.’ Furthermore, a preverbal object can be completely novel in the Kojiki. The preverbal objects in the following examples come into existence as a result of the event. Hence, they could not have been referred to in the preceding discourse. (31) a. 名 賜 曙立 王 (Kojiki, Suinin) name give Aketatu prince => [VP na=wo [V’ [Aketatu=no Opokimi]=ni tamapi-te name=Acc Aketatu=Gen prince=Dat give-Conj ‘(He) gave a name to Prince Aketatu, and….’ b. 爾 其 國造 火 著 其 野。 (Kojiki, Keiko) then Dem local.chief fire set Dem field => Kokoni sono kuni=no miyatukwo pwi=wo sono nwo=ni tuke-ki. then Dem local=Gen chief fire=Acc Dem field=Dat set-Past ‘Then the chief set fire to the field.’ 3.2. Ditransitive Analysis Aldridge (2001) accounts for the verb-medial positioning of the verb in Kojiki ditransitives by claiming that word order in this text reflects underlying head-initial word order in which the verb surfaces in its base position in VP. Following Miyagawa (1997), Aldridge assumes that the two objects can be base merged in either order, one as the verb’s complement and the other in the specifier position.

12

(32) a. VP DP=Acc V’ V DP=Dat b. VP DP=Dat V’ V DP=Acc However, there is an immediate concern with this structure in that it does not allow for the word order in the text to reflect verb-movement to v. There is compelling evidence in the text that verb-movement does taken place, given that verbs appear fully inflected. (33) shows the modal, which is written with the Classical Chinese character expressing future modality and encoding the old Japanese modal suffix –mu. (33) 将 入 海 時 (Kojiki, Keiko) Mod enter sea time => umi=ni iri-tamapa-mu-to-su-ru toki=ni sea=Dat enter-Hon-Mod-C-do-Adnom time=Dat ‘when (she) was about to enter the sea’ (34) shows an example of this modal in Classical Chinese. As in the Kojiki example in (33), the modal immediately precedes the VP. (34) 吾 將 問 之。 (Analects 1) Wu jiang wen zhi. I Mod ask 3.Obj ‘I will ask him.’ However, in the Kojiki, the modal always appears in immediate pre-verbal position and is never separated from the verb. (35) 自 出雲 將 上坐 倭國 而 from Idumo Mod go.Hon Yamato Conj => Idumo=ywori Yamato=no kuni=ni nobori-masa-mu to si-te Izumo=from Yamato=Gen country=Dat go-Hon-Mod C do-Conj ‘As (he) was about to go to Yamato from Izumo, ….’ Note in particular that the modal follows a PP in (35). (36) shows an Early Middle Chinese example with a similar type of PP. In both Middle and Classical Chinese, this modal was a free morpheme and occupied a specific syntactic position high within the TP structure (Wei 1999 and Aldridge 2010). Crucially, for the purposes at hand, this modal precedes a PP of the type seen in (33).

13

(36) 公 將 自 東 征 備。 (Sanguozhi 1.1) Gong jiang zi dong zheng Bei. lord Mod from east attack Liu Bei ‘Our lord was about to attack Liu Bei from the east.’ The contrast between (35) and (36) can be accounted for if we assume that the verb appears in the hentai kambun texts with inflections like the modal suffix attached to it. This entails that either the verb has moved to the position where the modal is base merged or that the verb has moved at least high enough in the structure where it is visible to the probe which can value inflectional features on it. I assume, then, that verbs in the Kojiki have moved to v, where they are located in the edge of the vP phase and are therefore accessible to probes in the inflectional domain immediately dominating vP. A second example comes from causative constructions. Recall from section 2.3 that causative verbs follow agentive causees. (37) 思金神 令思 而 (Kojiki, Amaterasu) Omoikanenokami Caus.think Conj => Omopikanenokamwi=ni omopa-sime-te O=Dat think-Caus-Conj ‘(They) had Omoikanenokami think, and….’ This is unexpected from Chinese, as mentioned before, since the causative marker is a free morpheme. In (38), the causee intervenes between the causative verb and the embedded predicate. (38) 云何 能 令 是 麥 茂 好？ (Baiyujing 82) Yunhe neng ling shi mai mao hao? how can make this wheat dense rich “How can you make this wheat dense and rich?” The discrepancy between Chinese and hentai kambun can again be accounted for if we assume that the causative marker in the Japanese text is a affix on the verb. This suggests in turn that the embedded verb has raised to the position where this morpheme is base merged, specifically causative v. An additional problem for the Aldridge (2001) approach is that it assumes that VP-internal constituents likewise do not dislocate. However, Yanagida (2006) and Yanagida and Whitman (2009) show convincingly that definite or specific objects and raise obligatorily to the edge of vP in Old Japanese. A bare object with a nonspecific interpretation appears in immediate preverbal position. Note that the object follows the genitive subject in (39a). Yanagida and Whitman (2009) assume that these objects are incorporated to the verb. The definite object in (39), however, raises to a position preceding the genitive subject. Yanagida (2006) proposes that this is an instance of object shift to the edge of vP.

