..... L9... 2 2 £11.; . 1mm 52. .m v , ‘ flu...“ dfim . - It. 2.. .. 3...... 3.. ~. I. . ._ u... nth”- I. 13.. h... . t. .v . l :5“. 13-.) . I9. k . 0- A 4L... 1.... .. :- :. .rmg‘yhi JR. 3 x a 7: v N»?! I .2. 35...: lass—V 3.9 5......15. .i.| v. rll'l 4'.quva 13‘ sauna...“ ) .2. 7... I . . .1. L. .. 4:04.. .e , 5...“... x??? .. 1 ; xx... . .. .3... . 3.1.3... c £11,.1Ir . .Xl I n «I? v... I. x..1t:uu Z, .x :lixvxa. 1 ii A?! )2 a a. foil. . is: :Lrflnn..rfixw FL. East Thisistocertifymatme dissertaflonentiued - - _._._._.-.-- JAPANESE LOANWORD PHONOLOGY IN OPTMMLITY THEORY: THE NATURE OF lNPUTS AND THE LOANWORD SUBLEXICON presentedby MASAHIKO MUTSUKAWA -n-c-o-I-Q-O-I-I-l--n-.-.—.-. .— hasbeenacoeptadtowardsmififlmem oftheraquirementsforthe PhD. dogma In LW 5;; n1: .2; ijoerfessor’sSbgnature 2 -- 2/ ’ 2 006 Date ----.-.-.--.-n_---—-.-.-.-.-V‘ -.-.— _ — MSUISMWWVEQWWWM LIBRARY Michigan State University PLACE IN RETURN BOX to remove this checkout from your record. To AVOID FINES return on or before date due. MAY BE RECALLED with earlier due date if requested. DATE DUE DATE DUE DATE DUE AUG 2 6 2007 FEB 2 42008 AUG 2 4 2008 61 2 1 1 U 2/05 p:/ClRC/DateDue.indd-p.1 JAPANESE LOANWORD PHONOLOGY IN OPTIMALITY THEORY: THE NATURE OF INPUTS AND THE LOANWORD SUBLEXICON By Masahiko Mutsukawa A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Department of Linguistics, Germanic, Slavic, Asian, and Afi'ican languages 2006 ABSTRACT JAPANESE LOANWORD PHONOLOGY IN OPTIMALITY THEORY: THE NATURE OF INPUTS AND THE LOANWORD SUBLEXICON By Masahiko Mutsukawa This dissertation deals with issues surrounding loanword phonology, more specifically issues surrounding the phonology of English loanwords in Japanese such as the nature of inputs to loanword adaptation and the structure of the Loarmord sublexicon. This dissertation discusses four phonological phenomena observed in English loanst in Japanese within the framework of Optimality Theory (01‘: Prince and Smolensky 1993). The four phonological phenomena are: accentuation of English loanwords (Chapter 3), the realization of English /r/ (Chapter 4), the realization of the English plural morpheme (Chapter 5), and English compound abbreviation (Chapter 6). The organization of this dissertation is as follows. Chapter 1 is a general introduction and Chapter 2 introduces the background relevant to this dissertation. Chapter3 argueswhatinformation is includedintheinput. Showingthatthemajorityof English loanwords in two major dialects of Japanese (Kansai Japanese and Tokyo Japanese) retain the English accent, i.e. the accent on the syllable stressed in English, I claimthatmeinformationonthelocusofEnglishstressis includedintheinputand Japaneseborrowershaveaccesstoit. InChapter4,Iclaimthattheinputstophonological processesareperoeivedsegments. Withrogardtomenanneofflreinpunitisnotclear whedrerthehrpmisbasedonthephmencrepresentafionofthesomoelanguage (Silverman 1992) or the phonological one (Paradis and LaCharité 1997). Based on the factthattheonset/r/andthecoda/r/inEnglisharerealized differentlyirrlapanese,l i f f . ¢ - . ' ’I ‘ I n ' § ‘ ' ‘ I u . . . . ‘ ' ‘ I : .' . ‘ . ‘. l 3." ‘ , : ' T ‘ . . b , '. ; i t g ‘ . ‘ t i ‘ . . ‘ I ’ i i '. I - - ' II'. I . ’ n a ' I ' J | . _. ‘ II N ‘ a ‘ . ‘ lu. ‘ H F I g i . . . r ‘ . I r '1; . , l '. I .0 ' ‘ I ‘ I v . U \ - . V! 1' , .1 I ‘ - I. . .1 k"."‘ 0‘ t 4. O) I 4 H ,3 .x . , ‘ n l -. . 'I '4‘. o ‘81,. I ' l " l . ' . . ,1 .Q 1 r “. '2 ‘ . a S ’ I I '7 A a e ‘7 I. . ,. . a ’f . . Z .‘ . ‘.3 ‘ I H l . r .6; . a!» .l ,“ 4 .fl . 5 l g ‘ i .‘ . . | t . , I , ‘ I ' A ' . b l‘ I‘ ‘ ‘ . 4 ‘ I . ’ _ t . I t . ‘ . 4i ' ’ I . , ‘ ~ . | . ‘ . ' . \ . , r ‘ - h , .4 . I .‘ I ' ’ J O '- V . , ” " . .9 v Masahiko Mutsukawa adopt Silverman’s model and further claim that an input segment can be perceived differently based on its location. Chapter 5 discusses who borrows loanwords. The English plural morpheme is treated differently from the stem in Japanese. This suggests that morphological information is included in the input and Japanese borrowers have access to it. Also, the different treatment of the plural morpheme and the stem indicates that Japanese borrowers have the knowledge of English morphology. This chapter also deals with the generation gap with respect to the accessibility to the morphological informationofEnglish.Chapter6aswellasChapter3and4concernsassimilation processes to the core part of the Japanese lexicon. Assimilation processes with regard to deaccentuation and the avoidance of the word-final [a:] are discussed in this dissertation. Thoseassimilationprocessesrevealthatprosodic feattuescanbeassimilatedtcthecore pm of the Japanese lexicon more easily. Finally, this dissertation is closed in Chapter 7 by summarizing the discussion and claims developed in this study. Major claims of this dissertation are as follows. First, the inputs to phonological processes are perceived segments, which supports Silverman (1992), and an input segment can be perceived differently based on its location. Second, English loanwords in Japanese are introduced by limited bilinguals, i.e. Japanese-English bilinguals with the knowledge of English morphology but not necessarily with the knowledge of English phonology (cf. Paradis and LaCharite (1997)). Third, the information on the locus of English stress and English morphology is included in the input and Japanese borrowers have access to it. Also, this dissertation shows that the accessrbility to English morphology has been changing over generations. Finally, with regard to the assimilation process, proMc features can be assimilated more easily. l . t i a it “a ' .__. .‘ ,. s . l V ' i i I r ‘ . i . .' s a 0 - V l t .q ‘l D, To my parents, Hiromi and Naomi Mutsukawa iv an n, ACKNOWLEDGEMENTS Thisdissertationrepresentstheend ofmygraduatesnrdentlife. Iwouldliketothank all ofthe people who have helped me in the course of my graduate studies. First and foremost,lam gratefirltoYen-Hwei Lin forreadingmorepartialdrat‘tsofthisworkdtan anyone else and accepting appointments at a short notice. This dissertation has benefited immenselyfiomherinspiration, feedback,andencouragement. Shehasbeenthebest adviser I could imagine throughout my graduate years at Michigan State University. I wouldalsolfltewdrmkflteodrermembersofmycommiuee:MutsukoEndoHudson, Grover Hudson, and Anne Violin-Wigent. Mutsuko Endo Hudson has provided invaluable comments from the perspective of a Japanese linguist. She also gave me oppmnnufiesofwachmglapamsemdfindingthejoyoflapaneselmguageteachmgl am grateful to Grover Hudson, and Anne Violin-Wigent for their insightful suggestions and invaluable comments. Besides my committee members, a number of people in the departmentdeserve specialthanks. Iwanttothankeverymemberofthefacultyandstaff and to my fellow students. In particular, I am grateful to Japanese TAs for their comments and friendship. IwouldliketodrankMarkKaufinarnmybestfi'iendinUS. lam luckytohavehim asmyfi'iend. HemademylifeinEastIansingenjoyable.lamalsogradthatlhavea fi'iendwholdonottalkaboutlinguisticswiflr.Hehasbeenafriendandhelpedmealot outsidethetmiversitytlnoughoutmydaysinEastLansing.Iamgratefulforhis friendship. I want to thank my wife, Yoko, and my sons, Miiru, Fuga, and Towa, for their patience and love during the past few years. They were always my most enthusiastic supporters. I cannot imagine having completed this dissertation without them. Last,butnotleast,Iwouldliketoexpressmydeepestgratitudetomyparents: Hiromi and Naomi Mutsukawa. They have given me unfailing love and a constant supportiamproudofbeingyomsonThisdissertationisdedicatedtoyou. vi TABLE OF CONTENTS CHAPTERI INTRODUCTION ............................................................................................................... I 1.1. Introduction - - ....... - I ..2 Organization of Dissertation ................ 3 CHAPTER2 BACKGROUND ................................................................................................................. 2..lIntroduction.- - -- _- - 2.2. Optimality Theory 2. 3. Models of Loanword Phonology-.--__-- : - - - _- : 2. 3. l. Silverman (1992).-- -- ....... @5003th 2. 3 ..2 Paradis and LaCharité (1997)" ...... 10 " ..... 11 2. 4. Japanese Phonology--- - ........ -- t 2 ..4 1. Japanese Phonetic Inventory A--ll 2 ..4.2 Adjustments Observed 1n English Loanwords 1n Japanese - 2.43. JapaneseLexicon — --13 --.-16 CHAPTER 3 AccentuationofEnglishLoanwords---- _ - .......................................................... 18 3..l Introduction - -- 3.2. Accentuation of Non-loanwords 1n Japanese 18 - -..---19 3 ..2 l. Accentuation of Non-loanwords 1n Tokyo Japanese ................... .20 3 ..2 2. Accentuation of Non-loanwords 1n Kansai Japanese - '23 3.3. Previous Studies .................................................................................. .......... 27 33..1 McCawley(l968) _____ -- - --.--- _ ........... -- 3..32 0m(1991).w ..... ---- 341 Data ..... -_ -28 - -. -29 3.3.3. Asano(1999) ............................................................................................. 31 3.4. Kansai Japanese ....................................................................................................... 34 1-411- Endisll'l‘ype ':'::------ __ ‘- 3..41.2. Non-English Type ........ 37 3.4.1.3. Unaccented Type - - 3....414 Summar'y-,--_- - A“__---- - ----34 ..--35 - 39 -.41 3.42. Analysis“ ...... 3.4.2..l TheLocusofAccent ........... 3.4.2.11. AccentedLoanwords —-=:v:=v -42 :~42 --42 3....4.212 UnaccentedLoanwords--- 3...4.22 PitchPattern -_ vii ‘AAAAA -_ --‘--- ‘--AA_ 7" 'vv 49 I r . ‘ .l . ' I..“1 .a . . n O 3 . ., . . l! . g ‘.r’ . , ‘ . I I I ’ ‘ - - . . - H ‘ — V I ;Ll ‘!‘/‘_,". ..' '1‘: . i 2 1 C l O I . . 1 ' '91 . . . t O. . I - -.. .. i ‘ . 4 I ’0 ‘, o ‘ y 3.5.2.1. The Locus ofAccent ................................................................ 61 3.5.2.2. Pitch Pattern ............................................................................. 65 3.52.3. Summary ......................................... - - - -67 3.6. Conclusion ............................................................................................................... 69 CHAPTER 4 The Realization of English lr/ ............................................................................................ 73 4.1. Introduction ............................................................................................................. 73 4.2. Data - .......... - - - ..... - ------..74 4.2.1. Onset/r/ ..................................................................................................... 75 4.2.2. Coda/r/ ...................................................................................................... 76 4.3. Analysis ................................................................................................................... 82 4.3.1. The Inputs to Phonological Processes ....................................................... 82 4.3.2. Analysis ............................................................. . -- - . 86 4. 3 .2. l. Word-final Coda /r/ .................................................................. 86 4. 3. 2.2. Word-medial Coda lr/---- -- - - . - -- - -- - - -94 4.3 ..2.3 Onset/r/ .................................................................................... 97 4.4. Conclusion ............................................................................................................... 98 CHAPTER 5 The Realization of the English Plural Morpheme ............................................................ 100 5.1. Introduction. ..................................... - -- - - - 100 5 .2 Data ....................................................................................................................... 102 5.3. Analysis ................................................................................................................. 105 5.3.1.1. Tateishi (2001, 2003) ............................................................. 105 5.3.1.2. Fukazawa and Kitahara (to appear) ....................................... 109 5.3.2. Analysis -- .................. -- -- -- -- -118 5.3.2.1. The Devoicing of the English Plural Morpheme ................... 118 5.3.2.2. Diachronic Change 111 the Realization of the English Plural Morpheme .............................................................................. 125 5.3.2.3. The Deletion of the English Plural “Morpheme ...................... 130 5. 4. Conclusion--. ----- . - - ------ - - ........ --136 CHAPTER 6 English Compound Abbreviation .................................................................................... 139 6..1 Introduction ........................................................................................................... 139 6.2. Sympathy Theory vs. Weakly Parallel Model -- -- -- - 141 6.2.1. Sympathy Theory. - -- - -- - 141 6.2.2. Weakly Parallel Model-.. -- - - - - - - . 1..-.44 6.3...Data - -- - . ------ 151 6. 4. Previous Studies .................................................................................................... 152 641 [to (1990) ...... - ....... -- -- -- - - -153 6...42 Nishiharactal. (2061)--- ............................................................................. 153 6..43. IthndMester (1997). - - - - - - - .158 6. 5. Analysis--- - ------ -- -- - . -- . - - - ----- 163 viii lu’. ark." I D o 1 . . . 1 a c 1' o 1 a! 1 Day . It I .1 . . O .0 '1. . o a J . 1|. 1 4 o o I 1 a . 1 1 1 . , ‘ ‘ 1 .l. r . . . . . . 11 . . o a p . t a v u n o 6.6. Conclusion ............................................................................................................. 188 CHAPTER 7 CONCLUSION ................................................................................................................ 192 7.1. The Nature of Inputs .............................................................................................. 192 7.2. Japanese phonology ............................................................................................... 194 7.3. Future Research ..................................................................................................... 196 APPENDIX ..................................................................................................................... 197 BIBLIOGRAPHY ............................................................................................................ 234 1. Introduction 1.1. Introduction This dissertation discusses issues surrounding loanword phonology, more specifically issues surrounding the phonology of English loanwords in Japanese. In the past decade, loanword phonology has become a major field of phonology, due to the conceptualshififi’ommlestoaconsn'aintsandrepmmodelofsoundchange (Kenstowicz and Suchato 2004). Especially, in Japanese phonology, the study of loanwords has been ofgreat importance, since most ofthe new words in the Japanese lexicon are loanwords, more specifically American English loanwords (Kay 1995). A number of issues have arisen in this field, but many of them are still under debate. Among them, discussing four phonological phenomena observed in English loanwords in Japmese within the framework of Optimality Theory (OT: Prince and Smolensky 1993), this dissertation especially focuses on issues regarding the nature of inputs to loanword adaptation and the Loanword sublexicon, one of the four sublexica in the Japanese lexicon. The four phonological phenomena are: accentuation of English loanwords (Chapter 3), the realimtion of English /r/ (Chapter 4), the realization of the English plural morpheme (Chapter 5), and English compound abbreviation (Chapter 6). This dissertation will discuss the following issues. With regard to the nature of inputs, it is not clear whether the input is based on the phonetic representation of the source language (Silverman 1992, Peperkamp and Dupoux 2003) or the phonological one (Paradis and LaCharite 1997, LaCharité and Paradis 2005). In this dissertation, adopting Silverman’s (1992) idea that there is an intermediate level (the Perceptual Level) between . I . . 'I '1' I . ’0 |-' 2.1 .’ I". ' l‘ 0'36 " - r .'L.l\ are input, i.e. the output of the source language, and the Operative Level, where phonological changes take place (see Chapter 2), I will claim that the input to the Operative Level is the perceived segment, i.e. the output of the Perceptual Level, and that it is closely related to the phonetic representation of the source language. As will be arguedindetailinChapter4,thisclaimisbasedontherealizationofEnglish{r}in Japanese. Chapter 4 will reveal that the purely phonetic or phonological representation as the input does not explain the realization of English [r] in Japanese. Second, regarding the input of loanwords, it is not clear what information is included intheinput. Inthisdissatafiomlwillm‘guethatlapaneseborrowershaveaccesstome information on the locus of English stress and English morphology. It has not been fully discussedinthelitaanuewhatinfonnafionismcludedmflreinputBasedonthe accentuation of English loanwords (Chapter 3) and realization of the English plural morpheme(Chapter5),Iclaimthattheinfonnationonthelocusofsu'essand morphology in the source language is included in the input. Third, with respect to loanword adaptation, it is not clear who borrows loanwords. Paradis md LaChar-ité (1997) claim that loanwords are borrowed by bilinguals. In this dissertation, I will elaborate their claim and argue that English loanst in Japanese are introduced by limited bilinguals, i.e. Japanese-English bilinguals with the knowledge of English morphology but not necessarily with the knowledge of English phonology. As willbediscussedinChapterS,thisclairnisbasedonthefactthatthestemandtheplural morpheme are treated differently. Furthermore, the fact that all the Japanese people study Englishatseconduyschoolwppmtstheideadenglishlomwmdstapmesem introduced by the Japanese speakers with some knowledge of English. u .. I a. . ,1 0 c C. n! ‘s .0 9' . $1. ‘a '- ’ J u n 1 . v. y\’ '. . ' 1 an. I a r I t I I. .I 1 . 1' L. a I I- . n .1 r; ..‘A ' .I 10 i a . «11 I. ’ n l. 1 r I . .v . 1 4 Fomth,thisdissertationwill discussthcassimilationprocesstothecorepartofme Japanese lexicon md the structure ofthe Loanword sublexicon in Japanese. It has been agreed in the literature (McCawley 1968, no and Mester 1995, 1999, among others) that the Japanese lexicon consists of sublexica. But the structure of the Loanword sublexicon, one of the four sublexica, has not been fully discussed. Analyzing four phonological phenomena of English loanwords, this dissertation discusses the structure of the Loanword sublexicon and how loanwords can be assimilated with regard to some particular aspects. Finally, this dissertation will also discuss issues regarding OT such as how phonological opacity can be accounted for within the fiamework of OT. English compound abbreviation reveals a case of phonological opacity. This dissertation will show that the Weakly Parallel Model (Ito and Mester 2001b, 2003a), a subtheory of OT, accounts for a case of phonological opacity observed in English compound abbreviation straightforwardly. 1.2. Organization of Dissertation This dissertation is organized as follows. Chapter 2 innoduces the bmkground relevant to this dissertation: Optimality Theory, models of loanword phonology, introduction to Japanese phonology including Japanese phonetic inventory, context-free phonological adjustments observed in English loanwords in Japanese, and the Japanese lexicon. Chapter 3 explores accentuation of English loanwords in two major dialects of Japanese, i.e. Kansai Japanese and Tokyo Japanese, within the framework of OT. In this chapter, I will show the following: (1) English loanwords in Japanese preserve English stress, which suggests that the information on the locus ofEnglish stress is included in the input, (ii) the pitch pattern, i.e. the contour, of loanwords in Kansai Japanese is highly predictable, although it has been claimed that the pitch pattern of non-loanwords is unpredictable (Pierrehumbert and Becltman 1988: 214), and (iii) Kansai Japanese and Tokyo Japanese are the same with respect to the accent assignment, although they are different with regard to the pitch pattern. Chapter 4 discusses the realization of English M in Japanese. English /r/ is a unique segmentinEnglishloanwordsinJapaneseinthesarsematitistheonlyconsonantthat can correspond to zero as well as a consonant [r] and vowels, [a] or [o], in the output. In this chapter, I will claim the following: (i) the perceived segment, which is closely related to the phonetic representation of the source language, is the input to loanword phonology, (ii) an input segment can be perceived differently based on the location it appears, and (iii) the diachronic change with regard to the realization of English M in Japanese can be understood as the result of the constraint reranking between two constraints ‘[a:]]pw, i.e. no word-final [az], and MAX, i.e. no deletion. Chapter 5 deals with the realization of the English plural morpheme in Japanese. Three phonological phenomena related to the English plm'al morpheme are observed in Japanese: devoicing, diachronic change, and deletion. In this chapter, my major claims will be the following: (i) the devoicing of the plural morpheme is a phenomenon in the aflixes in the Loanword sublexicon, (ii) Japanese borrowers have access to the morphological information of English, which means that morphological information is ' t. 1 ,. ,. .‘ u 1' l : . ._ -'l 3 :1. l." O .. J ‘ al'irr'l included in the input, and (iii) the accessibility to the morphological information of English is different between the older and younger generations. Chapter 6 concerns English compound abbreviation in Japanese. This chapter deals withacaseofphonological opacity, whichhasnotbcendiscussedinthe literatureand explains it within the framework of the Weakly Parallel Model of Ito and Mester (2001b, 2003a). In this chapter, I will show that there are two types of coda conditions in Japanese. Based on the structrne of trimoraic abbreviated compounds, this chapter also claimsthatanabbreviatedcompoundisnotapmsodicwordconsistingoftwo abbreviated prosodic words but a prosodic word by itself. This chapter will also reveal that the Light-Heavy structure is systematically disfavored in Japanese. Finally, this dissertation is closed in Chapter 7 by summarizing the discussion and claims developed in this study. Major claims of this dissertation will be as follows: (i) loanwords are adapted by limited bilinguals, (ii) the perceived segments, which are closely related to the phonetic representation but not to the phonological representation of the source language, are the inputs to loanword phonology, and (iii) the input includes the information on the locus of stress and morphology of the source language. . . 1 | ‘1' ' . . I ' . ,. 1 . 1 5 h ‘ ‘ ‘ \ I . t ' ’ . ' U I l ‘ I ~ ' I _ . V a ‘ l I . . u. , .' 4 ’ I D . O ‘ I ‘ I I “ (A . " . I ' 1" ,. r. r I 3 ‘4 . s ' ' 1 , r; -.2 ' ‘ j I V ' I. . . . ~1"')'.". . . r. m V “I . ‘ \ ' , U a ‘1 * I t 1 . . r | . ' l 1 , L I / . \ .. , H 1.:1‘)(J ,. . l X I I f n. .1 O O .‘r 2. Background 2.1. Introduction This chapter introduces the bmkground relevant to this dissertation. This chapter consists of three parts. In the first section, I will present a brief overview of Optimality Theory, which is adepted in this dissertation to analyze the four phonological phenomena observed in English loanwords in Japanese. Then, in the second section, I will illustrate two models of loanword phonology: one is phonetics-W (Silverman 1992) and the other is phonology-based (Paradis and LaCharité 1997). Finally, the last section is an introduction to Japanese phonology, where Japanese phonetic inventory, context-free phonological adjustments observed in English Loanwords in Japanese, as well as the Japmese lexicon, will be introduced. 2.2. Optimality Theory This section briefly overviews the basic concepts of Optimality Theory (OT: Prince and Smolensky 1993). OT is an output-oriented, constraint-based theory. In this theory, there are no rules and no serial derivations, and the set of constraints plays a central role. The set of constraints is provided by Universal Grammar. In OT, constraints are all considered to be universal whereas the ranking of constraints is language-specific. This concept is called ‘Richness of the Base’ (Prince and Smolensky I993: 191). The formulation in (1) is Smolensky’s (1996: 5). ‘Richness of the Base' means that there is no restriction on inputs and cross-linguistic variation attributed to the difference of constraint ranking. 1“ 1‘1 1 -.s- . . 1 | .I I! ' .1 . L J I l o ‘ 1., - ‘ 10 o a til i It .1 ‘- t it! .. II .1; ’0 . I t ‘ V 3.1' 11111 (1) Richness of the Base The source of all systematic cross-linguistic variation is constraint reranking. In particular,thesetofinputstothegrammarsofalllanguagesisthcsame.The grammatical inventories of a language are the outputs, which emerge from the grammar when it is fed the mriversal set of all possible inputs. There is another significant property of the OT grammar, which is relevant to language acquisition. That is called ‘Lexicon Optimization’. ‘Lexicon Optimization’ meansthatthe input formisassumedtobeidenticaltotheoutputwheneverthereisno overt evidence of existence of a specific lexical form, because it is the easiest strategy to build a lexicon. (2) Lexicon Optimization (Prince and Smolensky 1993: 192) Suppose that several different inputs 1,, 12..., In when parsed by a grammarG lead to corresponding outputs 0,, 02..., On, all ofwhich are realized as the same phonetic form 43 — these inputs are phonetically equivalent with respect to G. Now one of these outputs must be the most harmonic, by virtue of incurring the least significant marks: suppose this optimal one is labeled 0,. Then the learner should choose, as the 1mderlying form for (b, the input I... All the constraints in or, in principle, are violable, but the violation must be minimal. A candidate is considered to be optimal when it violates constraints minimally Q . . . A 1.. I ..11 pt I 11. v i . a r... ‘o 0‘1 a s \- s I ll 0 . . .1 .I (no . . s l‘ v a ‘1 . .o I, . 1 l . . . . . r u n . I. a. .- . 41 a s 1 .o. 0.. 4 1 I ( w u t. O . . . u . . e . . 1 A {u 1' and best satisfies the constraint hierarchy. Output candidates are evaluated and the best outputformisdeterminedinatableau,asin(3). (3) Input Constraint 1 Constraint 2 a. 6" Output Candidate A “ b. Output Candidate B ‘1 "' In a tableau, output candidates are shown vertically in the lefimost column in random order, while constraints are given horizontally in the top row in a descending ranking fiom left to right, i.e. Constraint 1 is ranked higher than Constraint 2 in the tableau (3). Candidate A satisfies Constraint 1 but violates Constraint 2 twice, whereas Candidate B violates both Constraint 1 and Constraint 2 once. Candidate A is the optimal output in the tableau (3), because Candidate B’s violation of Constraint 1 is fatal, which is indicated by the exclamation mark ‘1’. With respect to Constraint 2, Candidate A violates it more than Candidate B does. But Candidate A’s violation of Constraint 2 is irrelevant to the choice ofthe output form, since Constraint2 is ranked lowerthanConstraint 1. 2.3. Models of Loanword Phonology This section introduces two models of loanword phonology: Silverman (1992) and Paradis and LaCharite (1997). The nature ofinputs to loanword adaptation is still under debate in the literature: Silverman (1992) claims that the input is based on the phonetic representation of the source language whereas Paradis and LaCharité (1997) claim that it is phonology-based. -‘1 ." ... u U 11 u . a .4 i . l A 1 NJ 11' ‘1 1 '1 ‘a L! J-‘- 2.3.1. Silverman (1992) Silverman (1992) proposes a phonetics-based model of loanword phonology. In loanword phonology, the nature of inputs is unclear: for instance, it is not clear whether mempummphmologicalprocessesuephomlogicallyfaimfidmdrewmcelmgmgem theyaresornewhatafl‘ectedbythehostlanguage. Silverman(l992)assumesthatthe speakers of the host language have no access to the phonological representation of the source language and provides the following diagram. (4) Silverman’s Model (1992: 293) incoming Perceptual Operative acoustic .—-) Level 1—9 Level «) output signal representation representation A T Operative Level processes native segment native and tonal phonotactic inventory constraints and constraints preferences In Silverman’s model shown in (4), the input is merely a linguistically unanalyzed acmmficsignalanddrespeakasofdrehostlanguagecannmhaveaccesstoflre phonological representation of the source language. At the Perceptual LeveL the native segment and tonal inventory constraints apply, and restrict the representation of perceived segments.ThemitisatthereLevelthathedsegrnentsmaymdagoM VII. 1 tr . a s ’ I . O I) t‘ l l a P‘. 4 ' I 1’ 6 ”' lo . .. . 9 i ll 1 11 1 r 1 i J . . . I n a I i a ' r . a a. u '1 s' N . . .J a 1 1.1 1. . . . 2 . I . ~ ‘ . . . . . . 0.1 1 d 1 I .. is} 1 phonological operations, triggered by native phonotactic constraints” (Silverman 1992: 293). That is, the speakers of the host language perceive the inputs in accordance with the phonological system of the host language and fit “the superficial input into the native phonological system as closely as possible” (Silverman 1992: 289). This model accounts for why the identical input can be “perceived, represented, and ultimately produced in a distinct manner in each language it enters” (Silverman 1992: 289). There is another hypothesis that plays a role in Silverman’s model, which is the Perceptual Uniformity Hypothesis. This hypothesis means that, at the Perceptual Level, an incoming acoustic signal cannot have more than one acoustic correspondent. (5) Perceptual Uniformity Hypothesis At the Perceptual Level, the native segment inventory constrains segmental representation in a uniform fashion, regardless of string position. However, an input whose acoustic phonetic properties cannot be discerned due toitspresenceinanimpoverishedcontext(acontexttobedeterminedona language-specific basis) is not supplied representation at the Perceptual Level of the loanword phonology. 2.3.2. Paradis and LaCharlté (1997) Paradis and LaCharité (1997) take the opposite position of Silverman (1992) and propose a phonology-based model of loanword phonology. Paradis and LaCharite’s model of loanword phonology is shown in (6). 10 l . 'i‘ ' a ‘v‘ . '.. lgyr; e l I'Icc . ‘ I ). ,I -. e I ,' , . ”~- M‘nv (6) Paradis and IaChmité’s Model (1997: 394) L2 L1 I DICTIONARY DICTIONARY PHONOLOGICAL _ PHONOLOGICAL CONSTRAINTS: CONSTRAINTS: Lexical and Lexical and Postlexical Levels Postlexical Levels PHONETIC OUTPUT Paradis and LaCharité assume that loanwords are introduced by bilinguals who have competencies in both the source and host languages. In this model, the phonological Otupmofthesomcelanguage,whichisbasedonthephonenficrepresmtafion,isthcinput to the host language and the input is incorporated directly into the lexicon of the host language. 2.4. Japanese Phonology This section gives a brief introduction to Japanese phonology: Japanese phonetic inventory, context-free phonological adjustments observed in English Loanwords in Japanese, and the Japanese lexicon. 