Ahn - 1
Nominative-Genitive Conversion in Japanese A Consequence of the Syntax Byron Ahn1
0. INTRODUCTION In certain Japanese embedded clauses, there is a well-documented set of phenomena known as Nominative-Genitive Conversion (NGC)2.
In structures of this type, subjects3 of NP
complements or of relative clauses can seemingly freely4 alternate between the use of NOM case and GEN case. Examples of this can be seen below in (1) and (2). (1)
Complex NP Embedded Clauses John ga/no ki.ta koto wa sira.na.katta John NOM/GEN come.PERF fact TOP know.not.PERF “[I] didn't know that John came.”
(2)
Relative Clauses John ga/no kai.ta hon wa omosiro.i John NOM/GEN write.PERF book TOP interesting.IMPF “The book that John wrote is interesting.”
Contrasting the embedded clauses in (1) and (2) to the matrix clauses in (3) and the non finite clauses in (4), we can see clearly the point Harada makes when he states that NGC applies “only to…embedded sentences with finite main verbs.” (1971:27)
1
I would very much like to extend all my thanks and gratitude to all those who helped me in the writing of this paper. I would especially like to thank: Carol Rosen, for helping me get started with this idea two years ago; John Bowers for his guidance in Minimalist syntax, and for graciously donating his time to help me refine my first draft and to read on my honors committee; Wayne Harbert, for making the honors thesis process a smooth one, and for reading on my honors committee; Shigeru Miyagawa, Masao Ochi, Yuji Takano, Akira Watanabe and Tomoyuki Yoshida, for extending their extremely helpful hands in delving into Japanese linguistics; the Japanese linguists at Cornell, namely Yukiko Akasaka, Yumiko Nishi, Yasuhiro Shirai and Kazuha Watanabe, for their willingness to donate their time and thoughts; the others who gave native speaker data, without whose judgments I could not have formulated all my ideas; and my peers for giving me their unabashed opinions. Most importantly, however, I would like to thank John Whitman for taking me on as an honors thesis advisee; without his counsel and direction, I would still be on page one with this idea of an idea. Finally, I would like to thank the entire Cornell Department of Linguistics, for providing me with an irreplaceable undergraduate experience. 2 Also referred to in the literature as No-Ga Conversion (NGC), or Ga-No Conversion (GNC). However, it is worth noting that I am not pursuing any theory which would “convert” anything. 3 “Subject” is a slightly vague term; by it, I mean any DP that would “normally” be marked in the nominative case. This will also be able to account for so-called nominative objects being genitive marked at times. However, as Ochi (2001) points out, this does not include ga-marked PPs. 4 It should be mentioned that some scholars say that this alternation is purely optional in a certain kind of structure; others, myself included, would say there is a fundamental difference in structure.
Ahn - 2 (3) a)
Matrix Clauses John ga/*no ki.ta John NOM/GEN come.PERF Intended: “John came.”
b)
Udon ga/*no tsukur.are,ta udon NOM/GEN prepare.PASS.PERF Intended: “The udon was prepared.”
c)
Omae ga/*no ko.i you NOM/GEN come.IMPTV Intended: “You come!”
d)
Ware.ware ga/*no ik.oo ka? we NOM/GEN go.VOL Q Intended: “Shall we go?”
(4) a)
Other Embedded Clauses Boku wa [John ga/*no kai.te] hosi.i 1SG TOP John NOM/GEN write.CONT want.IMPF “I want John to write [it].”
b)
[John ga/*no su.reba] i.i John NOM/GEN do.COND good.IMPF “It will be good if John does [it].”
c)
[Mary ga/*no ik.u] kara boku ga ik.ana.i Mary NOM/GEN go.IMPF because I NOM go.NEG.IMPF Intended: “Because Mary will go, I will not go.”
d)
Kyoo wa Taro ga/*no mie.na.i to omo.u today TOP Taro NOM/GEN show-up.NEG.IMPF COMPL think.IMPF Intended: “I think Taro won’t show up today.”
Here we see that all matrix clauses, no matter their mood/voice, don’t allow NGC to take place. Furthermore, other embedded clauses besides relatives and NP complements are unable to be an environment for NGC, no matter if they are finite (4c-d) or not (4a-b). What I seek to achieve in this paper is to argue that NGC and its correlates are not independent phenomena of the Japanese language. Rather, I will do my best to show that the facts about NGC are consequences of other facts of the language. I will do so by structuring this paper as follows: Section 1 delves deeper into some of the basic facts about NGC, Section 2 covers generally the positions of several key papers in the literature of NGC, Section 3 explains
Ahn - 3 the deficiency of these inquiries, Section 4 presents a new theory of NGC, Section 5 presents areas that require further research, and Section 6 concludes this paper.
1. MORE FACTS `ON NGC 1.1 Other Verbal Forms NGC is possible with most any voice and mood on the finite verb of a clause embedded in a noun phrase. When interviewing native speakers, they rather unvaried showed acceptance of genitive subjects in unaccusative, unergative, active, passive5, causative, and potential (and any possible combination thereof) contexts. Furthermore, even adjectives and certain instances of the copula allow genitive subjects. Data of some of these kinds of situations can be seen below: (5)
Passive a) Boku wa keeki ga/no tabe.rare.ta inu o mi.ta I TOP cake NOM/GEN eat.PASS.IMPF want.IMPF “I saw the dog who the cake was eaten by.” Causative b) Ichiro ga/no musuko ni s.ase.ta shukudai wa yasashi.katta Ichiro NOM/GEN son DAT do.CAUS.PRF homework TOP easy.PRF “The homework that Ichiro made his son do was easy.” Unergative c) Mary ga/no yoku ne.na.i riyuu wa naze desu ka? Mary NOM/GEN well sleep.NEG.IMPF reason TOP why COP-IMPF-POL Q “What is the reason that Mary doesn’t sleep well?” Unaccusative d) Ithaca ni yuki ga/no toke.ru tsuki wa taitee sangatsu da Ithaca LOC snow NOM/GEN melt.IMPF month TOP usually March COP-IMPF “The month that the snow melts in Ithaca is usually March.” Potential e) Harry ga/no
oyog.e.ru
yoo ni mainichi YMCA ni
ik.ase,ru
Harry NOM/GEN swim.POT.IMPF way DAT every-day YMCA DAT go.CAUS.IMPF “[Someone] makes Harry go to the YMCA every day so that he will be able to swim” 5
However, the suffering passive cannot be found in any relative clause where the ni-marked subject is extracted, as below: * Musuko ga /no sin.are.ta titi wa zutto byooki datta. son NOM/GEN die.PASS.PRF father TOP always sick COP-PRF Intended: ‘The father who died on his son was always sick.’