14

(39) a. 佐欲比賣能 故何 比列 布利斯 夜麻 (MYS 868) Sayopimye=no kwo=ga pire puri-si yama Sayohime=Gen child=Gen scarf wave-Past.Adn hill ‘the hill where the girl Sayohime waved her scarf’

b. 蜻野叩6 人之 懸者 [CP [vP Akidu nwo=wo [v’ pito=no [VP tObj kakure-ba]]]] (MYS 1405)

Akizu field=Acc man=Gen speak.of-when ‘When a man speaks of the moorland of Akizu…’ Given the above considerations, ditransitives in the Kojiki in which the accusative object precedes the dative can easily be incorporated into the analysis developed so far. Since Japanese surface strings provide input to the encoding process, I assume that the verbal complex resides in v. The first step in the encoding process is the reordering of heads and complements, which I proposed in section 2. This will result in V+v being linearized to the left of VP. If no movement has taken place from VP, internal arguments will follow the verb. However, since specific objects must undergo object shift to the edge of vP, the direct object will appear preverbally in the text, since it is located in a specifier of vP, just to the left of the reordered verb. (40) a. 多 禄 給 其 老 女 (Kojiki, Oryaku) many thing give Dem old woman => Amatano mono=wo sono omina=ni tamapi-te many thing=Acc Dem old.woman=Dat give-Conj ‘He gave many things to the old woman, and….’ b. vP DPObj v’ tSubj v’ V+v VP tDP V’ <V> DP 4. Generalization of Raising7 In this section, I add a second step to the encoding process. This will make result in the following two step process.

6 The character in the text is slightly different, composed of a mouth and a blade. 7 This proposal was originally inspired by a suggestion by John Whitman (personal communication).

15

Encoding Process (41) 1. Heads linearized before complements 2. Optional raising from complement (as part of coding process) Crucially, I wish to suggest here that the second step was available to the author of the Kojiki text because this transformation was a process routinely employed in the derivation of natural utterances in Old Japanese. Specifically, this is the object shift process discussed in the preceding section. I demonstrate the need for the second step with one final type of anomalous word order in the text. There are a number of examples in which a verb or adposition appears within the complement it selects. In (42), an adposition appears to be inside of its complement DP, following a demonstrative in this DP in (42a) and a possessor in (42b). (42) a. 是 於 河 下 如 青 葉 山 者 [[Dem on river bank] be.like green leaf hill] Top 見 山 非 山。 (Kojiki, Suinin) appear hill not.be hill => [[kono kapa shimwo ni] awo ba=no yama=no goto-ki]=pa this river bank P green leaf=Gen mt.=Gen be.like-Adn=Top yama=to mie-te yama=ni ara-zu. hill=C appear=Conj hill=Dat be-Neg ‘What looks lie a hill of green leaves on this river bank is not a hill.’ b. 所殺 迦具土 神 之 於 頭 所成 神 [[[Pass.kill K god] Gen] from head] Pass.make god 名 正鹿山津見 神 (Kojiki, Izanagi) name Masakayamatumi god => [[[Korosa-e-si Kagututi=no kamwi=no kasira=ni na-re-ru kill-Pass-Past.Adn K=Gen god=Gen head=Dat bear-Pass-Adn kamwi=no na=pa masakayamatumi=no kamwi god=Gen name=Top Masakayamatumi=Gen god ‘The name of the god born [from the head of the slain Kagutsuchi god] was

Masakayamatumi.’ These orders can be accounted for by assuming a raising analysis. I propose that the author of the text co-opted the object shift transformation and applied it in the case of PPs. In effect, this process moved the specifier or adjunct from the complement DP into the specifier of the dominating PP. (43) PP DP P’ | Dem P DP tDP D’ D NP