2.4.1. Japanese Phonetic Inventory The following is the chart of Japanese consonants, which is the revised version of Tsujimura’s (1996: 16) (See also Bloch (1950: 107)). Among the consonants in (7), ll Japanese phonemes are /b/, /p/, /d/, N, /g/, lkl, lzl, Isl, /h/, M, /y/, lwl, lml, and ln/, and the others are allophones.l The chart in (8) shows Japanese vowels (Tsujimura 1996: 18, Nakajo 1989). (7) Japanese Consonants bilabial alveolar “V” palatal velar uvular glottal Stops: [+V] b d ‘ s 2 W] p t k Fricatives: [+V] z 22 [*V] «D s 5 9 h Amieateo:[+v1 clz I W] t‘ 6 Approximants: Liquid [+V] ' Glide [+V] y w Nasals: [+V] m n 113 p1 1] u '[mlisephonemebutitalsocanappearasanallophoneof/nl. 2Tettjitmtm (1996: 19) points out that “some native speakers seem to have [2] in rapid speech.” Bloch (1950)discussesthissomd inmoredetail. AccordingtoBloch (1950: 101), “Mostplnuescontaininglzl mpuflkledbyodmwinflmficflsynmymomphmcmmmingwflmmulhlmwmonm [d2],andinthespeechofmanypersonsdoesnotoccuratall. Examples: [mizlkailshom [alum] may, [sin-hm] thirty.” Bloch distinguishes mediovelar stop and nasal, i.e. [g] and [n], from prevelar stop and nasal. i.e. Is] and [01- Bloch (1950: 109) mentions that ‘inthmpain ofphom-{a nus. amt. ll—the membesmhpufidlyfiwvmiafimwifiemhotha,humuamvmlusbekeptaput”“shwedn altanationbetweenmgAfland[n.q,t]respectivelyarelimitedtocenainpluasesonly,andaincedleir commmvhommudonmfomaphmeficdlyapMmicanydefinabksetmedneepehsofpm mustbetreetedseparatelyinthephonemicanalysis.” ’wstlttepetdeto [fl] and [p], Tsujimura (1996: 20) explains as follows: “The eouaettletiott involvingthc nudemsmanbmdsobeohavedbefmedmmlahlmdpflaflwmmehdmemme 12 (8) Japanese Vowels front central back high 1 ul‘ mid e 0 low a 2.4.2. Adjustments Observed in English Loanwords in Japanese Some English phonemes do not exist in the Japanese phonemic inventory and Japanese possible syllable structures, i.e. (CXG)V(V)(C) where G is a glide [y] and coda consonantiseitheranasalorthefirsthalfofageminate,aredifl‘erentfromEnglish possible syllable structures. These facts lead to a number of adjustments having taken place in the borrowing English words into Japanese. The adjustments are divided into two categories: phonetic adjustments and vowel epenthesis. The major phonetic adjustments are shown in (9) - (l 1) (National Language Institute 1990, 01130 1991, Tsuchida 1995):“ nasalisrealizedasalveo—palatalnasel [n] andpalatalnasal [n],respectivcly. Examples includekalzya (mi) [keil la (nai)] ‘it is not a ticket’ and ken ya (kane) [ken ya (kanc)] ‘things like ticket and (money)’.” ‘Thesymbolfutlindicatesthemummdedhighbackwml.Whereasthelackofhpmlmdingismme pronrinentintlle'l‘okyodialecmhehighbackvoweltendstoberomdedintheWestemdialectsCI‘mjimma 1996: 18; snows I990: 161). Thesymbol forrmmdedhigh beck vowel in [u]. ’Theedjtnunetnorsngllehltlwmbediecmsedtndeatlincnaptet4. l3 I'e | . U 1 . - | 1 _' , I as ,‘I:" "'I I . ‘t c I l .“ I l .‘ (9) Consonants English 1W 8.11] ..., [‘13] (fur -’ [983]) b- [V] -’ [b] (Vitew -+ [by “'11) 019] -" [S] (Ithaca .... [isaka]) d- 15] "t [2] (1110mm -' (“182831) e- [I] -' If] (fly -* [IllwraiD (10) Lax Vowels“ English Japanese a- [t] -' [i] (pin -+ [piND b. [e] -* [e] (Pen —’ [FWD 9.1:] -+ [a] (rally --> [rari=]) ct [A] -+ [a] (cut —+ [kattoD e. [a] -+ [a] (top —» [toppwn £13] -» [0:] (call —+ 11:0:er s10] -* ["1] (book -+ [bwkka 6A:Inttttlltnow,tlteteoliznttlonofio]ItasnotgottttentlonInttteliteratuteentlnooneltee:tntlietllt systemtically. Interestingly, however, [a] in English words can correspond to the five vowels in Japanese, as illustrated below. In the examples below, the vowels corresponding to English [a] are shown in hold. The realization ofEnglish [9] seems relevant to English orthography. a. [a] —-> [i] (terminal -» [ta:mlmrut]) b. [a] ... [cl (W -+ 1W1) 9- [=1 -' [a] (my -* 19ml) (L [a] -’ [01 (vision -» Maul) e. [a] -> [In] (even -v [izbumD l4 a J 5 I i a ILL." l I 9ft. I ..-..t t’ {Imfl' l (11) Tense Vowels English Japanese 8- [i] -' [it] (key —' [kitD b- 10] -* [0:] (zone -’ [2034]) c. [u] -r [w:] (cue -—> [11’sz As mentioned above, Japanese possible syllable structures are (CXG)V(V)(C), and possible syllable structures in English and Japanese are different. This leads to the second type of the adjustment: vowel epenthesis. That is, since the sequence of consonants and coda consonants are highly restricted in Japanese, vowels are epenthesized to avoid the illegitimate structures. The epenthetic vowels are determined as in (12) (Ohso 1991, Kobayashi et a1 1991, Jorden et a1 1976, among others). (12) Epenthetic Vowels a. [o]isinsertedatter[t]and[d] (travel .. [toraberulD (road -' [rotdOD b. [i] is inserted afier [e], m, and [t] (bench -’ Ml) (brush -' lbwmfil) c. [In] is inserted elsewhere (three ~+ [swfiiD (noise -> [noidzuiD 15 I ‘ ' ' r‘ A z D I ll . g f l 1 l i A '0” -' II". n I"! .1 'st', i ‘ "j: , l! U .9 .0 In the rest of this dissertation, the orthography-based representation of loanwords, which is based on the Japanese orthography, will be adopted unless the phonetic representation is relevant to the discussion. 2.4.3. Japanese Lexicon It has been agreed in the literature that the Japanese lexicon consists of sub—lexica (McCawley 1968, Ito and Mester 1995, Ito and Mester 1999, Fukuzawa, Kitahara, and Ota 1998, among others). Ito and Mester (1995) claim that the Japanese lexicon consists of four sublexica, i.e. Yamato, Sim-Japanese, Mimetic, and Loarlword, and those sublexica are organized in a core-periphery structure. Sublexica are characterized by three constraints that apply. The relevant constraints are *P (no single [p]), ‘NT (no post- nasal voiceless obstruent), and ‘01) (no voiced obstruent geminate). As illustrated in (13), the Yamato sublexicon is the most restricted while the [oarlword sublexicon is the least. (13) (ltd and Mester 1995: 820) ‘P ’NT '00 a. Yamato 7 f t/ b. Sine-Japanese / t/ c. Mimetic / l d. Loanword In the core-periphery model, it has been assumed that the Yamato, Sine-Japanese, and Mimetic sublexica exist as strata and form the core part of the lexicon, whereas the Loanword sublexicon does not constitute a uniform stratum and loanwords exist in less centralareasofthelexiconwheremoreconstraintscanbeviolated 16 «Q o f‘.’l 9W9 Sublexica in the Japanese lexicon are not always defined by etymology. The Portuguese loanword karma ‘(playing) card’ is etymologically a loanword but it is considered as a Yamato item because it has phonological characteristics of Yamato items (e.g. Rendaku (Sequential Voicing): ham ‘flower’ + karma -’ hana-garura (‘hana- karuta) ‘flower card game’). Also, items in the Japanese lexicon, which are etymologically loanwords, can phonologically behave like the items in the core part of the lexicon with respect to some particular aspects. For example, the English word ‘bag’ behave like a core item with regard to the constraint ‘DD, which, as illustrated in (13), does not apply to the items in the Loanword Sublexieon (‘bag’ --9 babbl (or baggu); cf. ‘rod’ -» roddo (‘rotto)).7 Based on this, no and Mester (1999) divide the Loanword sublexicon into two constituents, “Assimilated” Loanword sublexicon and “Unassimilated” Loanword sublexicon. The items in the “Assimilated” Loanword sublexicon behave like core items with regard to some particular aspects, while the items in the “Unassimilated” Loanword sublexicon do not. The distinction between “Assimilated” and “Unassimilated” better explains the structure of the Loanword sublexicon. But, as ltd and Mester (1999: 70) point out, “many finer distinctions are hidden beneath this coarse classification: the less nativized an item is, the more it disobeys lexical constraints, i.e. the more it falls outside of various constraint domains and is located towards the periphery of the lexical space.” This dissertation further discusses the structure of the Japanese lexicon, especially the structure of the Loanword sublexicon and how a loanword can be assimilated. ’Theapplicationof‘DDto‘bag’isOpfionalandsubjecttoindividualdifl’erenoes,aldroughbakkaseems morecommonthanbaggn. 17 n~ 3. I l ' . t, e .1: J:_7 (‘1 r ‘l u .: ",l .' p .,, ,, “H ', ._.r.v\ i. {‘e 3. Accentuation of English Loanwords. 3.1. Introduction This chapter explores accentuation of English loanwords in two major dialects of Japmese, i.e. Kansai Japanese and Tokyo Japanese, within the fi'amework of Optimality Theory (OT: Prince and Smolensky 1993).l Accentuation is one of the major issues in Japanese loanword phonology and many studies have been conducted (McCawley 1968, Kubozono 1994, Labrune 2002, others). Most of the studies, however, are on the loanword accentuation in Tokyo Japanese and the loanword accentuation in other dialects suchasKansaiJapmesehasbeenstudiedlittle. This chapter examines the accentuation of English loanwords in Kansai Japanese andTokyo Japanesewiththreegoalsinmind. Thefirstgoalofthischapteristodiscuss the pitch pattern, i.e. the contour, of English loanwords in Kansai Japanese. In Tokyo Japanese, given the locus of the accent, the pitch pattern of the whole word is predictable. InKansaiJapanese,ondieoflrerhmdthepitchpatternprecedingtheaccentedmorahas not been explained in the literature (Pierrehumbert and Beckman 1988: 214). This chapter analyzes the pitch pattern of English loanwords in Kansai Japanese within the OT fiameworkandshowsdratdlepitchpanemprecedingflleaccentedmomismosdy predictable. ThesecondgoalistoexaminethelocusoftheaccentofEnglishloanwordsin Kansai Japanese. Regarding the location of the accent of loanwords in Kansai Japanese, few studies have been conducted. The data collected for this study shows that about 50% ‘BuliavasimorpmorthnenapterappeorinpennwwngpapeninLingtnenesIr.landCLS4o.l. ‘KamaiisanareainlapanwhereKobeOsaka,Nara,andeotoareinchrded. 18 v .- a r e 1 1' .“ . . ‘ . ‘ ’ . . ‘ r ' . f I. f n 1 . ' l I ' ‘ . if .I -' J - ”'H- ’~ ‘ 1 ' H) ..v ‘ t ’ ‘ . . I - ' ' ' I . ‘. ' ..2). l ‘. . ‘. 1 I n I ‘ 1 .' “. (:-.si)- 11 , if 3! l i new \. n s i t .' "‘ ' a i ' t L I J O - . - ' ' l | ‘ ' - p l . i l. ' 'h' ' 3 I. 1 9., n ' ‘ ‘\ . 4 .1 ,- .. ‘ I. - ‘W‘. ... ' I ' . l . ‘ a t {a ' U a I ‘ h . . I a . , ‘ A' ‘. ‘1 1 7 4 . O .1 ' .1, ‘. up. AH, ‘1 1 ' of 1090 English loanwords in Kansai Japanese have the antepenultimate accent, which is considered the default accent ofloanwords in Tokyo Japanese in the literature (Ono 1991, Tanaka 1992, Katayama 1995, Kubozono 1995b, Kubozono and Ohta 1998, others). This chapter reveals that most of English loanwords retain the English accent, i.e. the accent onthesyllablestressedinEnglish,whichindicatesthattheinformationonthelocusof Englishstressisincludedintheinput. Finally, this chapter reanalyzes English loanwords in Tokyo Japanese and compares KmsaiJapaneseandTokyoJapanese.TMschaptershowsmathrsaiJapanesemd Tokyo Japanese are the same with respect to the accent assignment, although they are difl‘erent with regard to the pitch pattern. This chapter is organized as follows. Afier illustrating an overview of the accentuation of non-loanwords in Tokyo Japanese and Kansai Japanese in section 2, section 3 reviews three previous studies of the loanword accentuation in Tokyo Japanese. Then,insection4andsectionS,EnglishloanwordsinKansaiJapaneseandTokyo Japaneseareanalyzed. Finally,thischaptereoncludesinsection6bysummarizingthe analysis and pointing to further issues. 3.2. Accentuation of Non-loanwords in Japanese Japanese is considered as a pitch-accent language, which means that the pitch pattern ofthe whole word is predictable given the locus ofthe accent ofthe word. Japmeseisdividedmmfombiggloupsbasedonwcenmafionsystems.Ammgthem, this chapter focuses only on Kansai Japanese and Tokyo Japanese. The accent in Japanese is marked by an abrupt falling pitch where, following the convention, the accent l9 me ill! is marked by an apostrophe placed immediately after the accented more and high- and low-pitched morae are indicated by ‘H’ and ‘L’ respectively. (1) a. tebu'kuro LHLL ‘glove’ (Tokyo Japanese) b. tebuku’ro LLHL ‘glove’ (Kansai Japanese) The rest of this section introduces the accentuation of non-loanwords in Tokyo Japanese and Kansai Japanese. 3.2.1. Accentuation of Non-loanwords in Tokyo Japanese The location of the accent of non-loanwords in Tokyo Japanese is lexically determined and impredictable, as illustrated in (2). (2) a. ha'si HL ‘chOpsticks’ (accent on the first syllable) b. hasi' LH ‘bridge’ (accent on the last syllable) c. hasi LH ‘edge’ (unaccented) Each word is either accented (e.g. (2a) and (2b)) or unaccented (e.g. (2c)). The word with the accent on the final more, i.e. (2b), and the lmaccented word, i.e. (2c), in isolation have the same pitch pattern. When they are followed by a particle such as the nominative marker ga, however, they show distinctive pitch patterns, as illustrated in (3). 20 (3) a. hasi' ‘bridge’ (accent on the last syllable) hasi’ga LHL b. hasi ‘edge’ (unaccented) hasiga LHH The possible pitch patterns of words of one to four light syllables are shown in (4) - (7), where both the pattern of the word in isolation and the pattern followed by the nominative marker ga are given. The examples below are from Sugito (1982: 3), except for (7b) and (7d), which were collected by the author. (4) Monomoraic Words a. as H ‘name’ (unaecented) naga LH b. na' H ‘eating greens’ (accented) na'ga HL (5) Bimoraic Words a. hasi LH ‘edge’ (unaccented) hasiga LHH b. ha'si HI. ‘chopsticks’ (accent on the first syllable) ha'siga HLL c. hasi' LH ‘bridge’ (accent on the last syllable) hasi'ga LHL 21 I. I I’ll! (6) Trimoraic Words a. kodomo LI-IH ‘child’ (rmaccented) kodomoga LHHH b. mi'dori HLL ‘green’ (accent on the first syllable) mi 'doriga HLLL c. koko'ro LI-IL ‘spirit’ (accent on the second syllable) koko'roga LHLL d. kagami' LHH ‘mirror’ (accent on the last syllable) kagami'ga LHHL (7) Quadrimoraic Words a. tomodati LHHH ‘fi'iend’ (unaccented) tomodatiga LHHHH b. ma'tibari HLLL ‘pin’ (accent on the first syllable) ma'tibariga HLLLL c. tebu'kuro LHLL ‘glove’ (accent on the second syllable) tebu’kuroga LHLLL d. zeita'ku LHHL ‘extravagance’ (accent on the penultimate syllable) zeita'kuga LHHLL e. kaminari' LHHH ‘thunder’ (accent on the last syllable) kaminari'ga LHHHL The following five generalizations can be made from (4) -— (7). First, the accented mora is high-pitched. Second, the morae following the accented mora are low-pitched throughout. 22 (1.0311. n e b ‘ «a Q I e ' _, I '( l i 1 it 4 Third, the morae preceding the accented mora are high-pitched, but the word-initial pitch and the pitch of the second mora must be distinct. That is, the first mora is low-pitched and the second more is high-pitched unless the first more is accented, otherwise the first more is high-pitched and the second mora is low-pitched. Fourth, a prosodic word cannot have more than one accented mora. Finally, a prosodic word must have at least one high- pitched mora. When a heavy syllable is accented, only the first mora of the syllable can be accented, m illustrated in (8). (8) a. oto'oto2 LHLL ‘younger brother’ (‘otoo'to LHLL) b. ge'ndai HLLL 'today’ (*gen'dai LHLL) c. ka'kkoo HLLL ‘cuckoo’ (’kak'koo LHLL) 3.2.2. Accentuation of Non-loanwords in Kansal Japanese The location of the accent of non-loanwords in Kansai Japanese is lexically determined and unpredictable, as in Tokyo Japanese (cf. (2)). (9) a. ha'si HL ‘bridge’ (accent on the first syllable) b. hasi HH ‘edge’ (unaccented) c. hasi LH ‘chopsticks’ (unaccented) ’ In this and the following chm innead of phomtic representation (e.g. [03). long vowels in examples are represented by the sequence oftwo vowels (e.g. ‘oo’) for convenience. 23 As illustrated in (9), each word is either accented (e.g. (9a)) or unaccented (e.g. (9b) and (9c)). Unlike Tokyo Japanese (cf. (2)), Kansai Japanese has two types of unaccented words: high-pitched unacccnted words (e.g. (9b)) and tmaccented words with the word- final high-pitched mora (e.g. (9c)). When an unaccented word of the second type (e.g. (9c)) is followed by a particle such as the nominative marker go, the high pitch moves to the word-final position, i.e. onto the nominative marker, as shown in (10c). (10) a. ha'si HL ‘bridge’ (accent on the first syllable) ha'siga HLL b. hasi HI-l ‘edge’ (unaccented) hasiga HHH c. hasi LH ‘chopsticks’ (unaccented) hasiga LLH The examples in (ID-(14) show the possible pitch patterns ofwords ofone to four light syllables. The accentuation system of Kansai Japanese is more complex than that of Tokyo Japanese (cf. (4) — (7)). In the examples below, both the pattern in isolation and the pattern followed by the nominative marker ga are given. The examples below are fiom Sugito (1982: 3) except for (14b), (14c), (14¢) and (140, which were collected by the author. 24 ‘a .t) ’9‘! ~V‘. (11) Monomoraic Words’ a. ko kosa b. ee ega c. ke'e ke'ga (l2) Bimoraic Words ahasi d. ame'e amega \ H HH 5 E LLH HL LHL ‘child’ ‘picture’ ‘hair, (unaccented) (unaccented, lengthened) (accented, lengthened) (miacccnted) (unaccented) (accent on the first syllable) (accented, lengthened) 3 In Kauai Japanese, unlike Tokyo Japanese, lexicallyomonomoraic-words can be lengthened (e.g. (11b) “‘1 (1 lo». 25 yf (13) Trimoraic Words a. kodomo HHH kodomoga HHHH b. tokai LLH tokaiga LLLH c. ko'koro HLL ko’koroga HLLL d. kaga'mi HHL kaga'miga HHLL e. kata'na LHL kata'naga LHLL (l4) Quadrimoraic Words a. tomodati HHHH tomodatiga HI-IHHH b. jagaimo LLLH jagaimoga LLLLH c. ma'tibari HLLL ma'tibariga HLLLL d. mago'koro HHLL mago'koroga HHLLL e. kitu'tuki LHLL kitu'tukiga LHLLL f. tetuda'i HHHL tetuda'iga HHHLL g. tebuku'ro LLHL tebuku'roga LLHLL ‘child’ ‘sword! ‘fi'iend’ ‘help’ ‘glove’ 26 (unaccented) (unaccented) (accent on the first syllable) (accent on the second syllable) (accent on the second syllable) (unaccented) (unaccented) (accent on the first syllable) (accent on the second syllable) (accent on the second syllable) (accent on the penultimate syllable) (accent on the penultimate syllable) The following six generalizations can be drawn from (ID—(14). First, the accented mora is high-pitched. Second, a prosodic word cannot have more than one accent, and a prosodic word must have at least one high-pitched mora. Third, the morae following the accented mora are low-pitched throughout. Fourth, the morae preceding the accented mora can be either high- or low-pitched throughout (e.g. (141) and (l4g)). Fifih, when a prosodic word has a low-pitched mora word-initially, only the accented mora of the accented word (e.g. (l4g)) or the last mora of unaccented word (e.g. (l4b)) can be high- pitched. Finally, the accent on the final mora is not allowed.4 When a heavy syllable is accented, lmlike Tokyo Japanese (cf. (8)), both the first mora (e.g. (15a)) and the second mora no. syllable, i.e. the second half ofa long vowel (e.g. (15b)) and the coda (e.g. (15c)), can be accented. (15) a. ge'ndai HLLL ‘modern times’ b. kyoo'dai LHLL ‘brother’ c. kon'ban LHLL ‘tonight’ 3.3. Previous Studies Accentuation is one of the main issues in Japanese loanword phonology and many studies have been conducted. But, most ofthem are about the loanword accentuation in Tokyo Japmese, and the loanword accentuation in Kansai Japanese has been studied little. ‘Yoshida deunmaaOOl: 216) give an example ofthe word-firm] accent, kald' ‘oyster’. As Yoshida demmapoimmhowVfl,mEBMsmficdlyprommceduWiwhemmefimlmmhlmgmem¢ mddnprommiafionwithomlengdnningisthensultofamcentdialectalchmge.Inthispaper,lwill ignorethistypeofexamples. 27 l" .5 'J 1"}. fa , J ..j This section gives an overview of three previous studies on the loanword accentuation in Tokyo Japanese: McCawley (1968), Ono (1991), and Asano (1999). 3.3.1. MeCawley (1968) McCawley (1968) is a milestone in the study of loanword accentuation in Tokyo Japanese. He observes the characteristics of the loanword accentuation in Tokyo Japanese given in (16). (16) McCawley (1968: 134 £11. 6) Loanwords fall into three classes based on the accentuation: i) unacccnted words, ii) words accented on the syllable containing the antepenultimate mora, iii) words accentedonthesyllablewhichwasstressedinthesoln'celanguage. Following the observation above, many researchers claim that the antepenultimate accent is the default loanword accent in Tokyo Japanese (Ono 1991, Tanaka 1992, Katayama 1995, among others). In addition, McCawley (1968: 134 iii. 6) refers to Tashiro’s (1953) interesting observation regarding the antepmultimate accent that “this ante~penultimate accent is also the normal pronunciation for lists of meaningless syllables; for example, the kana syllabary is recited: bflflr'kib, kalcilcu 'keko, tatitu'teto, 28 3.3.2. One (1991) Ono (1991) discusses the accentuation of loanwords in Tokyo Japanese, comparing with that in other languages: English, Russian, Turkish, Polish, and Macedonian. From the studies of the loanword accentuation in the five languages, he finds three common features in (17), and shows that these common features apply to the loanword accentuation in Tokyo Japanese as well. (17) Common reduces of Loanword Accentuation a. The accentuation of loanwords is different from that of native words. b. The accentuation of loanwords is as simple as that of non-loanwords (e.g. Polish and Macedonian), or simpler than that of non-loanwords (e.g. English, Russian, Turkish). c. The accent on loanwords can appear only on the syllable following the preantepenultimate syllable: on the last, penultimate, or antepenultimate syllable. Ono claims that the default loanword accent in Tokyo Japanese is the antepenultimate accent. He accOlmts for the exceptions to his claim as follows. First, some loanwords have the accent on the mora preceding the antepenultimate mora, e.g. preantepenultimate accent, as shown in (18). 29 (18) a. bo'onasu HLLL ‘bonus’ b. a'kusesu HLLL ‘access’ c. te'kisasu HLLL ‘Texas’ d. yu'nion HLLL ‘union’ e. si'glmaru HLLL ‘signal’ f. ra'gubii HLLL ‘rugby’ Exceptions of this type can be divided into two classes: class 1 ((18a) and (18b)) and class 2 ((18c) - (186). In class 1, either the second halfof a heavy syllable (e.g. (l8a)) or a devoieed vowel (e.g. (18b)) is the antepenultimate mora. The loanwords of class 1 do nothavetheaccentontheantepenultimatemorabecausetheaccentonthesecondhalfof a heavy syllable or a devoiced vowel is not allowed. Therefore, the accent in class 1 words moves one mora lefi fi'om the antepenult. Kanno (I971) observes that the loanwords ending with -su, -n, ~ru, or a long vowel have the accent not on the antepenultimate mora but on the preantepenultimate mora (e.g. (18c) — (181)). Those are the exceptions of class 2. Ono claims that they do not have the antepenultimate accent because ~su, -n, mo, and the second half of a long vowel are extrametrical elements in Japanese. Second, the loanwords in (19) have the accent not on the antepenultimate mora but onthepenultimatemora. Thelocusoftheaccentoftheloanwordsofthistype conespmdstothatofthesomcewords.Thatis,dlesewordsretahrdlemiginal accentin the source language. 30 (19) a. sups'i LHL ‘spy b. tora'i LHL ‘try’ c. gure'e LHL ‘gray’ Finally, the loanwords in (20) are unaccented. Following Kanno (1971), Ono claims thatunaccentedloanwordsaretheloanwordsbormwedlongtimeagoandtheyarethe result of nativization. (20) a. botan LHI-I ‘button’ b. sutamina LHHH ‘stamina’ c. guralmdo LHHHH ‘grolmd’ Based on the examples above, Ono hypothesizes that the nativization process of loanword accentuation has three stages. First, when a word is borrowed, the word is accentedonthesyllablesuessedinthesourcelanguage.‘lhen,inflresecondstagethe accent changes to the default pattern, i.e. the antepenultimate accent. Finally, the word loses the accent and becomes an unaccented word. 3.3.3. Assoc (1999) Asano(l999) clairnsthatthebasic loanword accentinTokyo Japaneseisthc accent on the syllable stressed in the source language. According to Asano, the locus of the accent is determined by the two groups ofnlles: rules offooting and rules ofaccent shift. 31 ‘lo or .1.! ‘J ...g Feet in Japanese are bimoraic (Poser 1984 and 1990, Tateishi 1989, ltd 1990, ltd and Mester 1992, Katayama 1995, Kubozono l995a, Asano 1999, among others). Feet are assigned by the three steps in (21). (21) Step 1: Assign a foot to the last two morae, regardless of their syllabic position. a....u u# —-> ...(p u)# l I l l O G O 0’ b....p. u# —-> ...(u u)# \/ V 0‘ 6 c was!” -+ ”.1101 u)# v I V I O 0’ 0’ 6 Step 2: Parse heavy syllables into feet. d....oo o...o oo...o# —. ...o o o...o oo...o# IAIAIIA l/\|/\l|/\ mum anus (1w) M11101: (111011110110 Step 3: Parse light syllables into feet iteratively fiom lefi to right. Degenerate feet arenotallowed. e o o o o...# -+ ...c o o o o...# AIIIAIIA Alli/\HA (samusuwwuuw) (rm ) (11 mm») Therearefomrulesthatmovethelocusoftheaccent,asgivenin(22).Inthe examples below, the accents are indicated by the apostrophes. 32 (22) a. Move the accent to the antepenultimate mora when it is in the last foot. i) ...o o# ...o o# I /\ l /\ 11(11'11) -+ 11'(1111) ii)...o out! ...6 cell /\ I l /\ ll 1111(11'11) —+ 111111111) b. Move the accent to the antepenultimate mora when it is on the foot preceding the penultimatefoot. i) ...o o oo# ...o o 001! l /\ H I/\ ll 11'(1111)(1111) -+ 11(11'11X1111) ii)...o oo o# ...o oo o# /\ ll /\ /\ ll /\ 11'11(1111)(1111) -' 1111(1111’X1111) c. Move the accent to the mora immediately preceding it when it is on the devoiced vowel. In the example below, /i/ in italic bold is the devoiced vowel. i) ofi'syaru -1 o'fisyaru ‘official’ d. Move the accent to the head of the syllable when it is in coda position. i) tinpanji'i ‘chimpanzee’ ii)e'rebeetaa ‘elevator’ tinpan'jii Rule (22a) erebee'taa Rule (22b) tinpa'njii Rule (22d) erebe'ctaa Rule (22d) In sum, Asano claims that, due to the rules in (21) and (22), the accent falls on the antepenultimate, the preantepenultirnate, or the filth mora from the last. 33 is.) In ‘. I . o s I. .1 ,, Iv 1’: .11? .‘ ~ 3.4. Kauai Japanese This section analyzes the location of accent and the pitch pattern of English loanwords in Kansai Japanese within the framework of OT. 3.4.1. Data For this study, 1090 unabbreviated English loanwords of three to eight morae (two to eight syllables) were collected.’ The main source of the data is Horiuchi (1996). Adjectives and verbs are not included in the data because the majority of the loanwords inJapanesearenounsandverbsaremadefl-omnounsbyaddingaverbsum ‘todo’ (e.g. tesuto ‘test’ +suru‘todo’ -1tesutosuru‘totcst’). The locusoftheaccentandthepitch pattern of the loanwords in Kansai Japanese are based on the author’s knowledge, who is a native speaker of Kansai Japanese. Bilnoraic loanwords are excluded from the data sincenoloanwordhastheaccentonthefinalmora,andthelocusoftheaccentandthe pitch pattern of accented bimoraic loanwords is always H'L (H = high pitch, L = low pitch). Also there is no monomoraic loanword in the data simply because monomoraic loanword does not exist in the Japanese lexicon. The present study focuses only on rmabbreviated English loanwords and ignores abbreviated ones. I will leave the accentuation of abbreviated loanwords for future study. Also, since the accentuation of compounds have different characteristics from that of non-compounds, compounds are excluded from the data.“ ’ThecompletelistisgivenintheAppendix. ‘nhuheenclelmedlndlemennnemmelocosonordeoccemeodmelengthordlesecondelemenror awmpmmdddamimthelmoftheMofhewmpmMmeedemileddiscussionofme accentuation of compounds in Japanese, see Kubozono’s series of studies (Kubozono 1994, Kubozono 1995b, KubozonoandOhta l998,arnongod1ers). 34 \ ‘ ’l.! .l' 11 .. ... The present study categorizes the unabbreviated English loanwords in Kansai Japanese intothreetypes,andeachtype isdivided intotwo subgroupsbasedontheword- initial pitch, as shown in (23). (23) (H = word-initial high pitch, L = word-initial low pitch) a. (i) English Type H (ii) English Type L b. (iii) Non-English Type H (iv) Non-English Type L c. (v) Unaccented Type 11 (vi) Unaccented Type L 3.4.1.1. English Type The accented loanwords that have the accent on the syllable stressed in English belong to English Type. The loanwords of this type are divided into two subgroups based on the word-initial pitch: high (see (24)) or low (see (25)). English Type H consists of 460 words of three to nine morae (two to six syllables), whereas English Type L consists of 272 words of three to eight morae (two to six syllables). In total, 732 words out of 1090, i.e. 67.2%, belong to English Type. This type of accentuation is most common in English loanwords in Kansai Japanese. 35 ‘ I 1"! (24) English Type I-I Kansai J. a. ba'taa HLL b. me'rodii c. pi'kunikku HLLLL d. konse'nsasu HHHLLL e. paasona'ritii HHHI-ILLL f. furasutore'eshon HHHHHLLL g. entaate'imento HI-II-IHHLLLL Tokyo J. (ha'taa) (HLL) (me'rodii) (HLLL) (pi 'kunikku) (ELI-LL) (konse'nsasu) (LHHLLL) (paasona'ritii) (LHHHLLL) (furasutore'eshon) (LHHHHLLL) (entaate'imento) (LHHHHLLLL) 36 ‘entertainment’ (3-mora, 2-syllable) (4-mora, 3-syllable) (S-mora, 3-syllable) (6-mora, 4—syllable) (7-mora, S-syllable) (8-mora, 6-syllable) (9-mora, 6-syllable) (25) English Type L Kansai J. Tokyo J. Gloss a. kuru'u (kuru'u) ‘crew’ (3-mora, 2-syllable) LHL (LHL) b. slma'kku (suna'kku) ‘snack’ (4-mora, 3-syllable) LHLL (LI-ILL) c. ame'nitii (ame’nitii) ‘amenity’ (S-mora, 4-syllable) LHLLL (LHLLL) d. wisuko'nsin (wisuko'nsin) ‘Wisconsin’ (6-mora, 4-syllable) LLHLLL (LHHLLL) e. imaj ine'eshon (imajine'eshon) ‘imagination’ (7-mora, 5~syllable) LLLHLLL (LHHHLLL) f. insutora'kutaa (insutora'kutaa) ‘instructor’ (8-mora, 6-syllable) LLLLHLLL (LHHHHLLL) 3.4.1.2. Noll-English Type The accented loanwords that do not have the accent on the syllable stressed in English are categorized into Non-English Type. The loanwords of this type are divided into two subgroups based on the word-initial pitch: high (see (26)) or low (see (27)). Non-English Type H consists of 150 words of three to eight morae (two to eight syllables), while Non-English Type L consists of 65 words of three to seven morae (two to six syllables). In total, 215 words out of 1090 English loanwords, i.e. 19.7%, belong to this category. Interestingly, 180 words out of 215 loanwords of this type, i.e. 83.7%, have the accent on the syllable containing the antepenultimate mora, i.e. on the antepenultimate mora when the syllable is light (e.g. (26g)) or on the preantepenultimate 37 _ 1L. .- when the syllable is heavy (e. g. (27e)). This strongly suggests that the accent moves onto the syllable containing the antepenultimate mora when a loanword does not have the English accent, i.e. the accent on the syllable stressed in English. I will further discuss this below. (26) Non-English Type H Kansai J. Tokyo J. Gloss a. ka'nuu (ka'nuu) ‘canoe’ (3-mora, 2-syllable) HLL (HLL) b. ro'mansu (ro'mansu) ‘romance’ (4-mora, 3-syllable) HLLL (HLLL) c. baruko'nii (baruko'nii) ‘balcony’ (S-mora, 4—syllable) I-IHHLL (LHHLL) d. jaanari'zumu (jaanari’zmnu) ‘journalism’ (6-mora, S-syllable) HHHHLL (LI-IIIHLL) e. koresutero'oru (koresutero'oru) ‘cholesterol’ (7-mora, 6—syllable) HHHHHLL (LHHHHLL) £ indianapo'risu (indianapo'risu) ‘Indianapolis’ (8—mora, 7-syllable) HHHHHHLL (LHHHHHLL) g. erekutoroni'kusu (erelmtoroni 'kusu) ‘electronics’ (8-mora, 8-syllable) 38 (0‘4, DO. I | 0“.*, l I V I ‘u :l’ ' ~ ‘ 1.1 i p . (27) Non-English Type L Kansai J. Tokyo J. Gloss a. gaurn (880m) (gown, LHL (HLL) b. ime'eji (ime'eji) ‘image’ LHLL (LHLL) c. edi'nbara (edi'nbara) ‘Edinburgh’ LHLLL (LHLLL) d. supeshari'suto (supeshari'suto) ‘specialist’ LLLHLL (LI-IIIHLL) e. konpure'kkusu (konpure'kkusu) ‘complex’ LLLHLLL (LHHHLLL) 3.4.1.3. Unaccented Type (3-mora, 2-syllable) (4.111013, 3-syllable) (S-mora, 4—syllable) (6-mora, 6—syilable) (7-mora, S-syllable) Finally, all the unaccented loanwords in Kansai Japanese belong to Unaccented Type. Unaccented Type is also divided into two subgroups based on the word-initial pitch: high (see (28)) or low (see (29)). Unaccented Type H consists of 134 words of three to six morae (two to six syllables), while Unaccented Type L consists of 9 words of three to six morae (two to four syllables). Among the six groups of accentuation in Kansai Japanese introduced so far, Unaccented TypeListheleastcommon. 