Ahn - 4
Passive-Causative f) Shinbun ga/no yom.ase.rare.ta kodomo wa joozuni yom.ana.i Newspaper NOM/GEN read.CAUS.PASS.PRF child TOP skilled read.NEG.IMPF "The child who was made to a newspaper does not read well.” (6)
Attributive Adjective a) Carol ga/no taka.i
riyuu wa miruku o
yoku nom.u
kara
da
Carol NOM/GEN tall.IMPF reason TOP milk ACC often drink.IMPF because COP-IMP
“The reason that Carol is tall is that she drinks milk often.”
“Transitive” Adjective b) Kare ga/no kowa.i monstaa wa gojira da He NOM/GEN scary.IMPF monster TOP Godzilla COP-IMPF “The monster that he is scared of is Godzilla.” “Transitive” Adjective with Copula c) Naomi ga/no suki na wain wa shanpen da Naomi NOM/GEN likeable COP-PA wine TOP Champagne COP-IMPF “The wine that Naomi likes is Champagne” It should be noted that these verbs/adjectives must be either perfect or imperfect in tense (that is, no untensed verbs/adjectives). This may not necessarily be a reliance on tense, instead, it may be due to the fact that only tensed verbs may appear in a clauses that are embedded in noun phrases. This may be evidence that tense isn’t what really matters – the external DP is. 1.2 Optionality Semantically speaking, it seems the choice between using nominative and genitive case may be close to null6. It has been claimed that genitive marked subjects may have a higher scope than nominative marked ones (Miyagawa 1993)7. In asking native speakers for judgments on scope interaction, I, too, have found that there is a difference between a subject being marked nominative Case and being marked genitive Case. The following is an example of that judgment.
6
Some native speakers report a difference in register, “sophistication,” focus and/or gentleness. However, the genesis of these facts most probably lies in a historical explanation of the (relatively) recent attrition of genitive subjects and/or the rise in default-ness of the focus subject marker ga. 7 This data has also been contested (Hiraiwa 2001, and others).
Ahn - 5 (7)
Scope Variation and NGC a) Gakusee-tachi ga yon.da yon.satsu no hon wa tsumarana.i Student-PL NOM read.PRF 4.CL GEN book TOP boring.IMPF “The four books that the students read were boring.” i. 3For a set group of four books, the students read them. Books > Students ii. * For the set of students, they each read 4 (different) books. Students > Books b) Gakusee-tachi no yon.da yon.satsu no hon wa tsumarana.i Student-PL GEN read.PRF 4.CL GEN book TOP boring.IMPF “The four books that the students read were boring.” i. ? For a set group of four books, the students read them. Books > Students ii. 3For the set of students, they each read 4 (different) books. Students > Books
At the very least, this should indicate that the genitive subjects let (make?) the scope interpret the subject in a higher position than the relativized object it is interacting with. Since even this kind of scope effect relies on the syntax, namely the height of the subject NP, I will depart from studying the scope further, and I will focus on the syntactic nature of this structure. 1.3 The Transitivity Restriction (TR) Japanese NGC has a very interesting additional fact to it. Just in the environment where the subject is marked genitive, accusative objects are disallowed8. This should seem to suggest that the syntactic conditions for use of ga and use of no are separate, and something about conditions for no don’t allow accusative objects. Determining the correct syntax of the TR should give us quite an insight on the syntax of NGC. Later, in section 4.4, I will introduce how the TR fits into my own theory. 1.3.1 TR in Action Examples of the TR are below in (8) and (9). (8) a)
8
TR in NP Complements [John ga (kuruma o) ka.u] koto ga yuumei da. John NOM (car ACC) buy.IMPF fact NOM well-known COP-IMPF “It is well known that John will buy (a car).”
At least at the phonological level. Perhaps a phonologically empty category, such as PRO, could enter the derivation. If we assume this, we must also assume that PRO does not have any features to be checked or Case that needs to be valued, as demonstrated by English Control constructions such as “I want [PRO to go].” The infinitive in English does not assign Case, so we must assume that PRO has no Case and that this is expected.
Ahn - 6 b)
(9) a)
b)
[John no (*kuruma o) ka.u] koto ga yuumei da. John GEN car ACC buy.IMPF fact NOM well-known COP-IMPF “It is well known that John will buy (a car).” TR in Relative Clauses [John ga sushi o tabe.ta] resutoran wa asoko da. John N OM sushi ACC eat.PERF restaurant TOP over-there COP-IMPF “The restaurant where John ate sushi is over there.” [John no (*sushi o) tabe.ta] resutoran wa asoko da. John GEN sushi ACC eat.PERF restaurant TOP over-there COP-IMPF “The restaurant where John ate (*sushi) is over there.”
I should be explicit about what is happening in (9b). Here, we have attempted to relativize something other than the object (making this a non-object relative clause) while leaving the possibility that object could be overt in the embedded clause. This is also the case with a goal datives, as shown below. (10) Relativized Goals and TR a) John no tomodati ni kasi.ta hon John GEN friend DAT lend.PERF book “The book that John lent to a friend” b) * John no hon o kasi.ta tomodati John GEN book ACC lend.PERF friend Intended: “The friend that John lent a book” c) John no PRO kasi.ta tomodati John GEN lend.PERF friend “The friend that John lent [something]” 1.3.2 TR as a Historical Change While the Transitivity Restriction is a crucial aspect of NGC, it is only a relatively new innovation of the language. In 1971, Harada noted that speakers split into two major groups when it came to the TR – speakers under 40 generally rejected clauses with both genitive-marked subjects and accusative objects, whereas most people over 40 accepted them. As would be expected, then, these days, the majority of speakers hold the TR as a hard and fast rule for grammaticality judgments.
Ahn - 7 Hiraiwa (2002) and Whitman (2006) claim this is part of a larger change that started as early as the 13th century. At that time, dialects of Japanese began the merger of rentaikei (attributive form, for clauses followed by a nominal element) and shūshikei (conclusive form, for sentence-final matrix clause verbs) verbal inflections9. Before this merger, only rentaikei allowed for marking subjects in any way – with genitive markers ga10 and no – and shūshikei subjects were unmarked. This merger meant a loss of affixal identity for triggering genitive marking on subjects. Also, as time went on, ga’s identity as a genitive marker was lost, and it was learned by new generations as a nominative marker, therefore, no became the only genitive marker. The loss of a visible trigger for genitive subjects caused them to be used less and less as no perceivable element was driving its use any more. In turn, contexts where no appeared as the subject marker also lessened, causing use to decline faster as well. According to data from Nambu (2005)’s research on a certain vast corpora of texts, the probability of using a genitive-marked subject has decreased from ~22% to ~7% over just the past 130 years. Whitman (2006) argues that the rentaikei was used to marker [+N] CPs, which made first language acquisition of the genitive subject possible and uncomplicated. In this way, when the marker of [+N] disappeared phonologically, it became more complicated as learners for learners of the languages, thus its use became more restricted, accounting for the decline noted in Nambu (2005). Moreover, Whitman notes that Middle Korean, which had an NGC pattern, lost it after losing an overt [+N] CP marking, which lead to a Transitivity Restriction in Late Middle Korean. Shortly thereafter, NGC in Korean was lost permanently. In this way, he says that “Korean provides a preview for developments in Japanese;” in other words, he expects NGC to continue to decline in use until it no longer exists. 9
This merger took place on all verbs, with one exception – the copula. The copula in its matrix form is da whereas the rentaikei form is na. We will see more on this later. 10 At this point, ga was a genitive marker – this is still seen in such frozen phrases such as wa ga ie – ‘my home.’