16

In the final two examples, we see the verb in the higher clause inside a relative clause modifying the complement of the higher verb. The relative clause in (44a) contains a coordinate structure. The first of these conjuncts appears before the matrix verb. In (44b), the subject of a relative clause precedes the matrix verb. (44) a. 作 筌 有 取 魚 人。 (Kojiki, Jimmu) [[set trap exist take fish] person] => Upe=wo puse-te, uwo=wo two-ru pito ari-ki. trap=Acc make-Conj fish=Acc take-Adnom person exist-Past ‘There was a person laying fish traps and fishing.’ b. 天下 者 汝 非 應 知 國。 (Kojiki, Chuai) world Top [[you not.be should govern] realm] => Ame=no sita=pa imasi=no Heaven=Gen below=Top you=Gen sirasu be-ki kuni=ni ara-zu. govern should-Adnom realm=Dat be-Neg ‘The world below is not a realm which you should govern.’ These orders can be accounted for on the current proposal if we assume that the preverbal constituent has been raised into the edge of the vP where the verb resides. First note that relative clauses appear in the text preceding the head they modify. I assume that this is due to influence from Chinese. Relatives in both Chinese and Japanese are prenominal in surface order. I assume the raising analysis of relative clauses, as proposed and developed by (Kayne 1994, Bianchi, Bhatt, and others). (45) is Honda’s (2002) proposal for modern Japanese.. The head nominal is located (base merged for Honda, but raised for Kayne) in a specifier in the CP layer of the clause. The TP portion of the relative clause raises to [Spec, DP], with the result that the head nominal follows it in surface order. (45) [DP [TP Taroo-ga tabeta proi] [D’ [CP ringo tTP]]] Taro-Nom ate apple ‘the apple that Taro ate’ The raising analysis of relative clause structure allows a straightforward analysis of the aberrant word orders in the Kojiki. In (44a), the embedded TP, which is a coordinate structure, has been raised to [Spec, DP]. The specifier within this constituent can subsequently be raised to [Spec, vP] in the matrix clause.

17

(46) vP TP v’ V+v VP V DP &P D’ tTP ... D CP NP C’ C tTP In (44b), it is the embedded subject which raises to the edge of matrix vP. Note that in both cases, it is the specifier of a specifier which moves. (47) vP DP v’ V+v VP tV DP TP D’ tDP ... D CP NP C’ C tTP 5. Conclusion In this paper, I have proposed that surface word order of Kojiki HTKB is derived by the following two-step process. The first step is obligatory, the second step optional. (48) a. Linearizing heads before complements. (deriving head-initial WO) b. Raising a specifier or adjunct from the complement. (deriving head-medial WO) What is striking is the similarity to existing processes assumed to be available in Universal Grammar. The first involves the linear ordering of constituents by fixing or altering the relative positions of head and complement. The second is a transformational process. I have suggested in

18

this paper that this process was available to the hentai kanbun authors because their native language (Old Japanese of the 8th century) regularly employed object shift to move specific direct objects out of VP. Reading of the text in Japanese involves reversal of the two-step algorithm in (48). (49) a. Reconstruction of any raised constituents. b. Linearizing heads after complements. (deriving Japanese surface head-final WO) Appendix Nakagawa (1995) proposes an alternative analysis of ditransitive clauses in the Kojiki. She proposes that verbs are placed between their two objects in order to distinguish the two internal arguments. One problem with this analysis is that it predicts that a verb should be able to intervene between any two clause-mate arguments in order to distinguish them. But as we have seen in section 2.2, verbs follow passive agents, with the result that the internal argument subject and demoted agent are adjacent to each other in preverbal position. (1) 汝 者 我 見欺。 (Kojiki, Okuninushi) you Top me Pass.deceive => Na=pa ware=ni azamuka-e-tu. you=Top me=Dat deceive-Pass-Past ‘You have been deceived by me.’ This is particular mysterious on Nakagawa’s account, given that agents in Chinese passives followed the verb, as shown in section 2.2. Nakagawa’s disambiguation approach has little to say about intransitive constructions in general. In section 2.3., I showed how my structural approach accounts for the asymmetry between agentive and non-agentive causees in causative constructions. The former reside in specifier positions and will consequently not be reordered with respect to the verb in generating HTKB word order, as shown in (2a). But the latter are complements and as such will surface following the verb, as in (2b). It is not at all clear how a functionalist approach could predict such an asymmetry. (2) a. 百 官 令拝。 (Kojiki, Richu) all official Caus.bow => Mwomwo=no tukasa=wo si-te worogama-sime-tamapu=ni all=Gen official=Acc do-Conj bow-Caus-Hon=Conj ‘(He) made all the officals bow (to them).’ b. 令為舞 其 嬢女。 (Kojiki, Oryaku) Caus.do.dance Dem maiden => Sono wotomye=ni mapi-se-sime-tamapi-ki. Dem maiden=Dat dance-do-Caus-Hon-Past

‘(He) made the girl dance.’