39 (28) Unaccented Type H Kansai J. a. baajon HHHH b.botoru HHH c. supuringu HHHHH d. bakuteria III-[HHH e. kariforunia HHHHHH (29) Unaccented Type L KansaiJ. akaree LLH b. marason LLLH c. sutookaa LLLLH d. rekoodingu LLLLLH Tokyo J. (ba'ajon) (HLLL) (botoru) (81111111111811) (LHHHH) (bakuteria) (kariforunia) Tokyo J. Gloss ‘version’ ‘bottle’ ‘California’ Gloss 6cm ‘marathon’ ‘recording’ (ll-mora, 2-syllable) (3-mora, 3-syilable) (S-mora, 4-syllable) (S-mora, S-syllable) (ti-mora, 6-syllable) (3-mora, 2-syllable) (4—mora, 3-syllable) (S-mora, 3-syllable) (6-mora, 4-syllable) ol 3.4.1.4. Summary As illustrated in the previous subsections, English loanwords in Kansai Japanese fall into three types and six groups, as summarized in (30). The table (30) shows two things. First, the majority of English loanwords, i.e. 67.2%, have the English accent. Second, the lowepitched mora in the word-initial position is less 00mm. (30) English Loanwords in Kansai Japanese a. English TypeH 460(42.2%) 0 b. English TypeL 272 (25.0%) 732 (67'2“) c. Non-English TypeH 150 (13.7%) 215 (19.7%) d. Non-English Type L 65 (6.0%) e. Unaccented Type H 134 (12.3%) 143 (13.1%) f. Unaccented Type L 9 (0.8%) Word-initial H 744 (68.3%) Word-initial L 346 (31.7%) 1090 (100%) Distribution ofaccented morae is summarized, as in(31). In (31),thetoprow shows thelocationoftheaccentedmora,i.e. ‘0’==unaccented, ‘1’=theaccentonthefirst mora..., while the leftmost column indicates the word-initial pitch (‘H’ = high and ‘L’ = low). As shown in bold in (31), among 346 English loanwords with the word-initial low pitch, 252 words, i.e. 72.8%, have the accent on the second mora and all the English loanst with the accent on the second mora have the word-initial low pitch. It has been said in the literature that the word-initial pitch of ncn~loanwords in Kansai Japanese is unpredictable (Pierrehumbert and Beckman 1988: 214). But the data suggests that the word-initial low pitch of English loanwords in Kansai Japanese is highly predictable and 41 ...- , x.s el .s e . e . a v v .. a a . y es . . . . r o % 1 l s . J . . v e ' u l . . .. I «l u - . v . . i . .‘l o a l . 1 1 . e I l. a J. r. a .4 i . I v by. .s . i. . I .s . . l. . a , ts“-‘- . . I. . pl l .. . . . . . I Q s . s as . e s. . . [I w t . . .. r . , 7 . I . 4. e l e i . . l :3. 4 . — r . - . . l. . l l i l . 1 . III. e i . e . Oars 1| I e s . .r I e e r y , e . . ell 1, . . . .... a I ll. 1' '( l\’ the accent on the second mora seems relevant to the word-initial low pitch. I will discuss this more in detail within the OT framework in the following section. (31) Distribution of Acccnted Morae 0 l 2 3 4 5 6 7 8 9 I H 134 455 0 91 48 I4 2 0 0 0 744 1090 L 9 0 252 62 20 3 0 0 0 0 346 143 455 252 153 68 17 2 0 0 0 1090 3.4.2. Analysis In the following subsections, I will account for the locus of the accent and the pitch pattern of English loanwords in Kansai Japanese within the OT framework. 3.4.2.1. The Locus of Accent 3.4.2.1.]. Aeeented Loanwords 947 words out of 1090 English loanwords in the data, i.e. 86.9%, are accented loanwords. As introduced in the previous subsection, accented loanwords fall into two groups: English Type, which has the English accent, i.e., the accent on the syllable stressed in English, and Non-English Type, which does not have the English accent. First, let us consider Non-English Type. 215 words out of 947 accented loanwords, i.e. 22.7%, fall into Non-English Type. 180 words out of 215 loanwords of Non-English Type, i.e. 83.7%, have the accent on the syllable containing the antepenultimate mora. Thissfionglnggeflsthatflreofiginalaccentmovesmmthesynabkmnminingdre antepenultimate mora when a loanword does not preserve the English accent. 42 .e.4s.. 1 /. ’ N1 ‘ | ’ Is ’ ' " el-u I ’ . ' 1 )1: . J -— ‘ ‘ ,. 1 e r l . ' « a 1 j 11,, ‘ s I ‘. .‘ .‘.' 3 Thenextquestiontobeconsidered is: WhattriggersdreaccentshiMTheaccent shift occurs to keep the accent within the last two feet. Among the loanwords belonging to Non-English Type, only 10 words out of 215, i.e. 4.7%, do not have the accent within the last two feet, although there are 120 words out of215, i.e. 55.8%, that do not have the moracorrespondingtotheEnglishoriginal stresswithinthelasttwofeet. Basedonthis,l claim that the restriction on the locus ofthe accent triggers the accent shift. The fact that most loanwords belonging to English Type (604 words out of 732, i.e. 82.5%) satisfy this restriction supports the claim. Myanalysisisbuedonthefootstructureinhpanese.Ithasbeenagreedinthe literature that foot size in Japanese is birnoraic (Poser 1984 and 1990, Tateishi 1989, ltd 1990, no and Mester 1992, Katayama 1995, Kubozono 1995a, Asano 1999, among others). As cited in (32), however, other questions related to the foot structure in Japanese have not been settled. In this chapter, therefore, I assume the following: (i) the foot distribution is right to left, (ii) the degenerate foot is not allowed, and (iii) following Poser (1990), the foot structure is independent of the syllable structure. That is, I assume that feet are assigned by the constraints FT-BIN (Prince 1980), PARSEoSYL (Hayes 1980), and ALL-FT-RIGHT (McCarthy and Prince 1993a), which are defined as in (33) ~ (35) and arerankedasin(36).Therankingin(36)determinesthefootstructlneinJapanese,as illustrated in (37). In this chapter, I will not include those constraints in tableaux for simplification. 43 . 1 . e r .v. I . p . alt 1. . e . s l \ r r ‘ I' . 1 .4 ll . . . s. . . s . u -‘ I i J ,I o o s .s 11 s .0 ...e ..r. . .s It s! I t 17‘ ob -' ‘f ' l e I . . . ll (32) Kubozono (1999: 57-58) The formation of “foot”, for example, raises many interesting questions: e.g. whether it proceeds from lefi to right or from right to left, whether (or when) it permits a monomoraic (i.e. degenerate) foot, whether an lmfooted syllable may be allowed, and whether it is entirely independent of syllable structure as assumed by Poser (1990). None of these questions has been settled in the literature. (3 3) FT-BIN: Feet are bimoraic or disyllabic. (34) PARSE-SYL: Syllables are parsed by feet (35) ALL-FT-RJGHT: Every foot is at the right edge of the prosodic word. (36) Constraint Ranking for Foot Assignment FT-BIN Passe-8Y1. ALL-Fr-Rlolrr (3 7) Fr-Bm >> PARSE-SYL >> ALLvFT-RIGHT o,o,o,o,o, [ Fr-BIN PARSE-SYL ALL-FT-R a. (“mouOuXOuoldl *1 $0,410.13 1). cyanogen“) l "'1’ c. Go,(o,,o,,)(aua,) I s as Within the OT framework, the accent shifi can be explained by six constraints: NONme (Prince and Smolensky 1993: 52), mom-most (mice and Smolensky 1993: 39), FAITI-ILOC(ACCENT) (Smith 1998), ALIGN—R(PENULT F, ACCENT), a l ‘c .. 3! I o I; o... I a 4 s I OI - e I I v i l 4.. . e l' . 1.. e .. ACCENNPROMINENT 11). and Lamosr (Prince and Smolensky 1993: 39). These constraints are defined as in (38) -- (43). (38) NONFINALITY: No accent falls on the word-final foot. (39) RIGHTMOST: The accented foot is rightmost of the word. (40) FAl'l‘I-IIDC(ACCENT): Output accent is faithful to its location in the input. (41) ALIGN-R(PENULT F, ACCENT): The last mora of the penultimate foot is accented. (42) ACCEN’KPROMINENT p): Accent is on the most prominent nucleus of the syllable, i.e. on the first part of a long vowel. (43) LBI-‘l'MOST: The accented foot is leftmost of the word. NONFINALITY prohibits the accent on the word-final foot, while RIGHTMOST requires the accented foot be rightmost of the word. As mentioned above, the accent that is not within the last two feet is avoided. The interaction of these two constraints accounts for this restriction on the locus ofthe accent, as illustrated in (44). In (44), the candidates that have the accent within the last two feet (e. g. (44a) and (44b)) are optimal, although they violate one of the two constraints once. The candidates (44c) and (44d), which have the accent in the antepenultimate foot or preantepenultimate foot, on the other hand, are ruled out, because they violate RIGHTMOST more than twice. 45 T 11 I l . .. l. s . . .J . e -1 J K .v 1‘ l - . J. . it s Os. '1 C. :0. Or a . e ‘1 l e l ,4 s. .l. e I. s 1.! n. a 0 cl . f. . .s v II. c . O I . l . . (44) “WW NONFINALITYE RIGHTMOST 11 0(1111X1111X1111X11'11) "' b-°'(1111)(1111)(11'11)(1111) t 9- (1111X11'11X1111X1111) “! 11 (11'11X1111X1111X1111) “l‘ FArrHLOC(ACCEN'r) (Smith 1998) says that the locus of the accent in the output is faithfilltothelocusoftheaccentinthe inputTthnglishaccentismostcommonin English loanwords in Kansai Japanese. They preserve the English accent when it is within the last two feet. This suggests that the constraints for the restriction on the locus of the accent, i.e. NONFINALITY and Riormrosr, are ranked higher than FAITl-IIDC(ACCENT). ALIGN-RCPENULT F, ACCENT) is a member of the constraint family ALIGN (McCarthy and Prince 1993a). This constraint requires that the last mora of the penultimate foot, i.e. the antepenultimate mora, be accented. The English accent shifts to the antepenultimate accent when English loanwords do not preserve the English accent. This suggests that FAITHIDC(ACCENT) is ranked higher than ALIGN-RO’ENULT F, ACCENT). ACCEmmromNENT 11) militatcs against the accent on the second half of the long vowel or on the coda. ACCENT(PROMINENT u) is the highest-ranked constraint because no loanword violates this constraint. Finally, LEFTMOST requires that the accented foot be leftmost of the word. When the antepenultimate mora is assigned to the second halfofa long vowel or a coda consonant, the accent on it moves leftward but not rightward. This accent shift is explained by LEFT'MOST, as illustrated in (45) where only the relevant constraints are shown. In (45), the candidate (45a), which has the accent on the antepenultimate mora, cannot be the optimal because of the violation of the highest- 46 'l r ranked constraint ACCENKPROMMENT n). The candidates (45b) and (45c) satisfy ACCEN'ImtOMntENT a). but the candidate (45b) loses to the candidate (45c), since the candidate (45b) violates LEFTMOST more than the candidate (45c) does. (45) ACCENKPROMINENT it) >> ALIGN-R(PENULT F, ACCENT) >> Lamosr comment ACCENT ALlGN-R L-MOST a. ko(n.pu.)(rek'.)(ku.su.) 'l ‘ b. ko(n.pu.)(rek.)(ku'.su.) "' “l c. ¢ko(n.pu.)(re'k.)(ku.su.) "‘ " Insum,fliesixconsh'aintsinh'oducedabovearetankedasin(46).Therankingin (46) produces the loanwords of Non-English Type, as exemplified in (47) and (48). The optimal candidates in (47) and (48) do not preserve the English accent because their English accents are not within the last two feet. (46) Constraint Ranking ACCENKPROMINENT a) NONFINALI’I‘Y RIGHTMOST FAITHLOC(ACCENT) ALIGILNPENULT F, ACCENT) LErmosr 47 (47) Non-English Type H journalism JACCENT NONFIN R-MOST Fxmmoc ALIGN-R L-MOST a. (jaa'.)(na.ri.)(zu.mu.)[ ‘! “ " ‘ b. (ia'a)(na.ri.)(zu.mu.l r ”l " c (jaa.)(na.ri.)(zu'.mu.)l ‘ "' ‘l ” d (jaa.)(na'.ri.)(zu.mu.1 : " * ‘l " e. <3"(.iaa.)(na.ri'.)(zu.mu.)1 " ‘ * (4s) Non-English Type L specialist ACCENT NONFtNgR-Mosr FAITHLOC ALIoN-R L-MOST a. (mpe'.)(sha.ri.)(su.to.) ”l s b. (supe.)(sha.ri.)(su'.to.) s . n n c. (supe.)(sha'.ri.)(su.to.) s s *2 s d. a”(supe.)(sha.ri'.)(su.to.) 7 s s a Next, consider English Type. 732 words out of 947 accented loanwords, i.e. 77.3%, belong to English Type. The constraint ranking in (46), which accounts for Non-English Type, explains English Type as well, as illustrated in (49) and (50). in (49) and (50), the optimal candidates preserve the English accent because they have the English accent within the last two feet. (49) English Type H Personality ACCENT NONF1N§ R-MOST FAITHLOC AuGN-R L-MOST 8. pa(a.so'.)(na.ri.)(tii.) * u! . . . b. pa(aso)(nari)(tii’) *! e e a .... c. pa(aso)(nariv)(tii) ‘ e s! u d. 5" pa(aso)(na’ri)(tii) L e . cc (50) English Type L snack ACCENT NONFlNg'R-MOST FAITHLOC ALIGN-R L-MOST 8. (811.1“le .101.) a! a E e a e b. (maximum) a a: s c. G"(su.na’)(k.lm) "‘ 3.4.2.1.2. Unaccented Loanwords 143 words out of 1090 English loanwords in the data, i.e. 13.1%, are unaccented loanwords. Unaccented loanwords are subcategorized: Unaccented Type H and Unaccented Type L. Kubozono and Ohta (1998: 39-42) analyze the accentuation of place name loanwords, and show that the majority of unaccented place name loanwords in Tokyo Japanese are quadrimoraic and their last two syllables are light. However, an examination of English loanwords in Kansai Japanese reveals that these generalizations do not explain the unaccented loanwords in Kansai Japanese. First, among 1090 English loanwords in Kansai Japanese, quadrimoraic words form the largest group, i.e. 445 words (40.8%). Among the quadrimoraic words, however, only 88 words, i.e. 19.8%, are unaccented. That is, the majority of quadrimoraic loanwords are not unaccented. Second, among 342 loanwords in Kansai Japanese whose last two syllables are light, only 58 words, i.e. 17.0%, are unaccented. That is, the majority of the loanwords whose last two syllables are light are not unaccented. These facts suggest that the generalizations by Kubozono and Ohta do not explain why the 145 words in Kansai Japanese do not have accents. How can unaccented loanwords be explained? The majority of non-loanwords in Tokyo Japanese are unaccented (Hayashi 1982) and unaccentedness is the default 49 I‘i fl \ hlt situation in Tokyo Japmese (Altinaga 1960, Katayama 1995). These observations seem to apply to Kansai Japanese as well, and the number of unaccented non-loanwords in Kansai Japanese seems to be increasing. Based on this, following Ono (1991), I assume that English loanwords become unaccented as a result of assimilation to the core part of the Japanese lexicon with respect to accentuation. Deaccentuation can be explained by the constraint *HL (Ito and Mester 2003b: 58) in (51). ’HL is ranked highest in the constraint ranking for the assimilated loanwords as shown in (52), while it is ranked lowest in the ranking for the accented loanwords as illustrated in (53). The ranking in (52) accomts for tmaccented loanwords, as exemplified in (54) and (55). In (54) and (55), the optimal candidates are tmaccented because of the highest-ranked constraint ‘HL. (51) ‘HL: No accent. (52) Constraint Ranking for Unaccented Loanwords ACCENTfPROMTNBNT a) *HL l:>—<:l NONFINALITY RIGHTMOST FAITHIDC(ACCENT) ALIGltII-RQENULT F, ACCENT) LsnT'tosr 50 (53) Constraint Ranking for Accented Loanwords ACCENT(PROM1NENT a) NONFINALITY Rlonmosr FAm~iLOC(ACCENT) ALIGN-Rfl’ENULT F, ACCENT) [Emosr ‘HL (54) Unaccented Type H California ‘HL ACCENTlNONFmgk-MOSTIFAITHLOC ALTON-R L- a. (ka.ri.)(fo'.m.)(ni.a.) I'! 0 b. (ka.ri.)(fo.ru'.)(ni.a.) ‘1 . c. “'(kflJ'i.Xf0.nl.)(ni.a.) fi. (55) Unaccented Type L marathon ‘HL ACCENTlNONFINER-Mosr FAITHLOC ALIGN-R a. (ma'raXson) ‘! . b. (mara'.)(son.) c. 0(ma.ra.)(son.) 3.4.2.2. Pitch Pattern I have introduced the pitch pattern of non-loanwords in Kansai Japanese in section 3.2.2. Non-loanwords and English loanwords in Kansai Japanese are basically the same withrespecttothepitchpattern,andtheysharethesamecharacteristiesin(56),whichis the summary of section 3.2.2. The significant difference between them is the word-initial pitch. With regard to non-loanwords, Pierrehumbert and Beckman (1988: 214) claim that 51 I ‘1 .ll'l 1% . , _ s I . I . n J . . t ’ e s v . ’l ~ .01 . ’ \ l. I! i u. . . . . I it a I O I I s . t . . . I e .s l - I - A \ l . I .l . . . I. . .. . . . .7. u I . . r u (A . _ ' , 9’ s n .7 Or . . l .1 a I u 0 . a n r v s l: . s A l 4. _ w - . i u I It . l. .. \ I I t l . the word-initial pitch is unpredictable. In English loanwords, on the other hand, it is highly predictable, as shown in (31) (repeated in (57)). The table in (57) suggests that the accent on the second mora triggers the word-initial low pitch. (56) a. The accent is marked by a falling pitch. b. The accented mora is high-pitched c. Aprosodicwordcannothavemorethanoneaccentedmora. d. A prosodic word must have at least one high-pitched mora. e. The morae following the accented mora are low—pitched throughout. f. The morae pmeding the accented mora can be either high— or low-pitched throughout. g. When a loanword has a low-pitched mora word-initially, only the accented mora ofthe accentedwordorthe lastmoraofthe maceentedwordcanbehigh-pitched. h. When the unaccented word with the word-initial low pitch is followed by a particle such as the nominative marker go, the location of the high pitch moves to the word-final, i.e. onto the nominative marker. (57) Distuhution of Accented Morse (= (31)) O l 2 3 4 5 6 7 8 9 I H 134 455 0 91 48 14 2 0 0 0 744 1090 L 9 0 252 62 20 3 0 0 0 0 346 143 455 252 153 68 V 17 2 0 0 0 1090 The pitch pattern of English loanwords in Kansai Japanese is determined by the following six markedness constraints. 52 (58) HEAD=H: Head mora should be high-pitched. (59) ‘NONHD/H: No High pitch on non-head mora. (60) ‘[HH: No word-initial HR. (61) ‘[LL: No word-initial LL. (62) '[LH: No wordoinitial LH. (63) LH': The mora immediately precedes the accented mora is low-pitched. HEAD==H (Yip 2002: 85) in (58) requires that the accented mora be high-pitched and ‘NONHD/H (Yip 2002: 98) in (59) militates against high pitch on non-accented mora. HEAD==H is ranked highest since no loanword violates this constraint, while ‘NONHD/l-I is ranked lowest. The constraints ‘[HH in (60) and ‘[LL in (61) prohlhit the word-initial HH and the word-initial LL, respectively. These constraints are members of the constraint family OBLIGATORY CONTOUR PRINCIPLE in the sense of McCarthy (1986), i.e. no adjacent identical elements except across morpheme boundaries. ‘[LH in (62) militates against the word-initial LH. This is a positional markedness version of ’LH. Versions of the constraint ‘LH play significant roles in Japanese. For example, the constraint ‘LH’, the self-conjoined version of I"L11, is a highest-ranked in Japanese, since a prosodic word cannot have more than one rising pitch. The sequence of low-high in the word-initial position is not allowed in Kansai Japanese unless the high is on the accented mora. As mentioned above, the word-initial low pitch is less common than the word-initial high pitch. This indicates that ‘[LL and ‘[LH are ranked higher than ‘[HH. The ranking between I"[LL and I'[LH is *[LL >> *[LI-I, because most of the loanwords with the word- initial low pitch have the accent on the second mora and the accented mora is high- 53 pitched. The constraint LH' in (63) requires that the mora immediately preceding the accented mora be low-pitched. In Kansai Japanese, the rising of the pitch is possible only in the position immediately preceding the accented mora. This constraint makes the accented mora the only high-pitched mora of the word, i.e. LH..L.. vs. ...I-IH'L.... That is, this constraint makes the accent of a prosodic word more prominent. The constraint LH' is ranked higher than I’[LIL The constraints in (60) - (63) determine the pitchpatternofthemoraeprecedingtheaccentedmora. Inmthesixconstmintsarerankedasin(64). Thisrankingdeterminesthepitch pattern of English loanwords in Kansai Japanese, as exemplified in (65) - (68). The optimal candidates in (65), (66), and (67) have the word-initial high pitch because of the high-ranked constraints HEAD=H and ‘[LL, whereas the optimal candidate in (68) have the word-initial low pitch due to the constraint LH'. (64) Constraint Ranking HEAD=H >> ‘[LL >> LH’ >> *[LH >> ‘[HH >> *NONHD/H (65) Unaccented Loanword bottle HEAD=H ‘(LL LH' ’[LH *[HH ’NONHD/H a. botoru LLL b.<' botoru HHH ‘! O 0‘. ‘4 K ’- (66) AccentedLoanword(theaccentonthefirstmora) canoe HEAD=H ‘[LL LH' ‘ILH ‘IHH ‘NONHD/H a. ka’nuu HHH *1 0. bx? ka'nuu HLL C. ka'nuu LLL ’! (67) Accented Loanword (the accent on the second mora) amenity HEAD==H *[LL LH' CILH *[HH ‘NONHD/H a. ame'nitii HHLLL *! b5" ame'nitii LHLLL (68) Accented Loanword (the accent on the third mora) 001133118113 HEAD=H ‘[LL LH' ‘U—H ‘IHH ‘NONHD/H konse'masu LHHLLL n ‘1 konse'nsasu LLHLLL *! ens" konse'nsasu HHHLLL O. 3.4.2.3. Summary In this section, I have discussed the locus of the accent and the pitch pattern of English loanwords in Kansai Japanese and showed that the rankings in (69) - (71) $5 account for them. The rankings in (69) and (71) and the rankings in (70) and (71) determine the accentuation and the pitch pattern of English accented loanwords in Kansai Japanese and the accentuation and the pitch pattern of English unaccented loanwords in Kansai Japanese, respectively. (69) Constraint Ranking for Accented Loanwords ACCEN‘KPROMINENT a) NONFINALITY RIGHTMOST FmLochcsm) Atari-Rosana F, ACCENT) LBFTMOST .1... (70) Constraint Ranking for Unaccented Loanwords ACCENT(PROMTNENT ll) *HL NONFINALITY Rroirmosr FurnLochcrsN-r) ALIGhIl-RWENULT F, ACCENT) LEFTMOST (71) Constraint Ranking for the Pitch Pattern HamH >> *[LL >> LH' >> *[LH >> *[HH >> ‘NONHD/H 56 3.5. Tokyo Japanese This section discusses the locus of the accent and the pitch pattern of English loanwords in Tokyo Japanese within the OT framework. 3.5.1. Data 1090 unabbreviated English loanwords of three to eight morae (two to eight syllables) were collected mainly from a loanword dictionary (Horiuchi 1996). Adjectives andverbsarenotincludedinthedatabecausemostoftheloanwordsinJapaneseare nouns and verbs are made from nouns by adding a verb sum ‘to do’ (e.g. tesuto ‘test’ + suru‘todo’ -ttesutosuru‘totest’).'l‘helocusoftheaccentandthepitchpattemofthe loanwords in Tokyo Japanese are based on two accent dictionaries of Tokyo Japanese (Nihon How Kyokai 1998 and Kindaichi and Akinaga 2001). In the Japanese accent dictionaries, there are words that have more than one possible location of the accent, as exemplified in (72). In such cases, the present study deals only with the most common accent (e.g. (72a)), which is indicated in the dictionaries. The difl‘erence in the loci of the accentwillbethesubjectoffiitm'eresearch. (72) ‘awordion’ (Kindaichi and Akinaga 2001: 10) a. ako'odion b. akoodi'on As in Kansai Japanese, English loanwords in Tokyo Japanese can be divided into three types: English Type, Non-English Type, and Unaccented Type. First, the accented 57 _ . l I; l V a I .l . r . I- r l s s r ‘ r . to 3. J . I! c u . .\ I. . . I. J at. a o; a s as f. v) a v. . u . _ . val loanwords that have the English accent, i.e. the accent on the syllable stressed in English, are categorized into English Type. This type consists of 749 words of three to nine morae (two to six syllables). (73) English Type Tokyo J. Kansai J. Gloss a. kuru'u (kuru'u) ‘crew’ (3-mora, 2-syllable) LHL (LHL) b. suna'kku (suna’kku) ‘snack’ (4-mora, 3-syllable) LHLL (LHLL) c. pi'kunikku (pi 'kunikku) ‘picnic’ (S-mora, 3-syllable) HLLLL (HLLLL) d. wisuko’nsin (wisuko'nsin) ‘Wisconsin’ (6—mora, 4-syllab1e) LHHLLL (LLHLLL) e. imajine'esyon (imajine'esyon) ‘imagination’ (7-mora, S-syllable) LHHHLLL (LLLHLLL) f. finasutore'esyon (fill'asutore'esyon) ‘frustration’ (8-mora, 6—syllab1e) LHHHHLLL (HHHHHLLL) g. entaate'imento (entaate'imento) ‘entertainment’ (9-mora, 6—syllable) LHHHHLLLL (HHHHHLLLL) Second, the accented loanwords that do not have the English accent are categorized into Non-English Type. This group consists of 216 words of three to eight morae (two to eight syllables). As in Kansai Japanese, most (79.6%, i.e. 172 words out of 216 loanwords of this type) of the loanwords belonging to this type have the accent on the 58 ,. "{,.v ‘5 ‘ , .55 3 '.A‘. J , .. < l t‘. . . 1 1 . w . n ' s O s ' ' o - \ l . a’ a ' . . . .-_ .r . p v D . ‘l t ..l ‘ f I i _. , 7‘ I, H ‘ 'I ‘ ’— 'iQii " fi‘t'lt"’.|:l syllable containing the antepenultimate mora, i.e. on the antepenultimate mora when the syllable is light or on the preantepenultimate when the syllable is heavy. This fact strongly suggests that the accent moves onto the syllable containing the antepenultimate more when a loanword does not preserve the English accent. (74) Non-English Type Tokyo J. Kansai J. Gloss a. ka'nuu (ka'nuu) ‘canoe’ HLL (HLL) b. ime'eji (ime'eji) ‘image’ LHLL (LHLL) c. edi'nbara (edi'nbara) ‘Edinburgh’ LHLLL (LHLLL) d. jaanari'zumu (iaanari'zumu) ‘jom'nalism’ LHHHLL (HHHHLL) e. koresutero'oru (koremtelo'oru) ‘cholesterol’ LHHHHLL (l-IHHHHLL) f. indianapo'risu (indianapo'risu) ‘Indianapolis’ LHHHHHLL (HHHI-IHHLL) g. erekutoroni'kusu (erekutoroni'kusu) ‘electronics’ Ll-IHHHHLL (HHHHHHLL) (3-mora, 2-syllable) (4-mora, 3-syllable) (S-mora, 4-syllable) (6-mora, 5-syllable) (7-mora, 6-syl1ab1e) (8-mora, 7-syllab1e) (B-mora, Myllable) Finally, all the unaccented loanwords in Tokyo Japanese fall into Unaccented Type. 59 Unaccented Type consists of 125 words of three to six morae (two to six syllables). (75) Unaccented Type Tokyo J. Kansai J. a. karee (karee) LHH (LLH) b. marason (marason) LHHH (LLLH) c. supuringu (supuringu) LHHHH (HHHHH) d. bakuteria (bakuteria) LHHHH (HHHHH) e. kariforlmia (karifonlnia) LHHHHH (HHHHI-IH) Gloss ‘bacteria’ ‘California’ (4nmora, 2-sy11ab1e) (3-mora, 3-syllable) (S-mora, 4-syllable) (S-mora, S-syllable) (6—mora, 6-syllable) In summary, English loanwords in Tokyo Japanese are divided into three types as shown in (76). In Tokyo Japanese, unlike Kansai Japanese, the word-initial pitch is totally predictable and each of these types needs not be divided into subcategories based on the wold-initial pitch. (76) English Loanwords in Tokyo Japanese a. English Type 749 (68.7%) b. Non-English Type 216 (19.8%) c. Unaccented Type 125 (11.5%) ‘ 1090(100.0°/.) ‘ ' - J as-ll a. . V .. ; l ‘ l ; - " l t l' '0 -3 --~v~v‘ 3.5.2. Analysis Inthis section,Iwill account forthe locus ofthe accentandthepitchpatternof ‘ English loanwords in Tokyo Japanese within the OT fi'amework. 3.5.2.1. The Locus of Accent 965 words out of 1090 English loanwords in the data, i.e. 88.5%, are accented. Accented loanwords fall into two groups: English Type and Non-English Type. With respecttothelocusoftlleaccenLTokyoJapaneseandKansaiJapanesesharethethree features: (i) English accent as the most common accent among accented loanwords, (ii) the antepenultimate accent as the most common accent among the loanwords of Non- English Type, and (iii) dispreference for the accent that is not within the last two feet. As a result, 891 words out of 965 accented loanwords in Tokyo Japanese, i.e. 92.3%, are identicalwiththeircorrespondingwordsinKansaiJapanesewithregardtothelocusof the accent." This means that the accentuation of accented loanwords in Tokyo Japanese can be explained by the constraint ranking for the accentuation ofaccented loanwords in Kansai Japanese in (69) (repeated here in (77)). The constraint ranking in (77) produces the correct outputs of English Type and Non-English Type, as exemplified in (78) and (79).” In (78), the optimal candidate has the English accent since it is in the penultimate foohwhereastheoptimalcandidatein(79)doesnothaveitbecauseitisinthe antepenultimate foot. ’ SeetheAppaflixfmwmdswhosewcemndmmTokyoJapamwhdiffuentfiommmmw Japanese. "l‘heconstraint‘HLisnotshowninflletablemmbecmnehplaysfifllemkindetumimngdBopfimal output. 61 C . . i - " ' ,. t . . v . I 6 . i , ‘ ' 1' ' - a . . l " I V. t t ." . ‘1’ .' " - . C t ‘ ,; 'I .' . ' .| I ' , 1‘ ‘ ' t . I - . -‘ )' ' . t . ‘ a . 4 ‘ " ‘ ' t, V I 1 l l. . l A v I. , ‘ . t ' . J. 1 . . . ._ l’.‘ . . .- : . . .‘ I‘ I‘." . 'I - ‘. . . I . 1.0" i l”"’ I .,.. ~31. ‘ ,'1 if 7|, J.) I . h .’ . ' 1!: " 'F’JI'y t l l .’ ill} J)‘ I J , . 'tt h : I I'l‘.0 “)4 l l 5J9 ...‘ (77) Constraint Ranking for Accented Loanwords ACCEN‘KPROWNENT p) NONFINALITY FAITHILOC(ACCENT) ALIGN-R(PENULT F, ACCENT) LEFTMOST ‘HL (78) English Type RIGHTMOST imagmatr’ 'on ACCENT NONFlN § R—MOST hmlLoc ALloN-R L-MOST a. i(maii)(nee 'Xsyon) ’! . b. i(ma’ji)(nee)(syon) O C. i(maji)(nee)(syo'n) ’l tit d. 9' i(maji)(ne'e)(syon) I. (79) Non-English Type journalism ACCENT NONFlN § R—MOST lFAmlLoc ALlGN-R L-MOST (jaa 'Xnarinumu) ‘I fit P"? (ia'aXnarinumu) Oi! (iaaXliarinu'mu) ‘! 0. d. (iaaXna'rinumu) I’l e. Cl"’(iaaXnari 'qumu) There is one more issue I need to address: the treatment of devoiced vowels. In Tokyo Japanese, high vowels are devoiced when preceded and followed by voiceless obstruents or when they are in the word-final position and preceded by a voiceless 62 obstrumt’Asano (1999:74) claims that the accent shift occurs to avoid the accent on a devoiced vowel (e.g. o'fisharu ‘oflicial’ (‘ofi'sharu)). But, the accent shifi Asano claims isnotohmetlinthetletewlniact, manystudiesshowthattheaccentshifitoavoidthe accent on the devoiced vowel has become rare (Vance 1987: 50, Kindaichi and Akinaga 2001: Appendix (8), among others). For examples, the words in (80), which are from Kindaichi and Akinaga (2001), have the accent on the devoiced vowel. Based on this, I claim that devoiced vowels in Tokyo Japanese do not trigger the accent shifi and, as discussedinthischapter,accentshifioccmsonlywhentheaccentisnotwithinthelast two feet. (80) Counterexammes to Asano (1999) (Devoiced vowels are italicized.) a. ti 'kin HLL ‘chickm’ b. ki'ttin HLLL ‘kitchen’ c. pi'kunikku HLLLL ‘picnic’ Finally, let us consider unaccented loanwords. 125 words out of 1090 English loanwords in the corpus, i.e. 11.5%, belong to Unaccented Type. Regarding the unaccented loanwords in Tokyo Japanese, Kubozono and Ohta (1998: 39-42) make two generalizations: (i) the majority of unaccented loanwords in Tokyo Japanese are quadrimoraic and (ii) their last two syllables are light. As in the case of Kansai Japmese, ’ Some ample. the shown below. In the exantplet, devoiced vowels ate shown in italic. fitkll (maecented) clothes ka’ki (accented) oyster ”Thewottieficltm ‘official’isnotinchldedinthedatabecauseitisanadjective. 63 however, these generalizations do not explain why the 125 words in Tokyo Japanese do not have the accent. First, as Kubozono and Ohta observe, the majority (77 words out of 125, i.e. 61.6%) of unaccented loanwords in Tokyo Japanese are quadrimoraic. Among 445 quadrimoraic loanwords in the data, however, only 77 words, i.e. 17.3%, are unaccented. That is, the majority of quadrimoraic loanwords are accented. Second, 342 loanwords in the data have the sequence of two light syllables word-finally. Among them, however, only 57 words, i.e. 16.7%, are unaccented. In other words, the majority of the loanwords whose last two syllables are light are accented. These facts suggest that the generalizations by Kubozono and Ohta do not explain unaccented loanwords in Tokyo Japanese. The majority of non-loanwords in Tokyo Japanese are unaccented (I-Iayashi 1982) and unaccentedness is the default situation in Tokyo Japanese (Akinaga 1960, Katayama 1995). Based on these, following Ono (1991), I assume that English loanwords become unaccented as a result of assimilation to the core part of the Japanese lexicon with respect to accentuation. The assimilated loanwords do not have the accent due to the highest- ranked constraint ‘HL. That is, the constraint ranking in (70) (repeated in (81)), which is developed for unaccented loanwords in Kansai Japanese, accounts for unaccented lomwonis in Tokyo Japanese as well, as exemplified in (82). In (82), the optimal candidate is unaccented due to the highest ranked constraint ’HL. ‘13:. ’l b ‘l d C .r‘) t ‘ ‘ re ‘- la :1 . it (81) Constraint Ranking for Unaccented Loanwords ACCENKPROMINENT p) ‘HL NONFINALIT'Y FAnrilIcc(ACCENr) RIGHTMOST ALIGN-R(PENULT F, ACCENT) Lamosr (82) Unaccented Type bacteria *HL EACCENT NONFIN R—MOST lFAmiLoc Anon-1h a. ba'(kute)(ria) *! .0 * b. ba(kute')(ria) ‘! c. Mutcxfi'a) '! d. 6' ba(kute)(ria) 3.5.2.2. Pitch Pattern With regard to the pitch pattern, there is no difference between Non-loanwords and English loanwords in Tokyo Japanese. They share the same characteristics in (83), which isthesummaryofsection 3.2.1. (83) a. The accent is marked by a falling pitch. b. The accented mora is high-pitched. c.Aprosodicwordcannothavemorethanoneaccentedmora. d. The morae following the accented mora are low-pitched throughout. e. The word-initial pitch and the pitch on the second mora must be distinct. 