Ahn - 8 1.4 Internal vs. External Theories The most widely accepted stances on NGC are generally able to be broken down as either relying on an internal licenser (e.g., C-T-v amalgamate in Hiraiwa 2001, 2002, and the [+N] CP in Whitman 2006) or relying on an external licenser (e.g., external D Agreement in Miyagawa 1993, 2003, Ochi 2001, 2005) for the use of genitive subjects. Along the lines of a [N] feature motivation (or a C-T-v amalgamate which is similar in nature), in earlier forms a Japanese a [+N] CP did not require an external DP to induce the usage, as in data from Hiraiwa (2002), though he claims such a pattern is still good today. I replicate some of his data below. (11) a) John wa [ame ga/no yam.u] made office ni i.ta. John TOP rain NOM/GEN stop.IMPF until office LOC exist.PRF ‘John was at his office until the rain stopped.’ b) John wa [Mary ga/no yon.da] yori takusan no hon o yon.da John TOP Mary NOM/GEN read.PRF than many GEN books ACC read.PRF ‘John read more books than Mary did.’ (Hiraiwa 2002:3)
Maki and Uchibori, however, note that many modern speakers of Japanese do not like this kind of example, and Whitman and many native speakers concur that these feel “archaic” in contemporary Tokyo Japanese. However these sentences are vastly improved when an explicit noun occurs after the relative clause, as in the data below. (12) a) John wa [ame ga/no yam.u] toki made office ni i.ta. John TOP rain NOM/GEN stop.IMPF time until office LOC exist.PRF ‘John was at his office until the time that the rain stopped.’ b) John wa [Mary ga/no yon.da] hon yori takusan no hon o yon.da John TOP Mary NOM/GEN read.PRF book than many GEN books ACC read.PRF ‘John read more books than the books that Mary read.’ This data suggests that speakers have shifted away from using an [+N] CP as the catalyst for NGC, and that instead, they rely on an DP. As a result of this kind of data, I will continue in
Ahn - 9 the tradition of Miyagawa and Ochi while still composing my own theory and resolving previous theoretical inconsistencies and shortcomings.
2. PREVIOUS LITERATURE Since Harada 1971, many have attempted to give a syntactic explanation for the alternation between nominative and genitive subjects in Japanese – most recently and notably, Hiraiwa 2001, 2002, Maki and Uchibori 2005, Miyagawa 1993, Ochi 2001, 2005, Watanabe 1996. In this section, I will review each of these papers, their theoretical contributions, and the problems specific to their paper. 2.1 Empirical Description Harada (1971, 1976) was the first to systematically set out to define and account for NGC. While he achieves these goals, he also cedes that “exact formulation of Ga-No Conversion is impossible at present,” so the amount of theoretical work is limited. He identifies three main facts: (1) the ga which alternates with no is not always a subject-marking ga, (2) this is restricted to embedded clauses with finite verbs, and (3) these embedded clauses must be embedded within a noun phrase. While he contributes greatly by his research, he “does not entertain a structural difference between the genitive and the nominative subject, instead assuming that both are possible in the regular subject position.” (Miyagawa 1993) 2.2 Wh-Movement Watanabe (1996), (rightfully) unsatisfied with the existing accounts of NGC, seeks to account for NGC with mechanisms that are cross linguistically attested. Specifically, he matches NGC to French Stylistic Inversion, using wh-movement as the driving force behind both. The following tree comes from Hiraiwa (2001)’s summary of Watanabe (1996).
Ahn - 10 (13) CP Op[wh]
C'
AgrsP TP VP
C[wh]
= wh-agreement
Agrs T
V'
DPsubj tOP
V
As Hiraiwa (2001) point out, it is unclear as to how Japanese wh-agreement manifests itself as genitive Case marking a subject. 2.3 Feature Checking by D Working to strengthen the more traditional approach to NGC, Miyagawa proposed a theory of feature checking and genitive case assignment via an external head, namely D. It had been previously thought that NGC is a side effect of the generalization that “all XP’s directly dominated by a projection of N or D must be marked with the genitive case marker.” (Miyagawa 1993) However, he showed full well that this is not the case through data showing the no marked subjects structurally acting like subjects. The foundation of Miyagawa (1993)’s theory is that the genitive marked subject covertly11 raises to Spec DP at LF. This accounts for data Miyagawa found whereby genitive subjects have a higher scope position than its nominative counterparts. This data is shown below. (14) Scope Differences between Nominative and Genitive a) [[[Rubii ka shinju] ga yasu.ku nar.u] kanousei] ga 50% ijou da ruby or pearl NOM cheap.CONT become.IMPF probability NOM 50% more COP-IMPF i. “The probability that rubies or pearls will become cheap is over 50%” ii. * “The probability that rubies will become cheap or the probability that pearls will become cheap is over 50%” 11
He posits covert movement because the DP in question does not phonologically cross over elements that intervene between the subject and the licensing Spec DP.
Ahn - 11
b) [[[Rubii ka shinju] no yasu.ku nar.u] kanousei] ga 50% ijou da ruby or pearl GEN cheap.CONT become.IMPF probability NOM 50% more COP-IMPF i. “The probability that rubies or pearls will become cheap is over 50%” ii. “The probability that rubies will become cheap or the probability that pearls will become cheap is over 50%” This data is crucial because it shows that genitive subjects do not interact in the same way as nominative marked subjects – necessitating that some other unique syntactic operation(s) to take place. Elaborating on Miyagawa (1993), Ochi (2001) stipulated that the genitive subject may covertly raise to Spec DP, but it ay also overtly raise to Spec DP. He uses the idea that the presence of the feature-checking D head, DAgr is only optionally in the tree. The following is an excerpt from Ochi (2001), formalizing his notion of optional DAgr. (15) “When the DAgr head is present, it triggers overt movement of the genitive subject into the spec of
DAgrP (another instance of the EPP). The Case of the genitive subject is checked against N within the projection of the DAgr head (Case checking is mediated by Agr). When it is absent, the genitive Case of the subject is checked against the N head (such as kanousei ‘probability’) after the formal features of the genitive subject are raised to the position of N in cover syntax.”
