An Optimality-Theoretic Alternative to the Apparent Wh-Movement in Old Japanese

An Optimality-Theoretic Alternative to the Apparent Wh-Movement in Old JapaneseChiharu Uda Kikutacuda@mail.doshisha.ac.jpThis paper proposes an Optimality-Theoretic alternative to the wh-movement analysis of Old Japanese (OJ: ?-8C). Diachronic aspects of Japanese have recently aroused strong interest in theoretical linguists, both functional and formal. Among the most remarkable works is Watanabe’s (2002) Minimalist analysis, which claims that OJ was in fact a wh-movement language. Japanese has always been a head-final (SOV), free-word order (scrambling) language. It was never expected that it went through a dramatic change in parameter setting. For this reason, Watanabe (2002) had a great impact on Japanese syntacticians, and it has been so influential that its claim is almost taken as a “discovery” of the unknown “fact” of the Japanese language (but cf. Tonoike (2002, 2003)). However, Watanabe’s (2002) analysis has some serious problems. This paper argues that Watanabe’s (2002) claim is wrong and that OJ did not have overt wh-movement. The observed word order pattern results from the interplay of several constraints, which basically have to do with morphological case marking.The proposed analysis not only accommodates the fact covered by the wh-movement analysis, but offers a more comprehensive picture of the development of the Japanese language during 8-10C.1. Word order restriction in OJ and Watanabe’s (2002) claimWatanabe’s (2002) claim is based on the word order restriction in OJ, which disappeared in early Middle Japanese (MJ: 9-12C). Through a thorough examination of data of Man’yosyu (c.750?: the only available text in OJ), regarding the distribution of the interrogative particle [ka], the topic marker [wa], and the subject marker [no/ga], Nomura (1993) and Sasaki (1992) conclude that [ka] almost always preceded [no/ga], while [wa] almost always preceded [ka]:thedataofnumber(1) I. Nominative subject: XP [ka] . . . Subj [no/ga] . . . approximately 90Subj [no/ga] . . . XP [ka] . . . 4 (or 5)II. Topicalized subject: XP [ka] . . . Subj [wa] . . . 2 (or 3)Subj [wa] . . . XP [ka] . . . approximately 50Based on this, Watanabe (2002) proposes the clause structure with split C system, which follows the insight of Rizzi (1997), and claims that XP[ka] is obligatorily moved to the Spec of FocP, a type of wh-movement:(2) [TopP Spec Top [FocP Spec Foc [IP Subj VP I ] ] ]So Watanabe’s (2002) wh-movement is in fact Focus-movement. The particle [ka] has a dual-function: (1) it attaches to either wh- or non-wh-word to make the clause interrogative, and (2) it places a focus on the phrase. As regards the second function, [ka] is with other focus-marking particles such as [ya] and [so] in triggering kakarimusubi, a special inflection form of the predicate (see below). All these focus particles obey the above word order restriction. Watanabe (2002) claims that the movement is triggered by the [-Interpretable] feature of the focus particle; kakarimusubi is a sign of wh-agreement, indicating that the clause includes a wh-trace in it.The wh-movement analysis suggests a new scenario of the decline of kakarimusubi. It is an established fact that the kakarimusubi completely disappeared around 14-15C, while the word order restriction in question was lifted around 9-10C. The difference of several hundred years may seem to reduce the credibility of the analysis. However, he demonstrates that the genuine interrogative function of [ka] was rapidly lost during early MJ, becoming confined to rhetorical questions. So it is suggested that the decline of kakarimusubi in fact started much earlier, along with the loss of wh-movement.2. Problems with the wh-movement analysisIn spite of its appeal, Watanabe’s (2002) analysis has the following problems.1. The pre-posing of a phrase does not necessarily require the [-Interpretable] feature of [ka]. Many instances of the pre-posed [ka]-phrase are a whole subordinate clause (conditional, etc.). However, the subordinate clause must precede the matrix clause in Japanese (=SOV) anyway, unless it is center-embedded, as evidenced by abundant data without [ka].2. Some of the pre-posed FocP cannot involve a wh-movement, if the predicate takes the attributive form. The [ka] phrases in (3), for instance, seem to be independent “presentational” phrases with no apparent grammatical function w.r.t. the matrix clause (i.e., considered as juxtaposition), which are unlikely to have moved out of the matrix clause. This type of [ka] examples is not uncommon:(3) Tahagoto-ka, oyodore-ka, komorikuno Hatuse-no yama-ni komori-seri-to-ihu.insane-word-Q, false-word-Q, (epigraph) Hatuse-no mountain-loc hide-do-C-say[attri.]‘(lit) A joke? Or a lie? They say that he has hidden himself in Mt. Hatuse.’3. The attributive form is not limited to kakarimusubi; the focus particle is not a crucial condition for the availability of the predicate form. It is the predicate form of a nominal clause and noun modificational clause. It was also common that the nominal clause occurred independently (expressing some kind of non-assertiveness.)4. The change in the behavior of [ka] may very well be lexical. As the particle [ka] lost its interrogative force, it is replaced by [ya] (for non-wh-word) and bare wh-word, which continued to trigger kakarimusubi through MJ. In other words, the word order restriction (and its loss) uniformly covered focus phrases ([so][ya][ka]), and kakarimusubi apparently flourished in MJ involving all focus phrases (with different membership), [so][ya] and bare wh-word. The proposed scenario fits only the history of [ka]. Besides, according to Watanabe (2002), only the kakarimusubi in OJ is the “real” one, while the one in MJ is merely a stylistic imitation, which somehow lasted very long. This distinction, however, has no support outside the theory.Due to the limitation of OJ data, such syntactic tests as unbounded dependency and cross over are not available. Given that Japanese has always allowed scrambling, the only solid motivation for the wh-movement analysis is that the word order pattern