65 The pitch pattern of English lomwords in Tokyo Japanese is determined by the six constraints introduced in the analysis of English loanwords in Kansai Japanese: HEAD-=11, ‘NONHD/H, LH', ‘[LL, '[LH, and ’[HPL The characteristic (83c) suggests that the constraints ‘[LL and ”'[HH are highest-ranked. The constraint HEAD=H is also highest- rankedbecausenoloanwordsinTokyoJapanese violate it. Inmthesixconstraintsare ranked as in (84). The constraint ranking in (84) determines the pitch pattern of English loanwords in Tokyo Japanese, as exemplified in (85) -- (88). In (85), (86), and (87), the optimal candidates have the word-initial low pitch because of me high-ranked constraints "[LL and ‘[HH, whereas the optimal candidate in (88) have the word-initial high pitch duetotheconstraint '[HI‘L There isno irregularitywithregardtothepitchpatternof English loanwords in Tokyo Japanese. The constraint ranking in (84) explains all the loanwords in the data. (84) Constraint Ranking ann=H r L ‘[HH ‘[LH LH' ‘Noan/H (85) Unaccented Loanword curry HEAD=H: ‘(LL ‘[HH ‘[LH LH' E‘Noan/H *! : : LLL ' ' a! E i hit (86) Accented Inanword (the accent on the first mora) HEAD=Hj *[LL ‘[HH ' a! E 3 so canoe *lLH LH' i'N‘dem c. ka'nuu I . HHH i E b5" ka'nuu . T i HLL : (87) Accented Loanword (the accent on the second mora) image HEAD=H *[LL ‘[HH ‘(LH LH' g‘Noan/H E ' § , d. ime'eji : : : *l “' HHLL g g bfi' ime'eji E LHLL (88) Accented Loanword (the accent on the third mora) Wisconsin HEAD-HE *[LL grunt *[LH LH' g‘Noan/I-I e. wisuko'nsin LLHLLL ‘l b. wisuko'nsin : ” M” E § § 5 LHHLLL g g ‘ 5 3.5.2.3. Summary Infllissecfiomlhavediscussedthattherankingsin(89)-(91)accormt forthelocus of the accent and the pitch pattern of English loanwords in Tokyo Japanese. The rankings (89) and (91) and the rankings (90) and (91) account for the accentuation and the pitch 67 ‘1 patternofEnglishaccentedlomwmdsinTokyoJapmesemdtheaccenmafionmdme pitch pattern of English maccentied loanwords in Tokyo Japmese, respectively. (89) Constraint Ranking for Accented Loanwords AWPROMINENT tr) NONFINALIT'Y RIGI-ITMOST FArrHLoqACCEN'r) ALloh'l-RQENULT F, ACCENT) Lamosr I'HL (90) Constraint Ranking for Unaccented Loanwords ACCENT(PROMINENT a) ‘l-IL NONFINALITY RIGHTMOST FAITHLOC(ACCENT) ALlGll-RGENULT F, ACCENT) ml... (91) Constraint Ranking for the Pitch Patter HEAD-=11 ‘ L ' I'[LH LH' I’Noan/H 68 3.6. Conclusion ThischapterdiscussedthelocusoftheaccemandthepitchpattemofEnghsh loanwords in Kansai Japanese and Tokyo Japanese within the framework of OT. The present study is summarized as follows. FhngansaiJapaneseandTokyoJapanesearethesamewithregardtoaccent assignment: (i) the English accent, i.e. the accent on the syllable stressed in English, is the most common, and (ii) the accent moves onto the syllable containing the antepenultimate mora when the English accent is not within the last two feet. Swoniflleinfonnafionondlelocrlsofaccentisincludedintheinpths mentioned above, the English accent is the most common and the non-English accent is duetotherestrictiononthelocusofthe accent, i.e.theaccentmustbewithinthelasttwo feet. ThiswggesmfllatthemfonnafiononfllelocmofEnglishsnessismcludedinthe inputandJapaneseborrowershavetheaccesstoflle information. Third,inOT,thedifl‘erencebetweendialectscanbeexplainedintermsofthe rankingdifi'erence. Tokyo JapaneseandKansdJapanesearedifl‘erentwithrespecttothe pitch pattern. In Kansai Japanese, the pitch pattern of English loanwords can be explained by the constraint ranking in (92), where constraints ‘[LL and LH' play significant roles. In Kansai Japanese, because of these constraints, the word-initial high pitch is more commonthanthelowpitch. InTokyoJapanese,ontheotherhand,thepitchpattemof English loanwords can be explained by the constraint ranking in (93), where the constraints ‘(I-IH and ‘(LL play crucial roles. In Tokyo Japanese, the pitch on the first momandthepitchonthesecondmoraaredistinctduetotheseconsnaints. 69 v“ I ‘5 ,. 0 L. I ..‘4 Q. i" o 1 ‘1 I. d 7. -, ‘ I ‘1. J . P. It'- A l .H’ (92) Constraint Ranking forthe Pitch Pattern in Kansai Japanese HEAD=H >> ‘[LL >> LH' >> *[LH >> ‘[HH >> ‘NONHo/H (93) Constraint Ranking for the Pitch Pattern in Tokyo Japanese HEAD==H ‘ L ‘ ‘fLH LH' ‘Noan/H Finally, the word-initial pitch of English loanwords in Kansai Japmlese is predictable. Because of the two possible word-initial pitches, high and low, the pitch pattern in Kansai Japanese is more complicated and it has been said that the pitch pattern of the whole word in Kansai Japanese is unpredictable. The present study revealed that the word-initial highpitchismorecommonandtheword-initiallowpitchistriggeredbytheaccenton the second mora, which is due tothe constraint LH’. mfilnnensearclhfllefollowdngissuesshmddbefimherstudiethsLfliemking for accented loanwords developed in this chapter accounts for most of English loanwords in both Kansai Japanese and Tokyo Japanese, but there are exceptions to it. Especially, the ranking does not explain the accentuation of short words belonging to Non-English Type consisting offour or fewer morae. In this study, I claimed that the restriction on the locusoftheaccenttriggersflleaccentshifi. Therankingaccounts fortheaccentshifiin the loanwords consisting of five or more morae but cannot explain the accent shifi in the loanwords consisting offour or fewer morae, since all the loanwords consisting offour or fewermoraehavetheEnglishaccentwithinthelasttwo feet.‘lhatis,therankingpredicts the wrong output, as illustrated in (94), and does not account for why they do not preserve the English accent. Among accented loanwords consisting of four or fewer 70 ‘0 ‘- 2 i 'l. l fl.“ 5 1. l 't .r' \ 1 LH' if: morae (504 words in Kansai Japanese and 530 in Tokyo Japanese), the analysis developed in this chapter does not explain 58 loanwords, i.e. 11.5%, in Kansai Japanese and 60 loanwords, i.e. 11.3%, in Tokyo Japanese. I will leave these exceptions for future research. (94) sapphire -+ safa'ia (Wrong Prediction) sapphire ACCENT NONF1N§ R-MOST lFAlTliLoc AuGN-R L-MOST a. ®(sa'fa)(ia) " " b. (straws) . ‘1 c. (safaXi'a) t a! t. . d. (safa)(ia'), A e i“ ‘l . . Second, the making in (92) explains the pitch pattern of most English loanwords in Kansai Japanese. But the word-initial low pitch has not been completely explained yet. The present study claims that the word-initial low pitch is triggered by the accent on the second mora. But, there are words that have the word-initial low pitch without having the meat on the second mora. Among 94 counterexamples, 54 words, i.e. 57.4%, have the await on the heavy syllable. The weight of the syllable might also be relevant to the Word-initial low pitch. This question needs to be further pursued. Thirdthischapterfowsedonlyonthemostcommonaccentoftheword. InTokyo Japanese, there are words that have more than one possible accent, as illustrated in (95), and thedifl‘erencebetweenthepossible accentswasnotdiscussed. Thisshouldbestudied in the future. 71 ‘ I d 4 ' I e 1’ I”! t I ,y‘ 'l ‘\ l is. J. ’1 (95) ‘accordion’ (Kindaichi and Akinaga 2001: 10) a. ako'odion b. akoodi’on Finally, this chapter dealt with unabbreviated English loanwords and ignored abbreviated ones. Labrune (2002) finds that apocope and aphaeresis generally occur just beforetheaccentedrnoraofabase.Thissuggeststhattheaccenmationofabbreviated words is different from that ofunabbreviated ones. The difi‘erence between abbreviated and lmabbreviated words with regard to the awentuation should be further researched. 4. The Realization of English lrl' 4.1. Introduction This chapter explores the realization of English /r/ in Japanese within the framework of Optimality Theory (OT: Prince and Smolensky 1993). English /r/ is a unique segment in English loanwords in Japanese in the sense that it is the only consonant that can conespond to zero, a consonant [r] or a vowel, i.e. [a] or [0], depending on its location. This chapter examines the realization ofEnglish /r/ in Japanese with three goals in mind. First, this chapter discusses the nature of inputs. With regard to the nature of inputs, it is not clear whether the input is based on the phonetic representation ofthe source language or the phonological one. In this study, adopting Silverman’s (1992) idea that there is an intermediate level (the Perceptual Level) between the input, i.e. the output of the source language, and the Operative Level (see Chapter 2 and section 4.3.1 below), I will claim that the input to the Operative Level is the perceived segment, i.e. the output of the Perceptual Level Thesecondgoal ofthisehapteristoarguethataninputsegmentcanbeperceived difi‘erently at the Perceptual Level based on its location, i.e. the English onset /r/, and word-medial coda /r/, and word-final coda /r/ are perceived as [r], [a] or [o], and [a], respectively. The onset /r/ and the coda M are treated differently in Japanese. This fact suggeststhatthemputsmphonologicalprocessesarebasedonthepemepfion,whichis closely related to the phonetic representation of the source language but not phonological 'AnearlierversionofpartsofthischapterappearsinUniversityofWashingtonWorkingPapersin Linguistics23. 73 u . I i x I a w n o a C 4" . v u .I (I t 'l . ’ a .. u I . I L a a i 2' e‘ d ' Iv v c t. . I. 4 . c s I is .1. .l 0 .0. n.. p. t . O . l v. I n w‘. ..I.I . Kl a. .l .t h n v I It It: . o a II n "v a} In, a , t c. i re a ‘ a . h l . , ‘4 J a. . . u ’ l a . J - ..— . . ' .J .I a . a ll . I . . l) a. . . w i . I. e n v . I, er . l is. .o .l is t . . I f t I r .- o I I n . . u 4 .. rt . 1 . . . 4 .. c . e o I is. .v . . l I a. Finally, this study explores the process of assimilation of loanwords to the core part of the Japmese lexicon. no and Mester (1995) claim that recent loanwords violate more constraints and are less assimilated. But this is not the case for the recent English loanwords with the word-final coda /r/. In recent loanwords, the English word-final coda /r/ in the input is deleted in Japanese. The deletion occurs to avoid the word-final long vowel [az], which is disfavored in the core strata of the Japanese lexicon. The fact that the English word-final coda M is deleted in recent loanwords indicates that they are more assimilatedtothecorepartofthelapaneselexiconwithrespecttotheu'eatmentofthe word-final [a:]. Thischapterisorganizedasfollows. Section2ptesentsthedataontherealizationof English /r/ in Japanese. Section 3 consists of two parts. The first part discusses the inputs to phonological processes. Then, in the second part, I accounts for the realization of English lr/ in Japanese within the OT fiamework. Finally, this study concludes in section 4. 4.2. Data The data in this chapter was collected from five Japanese dictionaries: Shimmura (1955, 1969, 1983, and 1991) and Akahori (1999). The examples below are represented phonemically and the location of accent is not shown unless it is relevant. English /r/ is a unique segment in Japanese loanword phonology in the sense that it is the only consonant that has more than one possible corresponding segment, depending on the location it occm's. English M has five possible corresponding patterns in Japanese, as illustrated in (1). Among them, only the corresponding pattern (la) applies to the English onset/r/ andthe others applyto the coda/rl. 74 a- *t (1) Five Possible Corresponding Patterns a. /r/ —-» [r]' b. /r/ -’ [o] (the second half of a long vowel [a:]) c. /r/ --> [at] (including the second half of a long vowel [a:]) d. /r/ -+ 0 (deletion) e. /r/ -> [a] or 0 (free variation) 4.2.1. Onset lr/ As introduced above, the English onset /r/ has only one possible correspondent in Japanese, namely [r]. Examples are given in (2). The corresponding pattern in (2) is simple. English /r/ corresponds to the closest sound in the Japanese phonemic inventory, namely [r]. This is the only corresponding pattern for the English onset /r/. The English onset/r/alwayscorrespondstotheonset[r]inJapanese. (2) Ir/ -’ [r] a. bureeki ‘break’ b. ribon ‘ribbon’ c. torakku ‘tmck’ d. kontorakuto ‘contract’ e. risaikuru ‘recyele’ ’InthisclnpterJnsteadofthe symbol [r] farthetap,[r] isusedforconvenience. 75 J ",’1‘t. -. _ 3 V I I .' i'v.’ "I.I‘I.l I" ll) 4.2.2. Coda Irl As illustrated in (1), the English coda /r/ never corresponds to a consonant but corresponds to vowels, or it is deleted. The English coda /r/ has three possible corresponding segments including the null segment, and four possible corresponding patterns in Japanese.‘ First, as shown in (lb), i.e. /r/ -+ [o] (the second half of a long vowel [o:]), the English coda /r/ can correspond to the second half of a long vowel [0:]. In the examples of this corresponding pattern given in (3), the nucleus preceding coda lr/ in the input correspondstothenucleus [o] inJapaneseandthecoda/r/appearsasthesecondhalfof [0:]. This corresponding pattern applies only to the word-medial coda M and the word- final coda lr/ does not correspond to the second half of [0:], even when the nucleus preceding the coda /r/ in English corresponds to [0] (see (6) below).3 This indicates that the word-medial coda lr/ and the word-final coda /r/ are treated differently in Japanese. ’Dr.MutsukoEndoHudson(personalcommrmication)suggcststtntsemanticsmigbtberelevanttothe deletion ofthe word-amt coda/r/ (including freevariation), since compmerqelated words (e.g. ‘scanner’ and‘computer’)tendtomdergodeletiondenglishwordawhoaemeaninglsrelatedtopeople(e.g. ‘catcher’and‘pitcber')seemtoresistthedeletion.lnthisdissertation.however,lwillexplainthe phanmemnphomhp‘cdly,bwmnethemrd—finflmdahlmEngfishwmdsmhu‘dhwtm’md ‘editor’eanbedeletedinJapmese.Exampleslike‘director’and‘editor’donotcompletely excludethe possibilitythatselnanticsplaysaroleinthedelction.lwillleavethisquestionforfiltmerescarch. ’Thcloanwordwoo‘war’istheonlyexceptionlfound. 76 i. ...l s,’.‘ 'ol ' . . J a l .U .f‘ P“ i ‘- ’0 r 'l D ~ r v d O t I “a" (3) Word-medial /r/ -+ [o] (a part of [a:]) a. pooku ‘pork’ b. hoosu ‘horse’ c. noosu ‘north’ d inpooto ‘import’ e. soosu ‘source’ Second, the corresponding pattern (1c), i.e. /r/ -—> [a] (including the second half of a long vowel [a:]), is the most common among the four corresponding patterns for me English coda /r/. This pattern is further divided into two groups: the English coda /r/ appears as a short vowel [a] when the nucleus preceding the English coda /r/ corresponds to a vowel other than [a], while it appears as the second half of a long vowel [a:] when the nucleus preceding the English coda /r/ corresponds to [a] in Japanese. Examples are given in (4) — (7). (4) Word-final Ir/ -* [a] (a part of [a:]) a. sea ‘sil" b. sutaa ‘star’ c. tawaa ‘tower’ d. sentaa ‘center’ e. furawaa ‘flower’ 77 k) (5) Word-medial lr/ -+ [a] (a part of [a:]) a. sutaato ‘start’ b. pataan ‘pattem’ c. kaatuun ‘cartoon’ d. haarnonii ‘hamlony’ e. chaati ‘church’ (6) Wordofinal lr/ —+ [a] (short vowel) a. doe ‘door’ (‘doo) b. furoa ‘floor’ (‘fur00) c. pia ‘pier’ (‘pii) d pyua ‘imm’ Wynn) e. kea ‘care’ (‘kee) (7) Word—medial M -+ [a] (short vowel) a. tiariidaa ‘cheerleader’ b. giasifuto ‘gearshifi’ c. eabasu ‘airbus’ As illustrated in (3), the word-medial coda lr/ corresponds to the second half of [0:] when thenucleusprecedingthelr/intheinputcorrespondsto[0]inJapanese.But,the examples (6a) and (6b) indicate that the word-final coda Ir/ does not correspond to the second half of [0:] even when the nucleus corresponds to [0] in Japanese. The English coda lr/ corresponds to a short vowel [a] when the nucleus preceding the English coda /r/ 78 D s C s. '- II‘ a I 4 I ' I. . .1 p D! 1 t o _. 'I ’ pl \I .‘ t P). f \ ’ .. ... n l o It I 'J! c a o n corresponds to a vowel other than [a], i.e. [i], [u], or [e] in the word-medial position and [i], [u], [e] or [0] in the word-final position. Third, in the corresponding pattern (1d), i.e. /r/ -’ 0 (deletion), the English coda /r/ doesnothaveanapparentconespondentintheoutput.Thispattemisobservedinthe loanwords that were recently brought into the Japanese lexicon.’ Examples of this type are shown in (8). (8) Word-final /r/ -* 0 (deletion) a. konekuta ‘connector’ ('konekutaa) b. weha ‘wafer” (‘wehaa) c. siikensa ‘sequenccr’ (’siikensaa) d. kurasuta ‘cluster’ (‘kurasutaa) e. sukyana ‘scanner’ (’sukyanaa) In the examples in (8), the nucleus preceding the word-final /r/ in the input corresponds to [a]andflwword-final/r/doesnothaveanappmentconespondentintheoutprmThe examples in (8) do not mean that the English coda /r/ in recent loanwords never has a counterpart in the output. As illustrated in (9), the word-medial coda /r/ in the input corresponds tothe second half0f[a:] inrecent loanwords, as inthe correspondingpattern (1c) (see (5)), and the word-medial coda /r/ is never deleted in the output. Fruthermore, ‘T'hcdefinitionof‘recentloanwords’inthischapteristheloanwordsthatarefcundonlyinAkahori(l999), whichhdnlatendicfiomamongmefivedicdmniesmedasthesomcefmdfissudy. ’ThkmrdwubmmwedfimnEnghshmaemmomc.mmhcau,meoummwehamems‘aconmomm ofacomputer’anddounmhavetheothameaning‘athinbisadt’.hthelatmrmse,dliswrdis pronouncedaswehaawiththelongvowelhzl 79 r . o v ttl. s‘ is I I. . ‘ r c . . I J In . l '- l \A - I I / (I T . n r . . II .II v e n . d J , I . s a . . l ' cw. : r r n ' . hi. I Lo - I’ I b . at I u I .l I a i a rd . . r I. l . a v I: c v . . -. 1 cl ' U ’ n I4 ' . 4 l I ,5 I . 's a. r \ n I - . . n O . . on.) K (as. . I ) . ‘0 r .. .. r u r . . . . . 4. t r r ' r l r v I a s \ l l p . . o r r. ‘I. r -r .s I O r I n I I L I (I v. . 0.. T I r or. r. r \- o the word-final long vowels other than [a:] are not prohibited in recent loanwords, as shown in (10). The examfles in (8) - (10) suggest that the only difference between the recent and less recent loanwords is the treatment of the English word-final /r/ and only the wordrfina1 English /r/ in recent loanwords does not have an apparent correspondent in the output. (9) Word-medial /r/ -* [a] (a part of [a:]) a. intaanetto ‘internet’ (*intanetto) b. intaariibu ‘interleave’ (‘intariibu) c. wizaado ‘wizard’ (‘wizado) d. komaasu ‘commerce’ (‘komasu) e. insaata ‘inserter’ (’insata) (10) No Shortening of Long Vowels a. fakutorii ‘factory’ (’fakutori) b- km ‘W’ (‘kyu) c- kyarii ‘cany’ ‘ (’kyari) diagii ‘iassy’ (‘iasfl Finally, the corresponding pattern (1e), i.e. /r/ --> [a] or O, is fiee variation. This is also a recent phenomenon (Inaba 2001). In the other corresponding patterns introduced thus far, English M has only one corresponding segment in an output including the null correspondent. In this pattern, on the other hand, English M has two possible correspondentsinJapanese. As illustrated in (11), thenucleusprecedingthecoda/r/in 80 the input always corresponds to [a] whereas the coda M has two possible correspondents in an output, i.e. the second half of [a:] or the null correspondent. (ll)Word-final/r/ —+ [a]or0(fiee variation) a. erebe'eta(a) ‘elevator’ b. kompyu‘uta(a) ‘computer’ c. puri'nta(a) ‘printer’ d. k00dine’eta(a) ‘coordinator’ e. kompure'ssa(a) ‘compressor’ The examples in (l l) have two common characteristics: (i) the accent in Japanese is on the syllable containing the antepenultimate mora, which is a favored landing site for the loanword accent as discussed in Chapter 3, and (ii) the syllable containing the antepenultimate mora is heavy. The flu variation introduced above is observed only word-finally. As illustrated in (12), the word-medial coda /r/ does not have more than one possible correspondent in an output and it always corresponds to the second part of [a:], even when the word has the two common characteristics above. (12) Word-medial /r/ -’ [a] (no deletion) 8~ ha'ma) ‘hufler’ (’Wfl) b- M'Ma) ‘maflter’ (‘mMaD 0- WM!) ‘Pfiflor’ 0M8» d slea'ato ‘spurt’ (’supato) e. taamine'eta(a) ‘terminator’ (*tamineeta(a)) 81 ‘ l l’:‘ I ":1: ..I "7‘ 7n As illustrated in (13), furthermore, other word-final consonants or vowels in the input neverhavemorethanonecorrespondentintheoutput, evenwhenthewordhasthetwo common characteristics observed in (l 1). (13) No Word-final Free Variation a. u'ddii ‘woody’ (‘uddi(i)) b. ku'kkii ‘cookie’ (*kukkifi» c. ta'akii ‘turkey’ (*taaki(i)) d. ge’kkoo ‘gecko’ (*gekko(0)) The examples in (l l) - (13) suggest that only the word-final /r/ can have more than one correspondent in Japanese. 4.3. Analysis This section discusses the realization of English /r/ in Japanese within the OT framework. Before analyzing the data, let us consider the inputs to phonological processes. 4.3.1. The Inputs to Phonological Processes As introduced in Chapter 2, Paradis and LaCharité (1997) propose a phonology- based model of loanword phonology where the inputs to phonological processes are based on the phonemic representation of the source language. In this model, all the allophones of a phoneme in the source language are treated identically in the host language. 82 The realization of English Irl in Japanese is problematic for Paradis and LaCharité’s model. In their model, both the onset /r/ and the coda /r/ should be treated identically in Japanese. As introduced in the previous section (see (14)), however, the onset M and the coda /r/ are treated differently in Japanese: the onset /r/ corresponds to [r], the closest sound in the Japanese phonemic inventory, whereas the coda /r/ never corresponds to /r/ but corresponds to vowels or it is deleted. (14) Five Possible Corresponding Patterns (== (1)) a. Onset /r/ -’ [r] b. Coda /r/ ---t [0] (the second half of a long vowel [0:]) c. Coda Ir/ -’ [a] (including the second half of a long vowel [a:]) d. Coda /r/ -’ 0 (deletion) e. Coda /r/ -’ [a] or 0 (flee variation) With regard to the realization of the coda /r/, one question arises: Why does the English coda Ir/ correspond to vowels in Japanese? As introduced in the previous section, the English coda /r/ can correspond to [a] or, [o]. For example, the English coda /r/ correspondsto[a] intheloanwordssuchaspaalcu ‘park’ anddoa‘door’ andcorresponds to [0] in the loanwords such as pooh: ‘pork’ and foam ‘form’. As illustrated in (15), however, all the other English consonants correspond to consonants in Japanese. That is, English/r/istheonlyconsonantthatcancorlespondtovowelsaswellasaconsonantin Japanese. [fall the allophones of a phoneme in the source language are treated identically intheinputtothephonologicalprocessinthehostlanguage,asParadisandLaCharité 83 .,.- -~A .‘l'. -J 3%. M claim, it is not clear why only the English coda [r] corresponds to vowels in Japanese and what distinguishes the onset [r] and the coda [r]. (15) Consonantal Adjustments Observed in English Loanwords English a. [a ... b. [v] _. c. [0] _. d [h] a e. n] ... 1W [t] (leaf [b] (love [8] (cloth [Z] (mother [r] (we! [fitthD [MD [MD (rim!) As discussed above, Paradis and LaCharite’s model does not explain the realization of English /r/ in Japanese. In this study, therefore, adopting Silverman’s (1992) model (see (16)), I will assmne that the inputs to the phonological processes are the perceived segmentsandthatthedifl’erentrealizationoftheonset/r/andthecoda/r/inJapaneseis dueMthepercepfion.Ashto¢wedhChapta2,Silvammpmposesaphmefics-based model of loanword phonology and hypothesizes that the speakers of the host language have no access to the phonological representation of the source language. In this model, the input is merely a linguistically unanalyzed acoustic signal and it is distinguished from the Perceptual Level, where the native segment and tonal inventory constraints apply and they restrict the representation of perceived segments. ‘t I, ... i at a ( p l t l ‘ . I. ‘I r f l w l. o 1. I 940 I s \ . 1 -.. I. .I ‘i s o It . (16) Silverman’s Model (1992: 293) incoming Perceptual Operative acoustic Level _> Level ...—...) output signal representation representation Operative Level processes native segment native and tonal phonotactic inventory constraints and constraints preferences Next, let us consider how English M is perceived in Japanese. With respect to the input status of English /r/, I assume that English loanwords in Japanese are borrowed from rhotic dialects of American English, since most of the English loanwords in Japanese are borrowed from American English (Kay 1995) and American English is rhotic in general (Downes 1998: Ch 3).6 In this Study, 1 hypothesize that an input segment canbeperceiveddifl’erentlyattheperceptuallevelbasedonitsloeationandthattheonset /r/ and the word-final coda M are perceived as [r] and [a] in Japanese, respectively. Regarding the perception of the English onset [r] in Japanese, it is plausible to assume diattheonset/r/isperceivedas [r] atthePerceptual Level, sinceitispronouncedas [r] in American English and it always corresponds to [r] in Japanese. With respect to the perception of the word-medial coda /r/, I hypothesize that the word-medial coda M can be perceived as [c] or [a]. As shown in the previous section, the word-medial coda /r/ ‘InvarietiesofAmericanandBritishEnglish,thconset/r/isdifi’erentfromtheeoda/rl. 85 r t .. or I .s 7 to a r. t a i . . l . ....o t V O I l t, i r t. I. It, H i i r . . . l .l 4 p . ‘ . .‘ c‘ . cl . e .l at - n. I, o x ’ I A t; t o toll \. o t x s. it i l.’ t . 5i Yr 1 I. I i’ ' K t . ..I r t/ ’l e . I I 9 .I t II it . s t o n! t . r . l‘ I I o ; t . n . v . e! I O t o s . I I I . .0 O I e O s t o corresponds to [0] when the nucleus preceding coda /r/ in the input corresponds to the nucleus [0] in Japanese (e.g. ‘pork’ -’ pooku). The output [0] in those examples is due to the perception. The word-medial coda /r/ in examples such as ‘pork’ is perceived as [0], probably because the word-medial post-vocalic (and pre-consonantal) position is less salient. In sum, the inputs, perceptions, and outputs of English /r/ are summarized as in (l7). (l7) Inputs, Perceptions, and Outputs of English /r/ lOnset lr/ Word-final Coda /r/ Word-medial Coda /r/ [RIMS [r] [r] [r] Pmtions [r] [a] [a] or [0] Outputs [r] [3] (including the second [a] (including the second part of part of [a:]) or zero [a:]) or [o] (the second part of [0:]) 4.3.2. Analysis In the previous subsection, I have argued that the different realization of the onset /r/ andthecoda/r/inJapaneseisduetothcperceptionandthatEnglish/r/canbeperceived difl‘erently in Japanese depending on the syllabic position it appears. In this subsection, I will discuss the realization of English M in Japanese within the OT framework. 4.3.2.]. Word-final Coda lrl As introduced in the previous section, the wordfinal coda M in English has four possible corresponding patterns, which are repeated in (18) - (21). The vowel [a] is the most common corresponding segment to the English coda /r/ and, among (18) — (21), the 86 corresponding pattern in (18) is the most common. The corresponding pattern in (19) is observed only when the nucleus preceding the coda /r/ in English corresponds to a vowel other than [a]. The deletion of /r/ is possible only in recait loanwords and the flee variation is also a recent phenomenon observed not only in recent loanwords but also in less recent loanwords. ( l 8) Word-final lr/ -’ [a] (a part of [a:]) (= (4)) a. sea ‘sir’ h. sutaa ’star' c. tawaa ‘tower’ d. sentaa ‘center’ e. furawaa ‘flower’ ( l 9) Word-final /r/ _. [a] (short vowel) (= (6)) a. doa ‘door’ (‘doo) 1). firms ‘floor’ (‘furoo) c. pia ‘pier’ (‘pii) d pyua ‘pum’ 0pm) e. kea ‘care’ (‘kee) 87 (20) Word-final M -+ 0 (deletion) (= (8)) a. konekuta ‘connector’ (*koneltutaa) b. weha ‘wafer’ (‘wchaa) c. siikensa ‘sequencer’ (’siikensaa) d. kurasuta ‘cluster’ (‘lrmasutaa) e. sukyana ‘seanner’ (’sukyanaa) (21)Word-final/r/ -> [a]or0(fieevariation) (=(1l)) a. erebe'eta(a) ‘elevator’ b. kompyu'uta(a) ‘computer’ c. puri'nta(a) ‘printer’ d. koodine'eta(a) ‘coodinatc’ e. kompure'ssa(a) ‘compressor’ Within the OT framework, the realization of the English word-final coda /r/ in Japanese can be explained by the six consu'aints MAX, Dar, IDENT(CONSONANTAL), ‘Comtnx-oitsm, CODA-Com, and ‘[a:]]pw, which are defined as in (22) - (27). (22) MAX: No deletion. (23) DEP: No epenthesis. (24) IDENKCONSONANTAL): An output segment has the identical feature values for [consonantal] as its input correspondent. (25) ‘COMPLEx-ONSET: No complex onset. (26) CODA-COM): Codas cannot have independent place of articulation. 88 It 5‘ (27) I"[a:]]pw: No word-final long vowel [a:]. MAX (McCarthy and Prince 1995) in (22) is the anti-deletion constraint that ensures every element in the input is preserved in the output, whereas DBP (McCarthy and Prince 1 995) in (23) is the anti-insertion constraint that requires every element in the output to have a correspondent in the input. IDENT(CONSONANTAL) in (24) is a member of the constraint family IDENT (Prince and Smolensky 1993). This constraint stipulates that an element ofthe input and its correspondent in the output must be identical with regard to the feature [consonantal]. *COMPLEx-ONSET (ltd 1986) in (25) does not allow complex onsets, while CODA-COND (ltd 1989) in (26) prohibits a coda with independent place of articulation. ‘[a:]]pw (Mutsukawa 2004) in (27) prohibits the word-final [a:]. The reasons only the word-final, but not the word-medial, [a:] is prohibited are because the final POSition is the least marked position for phonological changes crosslinguistically (e.g. Word-final devoicing in German and Russian) and the vowel [a] has the greatest sonorlty, or the longest, among the five vowels in Japanese (see Ladefoged 1993: Ch 10). That is, the distinction between a short vowel [a] and a long vowel [a:] makes little difl'erence in the word-final position. As will be shown later in this section, this constraint plays a 8iEllificant role in the core part of the Japanese lexicon. In Japanese, no words have a complex onset or a coda consonant with independent Place of articulation. This suggests that the constraints *ComLEx-Onssr and CODA- COND are highest-ranked. To avoid complex onset consonants or coda consonants, there are two possible ways: vowel epenthesis or consonant deletion. In Japanese, the former is adopted to avoid complex onset consonants and coda consonants. This indicates that 89 MAX is ranked higher than DEP (see (28)). The constraint IDENT'(CONSONANTAL) is also highest-ranked, since the word-final M in English, which is perceived as [a] in Japanese, never corresponds to [r]. (28) MAX >> DEP ‘foom (‘form’) --+ foomu’ foom7 MAX DEP a. G‘foomu “ b. foo *1 As introduced above, the English word-final coda M has four possible corresponding patterns: short vowel [a], the second half of [a:], null correspondent, and flee variation (the second half of [a:] or null correspondent). The realization of the English word-final coda /r/ suggests that the constraint ranking between MAX and ‘la:]]pw has been changing, as shown below. First, consider the corresponding patterns in (18) (e.g. ‘tower’ —-+tawaa) and in (19) (e.g. ‘door’ —+ den (‘doaa, ‘doo)). In these cases, English /r/ in the input, which is Pemeived as [a] at the Perceptual Level, corresponds to [a] in Japanese. In other words, the output [a], the counterpart of English lrl, is faithful to the perception. Within the OT fI‘lilneworlt, this pattern is explained by the constraint ranking MAX >> ‘Iadlrw in (29), 38 illustrated in (30). In (30), the candidate (30c), which has the word-final long vowel [a:], is the optimal candidate because the candidate (30b), which does not have the long vOwe] word-finally, violates the higher-ranked constraint MAX. The candidate (30a) also \ ’Intiteuhreetntinthischapm,theinputishasedonthepetception 90 .s 4. "I 's ._ "A. . J 'l. J-’ . Q.