19

Perhaps the most damaging evidence against Nakagawa’s approach are examples in which a head surfaces inside its own complement. Since only one argument is involved, no position adjustment should be necessary in order to identify that argument. (3) 是 於 河 下 (Kojiki, Suinin) Dem on river bank => kono kapa shimwo ni this river bank P ‘on this river bank’ References Aldridge, Edith. 2001. Hentai Kambun Perspective on Short Scrambling. Journal of East Asian

Linguistics 10:169-200. Aldridge, Edith. 2009. Short Wh-movement in Old Japanese. Japanese/Korean Linguistics,

Volume 17. S. Iwasaki, H. Hoji, P. Clancy, S. Sohn, eds, 549-563. Stanford: Center for the Study of Language and Information.

Aldridge, Edith. 2010. Clause-internal Wh-movement in Archaic Chinese. Journal of East Asian Linguistics 19.1:1-36.

Frellesvig, Bjarke and John Whitman. 2008. Introduction. In Bjarke Frellesvig & John Whitman (eds.), Proto-Japanese. John Benjamins.

Feng, Liangzhen. 1995. Gudai Hanyu he Gudai Riyu de Jiaorong: Gushiji de Yuyan Tese [Fusion of Old Chinese and Old Japanese: Linguistic Characteristics of the Kojiki]. Zhongguo Yuyanxue Bao 1994.7:114-120.

Honda, Kensuke. 2002a. An Antisymmetric Approach to Japanese Syntax [in Japanese]. In Jisedai no Gengo Kenkyu I, ed. Tsukuba Daigaku Gendai Gengogaku Kenkyukai, 175-224. Tsukuba, Japan.

Honda, Kensuke. 2002b. The Head Parameter and Related Matters [in Japanese]. Kogakuin Daigaku Kyotu Katei Kenkyu Ronso 40.1:109-122.

Miller, Roy A. 1967. The Japanese Language. Tokyo: Tuttle. Minegishi, Akira. 1986. Hentai Kambun. Tokyo: Tokyo-do Press. Miyagawa, Shigeru. 1997. Against Optional Scrambling. Linguistic Inquiry 28.1:1-25. Nakagawa, Yukari (1995) Kojiki no Kufuu: Mokutekigo (o-kaku) o Meiji Suru Tame ni

(Innovation in the Kojiki -- to Clarify the Direct Object (Accusative Case)), in Buntaishi, Gengo Seikatsushi, Kooza Kokugoshi, vol. 6, Taishuukan, Tokyo

Nishimiya, Kazutami. 1993. Kojiki no Kenkyu (Research on the Kojiki), Oofuusha, Tokyo Nomura, Takashi. 1993. Joodaigo no ‘no’ to ‘ga’ nituite [Old Japanese ‘no’ and ‘ga’]. Kokugo

Kokubun 62.2:1-17. Rabinovitch, Judith N. 1996. An Introduction to Hentai Kambun (Variant Chinese). Journal of

Chinese Linguistics 24.1:98-127. Sema, Masayuki. 1999. Kanji de Kakareta Kotoba: Kundokuteki Shiso o Megutte [Words

Written in Chinese Characters: On the Thought of Kundoku]. Kokugo to Kokubungaku 76.5:36-36.

Tsukishima, Hiroshi. 1963. Heian Jidai no Kambun Kundokugo nituite no Kenkyu [Research on reading kambun in the Heian Period]. Tokyo University Press.

20

Uchida, Kentoku. 1995. Kojiki no ‘Bun’ [‘Sentence’ in the Kojiki]. In Kojiki Kenkyu Taikei: Kojiki no Kotoba [Kojiki Research Series: Language], ed. by Kojiki Gakkai, 47-67. Tokyo: Takashina.

Wei, Pei-chuan. 1999. Lun Xian-Qin Hanyu Yunfu de Weizhi [On the Position of Operators in Pre-Qin Chinese]. In A. Peyaube and C. Sun, eds., Linguistic Essays in Honor of Mei Tsu-lin: Studies in Chinese Historical Syntax, 259-297. Paris: Center de Recherches Linguistiques sur l Asie Orientale.

Whitman, John (2000) KAYNE 1994: P. 143, FN. 3, to appear in Galina Alexandrova (ed.) The Minimalist Parameter, John Benjamins, Amsterdam

Yamaguchi, Yoshinori. 1995. Kojiki no Hyoki to Kundodu [Orthography and Translation of the Kojiki]. Tokyo: Yuseido.

Yanagida, Yuko. 2006. Word Order and Clause Structure in Early Old Japanese. Journal of East Asian Linguistics 15:37-67.

Yanagida, Yuko and John Whitman. 2009. Alignment and Word Order in Old Japanese. Journal of East Asian Linguistics 18.2:101-144.