is (almost) obligatory. I will argue, however, that the word order in question is obligatory for a different reason.3. Other important facts: for the proposalI take the following observations of the history of Japanese as significant.1. In OJ and MJ, both [no] and [ga] marked either nominative or genitive (underspecified or homonymous). More importantly, [no] and [ga] as nominative markers occurred only in the clause headed by the attributive predicate in OJ; otherwise, the subject is marked either with topic marker wa or without any marker. There is good reason to believe that the attributive form is a nominal/verbal, mixed category; namely, the predicate in the attributive form can directly head a nominal clause. The unique distribution of the nominative [no/ga] has been traditionally ascribed to the nominal nature of the attributive form in philology, and has found a clearer Stochastic OT account in Kikuta (2003) on a similar line.2. It has been recently observed that the word order [Subj[no/ga] . . . Obj[o] . . . V(attri.)] is prohibited in OJ (Kinsui 2001, Yanagida 2003). The [o]-marked object has to be pre-posed or has to occur as bare NP. This restriction apparently parallels the one in (1); Subj[no/ga] and V(attri.) does not allow an intervening phrase with a particle. However, wh-movement analysis cannot be extended to this pattern, in which the predicate form is not conditioned by a movement of any sort.3. It is well-documented that MJ saw the development of morphological case markers, particularly the accusative [o]. Although details are somewhat controversial, a general consensus is that in OJ: (1) bare NP was very common both for subject and object; (2) the marker [o] for a direct object often accompanied a special semantic overtone; (3) [o] marked a non-object, and even the subject in a special construction called mi-construction. It is also agreed that [o] was established as the accusative case marker during MJ; the mi-construction also disappeared in MJ.4. ProposalIn view of the above facts, I claim that the word order restriction is a reflection of a nominal nature of the phrase. The reason for the attributive form of the predicate in kakarimusubi is semantic rather than syntactic (wh-agreement), having to do with the presupposition (as opposed to focus) and the non-assertiveness of a nominal clause.Adopting the idea of Malouf (2000), I assume that the attributive form predicate is a [+n, +v] mixed category, and that a [no/ga]-marked argument appears at the left-edge of a [+nominal] phrase. I propose the four constraints in (4), with the dominance relation in (5):(4) (A) Argument Realization Faithfulness: lexically specified arguments should be syntactically realized.(B) Nominal Phrase Surface Tightness: *[YP NP[+Ncase] XP Y[+N]] where XP = XP[+mark](C) Case Transparency: the morphological case should properly reflect the abstract case.(D) Case Type Consistency: Case type of coarguments should be consistent (+N-type case, +V-type case),where [+Vcase] includes: nominative [no/ga], accusative [o], dative [ni]; and [+Ncase]: genitive [no/ga].(5) OJ: (A) >> (B) >> (C) >> (D)MJ: (A) >> (C) >> (D) >> (B)Namely, what caused the word order restriction in OJ is the dominance of the constraint (B), which means that in an endocentric nominal phrase, the link between the Spec and head should not be interrupted by a non-modificational phrase. The actual form of this constraint can vary depending on the type of language, regarding exactly what the [+Ncase] is. In most cases, it is equivalent to the genitive case. In OJ, where the abstract case and the morphological case are yet to find a complete correspondence, it is the morphological realization of case. Note that this word order is very natural for an NP. The situation with OJ is rather unique because YP is partly verbal, taking arguments as other verbs do. I assume that abstract case system (as some kind of argument licensing) is universal, while morphological case realization is not only language-particular, but is subject to diachronic change within a language (cf. Sigurdsson 2003).The apparent loss of wh-movement results from the relative demotion of the constraint (B). The demotion is caused by the promotion of (C) and (D), which reflects the development of the morphological case system in MJ. The establishment of the accusative marker [o], mentioned above, cannot merely be an idiosyncratic lexical change; it (re)organized the overall system of case marking. A relevant well-known fact is that the “nominative” [ga], which was sensitive to the nominality of the predicate in OJ, gradually came to be used with predicate in any inflectional form in MJ, indicating its gradual establishment as the nominative marker. Concomitantly, OJ’s rather unique mixture of case type is lost, and the nominal word order restriction is lifted from the verbal clause. As a consequence, the clause is free to behave just as other clauses.Thus the proposed analysis explains the word order restriction in OJ and its loss in MJ, without making a radical claim that Japanese was once a wh-movement language. The benefit of this analysis includes: that it accommodates the fact that kakarimusubi flourished both in OJ and MJ, yet it showed a difference; and that it has managed to relate the change in case marking morphemes, which is well-documented without being given a systematic explanation, to the global change in the Japanese syntax. Other points of discussion include: the relation between morphological case and abstract case; and the role of morphological case in syntax, which may be more significant than has been assumed (cf. Sigurdsson 2003, V ogel 2003). Important ReferencesKikuta, C.U. 2003. Subject Case Marking in Early Classical Japanese; A New Perspective from Stochastic Optimality Theory.In Japanese/Koreasn Linguistics 12, ed. W. McClure, 152-164. Stanford, CA: CSLI Publications & SLA.Malouf, R. P. 2000. Mixed Categories in the Hierarchical Lexicon. Stanford, CA: CSLI Publications.Watanabe, A. 2002. Loss of Overt Wh-Movement in Old Japanese. In Syntactic Effects of Morphological Change, ed. D.W.Lightfoot, 179-195. Oxford: Oxford UP.。