One obvious issue with these arguments is that covert raising has since been eliminated from Minimalism, with the loss of the syntactic operations being able to take place in LF but not PF. 2.4 The P-A Form Hiraiwa (2001) has proposed a distinctive theory, in which genitive case (in totally optional distribution with nominative case) is selected in relative and embedded clauses by a specific verbal form. The evidence he cites for this is the merger of rentaikei (which he calls the Predicate-Adnominal form, or the P-A) and shūshiskei, as we discussed earlier in section 1.3.2. Despite the fact that most verbs do not have a phonologically distinct form, the copula does maintain a distinction. Furthermore, it seems that the cases in which genitive is disallowed are those without this form of the verb – in this way, it is not an external D that drives genitive subject marking, rather the verbal agreement. Therefore, according to Hiraiwa’s logic, the
Ahn - 12 occurrences of na (the rentaikei form of the copula) coincide with ability to license NGC, as shown below. (16) a)
b)
P-A and Declaritive Forms of the Copula John ga/no suki na biiru wa kore da. John NOM/GEN like COP-IMPF-PA beer TOP this COP-IMPF “This is the beer that John likes.” Biiru wa John ga/*no suki da. Beer TOP John NOM/GEN like COP-IMPF “As for beer, John likes [it].”
While this does make the correct prediction for most all cases, one interesting case that it does not account for is the following, in (17). (17)
P-A Form without Genitive Subject John ga/*no suki na no da. John NOM/GEN like COP-IMPF-PA COMPL COP-IMPF “(It's that) John likes [it].”
This kind of construction uses no as the complementizer, which, importantly, cannot act like a normal Nominal in that this flavor of no never appears in a case position. However, it does catalyze the use of the P-A form on the proceeding copula. Significantly, this construction does not allow for genitive marking; something Hiraiwa might be hard-pressed to explain.12 2.5 Nouniness In their 2005 paper, Maki and Uchibori review the existing theories, deeming them all at least in one way inadequate. Instead, they try to relate the idea of NGC to a [+N] feature which is the licensing factor for genitive case. They arrive at two possible proposals, repeated here, below.
12
He does attempt to explain this in his 2002 paper as an effect of grammaticalizaiton of no with another element such as the copula and other particles. He cites data that in certain cases no + particle is grammaticalized to the point of being unable to be interpreted as a sum of parts. I do not necessarily find this convincing, as no da can certainly be interpreted as a sum of the parts as in “Kore wa John ga suki na no da” meaning “This is the one that John likes” as well as “(It’s that) John likes this.”
Ahn - 13 (18) [+N] Proposals a) Proposal 1: i. The rentai13 form of a predicate is induced by some element with [+N] features. ii. The GNC is allowed, only if the element has enough degree of [+N] feature to license the genitive case. GNC Degree of [+N] the rentai form P/Case Least OK * * Some OK OK * Most OK OK OK b)
Proposal 2: Agree in the GNC: [+N] feature is shared at least with the predicate, T, and C (or whatever heads the clause in which the genitive subject appears), at the rentai form of a predicate is regarded as materialization of the Agree relation.
There are a few problems with these proposals. Overall, the notion of [±N] having varying gradations of “+ness” is a little weak; how can one thing be more or less [±N] than another when [±N] should be a binary feature? Also, with Proposal 1, degree of [+N] determines if NGC is a possibility; but they use NGC as a factor of the degree of [+N] – this is circular. Furthermore, Proposal 2 does not specify any mechanism through which the [±N] feature is shared, and exactly which elements it is shared with. 2.6 Object Shift In a new attempt at refining an external motivation for NGC, Ochi (2005) posits that Object Shift (OS) plays a strong role. The main role of OS in NGC is that it induces the Transitivity Restriction. Ochi achieves this by having the object DP, when it remains in the same clause as the subject, block the assignment of genitive case to the subject. There are a few issues with this paper that deserve discussion. First, Ochi takes Chomsky (2000)’s Equidistance and also tries to incorporate Chomsky (2001)’s revision which disposes of Equidistance. In this way, Ochi uses Equidistance to apply at the first phase level, but specifies that, afterwards, only the highest (phonologically content-ful) specifier is accessible. In order for
13
rentai = P-A
Ahn - 14 this to work to his benefit, Ochi also has to identify DP as a phase boundry. To exemplify what Ochi used these mechanisms for, here are two examples from his 2005 paper. (19) OS as Intervention a) [TP T … [vP OB [vP SU [ v [VP … tOB …]]]]] |_____________↑ b) [D [TP T … [vP OB [vP SU [ v [VP … tOB …]]]]]] |_________*_______↑ Since D is a phase boundary and that is in the next phase level, the distance condition changes to only the phonological edge being available. If the OB moves higher, however, its trace will not intervene14. While his provisos work out the way he wants, this seems highly dubious as it does not present a clear picture as to why these things should be true. The next problem I see here is that there is not strong evidence that OS actually takes place everywhere in Japanese. While he attempts to prove this, neither his nor the opposing stances seem necessarily correct.
3. GENERAL INADEQUACIES OF PREVIOUS LITERATURE I find that across these theories, there tend to be inadequacies in several respects which are in need of more research. I will focus on three major areas which require further development. 3.1 Embedded Clause Structure First, the literature tends to gloss over the syntactic structure of the relative and embedded clauses. For example, some (Miyagawa 1993, Ochi 2001, 2005) take the view of Murasugi (1991) that relative and embedded clauses are TPs and are NP complements. Others (e.g., Hiraiwa) take a more traditional approach alternating between left and right branching in different places to account for word order. As the structure of embedded clauses is extremely 14
While I agree with this statement, it would seem that the OB would have to move before getting to D, otherwise this notion seems to be counter-cyclic.