‘ ‘ cannot be the optimal due to its violation of CODA-COND. Among the corresponding patterns for the word-final English lrl, this pattern is the most common. (29) Ranking for ‘/r/ —-> [a]’ ‘COMPLax-ONssr CODA—COND Inmflcous) MAX 1:331”! DE? (30) Tableau for ‘tawaa (‘tower’) -» tawaa’ tawaa 'COMPLBXT ConA- [DENT , , ONSET E CONT) i @0143) MA“ [“11” DE? a. tawar g ‘1 g r b. tawa j ’l c. G'tawaa 7 L * Second, the corresponding pattern in (20) (e.g. ‘cluster’ —-> kurasuta (‘kln'asutaa)) is Observed only in recent loanwords. This pattern occurs due to the constraint ‘laillrw. This constraint plays a crucial role in the core strata of the Japanese lexicon as well. The table (31) shows the result of a dictionary research I conducted“ As illustrated in (31), 1426 Japanese words have the long vowel [a:] word-finally. Among them, 1382 words, i.e. 96.9%, are loanwords and only 44 words, i.e. 3.1%, are non-loanwords. That is, the Word-final [a:] is highly disfavored in the core part of the Japanese lexicon, i.e. Yamato, Sfilo-Japanese, and Mimetic sub-lexica. The fact that recent loanwords do not have the x. ‘ The source is Shimmm (I991). 91 r t 7 ‘20 ‘tl . l s l ' ' I" - sl I . (k, , t ,. . , ., l 1 ‘ ' l '1 ...-o ‘3‘!“ ‘ "I " '- .. i ... - _ . in. l a 4 em ‘ ‘ l ‘ ...r t “ ' . . 5 '~’ ' ' " I . ' ‘ o 7 . I ‘ ‘ ‘ . t | . o ‘ up I. l I s it I. ‘h I .f ' V ‘_,. I at ’ i I ' ' ’ . . l l r 't ‘ .. . r 1 4.1' 4 .4.f. 0‘ I, _. 01L 1.. - word ltxic A t,: *4 word-final [a:] indicates that they are more assimilated to the core part of the Japanese lexicon with respect to the constraint ‘[a:]]pw. although ltd and Mester (1995) claim that recent loanwords violate more constraints and are less assimilated. (31) The Number of the Words with Word-final [a:] Total # # of Loanwords # of Non-loanwords 1426 1382 (96.9%) 44 (3.1%) (Non-loanwords include interjections, onomatopoeia, non-standard words, etc.) The fact that the deletion of the English word-final lr/ occurs only in recent loanwords suggeststhattherankingbetweenMAxand‘[a:]]rwhasbeenchangingasin(32) and that the ranking in (33), where ‘[a:]]pw is ranked higher than MAX, is the constraint ranking for recent loanwords. The constraint ranking for recent loanwords selects the Optimal candidate without the word-final [a:], as illustrated in (34). In the tableau in (34), the candidate (34d), which does not have the word-final long vowel [a:], is the optimal candidate because the candidate (34c), which has the long vowel word-finally, is ruled out due to the violation of ‘[a:]]pw. (32) MAX>>‘{atllrw --» ‘latllrw >>m (33) Ranking for ‘/r/ —+ e’ *COMPLax-Onsar CODA-COND Insurtcons) 'l'aillrw 5% l"? r. l ' ' ‘ d ' t ' s 0 I ~ f I I .K‘J ll“ , e.-. ...” 34) ' (34) Tableau for ‘ltrastaa (‘cluster’) W (‘kurasutaaY W $333": $311: (13;; shun”, Der a. lcrastaa ‘1‘ i * b. kurasutar 'r *1 r u c. kurasutaa * E *1 .. d.<"kurasuta % as Finally, the corresponding pattern in (21) (e.g. ‘computer’ -+ kompyuuta(a)) shows free variation. As Inaba (2001) points out, this flee variation is also a recent phenomenon. This phenomenon is observed both in recent and less recent loanwords. Free variation is due to the fleely-ranked constraints ‘[a:]]rw and MAX. The constraint ranking in (35), Where ’[azllpw and MAX are fleely ranked, predicts the optimal outputs, as illustrated in (36). In (36), both candidates (36b) and (36c) are optimal, since each of them violates one of the freely-ranked constraints once. 93 I I'll" I . .l he, ..,‘-. J . (35) Ranking for Free Variation a. Ranking for ‘lr/ ~+ [a]’ 'COMPLEX-ONSET CODA-COND Insprrtcous) MAX ‘[:a:]]pw DEP b. Ranking for ‘/r/ -» 0’ ‘COMPLEX-ONSET CODA-COND Inflows) ’l'aillrw MAX DEP (36) Tableau for ‘kornpyuutaa (‘cornputer’) -» kompyuuta(a)’ (‘laillyw and MAX are fleely ranked.) kompww ‘COMPLEXE CODA- E IDENT ‘. . -ONSET j COND j (cons) MA" ... [a-llrw DEP 8- kompyuutar E *l a t l : #- g b. a- kompyuuta . l c. a kompyuutaa ; , 4.3.2.2. Word-medial Coda lrl In the previous subsection, I have discussed the realization of the English word-final coda /r/. In this subsection, 1 will discuss the realization of the English word-medial coda ’11. The word-medial coda lr/ in English has three possible corresponding patterns, as in (37) — (39). I, “w. _‘ .l . ‘ 1: “, ,- . ,l , I I ‘ . l 5 C h ‘ I ‘1 (37) Word-medial lr/ -’ l 9001‘“ b. hoosu c. noosu d. inpooto 0.300811 (3 8) Word-medial lr/ -’ a. sutaato b. pataan c. kaatuun d. haamonii e.chaati (39) Word-medial lr/ -* a. tiariidaa b. giasifuto c.eabasu The only difference between the word-final coda /r/ and the word—medial coda /r/ is that the latter can correspond to [o] as in (37) while the former cannot. The word-medial c()(la/r'lcorrespondsto[o] whenthenucleusprecedingcoda/rlintheinputcorresponds ‘0 thenucleus [o] inJapanese. The correspondingpatternin(37)isduetotheperception. The only difference between [a] and [0] lies in the feature [toms]: [a] has [~round] and [0] (8 Part Of [01]) (=' (3)) ‘W’ .hom. ‘north’ ‘import’ ‘source [a] (a Part Of [8:]) (== (5)) .m. .m. .m. .mony. ‘church’ [a] (a Short vowel) (‘= (7)) warmer. .mm. “bus. 95 i,’ (37) Word-medial lr/ -’ [o] (a part of [0:]) (== (3)) 8- vook“ ‘Pork’ b. hoosu ‘horse’ c. noosu ‘north’ d. inpooto ‘irnport’ e. soosu ‘source’ (33) Word-medial It! —-> [a] (a part of [a:]) (= (5)) a. sutaato ‘start’ b. pataan ‘pattern’ c. kaatuun ‘cartoon’ d. haamonii ‘harmony’ e. chaati ‘church’ (39) Word-medial /r/ -’ [a] (a short vowel) (== (7)) a. tiariidaa ‘cheerleader’ b. giasifuto ‘gearshifi’ c. eabasu ‘airbus’ The only difference between the word-final coda M and the word-medial coda /r/ is that the latter can correspond to [o] as in (37) while the former cannot. The word-medial coda/r/correspondsto [o] whenthenucleusprewdingcoda/rlintheinputcorrecponds to the nucleus [0] in Japanese. The conesponding pattern in (37) is due to the perception. The only difference between [a] and [0] lies in the feature [round]: [a] has [~round] and 95 rra -‘ a . a y . . . I . . n . I . n ..- I . u. o v o ,- It. a? ‘4! . a e l‘ . v . ll ‘ V .b I . I4 . pit . I i]! . I P a n. . l a . n a . i c t. , l r 0-. . u, t 0 II. .- . a u . o Iv O. . . .- 1 J ‘ n . o n [0] has [+rormd]. The word-medial coda /r/ in examples such as ‘pork’ does not correspond to [a] in the output because of the highest-ranked constraint IDEN’KROUND) in (40). Within the framework of CT, the examples in (37) - (39) can be explained, as illustrated in (41) -- (43). In the tableaux in (41) - (43), the ranking between ‘[a:]]rw and MAX does not affect the selection of the optimal output, since ‘[a:]]rw does not apply to the word-rmdial segments. (40) IDENKROUND): An output segment has the identical feature values for [round] as its input correspondent. (41) Tableau for ‘pook (‘pork’) ... pooku’ (‘[a:]]pw and MAX are freely ranked.) Pook *Cowusx? CODA- I IDENT i Icem- ' . ONSET i COND i (cons) [(ROUND) M !‘[&]]w DE? a~ porku E O! i . E I . 1 t 5 ' b. poaku g g i i! I . ...'...... i r . (42) Tableau for ‘staat (‘start’) —+ sutaato’ (‘[a:]]rw and MAX are freely ranked.) staat ’Comuzxé CODA- E [DENT § loam -ONSET E COND 5(cous) Known) MAX i“[m]]pw DE? a. sutarto ‘ *I E ‘ '7 n b. asutaato : : I u L : 4 l c. sutaoto E 5 E *! a n : : : l '1 H", 1?. .l 5 [I . .1: t d tr r be I‘. II- I: -r r O . .. t‘. A. ‘ b‘ \ e . Io 1|! . x. . . a I . A a e - f. (43) Tableau for ‘tiariidaa (‘cheerleader’) -» tiariidaa’ (‘lafllrw and MAX are freely ranked.) tiariidaa ‘COWLBX?ConA- IDENTfiE IDENT H H T. . ONSET E COND E (cons) Known I [3'11" DEF a. tirriidaa ‘1 * ' i b. '3' tiariidaa . 'l c. tioriidaa f g *1 ; 4.3.2.3. Onset Irl Finally, this subsection deals with the realization of the English onset lrl. As illustrated in (44), the English onset lr/ always corresponds to [r] in Japanese. This conespmdhgpanmnisalmdmwthepercepfionJhecmmnaintmnldngdevelopedm the previous subsection accounts for the examples in (44) as well, as Shown in (45). In (45),English/r/doesnotcorrespondtoavowelintheoutputbecauseoftheconstraint IDENT(CONSONANTAL). In the case ofthe realization of the onset lrl, the ranking between MAX and ‘[a:]]rw does not affect the selection of the optimal output, since the onset /r/ is perceived as [r] and never corresponds to [a] in Japanese. (44) /r/ “’ [r] (= (2)) a. bureeki ‘break’ b. ribon ‘ribbon’ c. torakku ‘truck’ 1 kontorakuto ‘contract’ e. risaikuru ‘recycle’ 97 I . "H (4S) Tableau for ‘breek (‘break’) --r bureeki’ (‘[at]]pw and MAX are freely ranked.) break ’Cowwx? CODA- T Inwr [DENT MAX i. ONSET E Cwn E (cons) Known) [mllpw DE? l aG'bureeki 1| n b. buaeeki ] ? *l i “ c. buoeeki *l t ” : t L 4.4. Conclusion This chapter has discussed the realization of English /r/ in Japanese within the OT framework. The chapter is summarized as follows. First, this chapter revealed that the recent and less recent loanwords are different with respect to the realization of the word-final English [1‘]. In recent loanwords, the word-final English [r] is deleted, while in less recent loanwords it corresponds to [a]. This diachronic change can be explained in terms of the constraint reranking between I’[a:]]pw and MAX. The deletion of English /r/ in Japanese is triggered by l'[a:]]pw. Second, the recent English loanwords with the word-final coda M are more assimilated to the core part of the Japanese lexicon with respect to the constraint I"[a:]hvw. Ito and Mester (1995) claim that recent loanwords violate more constraints and are less assimilated. But this is not the case for the recent English loanwords with the word-final coda lrl. A dictionary research revealed that the constraint ‘[a:]]rw, which triggers the deletion of the English word-final coda lrI, plays a crucial role in the core part of the Japanese lexicon. The fact that the English word-final coda M is deleted in recent loanwordsindicatesthattheyaremoreassimilatedtothecorepartoftheJapanese lexicon with respect to the constraint ’[a:]]pw, which is contrary to Ito and Mester’s claim. 98 9 .I l o). o a IIIa . .. a . 1 . D- I it i . l‘- a I r. I v nlt I I r. t I [a to .3. 6|: . V r . .... .4 . t l 1,. . .. i r 1’ in. . o .a r) r . i ‘ ; . ., n I... s. bl. Third, the inputs to phonological processes in loanword adaptation are the perceived segments, which are closely related to the phonetic representation of the source language. The nature of the inputs is still under debate in the literature. Based on the realization of English M in Japanese, this study argued that the inputs to phonological processes in loanword adaptation are not the phonetic or phonemic representation of the somee language but are based on the perception. This study supports Silverman’s (1992) idea that the phonetic representation of the source language and the input to loanword phonologyaredifi‘erentandneedtobedistinguished. Finally, this chapter argued that the different realization of the onset M and the coda M in Japanese is due to the perception and that an input segruent can be perceived differently at the Perceptual Level based on its location; In this study, I hypothesized that English /r/ can be perceived differently in Japanese depending on the syllabic position it appears and that the English onset lrl, word-medial coda /r/, and word-final coda /r/ are perceived as [r], [a] or [o], and [a], respectively. I'm .4!) 5.," 5. The Realization of the English Plural Morpheme‘ 5.1. Introduction This chapter explores the realization of the English plural morpheme in Japanese within the framework of Optimality Theory (0T: Prince and Smolensky 1993). Three phonological phenomena related to the English plural morpheme are observed in Japanese: devoicing (e.g. tigers -+ taigaasu), diachronic change (e.g. raionsu -> raiongu ‘lions’), and deletion (e.g. the Lord of the Rings --> roodo obu za ringu). Among these, there are a few previous studies discussing the devoicing (Tateishi 2001, 2003, Fukazawa and Kitahara to appear) and the diachronic change (Tateishi 2003), but the deletion has been studied little in the literature. Also, the previous studies dealing with the devoicing of the plural morpheme have one common problem. They claim that the devoicing is a phenomenon in affixes (Fukazawa and Kitahara to appear) or in the Yamato sublexicon (T ateishi 2001, 2003), which is one of the four sublexica in the Japanese lexicon. These studies, however, do not explain why only the English plural morpheme undergoes devoicing in Japanese. The goals of this study are foru'fold. First, this chapter argues that the devoicing of the plural morpheme is a phenomenon in the affixes in the Loanword sublexicon. The previous studies claim that the devoicing is a phenomenon in the Yamato sublexicon or in affixes. Bm,basedonthefactsthattheEnglishplmalmorphemeisflreonlyaffixdlat undergoes devoicing in Japanese and the consonant devoicing is not observed in the 'AneuhavmionoffluschptaappeusumcwdEgsofme3MSemflhuemafidemfmm Fhonoloy. 100 I i,, »l I .h ., .(‘r .v‘ ‘iu \._ Yamato sublexicon, this study claims that both etymological and morphological categories are relevant to the devoicing of the English plural morpheme.’ ThesecondgoalistoshowthatJapaneseborrowershaveaccesstotbe morphological information of English in borrowing. The devoicing of the English plural mmphemeisaphenommmmmeaflinwhichisammphologicalcategoryjhisfaa suggests that Japanese borrowers have access to the morphological information and distinguish the stem and the affix. Third,flfissnrdydiscussesdlediachrmncchangewithregardmdlerealizafionofthe English plinal morpheme in Japanese. The older and younger generations are different with respect to the treatment ofthe plural morpheme in Japanese. In the older generation, the English plural morpheme is always devoiced when it is in the word whose singular formalsoexistsasaloanwordintheJapaneselexicon.Butitisfaithfultotheinputwhen it is in a word whose singular form does not exist in the Japanese lexicon. In the younger generation,ontheotherhand,theplmal morphemecanbedevoicedonlyintheword whoseinplnconmmsmomflimonevoicedobsmrentThepresemsmdyuguesflratthe accessibility to the morphological information of English has been changing: the older generation does not have access (or has limited access) to the morphological information of English whereas the younger generation has access to the morphological information of English. 'Thneammhneomomdmleingohomedmimwhiehhnphemmenmmthemm sublexicon (e.g. beddo—o betto ‘bed’). This devoicing occms to avoidthe sequerlee of voiced obstruents (NishhanMl).hdfischpmr,lMflignaemhphenomemnbecumhhopfimdmdmbjeum 'l"! ”.m . 101 ‘.I ‘v .0. I ll ’0‘. fl 4 I I f w l ‘5 r I- r. . l t r r n r I. til I t. (a e ‘4' o v u} . 1' . to al- o I... ..J 0 i, V r ‘ .a 9.. it. . . .\ l v} at 1 . > in! Finally, this study explores the difference between the word-level and the phrase— level with respect to the realization of the phual morpheme. The English plural morpheme does not have a counterpart in Japanese when it is in the phrase-final position in English, while it always has a correspondent at the word-level. This suggests that the plural morpheme at the word-level needs to be distinguished from the plural morpheme at the phrase-level. The present study claims that the deletion of the pllnal morpheme at the phrase-level is triggered by the constraint DEP-IO(PH AFF). This chapter is organized as follows. Section 2 presents the data on the devoicing of the English plural morpheme. In section 3, after illustrating an overview of three previous studies, I will analyze the three phenomena related to the English plural morpheme: devoicing, diachronic change, and deletion. Finally, this study concludes in section 4. 5.2. Data The phonological realizations of the plural morpheme in English are determined by the last segment of the stem, as summarized in (l).2 The phonological realizations ofthe phual morpheme in English loanwords in Japanese, however, do not necessarily follow English realizations, as illustrated in (2) - (4).3 ‘EnglishprommciationinthissmdyisbasedonxenyonandKnott(l944). ’T‘herealizationofllz]ianpaneseisnotshownhere,sincefewEnglisbwordswithitifany,havebeen takenirrtoJapsneseandIcouldnotfindexamplesofthistype. 102 n . ‘ . a. re . i. . h C. . t. ‘ Ill . .o c . u U .. 0 .. . e- l - . . .l . l I ‘ . l4 . .l . I Q\ It . \ . '0 a ’ | . . I. 0 v J I O I .1. s .. v I. v .r n t . ' tr. .1 I . .1 lo I is I o i a . 9 V I u. u o t 0: .4 t u U . . . e . [LJ’ (l) a. -s «[12]: b. -s —-» [s]: c. -s —+ [z]: (2) [8] -+ St! a. sokkusu b.supootsu c.hooltusu (3) [2] -r In a. shuuzu b. jiinzu c.indianzu (4) [2] -* 811 a. bumusu b. taigaasu c. kabusu when it follows a sibilant; e.g. crashes [krazsu], quizzes [kwrzrz], bridges [briJiz]. when it follows any other voiceless consonant; e.g. beliefs [bslifs], books [buks], plates [piers], . otherwise; e.g. dreams [drimz], hills [hllz], trees [triz]. 6 mks 9 C m 9 ‘ (Atlanta) Hawks ’ (an NBA team) 68h”, djm, ‘(Cleveland) Indians’ (an MLB team) Chm, ‘(Detroit) Tigers’ (an MLB team) ‘(Chica80) Cubs’ (an MLB team) The examples in (2) -— (4) show that English [8] always corresponds to an whereas English [2] correspondsto eitherzuorsu. Thecorrespondingpattern[z] -+su in(4) is not a simple consonant devoicing, since zu in the word-final position, which is faithfirl to the input [2] in terms of voicing, is possible as exemplified in (3). Moreover, voiced l03 Obstruents in other positions, i.e. the word-initial and word-medial positions, as well as the word-final non-derivational 224, which is not the counterpart of the English plural morpheme, do not undergo devoicing, as illustrated in (S) - (7). (5) Word-initial Voiced Obstruents a. gareeji I'ltareeji b. zukkiini ‘sukkiini c. juusu I"Shuusu (6) Word-medial Voiced Obstruents a. nozuru ‘mm b. rajio l’r'ashio c. puroguramu ‘purokuramu (7) Word~final zu (non-derivational) a. chiizu 1“chiisu b. saizu ‘saisu c. poozu ‘poosu ‘zucchini’ ‘0 Juice , Devoicing is a highly disfavored phonological phenomenon in Japanese.’ The examples above indicate that devoicing is allowed only in affixes, more precisely ill suffixes. L ’Vowei devoicinginJapaneseisnotaphonologicalphenomenonbruapboneticphenomenonmeit ‘lhommemuchuwwfisficofphomficmmhumawopfimfity’awmde 2003b: 276 en. 7). 104 5.3. Analysis 5.3.1. Previous Studies This subsection presents an overview of three previous studies on the devoicing of the English plural morpheme in Japanese: Tateishi (2001, 2003) and Fukazawa and Kitahars (to appear). 5.3.1.1. Tateishi (2001, 2003) Tateishi (2001) is the first study discussing the devoicing of the English plural morpheme in Japanese. In Tateishi (2001, 2003), he points out that the relevant constraints for this phenomenon are ‘NC and Lyman’s Law, which are defined as in (8) and (9). (8) ‘NC: Post-nasal Obstruents must be voiced. (9) Lyman’s Law: No two voiced Obstruents in a morpheme. lthasbeenagreedintheliteratlue (ItdandMester 1995, l999,andothers)thatthesetwo constraints play significant roles in the Yamato sublexicon ((10) - (11)) but not in the Loanword sublexicon ((12) - (13)). (10) ‘NC: Effective in the Yamato Sublexicon a. kaku -— kaita ‘write - wrote’ b. toru -- totta ‘take - took’ c. karnu - kanda, ‘kanta ‘bite - bit’ d. tobu - tonda, ‘tonta ‘fiy - flew’ 105 1" (l l) Lyman’s Law: Effective in the Yamato Sublexicon a. futa ‘lid’ b. buta ‘pig’ c. fuda ‘card’ (1. ‘buda (12) ’NC: Inefl‘ective in the Loanword Sublexicon a. rantan ‘randan ‘lantern’ b. atoranta ‘atoranda ‘Atlanta’ c. tento ’tendo ‘tent’ (13) Lyman’s Law: Ineffective in the Loanword Sublexicon a. bideo ‘biteo, ‘pideo ‘video’ b. doragon *dorakon, ‘toragon ‘dragon’ c. jazu l'jasu, ‘shazu ‘jazz’ However, these two constraints seem to apply to the loanwords with the English plural morpheme. In (14), the loanwords with the English plural morpheme immediately following the nasal never undergo devoicing, whereas the examples in (15) undergo devoicingbecauseeachofthemhasavoicedobstruentinthestem. 106 ..I‘i. (16) 2" (WW) a. shirazu ‘know-not’ b. kaerazu ‘retum-not’ c. yomazu ‘readonot’ (17) No Lyman’s Law Effects on zu (negative) a. tabezu 1'tabesu ‘eat-not’ b. shinjizu ‘shinjisu ‘believe-not’ c. kangaezu I'kangaesu ‘think-not’ IfTateishi’s analysis is correct, Lyman’s Law should apply to the negative suffix as well, sinceboththeEnglishpliual morphemeandthenegativesuflixareYamatosuffixesand are phonetically identical. But the examples in (17) show that Lyman’s Law does not apply to the negative suffix, although it applies to the English plural morpheme, as illustratedin(15). The factthattheEnglishplural morphemeandthcnegativesumxare different with respect to the application of Lyman’s Law suggests that they are not in the same sublexicon. Since the negative suffix 7.14 is undoubtedly a Yamato suffix, the plural morpheme zu should be in different sublexicon, presurnabiy in the Loanword sublexicon. Also, notice that devoicing is a highly disfavored phonological phenomenon in Japanese. It is plausible to assrmre that a highly disfavored phonological phenomenon occurs not in the core part of the lexicon, i.e. Yamato, but in the periphery of the lexicon, i.e. loanwordThisalsosuggeststhatthephrralmorphemeisnotaYamatosufi'lxbuta Loanword suffix. 108 'I (l4) I"NC Effects on Loanwords with the English Plrual Morpheme a. jiinzu ‘jnnsu” ‘jeans’ b. indianzu ‘indiansu ‘(Cleveland) Indians’ (an MLB team) c. penginzu ‘penginsu ‘(Pittsburgh) Penguins’ (an NHL team) (15) Lyman’s Law Effects on Loanwords with the English Pltu'al Morpheme (= (4)) a. buruusu ‘btu'uuzu ‘blues’ b. taigaasu 'taigaazu ‘(Detroit) Tigers’ (an MLB team) c. kabusu ‘kabuzu ‘(Chicago) Cubs’ (an MLB team) Based on the examples above, Tateishi concludes that the constraints ‘NC and Lyman’s Law are relevant to the devoicing of the pltu'al morpheme, although it has been agreedintheliteratlnethattheyareinactive inthe Loanwordsublexicomandthatthefact that these constraints apply to the English plural morpheme suggests that the plural morpheme falls not in the Loanword sublexicon but in the Yamato sublexicon despite its origin Tateislri’s analysis seems to account for the realizations of the English plural morpheme. But, his analysis has a problem. That is, it does not account for other suflixes in the Yamato sublexicon such as the negative suffix an in (16). As illustrated in (17), Lyman’s Law does not apply to the Yamato words with the negative suffix, although boththeEnglish plural morphemezu andthe negative suffixzuaresuffixesandappear word-finally. 107 0-;‘.\ ‘Al I Q )l ... . ht 5.3.1.2. Fakazawa and Kitahara (to appear) Fukazawa and Kitahara (to appear) account for the devoicing of the plural morpheme 234 within the OT framework. They claim that this devoicing reveals three paradoxical cases for the making for Rendaku (Sequential Voicing) developed in Ito and Mester (2001a), which is given in (18).’ The constraints in (18) are defined as in (19) - (23) (cited from Fukazawa and Kitahara (to appear) except (23), which is slightly modified by the author). (18) Ranking for Rendaku (slightly modified from Fukazawa and Kitahara (to appear))‘ IDENT(VOICE)-IDANWORD No-D’m IDENT(VOICE)-Snlo—JAPANESE (SJ) EXPRESsAmx IDENT(VOICE)-COMMON-SmO-JAPANESE (CSJ) ‘NC ID'ENNVOICE}YAMATO NO-D h ‘AccordingtorukmwarndxitrlnrndrnrmkingisaomitsmdMerter(2001atEmlwmdldester (zools)donctdiocurrltendalmandlcouldnotanddtecomunirnmltingroraendrlorintheptper, although I foundthe rankingsimilartothis in ltoaod Mester (2003b: 152). ‘ In Fukazawa and Kimhara (to appear), instead of ‘No-D, No-D’m lumvotcsyhomoun, and 'NC, the constraints ‘VorOBs, ‘VOIOBS‘m IDEN'rtvaEH‘anlw, and 'NT are used, respectively. in this chapter, I will use ‘No-D, No-D’m InmtvorcayLomwoan, and ‘NC for consistency. 109 .-. l I K a c i V. ' \ I f I ' \r e l .l o e: r . t a .. . . i c n c t a 20", I. u \ (l9) No-D’m: No double obstrucnt voicing in a stem.’ (=Lyman’s Law) (20) EXPRESSAFFDC Affixes must be realized in the output. (21) 'NC: No voiceless obstruent alter a nasal. (22) No-D: No voiced Obstruents. (23) IDENT(VOlCE)—X: In the sublexicon X, an output segment has the identical feature value for [voice] as its input correspondent. Among the constraints above, relevant here are ‘NC, No-Dzm, IDEN’KVOICE}YAMATO, and IDENT(VOICE)—IDANWORD. Ito and Mester (2001a) claim that these four constraints are ranked as in (24) and (25), which are extracted from (18). (24) Ranking for the Yamato Sublexicon No-D’m >> 'NC >> IDENT(VOICE)-YAMATO (25) Ranking for the Loanword Sublexicon IDENKVOICE}LOANWORD >> No-D’m >> ‘NC FukmwaandKitahmclaimthahalthoughhisetymologicaflyasuffixinthe Loanword sublexicon, the plural morpheme zu contradicts the ranking for the Lomword sublexicon in (25) in the following three respects. First, No-D’m must be ranked higher than IDEN1‘(VOICE)-LOANWORD to account for the examples in (26), although the ranking between them in no and Mester (2001a) is mmvolcnyLomvonn >> No-Dzm. The reversed ranking ill (27) explains the examples in (26), as illustrated in (28). ’No-o’mm istheself-conjoinedconstnintofthcmarkednessconsu'aim ’No-Dinstemdomain. 110 (26) No-D’m. Effects on Loanwords with the Plural Morpheme (= (4)) a. bumusu ‘blues’ b. taigaasu ‘(Det:roit) Tigers’ (an MLB team) c. kabusu ‘(Chicago) Cubs’ (an MLB team) (27) No-Dzm >> IDENT(VOICE)—LOANWORD (28) Tableau for ‘Cubs -t kabusu’ kabuzu No-D’sm. IoBNr(v01ca)-LW a. kabuzu ‘! b. 0' kabusu " Second, ‘NC needs to be ranked higher than No-D’m, as in (29). In Ito and Mester (2001a), No-D’m is ranked higher than me. But this ranking fails to account for the examples in (30) where the English plural morpheme immediately following a nasal is realizedaszudespiteanothervoicedobstruentinthestem. Theranldngin(29)con'ectly predicts the optimal output, as illustrated inthe tableau (31). (29) ‘NC >> No-l)’m (30) ‘NC Effects on Loanwords with the Plural Morpheme (= (14)) a. jiinzu ‘jiinsu ‘jeans’ b. indianzu *indiansu ‘(Cleveland) Indians’ (an MLB team) c. penginzu ‘penginsu ‘(Pittsburgh) Penguins’ (an NHL team) 11] 9.x! ‘DI (31) Tableau for ‘Indians -» indianzu’ indianzu ‘NC No-Dzm a. <5" indianzu * b. indiansu '! Third, ’NC is ranked higher than IDENT(VOICE)-LOANWORD. As shown in (27) and (29), the rankings between No.02“... and IDENT(V01CE)—L0ANW0RD and ‘NC and No.02“... are No-D’m >> IDENT(VOICE)—IDANWORD and *NC >> No-D’m... This leads to the ranking in (32), because dominance relations in OT are msitive. This ranking contradicts Ito and Mester’s ranking IDENltVOlca)L0ANw0RD >> ‘NC. (32) ‘NC >> IDsN1(v0lCl-:)-Loanword Thethreepuadoxicalcasesfmdlemnkingin(18)fllatFukazawaandKitahm(m appear) claim are summarized as in (33) (cf. Ito and Mester’s rankings in (34)). (33) Partial Rankings for as a. No-D’m >> InsmtvmcsyLomwonD b. *NC >> No-D’m. c. ’NC >> IDBNKVOICE}LOANWORD (34) Partial Rankings for Rendaku (Ito and Mester 2001a) a. IDEN‘TUIOICE}LOANWORD >> No-D’m b. No-D’m >> *NC c. InsN'r(v01ca)-L0ANWORD >> 'NC 112 Fukammdeimhmamappearhsannedlattwoofmedueepmadoxicalmnkingsin (33), i.e. (33a) and (33c), are due to mixing up etymological knowledge with phonological knowledge. Following Fukazawa, Kitahara, Ota (2002), F ukazawa and Kitaluira propose a phonology-based categorization of the lexicon, which is independent of etymological information. That is, in lieu of etymology-based categories such as Yamato and Loanword, they employ morphological categories, namely stem and amx. In this model, markedness constraints cannot be relativized while faithfulness constraints can. 'l‘herankingin(35)istheconstraintrankingFukazawa andKitaharadevelop. lnthis ranking, Fukazawa and Kitahara adopt No.1)z instead of No-D’m since the relativization of markedneas constraints is not allowed in their model. The definitions of these constraints are given in (36) - (41) (cited form Fukazawa and Kitahara (to appear». (35) Ranking for Consonant Voicing and Devoicing in Japanese No-D’ M|AX(VOICE)g1g. UNlroser(v01ca)srm EXPRESSAFFIX | In'limtvmcsm ’NC UNlromrv(v0lcs).,m Inlsmtvmce)mnr Max(vows)mm ll3 r . . . ,. . on . r . . t v - r P I t t . .. . l e \ 1 O . . . is o J .. t a. I r! . b ‘ .. a . rt . s t r . l. ‘ l. J .. z I .t I r . r. t t O . . . I , . 4 a o s. . . t c act (36) IDENT(VOICE)g—rm: The correspondent segments in a stem in the input and the output have identical values for the feature [voice]. (37) IDENKVOICB)mrx: The correspondent segments in an affix in the input and the output have identical values for the feature [voice]. (38) Uncommworcmm: No feature in a stem in the output has multiple correspondents in the input (i.e., no coalescence regarding the feature [voice] in a stem). (39) UNIFORMrrY(VOICE).ogD: No feature in a word in the output has multiple correspondents in the input (i.e., no coalescence regarding the feature [voice] in a word). (40) MAX(VOICE)gm.: Every feature [voice] attached to a segment in a stem in the input has a correspondent in the output. (41) MAX(VOICE)m: Every feature [voice] attached to a segment in a affix in the input hasacorrespondentintheoutput. The ranking in (35) accounts for consonant voicing and devoicing in Japanese, i.e. Rendaku and the realization of the English plural morpheme, as illustrated in (42) and (43).” In the tableau (42), the plural morpheme does not undergo devoicing because the candidate (42c), which has the devoiced plural morpheme, violates the constraint ‘NC, and it is fatal. In the tableau (43), Rendaku is considered the realization of a linking morpheme [+voice] (R in (43)) (see Ito and Mester 2003b: 81-87 for discussion). The candidates (43d) and (43a) lose to the candidate (43c), because Rendaku dose not take ' rheconstraintmvorce)misnotincludedinthereteblesrnr,sinceitplsyslittleroleinselectingthe optimrrlcrrndidste. 114 j . lr_ place in (43d) and (43e), resulting the violation of EXPRESSAFl-‘lx, which requires the affixRberealizedinthcoutput. (42) Tableau for ‘Indians -> indianzu’ indianzu No-Dz MAX“ UNI“ xrArri ID" ’NC UNI“ IDm a. G'indianzu [voi] b.indianzu [voi] [voi] ’! c. indiansu [voi] ‘! dintianzu [voi] ’! e. intian-su *! O. (43) Tableau for ‘oyako-genka’ (Rendaku) oyako+R+kenka No—Dz Max,r UNI“ ’NC UNI“, a. oyako- [voi]R I’l b’ ““1“?" [voi]R [voi] *! I‘ c. G"oyakoferlkri [VOila d. oyako-kenfa [voi] ‘! e. oyako—kenka ‘l 115 Thus far, I have introduced Fukazawa and Kitahara (to appear). Their analysis of the English plural morpheme has two problems. First, as Ito and Mester (2003b: 97.98) point out, their analysis is problematic because it is difficult to reconcile with Richness of the Base, a core tenet of OT. Fukazawa and Kitahara’s analysis makes crucial use of UNIFORMITY and accounts for both Rendaku and the realization of the English plural morpheme. But, this model is problematic because the constraint UNIFORMITY cannot rule out input representations as in (44). In (44), the feature [voi] is multiply linked in the input.'lhis isaproblem since forms like/buda/intheoutputdonotviolateUNlFORmTY or No.1)2 and it is predicted that they are well-formed Yamato items. As illustrated in (l l) (e.g. ‘buda), however, forms like lbuda/ cannot appear as Yamato items in the output. This is a basic problem that any UNIFORMI'I'Y-based approach to OCP encounters. This problem occurs because the constraint UNIFORMlTY militaws against fusion but has nothing to say about input-given multiple linking (44) [voi] /\ buda Second, Fukamwa and Kitahara’s analysis dose not account for other afiixes in the Japanese lexicon. Fukazawa and Kitahara propose phonology-based categorizations of the lexicon where etymological information plays little role. In this model, the plural morpheme 2:4 and other affixes fall into the same category, namely ‘afiix’, regardless of their etymological backgrormds. This model predicts that 22: (plural) and zu (negative) in (45), which is etymologically Yamato, behave phonologically alike. But, the plural 116 l,’ (is! V. morpheme and the negative morpheme behave differently with respect to the constraint No-D’, as illustrated in (46) and (47). (45) z“ (negative) (= (16)) a. shirazu ‘know—not’ b. kaerazu ‘return-not’ c. yomazu ‘read—not’ (4s) Lyman’s Law Efi‘ects on Loanwords with zu (plural) (-= (4)) a. buruusu ‘blues’ b. taigaasu ‘(Detroit) Tigers’ (an MLB team) c. kabusu ‘(Chicago) Cubs’ (an MLB team) (47) No Lyman’s Law Effects on zu (negative) (= (17)) a. tabezu ‘tabesu ‘eat-not’ b. shinjizu ‘shinjisu ‘believe—not’ c. kangsezu "kangaesu ‘tllink-not’ The afixes zit (plural) and zu (negative) are both suffixes and phonetically identical. If etymological information is irrelevant as Fukazawa and Kitahara claim, these suffixes should behave alike with regard to No-Dz. Interestingly, however, the constraint No-D2 applies to are (plural) but not to :14 (negative). Why does zu (plural) behave differently from zu (negative)? The only difference between them is etymological background. Then, it is plausible to assume that the etymological categorization is relevant and that the constraint No-D’ applies only to the items in the Loanword sublexicon. The examples 117 d U I i) 1 3 \ 3(.. ‘4’ (a ‘ a, l ‘ . - l ' "N | ’ ‘ 1' ' ‘ J. 1‘ r i, - _.. “ ri. above indicates that the ranking in (35) developed by Fukazawa and Kitahara does not account for the consonant devoicing in Japanese. 5.3.2. Analysis As discussed above, the previous studies reveal that the English plural morpheme does not fall in the Yamato sublexicon and that morphological categorization is not enough to account for the devoicing of the plural morpheme. In this study, therefore, I propose that both etymological information, i.e. Yamato or Loanword, and morphological categorimtion, i.e. stem or affix, are relevant to the devoicing of the English plural morpheme and that only aflixes in the Loanword sublexicon can be devoiced in Japanese. In the following subsection, I will analyze the devoicing, diachronic change, and deletion of the English plural morpheme within the OT framework. 5.3.2.1. The Devoicing of the English Plural Morpheme Within the OT framework, the devoicing of the English plural morpheme can be explained by the time constraints ‘NC, No-D’, and maratvorcshnmm me. These constraints are defined as in (48) - (50). (48) *NC: Postonasal Obstruents must be voiced. (= (8)) (49) No-D’pw: No double obstl'uent voicing in a prosodic word. (50) IDMVOICEMW m: In a loanword stem, an output segment has the identical feature value for [voice] as its input correspondent. 118 ... 1", The constraint IDENT(V01CE)wmm m is the conjtmction of the two faithfulness constraints IDMVOICBWM and IDENT(VOICE)m, whose definitions are given in (51) and (52). (51) IDENKVOICEMMM: In a loanword, an output segment has the identical feature value for [voice] as its input correspondent. (52) IDENT(VOICE)nm: In a stem, an output segment has the identical feature value for [voice] as its input correspondent. The five constraints introduced above are ranked as follows. The conjoined constraint [Damivorcahmm m outranks the component constraints manuvomshomm and lnamtvorcsh. It has been assmned in the literature that a conjoined constraint is universally ranked higher than its component constraints (Smolensky 1993). The ranking between *NC and No-D’pw is ‘NC >> No-D’pw. As illustrated in (53), this making accounts for why the English plural morpheme immediately following a nasal is not devoiced even when the word has another voiced obstruent. The constraint Iosm(vorcs)wmwm m. is the highest-ranked, since phonological changes with respect to the feature [voice] are not observed in loanword stems. The constraint 'NC, on the other hand, is ranked lower than mmVomEWm m because, as illustrated in (54), many loanwords violate ‘NC (Ito and Mester 1999). This leads to the ranking IDENKVOICEMMORD m... >> ‘NC, which selects the optimal output, as illustrated in (55). The ranking between No-D’pw and IDENKVOICEMW is 119 No-szw >> IDENKVOICEMMM because the devoicing occurs to avoid the violation of NO-szw, as shown in (56). (53) *NC >> No-D’pw Indians ‘NC NO-szw a. 9 indianzu " b. indiansu ‘! (54) *NC: Ineffective in the Loanword Sublexicon (= (12)) a. rantan ’randan ‘lantem’ b. atm'anta ‘atoranda ‘Atlanta’ c. tento ‘tendo ‘tento’ (S S) IDENT(VOICE)wmom m >> 'NC lantern Inemtvorcs) “61' ‘NC a. randan ’l b. <3" rantan " (56) No.01... >> IDENT(VOICE)wmortn (Chicago) Cubsl No-szw IDENT(VOICE)-LW a. kabuzu ] ‘l b. 6' kabusu I " 120 csor In sum, the constraints introduced above are ranked as in (SD.’This constraint ranking accounts for the realization of the English plural morpheme, as illustrated in (58) - (so). (57) IDENT(VOICE)wmonn mart ’NC NO-szw IDENT(VOICE)Lomonn In (58), the optimal output is the candidate (58a) where no phonological change is observed with respect to the feature [voice]. The candidate (58b), on the other hand, is ruled out due to its violation of IDENT(VOICE)mmmp. 6$m~w socks IDLwSr I"NC NO'DZPW IDLV a. <3" sokkusu b. sokkuzu ‘! In (59), the plural morpheme is not devoiced in the optimal output. The optimal candidate (59c), where no phonological changes with respect to the feature [voice] are observed, violates the constraint No.02”. But that is not fatal. The candidate (59b), where the plural morpheme is devoiced, is not optimal due to its violation of a higher-ranked "I’heconstraimlnfimtvmcshwillnotbeshownintheconstraintmkingsandtableatuintherestof thischspter,becaueitplayslitfleroleinselectingdleopfimalcmdidne. 121 'tala' constraint I"NC. The candidate (59a) is not the optimal candidate either, since it violates the highest-ranked constraint Inm(vorcs)wmm m. (59)[21~' Z“ jeans Dst-r *NC No-szw ID“ : siinzu ‘! b. jiinsu ‘! " c5" jiinzu " In (60), the plural morpheme in the optimal candidate (60c) is devoiced. The candidate (60b),whichisfaithfultotheinputtothe0perativebeve1withrespecttothefeatm'e [voice], is eliminated due to its violation of the constraint No-szw, while the candidate (60a) is ruled out because ofthe highest-ranked constraint mmuvowabm m. (60) [Z] "’ 8“ (Chicago) Cubs IDstT ’NC No-Dpr ID“ a. kafuzu l"l b. kabuzu *! c5" kabusu ‘ As mentioned above, affixes in the Loanword sublexicon can be devoiced whereas atfixesintheothersublexicacannot.Thisasymmetrycanbeexplainedintermsofthe constraint ranking between the three constraints lnsmworcs), ’NC, and No-D’pw. The constraint rankings in (62) and (63) are the rankings for the Yamato and Sim-Japanese sublexica, respectively. Compared with the ranking for the Loanword sublexicon in (61), the constraint IDENT(vorCE) is ranked higher than No.02", in both the Yamato and Sinc- 122 Japanese sublexica, which blocks the devoicing (see the tableau (64)). Yamato and Sino- Japaneseuedifi'erengonmeoflrahmdwithrespectmthermkingbetween IDENKVOICE) and ‘NC. Post-nasal voiceless Obstruents are not allowed in the Yamato, while they are allowed in the Sim-Japanese (Ito and Mester’s 1999). This leads to the ranking Inmrtvorcs)“ >> ‘NC >> IDENNVOICEhmm. The ranking between *NC and No-szw in the Yamato and Sim-Japanese is ‘NC >> No-Dz, as in Loanword. (61) Constraint Ranking for the Loanword Sublexicon (= (57)) IfiWVOICEho-om stem 'NC NO-szw IDENKVOICEW (62) Constraint Ranking for the Yamato Sublexiconlo ‘NC IDENTfVOICEhm NO-szw "Tlreconsu‘aintNo-D’nneedstobedistinguishedfiomNo-D’m. TheconstraintNo-D’n, accountsfor thedevoicingoftheEnglishphlralmorpheme,asdiscussedtlnrsfar,butnot1.yman’sLawefi‘ect,because miscomnaintdoesnotaflowmorethanonevoicedobsuueminacompomuegfirde+halto—tfudebako ('fudehako) ‘pencil case’). The constraint bro-Dz"... is the or version of Lymm’s Law and is ranked higherthan‘NC,asdiscussedinltbandMester(2003b)(see(18)).Intherestofthischapter,lwilligrrore No-n’msimeitirnotrelevarntothcrocnsorarirstndy. 123 t l‘H (63) Constraint Ranking for the Sim-Japanese Sublexicon IDIENT(VOICE)3.1 ‘NC NO-szw (64) Yamato Sublexicon tabezu mm. are my...» No-D’rv a. tabesu ‘! b.6'tabezu‘ "' The rankings in (61) — (63) can be combined as in (65). The ranking in (65) mounts for the devoicing of the English phtral morpheme. This ranking seems to conflict with the constraint ranking for Rendaku in (18) (Ito and Mester 2001a). The constraint IDENKVOICELDMM is ranked higher thm ’NC in ltd and Mester’s analysis whereas the ranking between them is reversed in (65). But, the ranking in (65), where mar-(vowsmm is divided into two constraints IDENT(VOICE)wmo.n and IDENT(VOICE)MM ...... and they are ranked as Ianorcsw ...... >> 'NC >> Insurtvorcshnmm does not affect Ito and Mester’s analysis of Rendaku, because RendakuisaphenomenonobservedinthestemoftheitemsintheYamatosublexicon and Ito andMester (2001a) focus onlyonthestem. 124 I . t" (65) Ranking for Devoicing IDENT(VOICE)s.j IDENttVOlcewm M, tTC IDENT(VOICE)YW NO-szw vaorcshm I have shown that both etymological information and morphological categorization are relevant to the devoicing of the English plural morpheme. The fact that morphological categorization is relevant to the devoicing indicates that Japanese borrowers have acwss to the morphological information of English and distinguish the stem and the aflix. 5.3.2.2. Diaehroale Change in the Realization of the English Plural Morpheme IntheprevioussubsecfimhlhavediscussedthedevoicingoftheEnglishplmd morphemewithintheOT framework. Inthis subsecfiomlwilldiscussdiachronic change with regard to the realization of the plural morpheme. As I have shown in the previous subsection, the devoicing of the English plural morphemeisobmedinthewordwhoseinputhasmorethanoneobsmlentThe devoicing of the plural morpheme is a phonological phenomenon where the markedness constraints No-D’pw and ‘NC play significant roles: the constraint No.92", triggers devoicing while the constraint ‘NC blocks it. With respect to the diachronic change in the realization of the English plural morpheme in Japanese, Tateishi (2003) introduces an interesting observation that, in the 125 7 ' ,’;s ' I . . . (o . ~ cl . o 'I I . 9 . 4) . i ’ I jl . t . 0’, .1 t t. . U .a ;t \ ,“ ..' ’ .‘ ,’ .‘ ‘ l 31‘ o \- . t ,v_ I f .. ' s I 1 . . ‘x‘ '\ ‘ t “ l I . B . I ‘. A older generation (in 60 or older ages) in Japan, the English plural morpheme is always devoiced (e.g. raionsu ‘lions’, taigaasu ‘tigers’), whereas in the younger generation, as discussed in the previous subsection, the plural morpheme can be devoiced only in the word whose input contains more than one voiced obstrucnt (e.g. raionzu ‘lions’, taigaasu ‘tigers’). Tateishi’s observation suggests that the realization of the English plural morpheme in Japanese has been changing. WhydoesthegenerationgapcxistwithrespecttotherealizationoftheEnglish plural morpheme? There are two possible explanations: (i) phonology, i.e. the constraint ranking,hasbeenchanging,and(ii)theinmrttothe0perativelevelhasbeenchanging. But, the explanation that phonology has been changing has a problem. As mentioned above, the constraints No.02PW triggers the devoicing of the English plural morpheme. In the case of ‘lions’, however, No—D’pe does not trigger devoicing because the word ‘lions’ doesnothavemoretlranonevoicedobsmrent'l‘hefactflrattheword ‘lions’ undergoes devoicing in the older generation suggests that there is another constraint that triggers devoicing and that the constraint is ranked higher than 'NC, which blocks devoicing, in the phonology of the older generation. The possible constraint that triggers devoicing in the older generation is No-D, the component of the self-conjoined constraint No.01”. But, the ranking No-D >> *NC contradicts the making *NC >> (IDENT(VOICE)-YAMATO >>) No-D, which is urrcontroversial in the literature (see (18) for example). In fact, there are many Yamato words with a voiced obstrucnt (e.g. kagi ‘key’) and those words do not undergo devoicing even in the older generation (e.g. ‘kaki ‘key’). In this study, therefore, ItakethelatterpositionthattheinputtotheOperative Levelhasbeenchanging. 126 l" Tateishi (2003) mainly analyzes the names of professional sports teama and concludes that the English plural morpheme is always devoiced in the older generation (e.g. raionsu ‘lions’). But, there are counterexamples to his analysis, as shown in (66). What distinguishes raiorsru ‘lions’ from the counterexamples is the fact that, in the case of raionsu ‘lions’, the singular form raion ‘lion’ also exists as a loanword in the Japanese lexicon, whereas the singular forms of the counterexamples do not. This fact suggests that raionru ‘lions’ and jiinzu ‘jeans’ are borrowed differently, as schernatized in (67) and (68): in the case of raionsu ‘lions’, the English phrral morpheme is realized as a morphemeinthcinputtotheOperative LeveLwhereastheword ‘jeans’ isborrowedas one word, i.e. ‘jeans’ is not treated as ‘jean + s’. (66) Counterexamples a. jeans -» jiinzu 1). shoes -> shuuzu (67) lions -v raionsu a. lion -+ raion b.raion+thepltn'almorpheme—+raionsu (68) jeans —+ jiinzu The example raionsu ‘lions’ reveals two things. First, the plrnal morpheme in the input to the Operative Level is [s]. The plural morpheme in ‘lions’ should not be devoiced, because the constraint No—szw, which triggers the devoicing, does not apply 127 . _ ,. .. . O. t It: a. - II II‘ ... t t’ _ r. e‘ r O o w l .... 0 .Or ...: ‘a/ to this word. The fact that the plural morpheme is realized as su in raionsu ‘lions’ indicatcsthattherealization oftheplural morphemeintheoutputintheoldergeneration isnottheresultofdevoicingbutfaithfitltotheinputtotheOperative Level withrespect to the feature [voice]. There are two possible explanations for the [s] in the input to the Operative Level. One explanation is that the plm'al morpheme in ‘lions’ in the input to the Operative Level is [3] because the older generation does not have access (or has limited access) to the morphological information of English. The plural morpheme always appears as ‘-s’ in English orthography irrespective of the pronunciation. In the older generation, the plural morpheme is realized as [s] in the input to the Operative Level probably because they know how it is written in English but they do not know how it is pronounced in English. The other explanation is that the realization of the plural morphemeintheinputtotheOperative Levelisduetotheperceptim'l‘heperceptual ability can be different between generations, since the younger and older generations have different set of phonemes (or possible syllables). For example, the younger generation have the distinctions such as ‘di‘ vs. ‘Ji', ‘ti' vs. '6i', and 'rh' vs. 'h', but the older generation do not. That is, the younger generation can have better perceptual ability (including the perceptual ability with regard to the voicing contrast), since they have more phonemes. It is not easy for Japanese speakers to perceive the word-final [2], since the word-final position is less salient position (Jun 1995, Beckrnan I998, Lombardi 1999, amongoflrem)mdcodacmrsonantsarehiglflyresnictedin1apanese.0neofthe explanations above or both ofthem might be relevant to the [s] in the input to the Operative Level. 128 I‘. t;_ .‘IIO \ Second,theconstraint ‘NCdoesnotapplytotheplural morphemeintheolder generation.Asshownintheprevioussubsection, ‘NC appliestotheitemsintheYamato sublexicon (Ito and Mester 1995, 1999, and others) as well as the English plural morpheme in the younger generation. But, ‘NC does not apply to the pltnal morpheme in words such as ‘lions’ in the older generation and the plural morpheme in ‘1ions’ is realizedassuintheoutput.Asillustratedin(69)and(70),‘NCdoesnotapplyto voiceless Obstruents immediately following a nasal in stems and compounds in the Loanword sublexicon. English loanwords in the older generation such as raionsu ‘lions’ medflmdrereismdisfincfionbaweenthestemmdfltesuffixmflieoldergamafion with respect to the application ofthe constraint ’NC and that the application of the constraint ‘NC has been changing over generations. (69) ‘NC: Ineffective in Stems in the Loanword Sublexicon (= (12)) a. rantan l‘randan ‘lantern’ b. atoranta *atoranda ‘Atlanta’ c. tento ‘tendo ‘tento’ (70) ‘NC: Inefl‘ective in Compounds in the Loanword Sublexicon a. man-pawns I'man-bawaa ‘man power’ b. ten-kaunto ’terhgarmto ‘ten count’ c. griin-kaado ’griin-gaado ‘green card’ The next question to be considered is: Why are the older and younger generations different with respect to the realization of the English pltn'al morpheme? That is because 129 ’ | ¢ , ~- 1 .tr. fit? I ')t. D! ...- ... they are difi‘erent with respect to the access to the morphological information of English: the older generation does not have access (or has limited access) to the morphological information of English, while the younger generation has access to it. In the older generation, as discussed above, the plural morpheme in words such as ‘lions’ is treated as a suffix, while the plural morpheme in words such as ‘shoes’ is not treated as a suffix but asapartofasinglewordinthe inputtotheOperative Level. Intheyoungergencration, mtheotherhmdflreloanwordswithflreplmalmorphemeisalwaysueatedas‘stem+ wflix’,asdiscussedintheprevioussubseaimh1heideathattheoldamdyomga generations are different with respect to the access to the morphological information of EnglishmightberelcvanttothefactsthatmostofthenewwordsinJapaneseareEnglish loanwords(Kay l995)andthatthosewordstendtobeusedbytheyormgermeration. Thediachronic changewithregardtotherealizationofthephrral morphemesuggeststhat the accessibility to the morphological information of English has been changing. 5.3.2.3. The Deletloa of the English Plural Morpheme This subsection discusses the deletion of the English plural morpheme. In the examples in the previous subsections, the English plural morpheme always has a corresponding segment in the output. In exampr such as those in (71), however, the plural morpheme does not have a counterpart in the output. 130 r o I I r I) a r . l . a el- - 9a. 4 r l - 5 a r. I ‘ . r t. ' I . I I r I. l ._ .r I r a. r. r I t \0 ll . It il Ir r D (1 l av . ..aa . n . rt I \ s . .I t a \II a J . . a I I . i 1. l , e I a 4. ts ' ‘ - .. It; . i. la y‘ r. , a (71) -z ——» 0 a. roodo obu za ringu ‘the Lord of the Rings’ (movie title) b. fiirudo obu doriimu ‘Field of Dreams’ (movie title) c. atakku obu za kiraa tomato ‘Attack of the Killer Tomatoes’ (movie title) Why is the plural morpheme deleted in (71)? And what is the difference between the examples in (71) and those in the previous subsections? The examples in (71) have two common features: (i) these are phrases containing function words such as the preposition ‘of or the definite article ‘the’, and (ii) the plural morpheme is in the phrase-final position. These common features suggest that Japanese borrowers distinguish the word- level and the phrase-level and that the deletion of the morpheme is a phonological phenomenon at the phrase-level. This idea is supported by the examples in (72). (72> rings a. rings --v ringusu (*ringu) (word-level) b.theLordoftheRings -+ roodoobuzaringu (phrase-level) (‘roodoobuzaringuztn’roodoobuzaringusro In (72a), the plural morpheme in the word ‘rings’ is devoiced in Japanese. In (72b), on theotherhand,theplural morpherneinthephrase‘theLordoftheRings’doesnot correspond to zu or 31: but it is deleted. Syntactically, the word ‘rings’ in (72a) can be a phrase by itself. However, the treatment of the plrnal morpheme in Japanese indicates that Japanese borrowers consider it as not a phrase but a word. The examples above mggestfliatthefimctionwordisthecuetodisfinguishwordsandphrases.'Ihatis, I31 If ,t on. S~O Japanese borrowers consider English phrases with function words as phrases and the phrases without them as words. The deletion of the plural morpheme at the phrase-level is observed only phrase-finally. As illustrated in (73), the phrase-medial plural morpheme is preserved in the output. (‘73) Phrase-Medial Plural Morpheme a. deizu obu sandaa ‘Days of Thunder’ (movie title) b. rediisu and jentorumcn ‘I.adys and Gnetlemen’ (song title) Within the OT framework, the difl‘erence between the word-level and the phrase- levelwithrespecttothetreatmentoftheEnglishpluralmorphemecanbeexplainedby the four faithfulness constraints in (74) - (77): DEF-100’}! APT), DEF-IO, MAX-IO, and I- CONTIG. As illustrated above, both syntactic categorization, i.e. phrase or word, and morphological categorization, i.e. stem or affix, are relevant to the deletion of the English plural morpheme. This suggests that the deletion of the plural morpheme is triggered by DEP-IO(PH APP) in (74), which is the conjunction of the constraints DEP-IO(PH) in (78) and DEP-IO(Ar-'F) in (79). The plural morpheme in the input cannot appear in the output byitselfbecausethecodaconsonantsarehighlyrestrictedinJapanese.Toprescrvethe plural morpheme in the input, a vowel needs to be epenthesized. But, since DEP-IO(PH AF?) prohibits vowel epenthcsis in affixes at the phrase-level, the plural morpheme can be deleted at the phrase level. 132 (74) DEP-IO(PH AFF): No epenthesis of segments in affixes at the phrase-level. (7S) DEF-IO: No epenthesis. (76) MAX-IO: No deletion. (77) I-CON'nG: No word-medial (or phrase-medial) deletion. (78) DEP-IO(PH): No epenthesis of segments at the phrase-level. (79) DEP—IO(AFF): No epenthesis of segments in affixes. The four constraints in (74) -— (77) are ranked as follows. First, the ranking between MAX-IO and DEP-IO is MAX-IO >> DEF-IO. In Japmese, the sequence of consonants md coda consonants is highly restricted (Ito 1986, 1989, among others). But, consonants in the input are preserved in the output by epenthetic vowels. This indicates that MAX-IO outranks DEP-IO. Second, the constraint DEP-IO(PH APP) is ranked higher than MAX-IO and Dar-IO. In Japanese, as mentioned above, the segments in the input are preserved in the output. But the English phnal morpheme in the phrase-final position is an exception. The English plural morpheme, which is an affix, is deleted when it is in me phrase-final position. The constraint DEP—IO(PH AFF) outranks MAX-IO and the deletion of the plural morpheme at the phrase-level is triggered by DEP-IO(PH AFF). Finally, the constraint I- CONTIG is ranked higher than DEP-IO(Pl-I AFF), since no deletion is observed word- or phrase-medially. In sum, the four constraints are ranked as in (80)." This ranking accounts for the difl’erence between the word-level and the phmselevel with respect to the treatment of the plural morpheme, as illustrated in (81) and (82). In the tableau (81), the phnal "rhecoratrtintsnsr-Iqru)mdner-10(Anv)nenotshownintrrerankingsincetheyrronotpray significantrolesinselectingtheoptimalmrtput. 133 ..re l‘la morphemeinthe inputispreservedintheoutputbecause oftheconstraintMAx-IO. In the tableau (82), on the other hand, the plural morpheme in the input is deleted in the output, since Japanese borrowers consider the input ‘the Lord of the Rings’ as a phrase and the plural morpheme in it cannot have a corresponding segment in the output due to the constraint DEP-IO(PH Arr). (30) (81) Tableau for ‘rings’ (word-level) . I- Dep-IO MAX- Dep- “ng‘ Comm (Prim) IO IO 3. ringu It! a b.<' ringu-su “ (82) Tableau for ‘the Lord ofthe Rings’n (phrase-level) . I- Dep-IO MAX- Dep- th“ 1”“ “the ““38 Como (PHAFFL IO IO a. 6' roodo obu za ringu "' *” b. roodo obu za ringu-su I"l "” "Thedeletionof‘the’isigmmdinthistahleatlnfliephrase‘theLordoftheRings’,mefimcfionword ‘the’inthephrase-initialpositionisdeletedintheormrtwhile ‘the’inthephrase-medialpositionisnot. Thisalsomightberelevanttothecomtraintl—Como. 134 The ranking in (80) also accounts for the realization of the phrase-medial English plural morpheme as illustrated in (83). In the tableau (83), the candidate (83a), where the phnal morpheme in the input is preserved, is the optimal candidate, since the candidate (83b), where the phnal morpheme is deleted, violates the highest—ranked constraint I-CONTIG. (83) Tableau for ‘Days of Thunder’ (phrase-level) I- Dep-IO MAX- Dep- Day’ “mm“ Cosmo (PHAFF) 10 10 a. 6' deizu obu sandaa " " b. dei obu sandaa *l " ‘ The next question to be considered is: Why is the English phnal morpheme deleted at the phrase-level? I believe it is deleted to avoid ‘long’ phrases. When foreign words are borrowed into the Japanese lexicon, since the sequence of consonants and coda consonants are highly restricted in Japanese, vowels are epenthesized to preserve all segments in the input. But, it is not economical to pronounce prolonged words with epenthetic vowels. As a result, long words tend to be abbreviated at the word-level, as will be discussed in the next chapter. The deletion ofthe plural morpheme is another example at the phrase level of avoidance of prolonged phrases (and words). The deletion of the plural morpheme at the phrase-level occurs only phrase-finally, because the final position is the least marked position for phonological changes crosslinguistically (e.g. wordofrnal deletion of English lr/ in Japanese discussed in Chapter 4, and word-final devoicing in German and Russian). At the phrase-level, only the phrase-initial definite articles and the phrase-final plural morpheme can be deleted. This suggests that only semantically-less-significant items in phrase-edge positions can be deleted. 135 I III... s V‘ a 4 .. ... .2 l r I i . fit I, ’ l r . I.» I O u . 4 ( . l t l , t i r r )5 «I .‘ .) .7 ... . . n ‘ - a , .\l. f I V . I 0 5.4. Conclusion The present study has discussed three phenomena related to the realization of the English plural morpheme in Japanese within the OT fiamework. The present study is summarized as follows. First, the devoicing of the English plural morpheme is a phenomenon within the affixes in the Loanword sublexicon. Previous studies claim that this devoicing is a phenomenon in the Yamato sublexicon or in affixes. However, those studies do not explain why only the English plural morpheme undergoes devoicing in Japanese. Also, since devoicing is highly restricted in Japanese, it is plausible to assume that it occurs in the periphery of the lexicon, i.e. in the Loanword sublexicon, where more constraints can be violated. This chapter revealed that both etymological ami morphological categories arerelevanttothisdevoicingandthatthedevoicingisaphmomenonthatcanoccuronly in the amxes in the Lomword sublexicon. Second, Japanese borrowers have access to the morphological information of EnglishwhentheyborrowEnglishwordsintotheJapanese lexicon. Asdiscussedinthis chapter, the devoicing is observed only in the affix, which is a morphological category. This indicates that Japanese borrowers have access to the morphological information of English. This study revealed that Japanese borrowers distinguish the English plural morpheme from the stem. But it is not clear yet how they treat other English amxes in borrowing. I leave this question for future research. Third, the accessibility to the morphological information of English has been changing over time. The older and younger generations are different with respect to the realization of the English phnal morpheme. In the older generation, the English plural 136 . a r, I p . (. r s a . . r - as . . . E: u ‘ a,ls p morpheme is always devoiced when it is in the word whose singular form also exists as a loanword in the Japanese lexicon. But it is faithful to the input when the English plural morphemeisinawordwhosesingular formdoesnotexistintheJapanese lexicon. lnthe younger generation, on the other hand, the plural morpheme can be devoiced only in a wordwhoseinputcontainsmorethanonevoicedobsu'uent. Thischapterrevealedthatthe generationgapwithrespecttotherealizationoftheEnglishplmalmorphemeisdueto thefactthattheoldergmerationdoesnothaveacwss(orhaslimitedaccess)tothe morphological information of English and the younger generation has access to it. The diachronic change with regard to the realization of the plural morpheme suggests that the accessibility to the morphological information of English has been changing. Next, in the older generation, the phnal morpheme in words such as ‘lions’ is realizedas[s]intheinputtothe0perative LeveLTheplm'almorphemeintheword ‘lions’ appears as su, although it is not in the position where the constraint No-D’n applies. This is because su in the output is not the result ofdevoicing but it is faithful to the input to the Operative Level. Also, the fact that the constraint ‘NC, which applies to the English plural morpheme in the younger generation and stems in the Loanword sublexicon, does not apply to ‘lions’ in the older generation indicates that there is no distinctionbetweenthestemandthesuffixintheoldergenemtionwithrespecttothe application of ‘NC and that the application of l’NC has been changing over generations Finally, Japanese borrowers distinguish the word level and the phrase level when they borrow English words into the Japanese lexicon. The plural morpheme does not have a counterpart in Japanese when it is in the phrase-final position in English, while it always has a corresponding segment at the word-level. This suggests Japanese borrowers 137 distinguish the word-level and the phrase-level.” The English plural morpheme is deleted at the phrase-level because of the constraint DBP-IOG’H AFF). This chapter focused only on the deletion of the plural morpheme and did not discuss the deletion of other elements such as the definite article, which is also observed at the phrase-level. The phonological differences between the word-level and the phrase-level would be an interesting topic for future study. ‘3 The distinction between the word-level and phrase-level phonology, i.e. the distinction between lexical phonologyandpostladcalpbonologyahoplaysssignficamMeinChaptew. 138 6. English Compound Abbreviation 6.1. Introduction This chapter explores English compound abbreviation in Japanese within the fiamework of Optimality Theory (01‘: Prince ml Smolensky 1993).l There are three possible ways of abbreviation: (i) file first element is retained as in (1), (ii) the second element is retained as in (2), and (iii) both elements are partially retained as in (3). (l)a. mini sukaato --» mini ‘mini slcirt’ b. konbiniensu sutoa -—t konbirli ‘convenience store’ c. shorudaa baggu -—+ shorudaa ‘shoulder bag’ (2) a. gooru kiipaa --t kiipaa ‘goal keeper’ b. fasshon moderu -> modern ‘fashion model’ c. nyuusu kyasutaa —-t kyasutaa ‘news caster (anchor)’ (3) a. waado purosessaa «t was-pure ‘word processor" b. dejitaru kamera —-t deji-karne ‘digital camera’ c. rimooto kontorooru —+ rimo-kon ‘remote control’ 'Commmdmmmlmnamoommmmmonmmamz) Muhabhevhfimfinon-wmpomdbmwmdsmhpmescandfindsmnapowpemdm generallyoccursjustbeforetheaccentedrnoraofsbase.Aswillbeshowninthillchapter,however,thisis nottllecaseforEnglishcompomdabbreviationinJapanese. 139 Among them, thepattern in (3) isthe mostproductive inJapanese (Nishiharaetal. 2001), whereasthepauanin(l)ismostproducfiveinlanguagessuchastch(Hamans1997: 1734), English (Szymanek 1989), French (Niikura et al. 1996: 384), and German (Wiese 1996: 62-63). This chapter deals with the pattern in (3) and discusses how abbreviated compounds of this type are formed. The goals of this study are fourfold. First, this chapter explains how u‘imoraic words can be formed in English compormd abbreviation. The most productive abbreviation pattern in Japanese is quadrimoraic consisting oftwo bimoraic components. But trimoraic wordsmeflsopossiblemddluemefllmctypesofuimoraicabbreviatedwmpmmds. The present study accounts for them within the OT framework. Thesecondgoalistoarguefllatabbreviatcduimoraicwordscanappearinflle outputasaresultofassimilationtothecorepartofthelexiconwiflrrespecttothe constraint ‘[a:]]pw. A previous study (Nishihara et al. 2001) claims that abbreviated English compounds become trimoraic as a result of the word-final vowel shortening and thisisaphenomenoninthe Loanwordsublexicon. But,basedontheresultofadictionary research, this study claims that it is due to *[a:]]pw, which plays a significant role in the core part of the Japanese lexicon. Third, this study reveals that English compormd abbreviation in Japanae involves a case of phonological opacity, which is similar to the opaque case observed in German truncation. Thistypeofcompound abbreviationinJapanesehasnotbeendiseussedinthe literature. The present study accounts for the opaque case within the fiamework ofthe Weakly Parallel Model, a subtheory of OT, proposed by MS and Mester (2001b and 2003s). This model explains the opaque case in English compound abbreviation ill 140 . Do . ‘ . r. s l. ' .l '. ’ ‘ t ' i ' l ' a v, . . y . t "‘ J l I" I " I‘ . . t i ' . , ‘ I I J V l.