Ahn - 15 relevant to the study of NGC, we must be sure to have a definite analysis of them. 3.2 What about the Case of Nominatives? Second, it is hard to find a theory which really answers the question, what are the specific conditions for determining nominative and genitive case? That is, these theories all do their best to explain when genitive case shows up on the subject, but only a few papers try to convincingly explain what conditions prompt the use nominative case instead. It should be noted, however, that Hiraiwa (2001) does deal with this by positing that distribution of nominative and genitive is completely optional. Even so, this is hard to believe if there are any differences at all of a syntactic consequence between the two.15 I will investigate deeper the structure of the nominative type clause in section 4.2. 3.3 The Transitivity Restriction While the issue of the Transitivity Restriction (TR) might be one of the more understudied facts about NGC, it is by far the most crucial of these three to a theory of NGC. I say that this is the most crucial of these three because, since the TR only occurs in the case of NGC, it would be best to theorize that the TR is actually an effect of the properties of NGC and not a property, in itself. In this vein, the specific conditions that cause the TR should also be the specific conditions that induce the use of genitive-marked subjects. Therefore, I will pay very close attention to the TR in terms of its relation to NGC. In the following subsections, I will go in depth into what different past papers have said on the subject of the TR. 3.3.1 Miyagawa Miyagawa (2003) tries to resolve the TR through a view that the no-type clauses are somehow impoverished in that they cannot assign morphological case; morphological genitive marking is licensed from an external D, but there is nothing to license morphological accusative 15
Perhaps what I mentioned earlier about possibilities of scope discrepancies could be enough to collapse this optionality. In fairness to Hiraiwa, he holds that these scope facts aren’t real. Also, the Transitivity Restriction is certainly a syntactic consequence – correlating use of no with the TR becomes purely stipulative for Hiraiwa.
Ahn - 16 case. Explicitly, the rule he posits is, “All morphological case marking is licensed by tense; Abstract Case on the object is licensed solely by the small v.” That is, he maintains that accusatives are not licensed by T, but the actual morpheme for ACC does originate in T. Under these assumptions, no structural arguments (dative, accusative, or nominative) should be allowed. However, with dative, Miyagawa admits that it can sometimes be hard to determine if the ni is actual structural case marking or a postposition. He refers to a test for case-hood that he developed in a 1989 paper, which is that structural datives allow for floated numeral quantifiers, whereas postpositional phrases do not. He then presents data that ni-marked phrases which are in no-type clauses are in fact postpositional and therefore are not subject to his case restriction. The data he uses in his 2003 paper is repeated below16. (20) Restriction on Dative ni a) * Gakusei ni san-nin John no kasi.ta hon students DAT 3-CL John GEN lend.PRF book Intended: “The book that John lent to 3 students” b) Gakusei ni san-nin John ga kasi.ta hon students DAT 3-CL John NOM lend.PRF book “The book that John lent to 3 students” The shortcoming here is, while Miyagawa presents a very clear picture of the TR, he relies on an unsupported fact – T is the locus of all morphological case. Perhaps we could rework this in such a way that this is not the case, while maintaining the more appropriate aspects of his theory. 3.3.2 Hiraiwa Hiraiwa (2002) claims that “[s]pell-out of morphological accusative case by v triggers nominative Case-checking on T in the next strong phase.” First, he does not support this with evidence and this is a little stipulative. Furthermore, if Hiraiwa is correct, then we expect only 16
I am unclear as to why the dative goal has been scrambled to the front of the clause. I intend to do some native speaker judgment tests with dative constructions that don’t involve scrambling.
Ahn - 17 cases of phrases with an overtly o-marked object to be restricted because Case-checking is not triggered otherwise. This seems to leave open the possibility of deleting just the o accusative marking at PF before the next strong phase, which is common in Japanese.17 It also seems to leave open the possibility of true datives in the embedded clause, which Miyagawa clearly showed is not true. However, it does rightfully allow for objects in the embedded/relative clause if they are specially marked, whereas Miyagawa might not.18 3.3.3 Maki and Uchibori The TR is only addressed in one of Maki and Uchibori’s two possible proposals. In their second proposal, they believe that “[t]he TR in the GNC is due to the lack of v in the relevant domain of a nominal element.” However, they don’t show any reason for us to believe that there is no vP. Furthermore, if we believe, as I do, that objects raise out of relative clauses, it is necessary that the structure to allow for objects exists. 3.3.4 Ochi 2005 The phenomenon that the argument for OS best captures is the TR. As I stated above in section 2.6, the shifted object intervenes between the licensing D and the subject just in case it stays in the vP with the subject. This argument is very well formed within the assumptions of the paper, but outside of the paper, it seems less likely to be true.
4. A NEW THEORY ON NGC 4.0 In Pre-theoretic Terms Before delving into the specifics of my theory, it is important to explain the direction in which I will move. First, I will be mostly assuming Kayne (1994)’s theory of Antisymmetry and will be applying it to Japanese embedded clasues. 17 At least in matrix clauses. 18 There are constructions in Japanese which have nominative marked objects, for example. More on this in Section 5.
Ahn - 18 Moreover, I assume that the syntax of the “normal” kind of embedded clause and the kind which leads to NGC are inherently different. For this reason, I will appropriate separate terms for both types; I will call the “normal” nominative-marked-subject clauses Nominative Clauses (NCs) and the genitive-marked-subject clauses Genitive Clauses (GCs). I mentioned that NCs and GCs are expected to be “inherently different,” but in what ways? Most specifically, NCs allow for Case valuing as we expect to see it – just like in a matrix clause. However, it seems (for reasons I will get into later) that GCs do not have any Case valuing at all, internal to themselves. For this reason, we (correctly) anticipate a restriction on other arguments inside a GC. I believe that this is the origin of all the facts of NGC in Modern Japanese. I expect that it will also lead us in the direction of Object Shift, which I later point out as one of the three necessary operations for NGC to occur; the other two being the lack of Case valuing and the D probe which assigns genitive Case. 4.1 Assumed Structure of the Embedded Clause I assume that relative clauses are CPs which are merged with the relativized N. However, to maintain proper antisymmetric relations between the relative clause and the relativized N, the clause must c-command the N. In a Kayne meets Chomsky kind of way, relative clauses are built such that the TP raises out of the CP into Spec CP, on its way to its final prenominal position. Under Kaynean Antisymmetry, relative clauses are always built as [NP [CP [TP ]]], and then rearranged as necessary to fit language-specific rules. So in the case of head-final language, the TP raises to above NP to obtain the correct word order. However, under Chomskian Minimalism, this would violate a phase boundary (CP’s, to be precise) if it occurred in one fell swoop. Therefore, to reconcile this inconsistency, we have the TP move successive-cyclically, stopping at the Spec CP on the way to its final destination. While it might seem stipulative to posit movement of TP for purely phonological purposes, others have argued for “PF Purposes” being able to take part in
Ahn - 19 driving movement. (Fox, 2000, Richards, 2005) Furthermore, the object must raise out of the original clause in order to be relativized; I believe this happens when the TP sits at Spec CP. In a way, the TP is smuggling the object to the phase boundary. Collins (2005) discusses smuggling and cyclicity, saying when smuggling across a phase boundary, “…[the spec of a phase boundary] is providing the escape hatch (via smuggling) for the movement … to [a position higher than the phase boundary].” After the object is smuggled to the phase boundary and moved out, then the TP double-moves over the object to sit in the position for correct antisymmetric relations (including word order). All of this is captured in the tree below. (21) DP D
NP
TP
N' OBJ '
CP
TP
C'
…OBJ… C
TP
As for NP complements, I believe that they are much the same as relative clauses with the exception that there is no raising of an N out of the TP; instead, the N is base generated at that position. I demonstrate this below. (22) DP D
NP
TP
N' N
PrdP Prd
'
CP TP
C' C
TP
Ahn - 20 4.2 The Nominative Case To help better understand what I will get into with the clauses with GCs, I will first explore the NC’s case. Below, I construct a tree for a prototypical object relative NC. (Single directional arrows indicate feature matching, bidirectional arrows indicate feature matching and Case valuing)
4.2.1 Object Relative NC (23)
Prototypical Object Relative NC DP D
NP
TP
N' OBJ
CP TP
C' C
TP
…… SUB
T'
T
vP
v' v
voiP OBJ
voi'
voi
VP V
In this NC, first what happens is that voi19 matches the φ-features20 of the object, OBJ. Just following that, the OBJ checks voi’s EPP feature and moves to Spec voiP. The derivation continues, and after the subject, SUB, is base generated at Spec vP, T matches its φ-features and
19
voi here refers to the voice head, which is theoretically comparable to the Tr head proposed by Bowers (2002). It is the locus of the Accusative Case Agree, as well as several other morphemes and relations. 20 I am not sure whether OBJ is Case valued here or not. It is possible that it is Case valued and that when relativized, only the head N moves out, stranding the Case valued DP. However, it is also a possibility that Case is not valued and that this voi has its Case valuer suppressed. I assume the latter due to the fact that OBJ stays an active goal. I am unsure, however, as to why this should be the case. Ultimately, however, this is a question for Kaynean relative clause syntax. When an argument is relativized, what is moved? Is it an entire DP or just an N or something else?