‘ l .‘l I. I t-‘V ' i 3 y . r! . " 'i' H i 0‘ - 0‘ ~ ' ’ i ‘ ‘II 01 ' r‘l “‘I . I , I i .: -. " ‘t ' . 5 s i‘" g '. Japanese straightforwardly. This chapter shows how successfully a case of phonological opacity observed in English compound abbreviation in Japanese can be explained within the framework of the Weakly Parallel Model. Finally, drischapterdiscussesthatthaemtwotypesofcodawndifionsinhpanese. mulefiteraune,onlyoneofthemhasbeendiscussed8ugflleopaquecaseandvab stemsmvealthatthereamacmallytwotypesofcodacondifionsmdthatthecoda condition discussed by 1th (1986) applies only to prosodic words while the other type of coda condition applies to both bound morphernes and prosodic words. This chapter is organized as follows. After introducing two subtheories of OT, Sympathy Theory (McCarthy 1999) and the Weakly Parallel Model in section 2, section 3 presents the data. Section 4 overviews three previous studies relevant to English compormd abbreviation in Japanese. Then, in section 5, [analyze four patterns of English compound abbreviation within the OT framework. Finally, the present study concludes in section 6. 6.2. Sympathy Theory vs. Weakly Parallel Model This section introduces two subtheories of OT, Sympathy Theory (McCarthy 1999) and the Weakly Parallel Model (Ito and Mester 2001b, 2003a), which are proposed to deal with phonological opacity within the framework of 01‘. 6.2.1. Sympathy Theory Sympathy Theory (McCarthy 1999) is a subtheory of OT, which is developed to cape with phonological opacity. McCarthy (1999) proposes a new type of 141 t‘ ‘t-l t I" Ill 1 l D . r ,. ’ A.“ it 4'... correspondence relation: ‘sympathy’. Sympathy is a relation between candidates. A candidatecanbemopfimaloutpmbecauseoffllesympathymlafionwidlapufiaflar failed co-candidate, i.e. a candidate that is optimal with respect to a specific lower-ranked constraint. McCarthy (1999) develops Sympathy Theory, discussing Tiberian Hebrew. Tiberian Hebrew reveals a case of phonological opacity, which is caused by the interaction of e-epenthesis and P-deletion: a vowel [e] is inserted between the word-final consonants whereas the [7] is deleted in the coda position. In derivational model, the case in Tiberian Hebrew can be explained as in (4). But, as illustrated in (5), this is problematic in the framework of OT, because it is not clear why the vowel [e] is epenthesizedinthe outputform,i.e. itisnotfollowedbyaconsonantintheoutput. Inthe tableau(5),theconstlaintrankingpredictsthewrongwinnerdd. (4) /ded?/ -t [dese] Input: Ides” eepenthesis dose? I P-deletion 00W: [Me] (5) Tableau for /de§?/ —-t [dose] (Wrong Prediction) Input: /de§?/ Coot-Conn; 'Comrex MAX-C Dar-V a. dei? ‘1 "' b. deter ‘! ‘ c. 8 des " d. dede : s s! 142 I o l . 7.? . ll . o. A J n 0" I, I .Il ' h a 3 t 01 I. .4 u a. v I ' a! .0 I‘lv. " I cl . I ' \ .I u McCarthy (1999) claims that the candidate date is selected as the optimal output because it is in the sympathy relation with the candidate date}: which is the intermediate form in derivational model. The candidate date? cannot be the optimal output because of the violation of CODA—Corn). But it serves as the sympathy candidate. The sympathy candidate is selected among the candidates that satisfy the selector constraint MAX-C. In the tableau (S), the possible candidates for the sympathy candidate are (5a) and (5b). The candidate dese7in (5b) is selected as the sympathy candidate, since it best satisfies other constraints. 'I‘hen,thecandidatedese is selectedastheoptimaloutptnbythesympathy constraint, i.e. the faithfulness constraint between a candidate and the sympathy candidate. In the tableau (5), the output form date and the sympathy candidate date? have the epentheticvowelbuttheothersdonot.Thissuggeststhatthesympathyconstraintis MAX-OO-V, which requires that all vowels in the sympfly candidate be preserved in theoutput form. Thisconstraintisranked higherthanMAx-C. Thisconstraintranking selects the optimal output defe, as illustrated in (6). In the tableau (6), the candidate (6c), the optimal output in the tableau (5), is ruled out by the sympathy constraint MAX-CON. (6) Tableau for /de§?/ ... [dese] Input: mam Com-Conn; 'Comax {Mix-eov MAX-C Dar-V .. a... n - . b. dese? ‘! ‘ c. deg a! e d. a dese a *2 143 I" 6.2.2. Weakly Parallel Model The Weakly Parallel Model (ltd and Mester 2001b, 2003a) is another subtheory of OT to handle phonological opacity. In this model, “the traditional distinction between lexical phonology and postlexical phonology persists as a serial interface between two separate modules of grammar” (Ito and Mester 2003a). Lexical phonology and postlexical phonology are characterized by the three properties in (7) (cited from ltd and Mester (2003a)). (7) a. The lexical and postlexical modules constitute separate constraint systems. b.'Ihey share many (not my an)2 constraints, but rankings can differ in limited ways. c. The two modules interact serially, with the output of the lexical module serving as the input to the pestlexical module. ltd and Mester (2001b and 2003a) argue against Sympathy Theory and develop this alternative model. Their argtunent against Sympathy Theory is that it is difficult to reconcile with a core tenet of OT, i.e. Richness of the Base, and is problematic for the treatment of phonological opacity. In Ito and Mester (2003a), they argue against Sympathy Theory as follows. 1"nunnantmtmmammwormcmmanmgrmmmwmduor evaluation, are literally not part of the lexical module (Ito and Mester 2003a).” Ito and Mme: (2003a) findermemOb):“[W]edwdondnchflemepmdbifitydmmeposflexkdmwmmigMbe quitedifiermtincharacterfromthelexicalone...Forexample,onlythelattermightunnouttobeastrict OTeyuemwhammefmmumigMbeamonhoadlyopfimimfiombudqumfimfiwsynem.”Mmy quufimsngudingmenuunmdsmmofthelexicdmodtdehawnmbeenmmdfis dissertation,however.lwiflignorethemandleavedremforfimueresearch. 144 ('l (,1 Japanese reveals a case of phonological opacity, which is caused by the interaction of two well—known processes: Rendaku and g-weakening. Rendaku is a phonological process observed in compormds that replaces voiceless obstruents with their voiced counterparts at the beginning of second members. But, it is systematically blocked whenever second members already contain a voiced ownent (e.g. hana + sono --» hanagono ‘flower garden’, furi + sode -v furigode (‘furigode) Tong-sleeved kimono’). The other process g—weakening is an allophonic process where non-initial [g] is replaced by [0] (es /sai/ -’ [nail in [koku + nail abroad’ V8- word-initial [sail in [sai + fin] ‘foreigner’). The properties of g-weakening, i.e. phonetic gradiency, non-contrastiveness, and sociolinguistic variation, suggest that this is a postlexical process. In derivational terms,theinteractionoftheseprocessescanbeshownasin(8): In(8b),theseprocesses are in a counterfeeding relation and the reversed order of application would produce the wrong outpm ’[saka + dose]. (8) The Interaction of Rendaku and g-weakening a. ‘folding paper’ b. ‘reverse thorn’ /ori + kami/ /saks + toge/ Rendaku ori gami n/a g-weakening ori gami saka tone [cti Hamil [8813 mm] InOTterms,relevantconstr-aintrankings formeseprocessesaretherankingsin(9) and(10):theranldngin(9)isforRendaku whiletheranltinginOO) is fag-weakening. 145 l . I. "r '. ‘1 I ' .- )1‘ -l o " .b In. ,’ . ' ‘i :‘K! s (9) Constraint Ranking for Rendaku (Based on ltd and Mester (1998)) Ranking Definitions OCP(vor) OCP-type markedness constraint, here understood as ruling out multiple obstruent voicing within a stem REALMORPH RealizeMorpheme (here, enforcing the realization of the compound voicing morpheme) IDENT-IO(V01) Faithfulness constraint militating against changes in voicing (10) Constraint Ranking for g-weakening (Based on ltd and Mester (1997a, c)) Ranking Definitions ‘mvdm Positional markedness constraint against PrWd-initial r3 ‘g Markedness constraint prohibiting voiced dorsal Obstruents lDENT-IO(NAS) Faithfulness constraint against changes in nasality The constraint ranking in (10) selects the optimal outputs regardless ofthe input status of lgl, as illustrated in (11) and (12). This will be a significant point in their argument against Sympathy Theory. Richness of the Base should allow either of them as a viable mptmbecausemtwosegrnentsdonotsmndmwnnastandmehdism’bufimis allophonically determined (ltd and Mester 2003a).” 146 J '7 (11) /g/ as input lgeta/ ‘clogs’ ‘nmm I’g [DENT-10(NAS) a. <7 geta " b. nets *1 "' W ‘ch’ ’thDJ ’8 IDENT‘IWNAS) a. kagi *1 b. 3 kani " (12) In’ as inpm lgeta/ ‘clogs’ "nwaln *g IDENT-IO(NAS) a. (7 gm . O b. nets ‘! /kagi/ ‘key’ *rtwaln ‘ g IDENT-IO(NAS) a. kagi t! e b. <5" karJi Theconstraintrankingsin(9)and(10)canbecombinedasin(13).Therankingin (l3) accounts for the phonological process lori + kami/ ... [ori gami] ‘folding paper’ in (8a), but not Isaka + toge/ -» [saka tone] ‘reverse thorn’, as illustrated in (14) and (15). 147 u— >..n V .-. p n , .. ~~-. .‘Ou’ - ”Mr. 0 - " ‘ r. ' u M ‘. ' Hi‘ ‘ ' I ’ fl . . ' ~ ‘ l ‘ V l , ‘ ‘ .i ‘ ’ ' . I ‘7 ‘ .I‘ . I l- ‘ - I ' T Q I . s & -- . O .0 - ..- .‘.a. a.“I.-‘”‘ . .w.O.-‘.- .. A d new a -" ' ' ‘ 0' H, 'J tn“. 1 A ‘ 3 .tr - It .I I I, I ‘ v. ‘ “‘ v - 1‘ .‘ I r v av - . ~-, .- '. . - ‘ ‘ _ _ - ~. . ; .. t ’ t . .. . " 'n ' 'L ' l" 4 r i q ‘ f- I 1 ’ '- 1 I I' ' . ‘ : I. ’4 1 (13) ’rwal‘l OCP(vor) RIISALMORPH ‘s IDENr-IO(NAs) IlJENT-IOWOI) (14) /ori + kami/ --> [ori gami] ‘folding paper’ [ad-W ‘meIJ OCP(vor) REAL ‘8 [DENT-IO IDENT-IO Mom (NAS) (vor) a. ori-kami ‘! b. ori-gami ‘1 ' c. 6' iori-nami "' " (15) Isaka + toge/ -» [saka tone] ‘temse thorn’ (Wrong Prediction) REAL tum-IO [DENT-IO Isaka-toge/ ‘mvdlll 0CP(VOI) ‘s Moapu (ms) (vor) a. saka-toge ‘! * b. saka-doge ‘! ' " c. saka-tone t! e d. ®saka-dor3e ‘ "' Theranldngin(l3)doesnotaccountfor/saka+toge/—+ [sakatone],becausethisisa case of phonological opacity. With the help of the sympathy constraint IDENT-OOWOI), which is ranked higher than REALMORPH, the phonological process lsaka + toge/ -» [saka I48 tone] can be explained as in (16) where the selector constraint and the sympathy candidate are marked by “i”. (16) Isaka + toge/ -+ [sake tone] ‘reverse thorn’ Isaka-toge/ ’rmltj OCP *O~ REAL '8 10- IO- (vor) (vor) Mom mas)" (vor) a. Osaka-toge "' ’l b. saka-doge ‘! " " ‘ c. ”sake-tone " " (1. mm at! e e The Sympathy-based analysis in (16) seems to accounts for [sake tone] successfully. But, itcontainsaseriousproblem. Thatis, itpresupposesthattheinputis/saka-toge/but not lsaka-tonel. As mentioned above (see tableaux (l 1) and (12)), the surface distribution of[g]and[n]ispredictedbytheconstraintsystem.ThatmeansacoretenetofOT,i.e. Richness oftheBaserequhesthat‘thermkingofoutpmconsnamtsaloneberespmsible for the derivation of the distribution of the two variant” (ltd and Mester 2003s). As illustrated in (17), however, the Sympathy-based analysis does not explain a possible input variant leaks-tonel. Based on this, ltd and Mester (2003a) conchrde that “Sympathy cannotcopewiththerichinputsdemandedbyRichnessoftheBasewheneverthe masking process of an opaque interaction is allophonic” and propose the Weakly Parallel Model to handle phonological opacity. 149 (17) lsaka + tone! -+ [saka tone] ‘reverse thorn’ (Wrong Prediction) lsaka-tone/ .Hmm OCP *0- REAL ’8 10 IO- (vor) (vor) Mom (NAs)' (v01) & saka-toge e! e e e b. saka-doge ‘! " " "‘ c. saka-tone ‘l ‘ d. ®saka—done * In the framework of the Weekly Parallel Model, the case of [saka tone] can be accounted for as illustrated in (18) and (l9).3 In this model, the lexical and postlexical modules constitute separate constraint systems, where rankings can differ in limited ways, andthetwomodulesinteractserially,withtheoutputofthelexicalmoduleservingasthe input to the postlexical module (see (7)). In (18), the ranking for the lexical module produces saka-toge, which is the input to the postlexical modules. Then, in (19), the ranking for the postlexical module selects mica-tone as the optimal output. In (18) and (19), I’VgV, i.e. no intervoalic [g], and [DENT-10(vor), i.e. no changes in voicing, are the constraints that are ranked differently in the lexical and postlexical modules. ’ ltd and Mester (2003a) employ the constraints 'n and *ng instead of 5mm, which is adopted in their earlier study. 150 Q... .1! (18) Isaka + tone! -+ [saka tone] ‘reverse thorn’ (lexical) Isaka-tone/ .0 + OCP REAL *ng ‘8 IO- 10- (vor) MORPH (NAs) (vor) a. asaka-toge a e e b. saka—doge - *1 " ' ‘ c. saka-tone ‘! " ‘ d. saka-done ‘! ' (l9) Isaka + tone! -t [saka tone] ‘reverse thorn’ (postlexical) IO- OCP REAL 10 “(Hose ‘ng (vor) ‘9 (var) Mom .8 (ms) a. saka-toge *! " ‘ b. saka-doge ‘! ‘ " c. <"salts-tone " ‘ " d. sake-done *! ‘ " 6.3. Data The majority of abbreviated loanword compounds in Japanese are four-mora long, consisting of two bimoraic components (ltd 1990, Nishihara et al. 2001, Labrune 2002), as shown in (20). But, trimoraic words are also possible and have been becoming more common (Nishihara et al. 2001). The trimoraic words fall into three groups ((21) - (23)). (20) (= (3)) a. waado purosessaa -+ was-pure ‘word processor’ b. dejitaru kamera -> deji-kame ‘digital camera’ c. rimooto kontorooru -+ rimo—kon ‘remote control’ 15] (21) a. dansupaatii -> dart-pa “dance party’ b. furii maaketto -> furl-ma ‘fiea market’ c. terefon kaado -> tere-ka ‘telephorle card’ (22) a. buraddo pitto -> burs-pi ‘Brad Pitt’ b. potato chippu -—+ pate-chi ‘potato chip’ c. netto sukeepu --> ne-suke ‘Netscape’ (23) a. purasutikku modern -> para-mo ‘plastic model’ b. roiyaru hosuto --> roi-ho ‘Royal Host (a restaurant chain)’ c. burakku bisuketto ... burs-bi ‘Black Biscuit (a band name)’ The three types of trimoraic words have the following common features. In (21), the unabbreviated form of the monomoraic component has along vowel in the first syllable (e.g. mtii (21a)). In (22), the unabbreviated form of the monomoraic component has a coda consonant in the first syllable (e.g. pitto (2211)). In (23), the second syllable of the unabbreviated form of the monomoraic component starts with an obstr'uent (e.g. madam (23a)). 6.4. Previous Studies This section overviews three previous studies on abbreviation: ltd (1990), Nishihara et al. (2001), and ltd and Mester (1997b). 152 6.4.1. ltd (1990) ltd (1990) is the first comprehensive study about loanword abbreviation in Japanese. This study deals with compound abbreviation, as well as non-compound abbreviation, and reveals that the majority of abbreviated loanword compounds in Japanese are four-mora long, consisting of two bimoraic components, as illustrated in (24). (24) Typology of Loanword Compound Abbreviation (ltd 1990: 221) u-Pattern Count Example 2s= [lit+ In] 4 Heathen!) ‘base unpay mise’ 3n = [2n + 1n] 4 dore(su) me(ekaa) ‘dressmaker’ 4"” ”“2” 134 wkdetowoomo) ‘skawboard’ ltd claims that individual members of the compound cannot be monomoraic because ofMinimal StemRequiremerlehichrequiresthattheminimalprosodicstembe bimoraic. But, this study does not explain why trimoraic forms such as dorenle can be selected over quadrimoraic ones consisting of two bimoraic components, dorm for example. 6.4.2. lehlhara et al. (2001) Nishihara et al. (2001) argue against headedness in compound abbreviation and develop a phonological accomt for compound abbreviation. Nishihmetal.inuoducethreetypesofcompoundabbmviafioninJapaneseand showthatdoubleMmcafionismostpmducfivepaueminJapanese.Amongdlethree 153 patterns in (25), Nishihara et al. account for file pattern (25a) within the framework of OT, as shown below.’ (25) Nishihara et al. (2001: 308) a. double truncation (47%) maikuro konpyuutaa -> mai-kon ‘micro computer’ sekusharu harasumento --9 seku-hara ‘sexual harassment’ purofeshonaru resuringu -+ puro-resu ‘professional wrestling’ pawaa sutearingu --> pawa~sute ‘power steering’ afutaa rekoodingu .... afu-reko ‘post-recording’ b. back truncation (39%) paato taimu "" paato ‘part-time JOb’ homo sekusharu -+ homo ‘homosexual’ bideo dekki —. bideo ‘video deck’ tekisuto bukku —-> tekisuto ‘textbook’ shorudaa bakku -+ shorudaa ‘shoulder bag’ 0. front truncation (14%) nyuusu kyasutaa ——r kyasutaa ‘news caster’ mootaa haiku -+ bdku ‘motor bike’ ’fasshonmoderu -. moderu ‘fashion model’ kafe oore -+ oore ‘café an lait’ ‘ Nkhihraaddgnmefiommnwafimbecmutheabhefiuedlmwmdsofmiswpememodya mmpacmmgeofmewmlnumbaofmaedforms”and“swmlascemdmaprowdicmm” (Nishiharaetal.(2001: 310)). Nishiharaetal.donotexphin explicitly why back-tnmcatcdfirrmssuchas paarocanbeselectedoverdouble-ulmcated formssuchaspaa—tai.Blnitseemsthattheyassumethe preferenceforbackulmcafionoverdoubleumcafionislexicallydetermined 154 ‘h I: ... \I_|_' .’ V . )1 Most of the double-tnmcated compounds are four-mora long consisting of two bimoraic components as in (25s). The quadrimoraic form is selected by the constraints in (26) — (28), which are ranked as in (29). The constraint ranking in (29) predicts for the correct output, as illustrated in (30). (26) MmWD: Truncated outputs are minimal words, i.e. bimoraic. (27) PARSB-PRWD: Even a small fraction of every prosodic word which composes the compound mustbeparsed intothetrrmcatedoutput. (28) LEFTMOST: The leftmost element of constituent is retained in truncation. (29) Lsmosr >> MINWD, PARSB-PRWD (30) Tableau for ‘pawaa sutearingu (‘power steering’) -+ paws-sute’ pawaa sutearingu Lsrmosr MINWD iPARSB-PRWD a. 6' pawa-sute r b. was-sute *! c. paws-su ‘! d. paws "! Next, Nishihara et al. explain two types of trimoraic abbreviated compounds in (31) and (32) as follows. (31) a. buraddopitto b. potato chippu —+ burs-pi -+ pote-chi ‘Brad Pitt’ ‘potato chip’ 155 (32) a. furii maaketoo --> furi-ma ‘flea market’ b. depaato gaaru -> depaoga ‘department store girl’ c. haiwei kaado --> hai-ka ‘highway card’ d. terefon kaado --» tere-ka ‘telephone card’ e. misutaa doonatu —-> misu-do ‘Mr. Donuts (a donuts shop chain)’ ln(3l), theunabbreviated form ofthe secondcomponenthasacodaconsonantinthefirst syllable (e.g. pitto (31a)). But, the coda consonant cannot be retained in the output due to the constraint CODA—COND in (33) (ltd 1986). A possible repair strategy for the ill form likepitistoskipthesecondmmwhichresultsinpito.Butthisalsocannotbethe opfimaloutpmbecauseofdleconsuaimCONnGurrYin(34)(Mchfllyandence 1993a). In (32), the unabbreviated form of the second component has a long vowel in the first syllable (cg memo (32a)). But the long vowel cannot appear in the output due to the constraint NoFmALLonoVowsLs in (35). (33) CODA-COND: ‘Place] (34) CONTTGUITY: Segmental material that is contiguous in the input must also be contiguous in the output. (35) NoFmALLONGVowsLs (NFLV): ‘VWW m ’ ’ thihmfld.asnmedntabhwifledconmomdshlapmeuhkethefoflowhgmm.lkymfl the word-level PrWd, i.e. Prth, and PrWd2, as minorPrWdand the compound-level PI'Wd as major PrWd. [[‘Prde [PrWdfllhm (both PrWdt and PrWdz consist of two morae) 156 'O' ~'\\ 1“ y'. i 17"" I ‘ I l. 7 1v rd 1, '\ ,. : 1‘ ’. I ' I It ' i'..- t- . . I 3' l s a 3‘ . “ ,4 ’ ... I Nishihara et al. assume that the constraints in (33) - (35) are ranked higher than MINWD and PARSE-PRWD and pr0pose the constraint ranking in (36). The constraint ranking in (36) accounts for the trimoraic abbreviated compounds in (31) and (32) as well as the quadrimoraic ones, as illustrated in (37) - (39). (36) CODA-COND, Cormourrv, NFLV, Lsrmosr >> MINWD, PARSE-PRWD (37) Tableau for ‘buraddo pitto (‘Brad Pitt’) -+ bura-pi’ buraddopitto CODACD § Comm § NFLV § L-MOST MmWn i Purse a. <7 burs-pit ‘! b. burs-pi c. burs-pita ‘! (38) Tableau for ‘furii maaketoo (‘flea market’) —» furi-ma’ furiimaaketoo CODACD g Como g NFLV g L-MOST MINWD PARSE a. furi-maa *7 v— : :tg fi b. 6" furl-ma L '— I c. furl-make I'l (39) Tableau for ‘pawaa sutearingu (‘power steering’) -+ paws-sute’ pawaasutean'ngu ConACn g Como g J v NFLV g L—MOS‘I‘ MINWD g PARSE a. d" paws—sure T V I L b. was-sute ‘! c. paws-su ’! d. pawa 157 . .f b . I t t n . , o .. x v , t u .. o l a r l s t t l c , . I x .11 .r y a\ t .I l J n u L t I“ . . ‘t a all I | t I 1 l. I . .. a n a J. 1' ( . AI . h I w a r l r .v t I. . a . . VI .. d I a . ‘ l u (‘1 d t u t l . .I J . a t . l . \ r '. u r. I D. t u. s . I l | o . r . I . p A. t a v 9 . \ . . 4f . I . t . . l , t . s , .I‘ 's r . s Q I t (- r o . . t o r . . ll 1 t V t , a l I r t . l J l g . 1' cl . . l ... n s . ’l . n . t a. s Nishihara et al. introduce the third pattern of trimoraic abbreviated compounds such as those in (23) (repeated in (40)). But they ignore this pattern because “there is no phonological reason to avoid [quadrimoraic formsl” (Nishihara et al. (2001: 311)). (40) (= (23)) a. purasutildtu model-u .—. pura-mo ‘plastic model’ b. roiyaru hosuto -+ lei-ho ‘Royal Host (a restaurant chain)’ c. burakku bisuketto -+ burs-bi ‘Black Biscuit (a band name)’ 6.4.3. ltd and Master (1997b) ltd and Mester (1997b) adopt Sympathy Theory (McCarthy 1999) to deal with German truncations within the OT fi'amework. ltd and Mester introduce a productive pattern of truncation deriving hypocoristics andotherkindsofshortenings in(41)anddrawthegeneral form ofGermantruncationin (42), i.e. the whole first syllable and the first onset consonant of the second syllable of the base word followed by the sufix -i. 158 (41) a. Personal Names Gabriele Gabi Eva Evi Wildemdr Wildi b. Surnames Gdrbatschdw Gdrbi Kllnsmdnn Klinsi Wismeier Wési c. Common Nouns (mostly denoting persons) Alkohdliker Alki Prdletérier Prdli Erdkhode Erdi (42) The General Form of German Truncations o c A. A. CoV(C) C+i \___~____z frombase c 0 AA e.g.Gdrbatschdw -—-> 661' b+i In this study, ltd and Mester adopt the model of trrmcation in (43), which is a modified version of Benua’s (1995). This model of truncation accomts for examples like 159 Gérbatscho‘w -> Gérbi, as illustrated in (44). Constraints in (44) are defined as in (45) - (43)}5 (43) ltd and Mester’s Model of Truncation Input: lgorbadof/ Input: ITRUNC + i/ lO-Faithfulness 1 t BT-Faithfulness (44) Tableau for ‘Gdrbatschdw —. Gdrbi’ Base: [(.gor.ba).(ébw.)] MAX-IO ALL‘FT‘ pm” Max-ET Input: ITRUNC + i/ LE" 3. (.gdr.ba).(ddf-i.) ‘1 b. (.gdr.ba).d-i. " 9f c. 6’ (.gdr.b-i) “6f d. (.g6.r-i.) Wfl e. (.gdr.ba.) i! 6“ f. (.gdrb.) i! “‘5“ g. (.gdr.) 1! WM (45) MAX-IO: Every element in the input has a correspondent in the output. (‘No deletion’) (46) ALL-FT—LEFr: Every foot stands at the left edge of the PrWd. (47) PARSE-o: Syllables are parsed by feet. (48) MAX-BT: Every element in B has a correspondent in T. (‘No truncation’) ‘ ItdandMesterdonotdefinethosecomtraintsintheirstudy.Thedefinitionsin(29)-(32)arefroml(ager (1999). 160 The constraint ranking in (44) explains examples like Gérbatschdw «t Grit-bi but not examples such as Gdbriéle —9 Gébi. As illustrated in (49), the constraint making predicts the wrong winner Gdbrl'. This problem occurs due to the constraint MAX-8T. MAX-BT has the effect of maximally preserving elements in the base. This constraint copes with the word with a simplex onset in the second syllable of the base such as Gdrbatschdw but not with the word with a complex onset in the second syllable of the base like Gabriele. (49) Tableau for ‘Gdbriéle --> Gabi’ (Wrong Prediction) Base: [(.gttbri).(é-le.)] Mum “VF“ pm“ mm“ Input: ITRUNC + i/ LEFT a (.ga.bri).(é.le.)-i. ‘3 ’ h. (.gabri).(é.l-i.) ’1 ° c. ® (.gdbr-iJ iéle d. (.gab-i.) riéle! e. 0839-) i! riéle Gérbatrchdw and Gdbriéle are truncated as Gérbi and Gébi, respectively. Two generalizations can be drawn from these truncated words: (i) the bare truncatum, i.e. the tnmcated form without the suffix -i, is a possible syllable of German (see (50)) and (ii) the tmncatum is the maximal syllable abstracted from the base (see (51)). 161 I‘V (50) fa Gdrbatschbw --> Gdrbi A .gdrb. (a ‘0 Gabriele ->Gabi A A $35, .gabr. (51) 0’ , A Gdrbatschbw -’ 06be /.g6rb. (Goth-i) team» ('Gér-i) O . _ A Gabriele -+ Gabi l.gdb. (Gab—i) age. (rose) What is significant here is that the syllables [.gorb.] and [.gdb.) are not found anywhere in the input or the output in spite of the fact that they play significant roles in selecting the optimal output. In that sense, this is a case of phonological opacity. Ito and Mester (1997b) account for the case of phonological opacity by adopting the expanded Sympathy Theory. Ito and Mester expand Sympathy Theory by removing the stipulation made by McCarthy (1999) that the constraint that selects the sympathy candidate must be afaithfulness constraint. In Ito and Mester (2001b and 2003a)’, however, they argue against Sympathy Theory and propose the Weakly Parallel Model, since Sympathy Theory is difficult to reconcile with a core tenet of OT, i.e. Richness of the Base, and is problematic for the ’mmwmmmwmmmmmopmcx-vowmammommm interaction of g-weakening with compound voicing in Japanese, respectively. 162 !|\'.. treatment of phonological opacity, as illustrated in section 6.2.2. As will be revealed in the following section, English compound abbreviation in Japanese has a case of phonological opacity that is similar to the case of German truncation introduced above. I will account for English compound abbreviation in Japanese within the framework of the Weakly Parallel Model. 6.5. Analysis As introduced in section 6.2., abbreviated English commands in Japanese are trimoraic or quadrimoraic. Among them, quadrimoraic words (e.g. (52)) are the most common, and trimoraic words are flu-ther divided intro three subgroups as in (53) - (55). (52) (== (3)) a. waadc purosessaa -+ waa-puro ‘word processor’ b. dejitaru kamera -+ deji-kame ‘digital camera’ c. rimooto kontorooru -+ rimo-kon ‘remote control’ (53) (= (21)) a. damn paatii -+ dan-pa ‘dance party’ b. furii maaketto -+ fini~ma ‘flea market’ c. terefon kaado -» tere—ka ‘telephone card’ 163 u“ l C l 0‘: f 7| t $.r -J“.l(\| \ (54) ('-= (22)) a. binaddo pitta -+ burn-pi ‘Brad Pitt’ b. potetc chippu -+ pote—chi ‘potato chip’ c. netto sukeepu -+ ne-suke ‘Netscape’ (55) (=== (23)) a. purasutikku modern -+ pura-mo ‘plastic model’ b. roiyaru hosuto ——> roi-ho ‘Royal Host (a restaurant chain)’ h c. bmakku bisuketto -o btn'a-bi ‘Black Biscuit (a band name)’ The four constraints in (56) — (59), i.e. TRUNC‘-=F, CON'I'IGUITY, ANCHOR-L-BT, and MAX-u-BT, are relevant to the quadrimoraic abbreviated compounds in (52).8 (56) Tumor-F: Truncated form must be a foot, i.e. bimoraic. (57) CONTIGUITY: Segmental material that is contiguous in the input must also be contiguous in the output. (58) ANCHOR-L-BT: The left peripheral element of the truncated form corresponds to the left peripheral element of the base. (59) MAX-u-BT: Every mora in the base has a correspondent in the truncated form. TRUNC=F in (56) is a mar-Redness constraint that requires the truncated form he a foot, i.e. bimoraic. Due to this constraint, the output form consists of bimoraic components. ' The constraint MINWD in (26) and Lamosr in (28), which are adopmd by Nishihara «4 al. (2001), miglnmrkmmecaseuhmdhmtudofdmhoweva,mepruentmidyadopuTwNC=Fand ANCHOR-LBT,simemeyuemmecommonlymedinmcansmdiesinthefitermne 164 Or- no r. 00me in (57) (McCarthy and Prince 1993a) and ANCHOR-L-BT in (58), which is a member of the constraint family ANCHORJNG (McCarthy and Prince 1995), are faithfirlness constraints. Cormourrv prohibits intrusion or skipping while ANCHOR-128T prohibits deletion or epenthesis at the left edge. The first two morae of the base are preserved in the truncated form because of these constraints. MAX-3T in (59), which is a member of the constraint family MAX (McCarthy and Prince 1995), militates against trrmcation. These four constraints, which are ranked as in (60), accounts for the quadrimoraic abbreviated compounds in (52), as illustrated in (61). The candidate (61a), where the first andthirdmoraeofthefirstcomponentofthebasecompoundispreservedintheoutput, cannot be the optimal candidate due to the constraint CONTIGUITY. The candidate (61b), wherethesecondandthirdmoraeofthesecondcomponentcfthebasecompoundis preserved in the output, is not optimal either, because of ANCHOR-LET. The candidate (61c) is ruled out since it violates TRUNO==F. The candidate (61d) is the optimal candidate because it satisfies all the highest-ranked constraints. The candidates (61a) — (61d) violate MAX-ET. But it does not affect the selection ofthe optimal candidate because it is ranked lowest. Finally, the candidate (61c), which is maximally faithful to the input, is ruled out by T'RUNO=F, one of the highest ranked constraints. (60) Constraint Ranking 1 Cormourrv ANCHOR-LET 139er MAX—8T 165 .o (61) Tableau for ‘waado purosessaa (‘word prowssor’) + T —+ waa-puro’ Input: WPW” Comm iAncnoa-Li T‘RUNt:=F MAX-ET Base: waado purosessaa 5 g a. wado puro ‘! 7 a, use b. was. rose 3 *1 s, can C. m p“ E t! O, ....l d. a. m M . a, can e. waado purosessaa T up Next consider the trimoraic abbreviated compounds in (53). In (53), the base form of the monomoraic component has a long vowel [a:] in the first syllable (e.g. pantii (53a)). The examples in (53) cannot be quadrimoraic due to the constraint *[azllrw in (62). ‘[a:l]rw is ranked higher than TRUNO=F, as illustrated in (63). Based on this, the constraintrankingin(60)canbemodifiedasin(64).Themodifiedconstraintranldng accounts for the examples in (53) as well as those in (52), as illustrated in (65) and (66). (62) I"[a:]hwy: No word-final long vowel [a:]. (63) ‘[a:]]pw >> TRUNC=F : dmsu paatrr" + T Input .. ‘(atllrw TRUNC=F Base: dansu paatn a. dan paa ‘! b. 6' dan pa * 9 FollowingltdandMester(l997b),lassmnethfltruncationisniggeredbythetrmationmorphemea‘ inthetableau).SeealsoltdmdMeuer’s(2003b:Ch4)argunmuthatRendakuisaiggeredbya morpheme. 166 o pato-kaa (‘pato-ka) ‘lit. patrol car’ (‘police car’) b. gei bu -+ *geirba ‘gay bar’ As introduced above, Nishihara et al. (2001) adopt the constraint NFLV in (68) to explain the vowel shortening in examples in (69). But, this constraint is too strong for two reasons. First, the vowel that undergoes shortening is are long vowel [a:] in most cases and many of other long vowels resist shmtening, as Nishihara et al. (2001) notice m Theexampleskaa‘car’andbaa‘bar’in(67)methemuycomtemxampleslfound. " AsmenfionedhChaptaSQflmeomaichmwmddoesnMexinmtheJapmelexiem. 168 (see (70)).12 In Nishihara et al. (2001), nrisu-do in (69c) is the only example of the shortening of a long vowel other than [a:]. This shortening might occur due to the analogy to mkudo (makudonarudo -> makudo ‘Mcdonald’s’) because they are both fast food chains selling American food. (68) NOFINALLONGVOWELS (NFLV): WWW“ pm (= (35)) (69) ( = (32)) a. fin-ii maaketoo -+ furl-ma ‘flea market’ b. depaato guru -> depth-88 ‘depment store sirl’ c. haiwei kaado --t hai-lta ‘highway card’ d. terefon kaado -+ tare-ks ‘telephone card’ e. misutaa doonatu -—-» misu-do ‘Mr. Donuts (a donuts shop chain)’ (70) a. sukeeto boodo —t sake-boo (‘suke—bo) ‘skateboard’ b. rondon buutu --> non-bun (‘ron-bu) ‘London Boots (comic duo)’ c. bataa piinattu --> bata~pii (‘bata-pi) ‘buttered peanut’ Second, Nishihara et al. (2001) assume that the vowel shortening in (69) is a phenomenon in the Loanword sublexicon. But this is not correct. As illustrated in (71), the word-final long vowel [a:] is highly restricted in the core sublexicon of the Japanese lexicon, but not in the Loanword sublexicon. In this paper, therefore, I propose that the long vowel '2 Nishiharaetal. (2001)refertotheexamples in(69) intheendnote as exceptionstotheirtmalysis. 