Ahn - 21 Case values it nominative – rendering SUB inactive as a goal. SUB checks T’s EPP feature and moves to Spec TP. As in (17), the TP moves successive-cyclically to Spec NP, stopping along the way at Spec CP, to honor the Phase Impenetrability Condition (PIC). While sitting in Spec TP, N’s probe matches features with OBJ, the closest active goal, and raises it to the relativized position, where it will later get its case from a matrix-clause licenser. The TP finally makes its way to its final position in Spec NP for correct c-command relations and linear word order. 4.2.2 NP Complement or Non-Object/Non-Subject Relative NCs The case of a relative NC whereby an argument that is neither the subject nor the object (e.g., a goal or a locative) is relativized is the virtually the same as an NP complement NC21. For this reason, I will only create a tree for the NP complement NC. I will assume there is an object in the embedded clause. (24)
Prototypical NP Complement NC DP D
NP
TP
N' N
CP TP
C' C
TP SUB
T'
T
vP
v' v
voiP OBJ
voi'
voi
VP V
In this NC, first what happens is that voi matches the φ-features of and Case values the 21
The only difference being that a relative NC moves something (like a goal or locative) out from within the embedded clause, whereas a NP complement NC base generates the head N.
Ahn - 22 object, OBJ, accusative – rendering it an inactive goal. Just following that, the OBJ checks voi’s EPP feature and moves to Spec voiP. The derivation continues, and after the subject, SUB, is base generated at Spec vP, T matches its φ-features and Case values it nominative – rendering SUB inactive as a goal. SUB checks T’s EPP feature and moves to Spec TP. As in (17) and (19), the TP moves successive-cyclically to Spec NP, stopping along the way at Spec CP (again, due to PIC). The noun is base-generated at the N head.22 The TP finally makes its way to its final position in Spec NP for correct c-command relations/linear word order. 4.3 The Genitive Case My theory of NGC employs theoretical apparatuses from almost all of the theories discussed above. Before going in depth, each section is opened with a tree of a prototypical example of the structure. The structures described below are only applicable to GCs. Detailed explanations of each of the syntactic operations follow the derivations. 4.3.1 Relative Clause (25)
Prototypical Relative GC DP D
NP
TP T
N' vP
OBJ vP
SUB
CP TP
v'
C' C
T'
TP OBJ
T' T
vP
vP
SUB
v' v
voiP voi
VP V
22
Alternatively, in a non-subject/non-object relative NC, the noun to be relativized raises to this position, as in (14).
Ahn - 23 The first operation of note that takes place, rather that doesn’t take place, is that voi does not Agree with the object DP (henceforth, OBJ). Therefore, no Case is assigned by voi; we can assume that GCs suppress the voi head from assigning Case, which is not so unusual (e.g., English passive constructions). Since there is no Case assigned, OBJ is still active as a goal. The next atypical operation to take place is Object Shift (OS), as with Ochi (2005). However, unlike Ochi (2005), I assume that OS only takes place in GCs – not in normal matrix clause constructions, anyway (for reasons I explained above in my criticism of his paper). While I currently lack the theoretical reasoning for it, I believe that this is motivated by the fact that voi does not assign Case. This may seem to be a rather extemporaneous use of OS as I hold that it doesn’t occur anywhere else in Japanese Syntax (to my knowledge), but it could be argued that the whole GC construction is exceptional and relies on exceptional use of syntactic mechanisms. My arguments are weak, but I am currently researching other potential motivating factors for OS in GCs, such as scope. Next, the T probe matches φ-features with OBJ, as it is the nearest matching DP. Again, as with voi, T does not assign Case in its Agree with OBJ (again, this is not completely atypical, as it happens in English ECM ‘want’-type constructions), so OBJ is still an active goal as it is still Case-less. Also, OBJ matches T’s EPP feature and is raised to Spec TP. Just as in (17)-(20), to abide by the PIC, the TP raises successive-cyclically to the Spec NP, stopping along the way at Spec CP. While the TP is in Spec CP we continue building the tree, and when we want to merge N with CP, we find that the appropriate N, OBJ, is the Spec TP. We then move it out up into the N head, where OBJ will remain for the rest of the derivation, and where it will Agree with something else to get Case. Completing its journey, the TP moves to Spec NP, giving us the correct word order as in (22), below. (5) [DP kono [TP boku no
kat.ta [NP pen [CP tTP]]]]
Ahn - 24 this 1SG GEN buy.PERF pen “This pen that I bought” Then, after D is merged with NP, D’s probe matches the φ-features on the closest active goal, the subject, SUB. This checks SUB’s uninterpretable features and assigns genitive case to the subject.
4.3.2 NP Complement (26)
Prototypical NP Complement GC DP D
NP
TP SUB
N' T'
N
CP TP
C' C
TP SUB
T' T
vP
v'
v
voiP voi
V
The NP complement GC structure is very similar to but simpler than that of its relative clause counterpart. First, T’s probe agrees with the φ-features of SUB, raising it to Spec TP. Just like the relative clause, the T here (as well as the voi, which I will get into in the next section) does not assign Case, leaving SUB as an active goal.23 Again like the relative clause, following the PIC, the TP moves successive-cyclically to Spec NP, stopping at Spec CP. After arriving in its final position at Spec NP, the D probes for a head with matching φ-features, finds and Agrees with SUB, giving it genitive Case.
23
Note that there is no object in this derivation. That is because if there were one, it would crash, as I explain in section 4.4.
Ahn - 25 4.4 Transitivity Restriction One extraordinarily desirable consequence of this theory is that the TR falls into place on its own.24 The first issue with accusative objects in a relative GC is that the voi’s Case assigning ability is suppressed. Even if this were not the case and OBJ were able to be marked Accusative Case, there would be no OS since the object in question has already become inactive as a goal (its Case and φ-features being assigned and checked). Thus, we would expect SUB to raise as the relativized N. Furthermore, D would not be able to check its uninterpretable features, crashing the derivation. This is also evidence for voi not being able to assign Case; if it were able to, the derivation would crash as I just demonstrated. Following the idea that voi cannot assign Case as I just confirmed, it is clear that accusative objects cannot occur within a NP complement GC. If there were an object in a NP complement GC, it would crash the derivation as the features of the subject would never get checked – in fact, there would be a genitive marked object. This would be due to OS taking place without relativization – leaving OBJ as the closest feature-matching DP to the genitive case assigning D.
5. FURTHER RESEARCH There remains much work to be done in researching this phenomenon; however, my research along with those of past researchers has been able to successfully capture many of the syntactic properties of NGC.
24
That is, restriction on the use of Accusative Case falls into place on its own. If Miyagawa (1989)’s test for Casehood – namely whether a numeral quantifier may be floated from the noun in question – is accurate, then Dative Case should also be restricted in GCs. If so, we would want to assume that all Case is disallowed in GCs, due to some mechanism – for Miyagawa (1993) that mechanism is his idea that all morphological case comes from T. However, strangely, as I note later on in the Conclusion, Nominative Case is not always restricted. This creates a problem if, as I assume, Genitive Case is assigned as a result of the DP in question not being able to get Nominative Case.
Ahn - 26 That being said, further research is necessary in several areas, some of which I have already discussed. 5.1 MNCs and NGC There is a structure in Japanese in which constituents which look like objects (at least semantically) are marked Nominative while the subject of that sentence is also marked Nominative – these structures are known as Multiple Nominative Constructions (MNCs). As I discussed towards the beginning of this paper in footnote (3), “subject” is a vague term, and I am defining it here as something that is “normally” marked nominative. In this case of MNCs, then, how does the fact that two elements are marked nominative (thus “subjects”) interact with their ability to undergo NGC? Below, (27a) exemplifies an MNC, and (27b-d) exemplify how NGC may apply. (27) NGC with an MNC (a) Boku ga kanji ga yome.na.i koto, sit.te i.ru? 1SG NOM Chinese-characters NOM read.NEG.IMPF fact, know.CONT be.IMPF ‘Do you know that I can’t read Kanji?’ (b) Boku no kanji no yome.na.i koto, sit.te i.ru? 1SG GEN Chinese-characters GEN read.NEG.IMPF fact, know.CONT be.IMPF Do you know that I can’t read Kanji? (c) Boku ga kanji no yome.na.i koto, sit.te i.ru? 1SG NOM Chinese-characters GEN read.NEG.IMPF fact, know.CONT be.IMPF ‘Do you know that I can’t read Kanji?’ (d) Boku no kanji ga yome.na.i koto, sit.te i.ru? 1SG GEN Chinese-characters NOM read.NEG.IMPF fact, know.CONT be.IMPF ‘Do you know that I can’t read Kanji?’ There are definitely questions raised by this kind of data. For (27b)25, how is it that Case is assigned more than once? Perhaps genitive Case is assigned here more than once in the same way that nominative Case is assigned more than once in MNCs. For (27c), how is Case being
25
It should be noted that (24b) has a strong genitive reading (that is, “Do you know that [someone] cannot read my Kanji?”) first, unless the speaker is prompted with the proper semantics
Ahn - 27 assigned to kanji when T is supposedly defective?26 Perhaps this “nominative” marking is really more of a focus marker without Case, or some kind of inherent case marking. 5.3 Hachijōjima Dialect There is an island, about 150 miles south of Tokyo, called Hachiōjima. Perfect as a comparison to the Modern Standard Tokyo dialect, the dialect spoken there has maintained an morphemic difference between the rentaikei and shūshikei verbal endings on all verbs. An example of this can be seen below, as extracted from a 1950 linguistic survey of the island by Kokuritsu Kokugo Kenkyūjo. (28) NGC with an MNC a) ara ik.u I go.SS “I go” b)
waga ik.o toki I(GEN?) go.RT time “The time that I will go”
However, at this point in my short look into the dialect of Hachijōjima, I cannot say for sure that they have genitive subjects27, although, it seems that the ga of waga in (28b) is the genitive marker. If that is truly a genitive marked subject, it would seem that the data in (28a) lacks genitive (or any, for that matter) subject marking, demonstrating that subject marking is done the same way now in Hachijōjima as it was done centuries ago across all dialects of Japanese. Futhermore, I would like to investigate to see if a TR has ever come about without the merger between rentaikei and shūshikei. If not, this implies that the TR is dependent on the loss of a this inflectional distinction, supporting the position I have taken on the TR as a historical change. 5.4 Further evidence for Object Shift in GCs I have admitted that my argument for Object Shift in GCs is weak; however, I think if the 26 27
The answer to this question undeniably relies on our analysis of MNCs and how Case is assigned then. Based on the historical data, I would presume that if there is case marking at all, it will be genitive subjects with the rentaikei.
Ahn - 28 proper kinds of tests were conducted, it might be able to shed more light on this aspect of my theory. I attempted to perform such a test with native speakers, by utilizing the notion that Floated Quantifiers (FQs) can only exist in a (former) position of the noun being quantified. Therefore, if objects are shifted, then there should be a new position available for FQs that is not available for non-OSed objects. I tested this using sentence that varied on position of the FQ and whether the subject was in nominative or genitive Case. A sample of this is shown below. (29) NGC with an MNC a) Ik.ko Taroo no tabe.ru ringo wa fuji desu ka? 1.CL Taro GEN eat.IMPF apple TOP Fuji COP-IMPF Q “Is the one apple that Taro ate a Fuji apple?” b) Ik.ko Taroo ga tabe.ru ringo wa fuji desu ka? 1.CL Taro NOM eat.IMPF apple TOP Fuji COP-IMPF Q “Is the one apple that Taro ate a Fuji apple?” c) Ni.satsu Mary no yon.da hon wa dochira? 2.CL Mary GEN read.PRF book TOP which “Which are the two books that Mary read?” d) Ni.satsu Mary ga yon.da hon wa dochira? 2.CL Mary NOM read.PRF book TOP which “Which are the two books that Mary read?” While these tests should have worked, they elicited strongly varying responses from person to person. Perhaps I did not explain the contexts clearly, or perhaps people vary on how freely they can float quantifiers. Either way, new tests should be formed to research the possibility of OS in GCs to either support or deny my use of it. 5.5 Cause of Defective T/voi While the arguments presented above in favor of T and voi not assigning Case in GCs are sound, we should also like a syntactically-based reason for this to be true. Perhaps the [N] feature as discussed by Whitman (see above in sections 1.3-1.4) is no longer being interpreted onthe C, but on the T and voi, thus impairing them in such a way that they cannot assign Case.
Ahn - 29 5.6 Cross-Linguistic Research Finally, more light would be shed on Japanese NGC by comparing it to comparable phenomena as they occur in other languages such as Middle Korean, Turkish and Quechua. In doing so, the slight variations should be exploited showing that NGC is not really a language specific or cross-linguistic phenomenon, rather one that is a result of other, independent language specific facts. One thing to pay check for cross-linguistically is to see if all languages that allow for genitive subjects are verb-final, as the ones mentioned above. If this is the case, then perhaps the principals of genitive subject marking as defined in this paper can be extended to those languages. After all, they all have a relative clause structure whereby a TP has to move under a DP anyway. Also, do these other languages have something like the Transitivity Restriction? If they don’t, why not (conversely, if they do, why do they)? Perhaps the answers to these questions will determine facts about the Japanese TR. To be specific, if another language that has NGC that is triggered by an overt verbal change and that language has a TR, then the explanation of the TR as a historical change as I have given here might be in trouble. However, if the other languages are in line with my generalization, then this research is even further supported.
6. CONCLUSION I believe that NGC can be summed up as a result of three factors: (1) Object Shift, (2) suppressed Case marking by voi and T, and (3) D’s genitive case marking probe. The benefit of using only three syntactic operations to encapsulate NGC is that it de-mystifies the process and gives it a foundation in Japanese Syntax that is not purely stipulative. Also, by understanding NGC as a result of these relatively few and simple mechanisms, we have derived the most seemingly complex and mysterious facet of NGC – the Transitivity Restriction. Finally, in
Ahn - 30 thinking about NGC in this way, we understand that NGC is not an independent principle or parameter of a language, but rather we understand it as the consequence of other aspects of the language.
ABBREVIATIONS USED: ACC
Accusative Case Marker
CAUS
Causative Maker
CL
Numeral Classifier
COND
Conditional Form Marker
CONT
Continuative Form Marker
COP
Copula
DAT
Dative Case Marker
GEN
Genitive Case Marker
GOAL
Goal Marker
IMPF
Imperfective Tense (non-past) Marker
LOC
Locative Marker
NOM
Nominative Case Marker
PA
Predicate-Adnominal Form (aka rentaikei)
PASS
Passive Voice Marker
PERF
Perfective Tense (past) Marker
POT
Potential Mood Marker
RT
Rentaikei
SS
Shūshikei
TOP
Topic Marker
VOL
Volitional
WORKS CITED Bowers, J. 2002. Transitivity. Linguistics Inquiry 33:2. 183-224. Chomsky, N. 2000. Minimalist Inquiries: the Framework. In Step By Step: Essays on Minimalism in Honor of Howard Lasnik. 89-155. Chomsky, N. 2001. Derivation by Phase. In Ken Hale: A Life in Language, ed. by Michael Kenstowicz. 1-52. Collins, C. 2005. A Smuggling Approach to the Passive in English. Syntax 8, 81-120. Fox, D. 2000. Economy and semantic interpretation. MIT Press, Cambridge, MA. Harada, S.I. 1971. Ga-No Conversion and Idiolectal Variation in Japanese. Gengo Kenkyu 60, 25-38. Harada, S.I. 1976. Ga-No Conversion Revisited: A Reply too Shibatani. Gengo Kenkyu 70, 23-38. Hiraiwa, K. 2001. On Nominative-Genitive Conversion. MIT Working Papers in Linguistics 39, 66-124. Hiraiwa, K. 2002. Nominative-Genitive Conversion Revisited. Japanese/Korean Linguistics 10. Kayne, R. 1994. Antisymmetry of Syntax. Cambridge, MA: MIT Press. Kokuritsu Kokugo Kenkyūjo. 1950. Hachijōjima no gengo chōsa (A Linguistic Survey of Hachijōjima). Tokyo, Japan. Maki, H and A Uchibori. 2005. Ga/No Conversion. Paper presented at Harvard University for the Workshop on Linguistic Theory and the Japanese Language. July 29, 2005. Miyagawa, S. 1989. Syntax and Semantics 22: Structure and Case Marking in Japanese. Academic Press. Miyagawa, S. 1993. LF Case-checking and minimal link condition. MIT Working Papers in Linguistics 19, 213-254. Miyagawa, S. 2003. A-movement scrambling and options without optionality. Word order and scrambling. 177-200. Miyagawa, S and T. Tsujioka. 2004. Argument Structure and Ditransitive Verbs in Japanese. JEAL 13. 1-38. Murasugi, K. 1991. Noun phrases in Japanese and English: A study in syntax, learnability and acquisition. Ph.D. Dissertation, University of Connecticut. Nambu, S. 2005. Corpus-based Study of The Change in Ga/No Conversion. MA Thesis, Kobe Shoin Graduate School of Letters.
Ochi, M. 2001. Move F and Ga-No Conversion in Japanese. JEAL 10. 247-286. Ochi, M. 2005. Ga-No Conversion and Overt Object Shift in Japanese. Nanzan Linguistics 2. 6180. Richards, N. 2005. Beyond Strength and Weakness. Paper presented at Cornell University. December 1, 2005. Saito, M. 1985. Some Asymmetries in Japanese and Their Theoretical Implications. Ph.D. Dissertation, MIT. Watanabe, A. 1996. Nominative-Genitive Conversion and Agreement in Japanese: a CrossLinguistic Perspective. JEAL 5. 373-410. Whitman, J. 2006. The Attrition of Genitive Subjects in Korean and Japanese. Syracuse/Cornell Workshop on the Internal Syntax of Nominalized Clauses. February 28, 2006.