169 d _‘l . l ..l‘ shuteningobsavedhEnglishcompmmdabbreviflionisaphanmmonhdrecme sublexicon of the Japanese lexicon triggered by the constraint ‘[a:]}pw. (71) The Number of the Words with Word-final [a:] (the result of a dictionary research) Total# #oflnanwordsl #ofNon—loanwords 1426 1382 (96.9%) 44 (3.1%) (Non—loanwords include intericctions, onomatopoeia, non-standard words, etc.) Next, letusconsiderthenimomicabbreviatedcompotmdsin(54).ln(54),the mabbreviatedfmmofthemmmnmaiccmnpmanhasacodaconsmantindrefirst syllable (e.g. pitto (54a)). In these words, the coda consonant cannot be preserved in the output because of a highest-ranked constraint CODA-COND in (72) (ltd 1986). There are two possible ways to avoid the ill form like pit and maintain the quadrimoraic form consisting of two bimoraic components: (i) skipping the second mora or (ii) vowel epenthesis. But neither of them works, because the former is blocked by the constraint CONTIGUITY while the latter by the constraim DEP-BT in (73), which is a member of the constraint family DEP (McCarthy and Prince 1995), as illustrated in (74). In (74), ‘o,’ and ‘oj’ indicate non-epenthetic and epenthetic vowels respectively. (72) CODACOND: ‘Place] (73)DEP-BT:Eva'yelemanhrflremmcatedformhasacureqrmdanhlfllebase. 170 (74) Comramrr, Coos-Conn, Dar-8T >> TRUNC=F WWW” Carma CODA. iDEP—BT ThUNc=F amunaddopitto. COND g a. burapito; ’! b. bmpitoj '! c. burapit ‘l d. Imps '! eaburapi : A s Basedonthetableau(74),theconstraintrmkingin(64)cmbemodifiedasin(75). Themodifiedconstraintrankingaccormtsfortheexamplesin(54)aswellasthosein(52) and (53), as illustrated in (76) - (78). (75)ConstraintRankingIII Cormourrv AnaloR-LBT ‘[a: CODA-COND Dar-8T ’___________________ ; Tsunc=F MAX-8T (76) Tableau for ‘buraddo pitto (‘Brad Pitt’) + T -+ bura pi’ v v v WWW SSE eeés :3 as W=WPM § =33 53% ‘3'. E G? en : t" : ’7‘ "l a. buraprtoy *1 E 5 1st,. b btn'apitoj r . fl . es’s C. burnpit 0!: age (I. burapii f Er 1'! ' exec we... . t: bmaddopino T "! 171 h(76Ldnqua¢immaiccmdidflacanisfingoftwobhnoraiccomponans06a)-(76d) nemledanbyCONnGUlTY,DEP—BTandCODA-COND.Thecandidate(76e),whichisnot qmrkhnmeichttimmaiqisseleaedasmeopfimalcmdidatesinceimviolafimof TkUNC=FisnmfataLTheurdidate060isfllemostfaiflrfidmthempmhncmmotbe opfinnlduemhsdoublyvidafimofTRUNFF.h(77)md(78),mflreoduhmd,flre cursuahusCONTIGUm,CODA-COND,mdDEP-3Tdonotafl‘ectflreselecfimofflre opfimflcmdidflemdthecmdidfles(77d)md08d)uesdededasdmopfimflmnpum. (77) Tableau for ‘dansupaatii(‘danceparty’)+T—+danpa’ “mm” 9 5 99 E t; ’3 is: Basezdlnsnpflfii a . a “é? a . bf 31 :3 a. dmpati t: t s s a t," I». dam in c. dmpaa J E f it! use tan. s 2 E E . 9» .. mm T (78) Tableaufor‘waadopmesessaa(‘wnrdprocessor’)+T—+waa-puro’ Inputwordprocessrr 8 2 ion 9 ET. g1 g == 208 r =—- E Basezwaadopurosessaa o g :5 if '13 E 3" ‘3 I'll r'-* a ”a "*1 *i a. wadopuro ‘! E 3 o’eeee b. waarose ‘! E T o’eeee C. wanpu Er c! o’ecece iamm g E E 4; .’.‘.’ 172 Finally, consider the examples in (55). In (55), the second syllable of the mabbrwiatedfmmofdremmommaiccompmentstmtswithmobsmwwgmogem (55a)).Intheseexamples,drefirstsyflableofflrebaseformofflremonomaaic component does not have a long vowel [a:] as in (S3) (e.g. mtii (53a)) or a coda corrsonantasin(54)(e.g.pi_t_.to(54a)).Inotherwords,thereseemsnomotivationfor Why.Bm,asifllmuedm(55),mesbhreviaedfmmisthnaaicmmisfingof onebimaiccomponartmdonemonomaaiccomponaiWhyueflreabbreviated formsin(55)notquadrimoraic?Aswillbeshownbelow,thisisaeaseofpbonological opacity. TheEnglishcompormdpurnsufiflunroda-a‘phsticmodel’isabhreviatedas pmu—m(’prnu—nnde)1hemapmfamarggestsdratthemmomuaiccndidatem usmmmmmgmmmbmmmmmmas menfimedaboveflraeseemsmmofivafionforthemmommaicity.Whyisflrebimmaic candidflenwdendseleaedasmeopfimdcmdidmfllbelievethatisbecmnedre bimmaiccmdidatenndeiswmseflnnmodlabimaaiccmdidatemdintamsofflle numbaofsyllable.Thebasicmofivatimfmabheviafimismmakewmdsshorter.hr thusmsemedisylhbiccmdidmemdeiswasedmthemmyllabiccmdidatennd. TummylhbiccmdidatemodJowwmcmmabeflnopfimalmeidlensmcefi violatesfllecmsuathODA-COND.AsareallLdremmomaaiccmdidaemcmbeflle optimalcmdidate.hlderivafimnlmodel,flusprocesscmbeshownasin(79). 173 (79) ‘moderu (‘model’) -+ mo’ Input: mo.de.ru. resyllabification, TRUNC=F mod. I CODA—COND Output: mo. In (79), the intermediate form mod plays a significant role in determining the output form. Thetnlncationmoderuamisacaseofphonologicalopacityinthesensethatthe mmcatedformmodisnotfomldanywhereintheinpmortheoutpmmeaysa significmtmlemselecfingmeopfimflmnplmThecaseofphmobgicdopacityathand issimflm'totheoueinGamanmmcafiondiswssedbyltdandMester(l997b)inthc sensethattheresyllabifiedformplaysasignificantroleindeterminingtheoutputform.” hdlissmdy,1winaccmmtfmflrecaseofphmdogicalopacityathandwidlinflre fiameworkoftheWealdyParallelModethdandMesterZOOlb,2003a).Inthismodel, dreabhreviationpurasutimnodaw-apmu-mcanbeexphinedasfoflows. Asmalfimedabovedrebimomicformmdseemstoplayasignificantmlein ddammingdremnpruformfihisisbecmsednbhnaaicfixmmodisflremnpmofflre lexicalmochde(mddlemp\nmdrepostlexiealmodule)1heconsuahnmking devdopeddrmfaristhemefmpoaleximlphmobgyflhisrmkingpediasflncmma output,asilhrstratedin(80).hl(80),thebimoraiccmdidatenro.dein(80a)isruledoutby ”Arrtsanduater(1997b)pointont,thirtypeorphonologrerl' qneityis“detaehedfmmtherule W " sinkingornadrnonrl“ nrle-buedphnnologymeedmg,‘ bleedmg,’ W“ cmflablwdhg’;“opaqmvstupuuumkhlerwfim”)mTHseueahmdxisamflnmidaof womdicmaphoby,whaephanmmofdfishndwuehea¢dproce¢nflybymamomeodic Circumscription (McCuthyandPrince 1990)” (Itdrlld Mester l997b: 9). 174 DEP-BT, while the candidate mod. in (80c), which is faithfirl to the input, is ruled out by CODA-COND. (80) Tableau for ‘purasutikku moderu (‘plastic model’) T -—+ pura mo’ (postlexical) Ianzpmastlfikkumoderu+T 9g '3 8§ g t; g E Base: puma. mod. 6% g '5 if é a Z: ‘R a : : "1 "5 a. pu.ra. mode. ‘l 4' b. pure. mode. T j ‘l f " " c. pu.ra. mod. i *l d.‘="pu.ra. mo. T e a e. pur. mo. 5 t! r e u The next question to be considered is: How is the bimoraic form modselected by the lexical module? In the lexical phonology, resyllabificaton occurs when truncated Mm -+ mod.). This resyllabificaton is triggered by the constraint ALL-04.5}? in (81), which is a version of ALL-Fr-LEFT by McCarthy and Prince (1993a). This constraint explainswhyflrebimoraiccandidatenrodcisworsethanthecandidatemod (81) ALL-o-LEr-T: Every syllable stands at the lefi edge ofthe truncated form. Thereisonemoreissueweneedtoconsider.Thatisthetreatmentofcoda mainlapmeseobsmrenmwithmdmdepmdanplweofmfiudafimaswellu 175 the placeless nasal, can be in the coda position.M In the lexical phonology, the optimal output mod has a coda consonant [(1]. But this coda consonant seems illegitimate, since it has independent place of articulation. Why is the [d] with independent place of articulation possible in the form mod? That is because mod is a bound morpheme. ltd (1990) discusses the phonological difference between prosodic words and bound morphemes in compound loanword abbreviation in Japanese. For examme, consider dansu paatii -» dart-pa. The abbreviated component dan is a well-formed component but it cannot worn: in isolation. This suggests that different set ofconstraints apply to bound momhm. Inthisdissertation,IproposethatdrereareactuallytwotypesofCODA-COND in Japanese: (i) only nasals and Obstruents can be in the coda position (CODA-Comm in (82)), and (ii) they cannot have independent place of articulation (henceforth CODA-CONDpw in (83)). The constraint CODA-Comm Plays an active role in both lexical phonology and postlexical phonology and applies to bound morphemes as well as prosodic words, whereas CODA-CONDpw plays an active role only in postlexical phonology and applies only to prosodic words. That is, CODA-CONDpw is the constraint that needs to be ranked differently in lexical phonology and postlexical phonology. (82) CODA-Calm“: Coda consonants are Obstruents or nasals. (83) CODA-CONDpw: *Place] "rnrrpateeementsumeodaireomideredmtheplneelesrmmp1991).rheplroeletrnneol assimilates in place when followed by a consonant (e.g. ni[r)].ki. ‘popularity’), while it appetu's as [u] when followed by a vowel (e.g. ni[u].i. ‘optional’) or in unr'dofinal position (e.g. ho[lt]. ‘book’). 176 The idea that CODA-CONDpw plays an active role only in postlexical phonology and applies only to prosodic words is supported by verb stems, which are also bound morphemes. As illustrated in (84), verb stems in Japanese can have coda consonants with independent place of articulation, although the output forms, i.e. the outputs of the postlexical phonology, never have coda consonants with independent place of articulation.” (84) Verb Stems (Coda Consonants with Independent Place of Articulation) Stem Past Causative a. mat matta. matasexu. ‘to wait’ b. kat katta. kataseau. ‘to win’ c. yom yon.da. yo.ma.se.ru. ‘to read’ '5 Mmmmmvabmmmmmmmhumdrmminimum: appearstem-finaflywhfleflieydoindiecodapositionofmmcatedforms(e.g.modem—vino“. Avoidanceof[d]and[zlinthestem-finalposifionmayberelevunwthefactflntvoicedobmmtsm mm‘kedinJapanesc.Second,nuncatedformsneverhaveapproximaIns,i.e.[r],[w],and[y],intbecoda position,whereasverbstemscanhave[r]incoda.(Someverbstemsseemtohave[w]incoda(e.g.wa.ra, wa.rat.ta., wa.ra.}ga.se.ru. ‘to laugh’ (cf. (84)». But it is not considered as a coda consonant in Modem Japanesebccausethisistheresultofphonologicalchmges([p]—+[¢]~[w]-+0)(McCawleyl968:82, Vancel987:182,amongothers).)1hecoda[r]ismiqueinJapaneaeinthesensethst(i)itistheonly approximmtthatcanappearinthecodapositionmd(ii)itistheonlyvoicedcodaconsomntthatdoesnot trigger voicing assimilation (e.g. kar-u --v tat-(a ‘cut—past’ (cf. ton-dc ‘fly-past’, mat-(a ‘wait-past’» (McCawleyl968:96).T‘hepeculiarityofcoda[r]mggeststhatitneedstobetreateddifi'erentlyfi'omthe othercodaconsonmtsinthelexicalphonology,althoughitdoesnotnecdtobeneateddifi‘erenflyindie postlexicalphonology,becausenoconsonamsincluding[r]canappearinthecodapositioninthe postlexical phonology. One possible explanation I have for the peculiarity ofcoda [r] in verb stems is that thestem-final[r]inverbstemsisnotinthecodapositionbutinthemicleusposition.Theasymmen'y betweenthecodacomonantofverbstemsandthatofn'mcatedformsisinterwfinginitsownfighhbutitis beyondthefocusofthisstudyJwillleaveitforfimuemearch. 177 'la. The constraint CODA-CONDBM is a highest-ranked constraint and prohibits coda wnsonmtsotherthmobstumtsmmsalsflhisconshahtaccomtsfmwhymesecond component of the examples in (85), which has an approximant [r] in the onset of the second syllable, never appears as the monomoraic form in compound abbreviation. (85) a. sarada doressingu —. sara-dore ('sara-do) ‘salad dressing’ b. pari korekushon -—+ pari-kore (‘pari-ko) ‘Paris collection’ c. imeeji toreeningu -> ime-tore (‘ime-to) ‘image training’ d. intaa karezzi -—+ in-kare (‘in-ka) ‘Iit. intercollege‘ (intercollegiate game’) Now, let us get back to the main point: How is the monosyllabic and bimoraic form mod selected by the lexical module? The constraint makings in (86) and (87) are the rankings for English compound abbreviation in Japanese, where the constraint CODA-CONDpw is ranked differently. These rankings explain why model is tmncated as mo and why domaingu cannot be do, as illustrated in (88) - (91). (86) Constraint Ranking for English Compound Abbreviation (Lexical Phonology) Comm ANCHOR ‘[a:]]pw DEP-BT Cong-Comm TRUNc=F ALL-e-L MAX-3T CODA-CONDpw 178 ’Q 3‘. CONTIG ANCHOR TRUNFF ALL-o-L MAX-HT (as) Tableau for ‘purasutikku modern (‘plastic model') T -+ pura mo’ oexical)“ (87) Constraint Ranking for English Compound Abbreviation (Postlexical Phonology) ‘[a:]]pw DEP—BT Cong-Comm Cogs—Comm Input: masutikkumoderu+T 0 . n; U A S] 0 Base':urasutikkumoderu agagggz % E Egg ' age-wises- a. puma. made. 5 T gages", 5 E E o b. pu.ra. mod.e. T 5 5 a! 0,. use a c.°'pu.ra. mod. I! run t ' : ' - e (1. para. mo. a! a nu E n e. pur. mo. 4N? o ”u a : g : : " " Innaditionallexicalphonologymodel,ithasbeenassmncdthattheompmsofthclexicalphonologyare prosodic words. Inthis dissertation, however, following as and Mester (2001b, 2003a), I assmne that this doesnot apply in the Weakly Parallel Model. 179 (89) Tableau for ‘purasutikku modern (‘plastic model’) T —-> pura mo’ (postlexical) Input:pm'asutikkumoderu+’l‘ Q; ggé) a??? g E E Basezpu.ra.mod. §:§ ST§TEE§ i {a}: :9; a. pu.ra.mo.de. i Tth h! b. pu.ra.mod.e. ‘l " I! fit .. ...... . d.<3'pu.ra.mo. T t o e .. mm... gs .3 . hithelawn:ofgnnuuufikkulnoukru,thelunufidauetnqunuxh:hmustualnnuhmuuifinthe lexical module due to the constraint ALL-o-LEFT, and the candidate para-mo is the optimal candidate in the postlexical module because the candidate paramlod is ruled out due to its violation of CODA—CONDpw. In the case of sarada doressingu, on the other hand, the candidate sans-dare is the optimal in both lexical and postlexical modules, since the candidates sara-dor and sara-do violate CODA-COND (both CODA-CONDpw and CODA-CONE“) and TRUNC==F, respectively. (90) Tableau for ‘sarada doressingu (‘salad dressing’) 1‘ -v sara dore’ (lexical) Input:saradadoressingu+T QTE§§OE:§ 3 Eng I l o o as r Basezsaradadoressin 33:: >5b5= % é ' > 3“ o g% a: e g3 s, e E ' a?sa.m.do.re. " *, : I E : .... b. sarador. g in? i t *. ’ : ' : E fitii c. sarado. E i 5 7 ’l ‘ fl 5 t ' ' asses 180 (91) Tableau for ‘sarada doressingu (‘salad dressing’) T —+ sara dore’ (postlexical) Inputzsaradadoressingu+T g ‘ i :8 $3568) g .1; g E E . : E ' E: d I Base.sa.ra.do.rc. o 523:“: :f g 53 31 6‘ a a¢sa.ra.do.re. T 1;: b. sa.ra.dor. ‘ *l T * ‘ s r c. sa.ra.do. ‘! "' " The constraint rankings in (86) and (87) account for the other types of compound abbreviation in (52) - (54) as well, as illustrated in (92) - (97). (92) Tableau for ‘bln-addo pitto (‘Brad Pitt’) + T —+ bura pi’ (lexical) Input:buraddopitto,+T 9 5 Egg: E5 .5; '7? E g 95 WWW 5:5 55 5,, s: :5: a. bu.ra.pi.t0;. ‘! T T u a} b. burs. pi.to}. T *l " *:, c.°'bu.ra.pit. T A t a} t d. bu.ra.pi. T at! e n, e. bur.pi. *l a x, t E as 181 I: (93) Tableau for ‘buraddo pitto (‘Brad Pitt’) + T -> bura pi’ (postalexical) WWW,” 8%538995 53:? 5 E 5 WM?“- 53535555535“ 5: 5- 5 a. bu.ra.pi.to. : i T T’! ‘3" b. bumpit. T‘!T "' c.9bu.ra.pi. ' T ' ' " ’ " In the case of buraddo pitta, the candidate bum-pit is selected by the lexical module because the quadrimoraic candidates (92a) and (92b) violate the highest-ranked constraints Comoum and DEP-BT, respectively. In the postlexical module, on the other hand, the candidate bum-pit loses to the candidate bum-pi because of the violation of CODA‘CONDpw. (94) Tableau for ‘dansu paatii (‘dance party’) + T --9 dan pa’ (lexical) In :dansu ’i+T 0: Hon TA :1 Bwfm‘" gégégeégéaggégé‘ m M «3:555:53: sue-.55? a. dan.pa.ti. "T E E E ’ 5 5 : E E “ b. «mm. i *1 a s s t, i i " c. dan.paa. i 5 E 5"! ‘, i 3 i i " i‘dan-m 2’ § § 5 * *, i 3 i i no 182 (y! (95) Tableau for ‘dansu paatii (‘dance party’) + T -+ dan pa’ (postslexical) :dansu "+ Ci; 0 O :Uipfi : if”... .... 3 935% 2:; a E? ”e' p" 0 E9 ?:?...,§2 31!!" 5’1 a. dan.pa.ti. I" b- 4mm 3“?" e.gdan.pa. f 4‘ " In the case of dansu paatii, the bimoraic candidate dun-pa is selected by the lexical module because the quadrimoraic candidate (94¢) violates the highest-ranked constraint ’[a:]]pw, which is fatal. In the postlexical module, the candidate danopa is selected again as the optimal output, since the other candidates violate the highest-ranked constraints. (96) Tableau for ‘waado purosessaa (‘word processor’) + T —+ waa-puro’ (lexical) Input:waadopm'osessaa+'l‘ i8 0% E; 3'3 :1 3 E 8 g Base:waadopurosessaa ES :78; w E? % é ti: g :7 "a ~21 1" -l a. wado.pu.ro. u e, L, i; 1 J aces b. waamse. I'l * *, . ' . . 0“. C. W88. pu‘ g WE g E .! .9 E 3 5 : paced d. 6' waa. pu.ro. ’ L i ‘ ‘3 e. weapur. *l ‘, * i i 3 t... 183 (97) Tableau for ‘waado purosessaa (‘word processor’) + T -+ waa-puro’ (postslexical) r I WWW” a “:9: as: a a a Basezwaapuro. g??? 5i? 31 a 2 a. waapu. ; g : 'l * b.0waa.pu.ro. j: ‘ c. waapur. E’! *i " In the case of waado purosessaa, the quadrimoraic candidate wan-puma is selected by both the lexical and postlexical modules because the trimoraic candidates waaepu violates the constraint TRUNC==F and it is fatal. There is one more issue we need to consider. That is asymmetric word structures. As introduced above, there are three types of trimoraic abbreviated compmmds. Among them, almost all of them have the structure like para-mo, where the first component is truncated as a bimoraic form and the second as a monomoraic form, i.e. [Watt + pm]. But, the structm'e like rte-sake in (98), where the first component is truncated as a . monomoraic form and the second as a bimoraic form, i.e. [um + pun“), is quite uncommon." Among the three types of trimoraic abbreviated compounds discussed thus far, one type is triggered by the constraint ‘[a:]]pw whereas the others by two types of Com-Conn, i.e. CODA-CONDpw and CODA—CONDBM. The trimoraic words triggered by ‘[a:]]pw do not have the structure [mm + ppm], since the constraint applies only to prosodic word-finally. But, the other two types such as those in (99) should be able to '7 ne-suke‘Netscape’andme(ru)-ado‘emailaddress‘mtheonlyexampleslfom'l'heabbreviation procusof‘amfladdreu’hmMonepszfimemadorm‘emailm’qmmadarmfi memo-ado. 184 s tvh bl .'_ .h “I" have the structure [um + MM]: because CODA-COND is a positional constraint for the coda, but not for the prosodic word-final position. Why is the structure [mm + mum] disfavored? That is because the Light-Heavy structure is marked in Japanese, as shown below. (98) netto sukeepu —-+ ne-suke ‘Netscape’ (99) a. dejitaru kamera ~+ deji-kame (’de—kame) ‘digital camera’ h. potetc sarada --> pots-sara (‘po-sara) ‘potato salad’ c. sutaatingu menbaa -> suta—men (‘su-men) ‘Iir. starting member’ (‘stmfins lineup’) Kubozono (1988) discusses the dispreference for the Light-Heavy structure at the prosodic word level (see also Kubozono 2003). Kubozono (1988) analyzes the structure of Japanese compound nouns and finds that the kit-branching structure is more common than the right-branching structure. In other words, the left-branching component tends to be heavier than the right-branching component, as schematized in (100). The morphological asymmetry in (100) is reflected in the prosodic asymmetry. The right-brmrching structure is generally realized as two prosodic words, i.e. the output has two accented morae, whereas the left-branching structure is realized as one word, i.e. the output has only one accented mora, as illustrated in (101). (100) [[ABIICH > [[AHBCH 185 ’l (101) (Kubozono 2003: 117) 8- [lABHCD -’ {ABC} [[[do'itu][bu'ngaku]][kyookai] --> {doitu-btmgaku-kyo'oltai} ‘Germany-literatme-association’ ‘Association of German literature’ b. [[AIIBCD -* {A} {BC} [[dO'itullfth'ngflkullkyookailll *{dO'imH W-kydokai} ‘Germany-literatm'e-association’ ‘German Association of literature’ ltd (1990) discusses the avoidance of the Light-Heavy structure at the syllable level. ltd (1990) studies non-compound loanword abbreviation in Japmese and finds the truncation patterns in (102) (Ito 1990:217, see also Labrune 2002). As illustrated in (102), where nonproductive abbreviation patterns are marked by the symbol “It”, the sequence of [0,. cm] in the word-initial position is highly disfivored in Japanese. (102) u—Pattern o—Pattem lu # [on] 2n {on 0:1 # [Gm-J 3“ [0» on] [0,. 0,. on] # [on 0w] 4” [can 51m] [0w 0,. on] [Us 0,. arm] [on 0,. an 0,] # [on on, ad Example choko(reeto) saike(dcrikku) anime(eshon) baaten(daa) intoro(dakushon) eakon(dishonaa) finasuto(reeshon) I86 ‘;¢z" The avoidance of the word-initial [trF am] has been explained in terms of foot parsing. [to (1990) claims that this is due to Lefi Edge Requirement (the interaction of ALIGN-La’w, FT) (Every foot stands at the lefi edge of the prosodic words.) and PARSE-HEAVYSYL (Heavy syllables are properly parsed by feet.) in OT terms), which requires the lefi edge of a prosodic word matches the lefi edge of a foot (see also Kubozono 2003). In other words, ltd (1990) claims that the word-initial [0,, cm] is not allowed because the possible foot parsing ([(tr,, 0.0"] or [on (0,...)]) violates one of the higher-ranked constraints ALIGN-L(Pw, FT) and PARSE-HEAVYSYL. The dispreference for [um + gum] observed in English compound abbreviation is another example of avoidance of the Light-Heavy structure, i.e. the avoidance of the Light-Heavy structure at the mora level. The structure [um + Film] is disfavored because it is prosodically worse than [ppm + um]. The avoidance of [um + ppm] can be explained in terms of foot parsing. As mentioned above, ltd (1990) claims that the avoidance of Light-Heavy structure is due to Lefi Edge Requirement (the interaction of ALIGN-LG’W, Fr) and PARSE-HEAVYSYL in OT terms). But, if this is the correct analysis, thewordsin(99)shouldbeabletohavethcstr'uctrn'e[uau+uum]becausethe constraints ALIGN-LG’W, FT) and PARSE-HEAVYSYL do not prohibit the foot parsing [01314 + u) um]. The fact that the foot parsing such as [(tlasr + u) m] is not allowed suggests that the foot parsing across stem boundaries is not allowed in Japanese. In other words, the morphological structure plays a significant role in foot parsing. As mentioned above, Kubozono (1988) finds that the morphological asymmetry is reflected in the proMc asymmetry. But the prosodic asymmetry pointed out in Kubozono (1988) is not relevant to the asymmetric word structures observed in English commund abbreviation, 187 ‘. r V '. . a J' t 'i ' 0 fl. - t, . n. _( r ‘1 'l‘. . 0‘4. J ‘ l a ‘ f . e 0 t. O - i ' . ,‘J" Al ' '. . t since the examples in this chapter do not have more than one accented mora in the abbreviated form. In other words, the abbreviamd words are all single prosodic words. Insunhtheavoidanceofthe structure [ugM+uugM] isnotasporadicphenomenonin Japanese. But, it is a part of the systematic dispreference for the Light-Heavy structure observed at the levels ofthe prosodic hierarchy in (103)." (103) Prosodic Hierarchy Prosodic Word Foot Syllable Mora 6.6. Conclusion This chapter has discussed English compound abbreviation in Japanese within the OT framework. This chapter is summarized as follows. The majority of abbreviated loanword compounds in Japanese are quadrimoraic consisting of two bimoraic components. But, trimoraic words are also possible and can be divided into three groups. The trimoraic words can be explained as follows. The first type of the trimoraic words are triggered by the constraint ‘[a:]]pw. In those words, the tmabbreviated form of the monomoraic component has a long vowel in the first syllable. Nishihara et al. (2001) claims that they are trimoraic due to the constraint NOFINALLONGVOWELS and this is a phenomenon in the Loanword sublexicon. But, this " The avoidance of the Light-Heavy stt'uctm'e is not observed at the foot level probably because degeneratedfeetarenotallowedinlapanese. 188 study revealed that it is only the long vowel [a:] that undergoes shortening and that the constraint ‘[a:]]pw triggers it. As discussed in Chapter 4, this constraint plays a significant role in the core part of the Japanese lexicon. It suggests that the shortening of [a:] observed in English compound abbreviation is a phenomenon in the core part of the lexiconandthetrimoraicwordsofthistypeareassimilatedtothecorepartofthe Japanese lexicon with respect to the constraint *[a:]]pw. 'Ihesecondtypeoftheu'imoraicwordsaren'iggeredbytheconstraint CODA—CONDpw. In those words, the unabbreviated form of the monomoraic component hasacodaconsonantinthefirstsyllable. Butthecodaconsonantcannotappearinthe output because the coda consonant with independent place of articulation is not allowed in the output. 'I‘hethirdtypeofthetrimoraicwordsareacaseofphonological opacity. Inthose words, the second syllable of the unabbreviated form of the monomoraic component startswithanobstruent. 'l‘histypeofcomponmdabbreviationhasnotbeendiscussedin the literature. The present study revealed that this is a case of phonological opacity where the constraint ALL-O-LEFT plays a sigrnificarnt role. In this chapter, I have acconmted for the opaque case within the framework of the Weakly Parallel Model. I have argued that there are actually two types of Com-Corn), i.e. CODA-CONDpw and CODA-CONDBM, and that CODA-CONDpw plays an active role only in the postlexical phonology. This idea is supported by verb stems in Japanese. The Weakly Parallel Model accounts for the opaque case observed in English componmd abbreviation straightforwardly, since it handles the codacmsomntsmflneopaqmcaseandflnosemverbstemsmarmifiedmmmerztheym both bonmd morphemes and are allowed in lexical phonology. The asymmetries between 189 I .' ,- thecodaconsonantsintheopaquecaseandthoseinverb stemswereintroducedbutnot discussedindetail.Thatneedstobefirrtherstudiedinfinnueresearch. This chapter has provided supporting evidence for the Weakly Parallel Model, a subtheory of OT. Phonological opacity is problematic for OT, because markedrness constraints apply only to surface representations of candidates and the surface representations do not show the motivation for opaque phonological processes, and OT has not resolved it successfully. But, the Weakly Parallel Model accounts for it successfully by adopting the traditional distinction between lexical phonology and postlexical phonology. This chapter has shown how a case of phonological opacity observed in English cornpound abbreviation in Japanese can be explained within the framework of OT. Also, the Weakly Parallel Model gives a straightforward explarnation for the difference between the two types of CODA-COND, i.e. CODA-CONDBM and CODA-CONDpw. In this chapter, I have argued that CODA-Comm applies to both bournd morphemes and prosodic words whereas CODA-CONDpw applies only to prosodic words. The difi‘erence between these two constraints can be explained straightforwardly within the fi'arnework of the Weakly Parallel Model, since the lexical and postlexical modules can have separate constraint systmns irn this framework. Finally, this chapter has revealed that Japanese disfavors the Light-Heavy structure systematically. When English compounds are abbreviated into trimoraic forms, almost all the abbreviated forms have the structure [tillers + m]. This is because of the dispreferencc for the Light-Heavy structure. Previous studies discussed avoidance of tire Light-Heavy structure at the syllable and prosodic word levels. But they have not shown I90 at“ ll:’ that it is a general characteristic of Japanese. This chapter has revealed that avoidance of the Light-Heavy structure is observed at any level of the prosodic hierarchy and it is a general characteristic of Japanese. I91 7. Conclusion In this dissertation, I have discussed issues surrounding the phonology of English loanwords in Japanese within the fiamework of OT. OT can account for dialect differences (see Chapter 3) and diachronic changes (see Chapter 4 md 5) in terms of different rankings of universal constraints. Furthermore, OT can deal with phonological opacity by weakening one of the core tenets of OT, i.e. no intermediate levels, and by adopting the traditional distinction between lexical and postlexical phonology (see Chapter 6). To close this dissertation, I will summarize the major findings below. The major findings of this dissertation can be categorized into two groups: findings regarding the nature of inputs to loanword adaptation and findings regarding Japanese phonology including the structure of the Loanword sublexicon. 7.1. The Nature of Inputs Withrespeamthenatmeofmpnnsmlomnwordadaptafiomflnepresentsmdy reached three major findings. First, this dissertation revealed that the inputs to phonological processes in loanword adaptation are the perceived segments. In the previous studies, Silverman (1992) claims that the irnput is based on the phonetic representation of the source language while Paradis and LaCharité (1997) claims that it is based on the phonological representation. Based on the realization of English /r/ in Japanese(Chapter4),thisstudyarguedthatthe inputtophonologicalprocessesin loanword adaptation is not the phonetic or phonemic representation ofthe source language but it is based on the perception. The present study supports Silverman’s (1992) l92 “it" idea that the phonetic representation of the source language and the input to lomword adaptationaredifi‘erentandneedtobedistinguished. Also,Iarguedthatthedifl‘erent realizationoftheonset/r/andthecoda/r/inJapaneseisduetotheperceptionandthatan irnput segment can be perceived differently at the Perceptual Level based on the location it appears. Second, this dissertation argued that Japanese borrowers have access to the information on the locus of English stress and English morphology. It has not been fully discussedmmeliterauuewhmmformafionismcludedmflnemptmwmregmdmthe phonological irnformation of the source language, Silverman (1992) claims that the speakers of the host language cannot have access to it while Paradis and LaCharité (1997) claim that they can. The four phonological phernomena arnalyzed in this study did not give arnyevidencedirectlysupportingtheclaimthatthespeakersofthehostlanguagehave access to the phonological representation of the source language. But, tlne accentuation of English loanwords (Chapter 3) and the realization of the English plural morpheme (Chapter 5) showed that the information on the locus of English stress and English morphology is included in the inpnrt. Third, in this dissertation, I claimed that English loanwords in Japanese are introduced by limited bilinguals, i.e. Japanese-English bilinguals with the knowledge of English morphology but not necessarily with the knowledge of English phonology. Paradis and LaCharite (1997) claim that loanwords are borrowed by bilinguals and that tlnespeakasofmehostlmguagehaveaccessmthephonologicdrepresenmfionofthe source language. As mentioned above, however, the four phonological phenomena analyzed in this study did not give evidence supporting their claim that the phonological I93 ‘l|0" -.. a“ representation of the source language is accessible, although the realization of the English plural morpheme (Chapter 5) indicates that morphological information is accessible. BasedmdualelabomtedeadisdeaCharité’sclaimmdclaimedthmEnglish loanwords in Japanese are introduced by limited bilinguals. This claim is supported by the facts that all the Japanese people study English at secondary school and that they have some knowledge of English.‘ With regard to the accessibility to the morphological information of English, the present study revealed that there is a difference betweern generations (Chapter 5). The Englishphnalmorphemeisneateddifi'erenflymmeolda'genmfimmddneymmger genemfioanmisdissatafiomlarguedmatflnegenemfiongapudthrespectmflne realization of the English plural morpheme is due to the difference in the accessibility to the morphological information of English. That is, the older generation does not have access (or has limited access) to the morphological information of English, whereas the younger generation has access to the morphological information of English. 7.2. Japanese phonology WithmgardeapanesephmnologymcludmgthesmuneofflneIpmword sublexicon, tlnis dissertation reached four major findings. First, the present study revealed that Japanese disfavors the Light-Heavy structure systematically (Chapter 6). ltd (1990) and Kubozono (1988) introduce the dispreference for the Light-Heavy structure at the syllable and prosodic word levels, respectively. But they do not show that it is a general lThisdoesnotnecessarilymeanthattlneJapanesepeoplearefinllycompetentinEnglish.Onthecowary, monofmeJapmmepwpkdonmhavephomlogicdwnmamofEnglhhsimewhmmeylwnn mostlyEnglishgrammarbutnotEnglishconversation. 194 f s o , 1 r], , J.: a. J 9 ' . ‘ ' | ""‘ 01. . ,I.,l characteristic of Japanese. Discussing the dispreference for the Light-Heavy structure at the mora level, this dissertation showed that avoidance of the Light-Heavy structure is observed at any level of the prosodic hierarchy and that it is a general characteristic of Japanese. Second, this study argued that there are actually two types of CODA-COND, i.e. CODA-CONDpw and Com-Conn)», in Japanese (Chapter 6). English compound abbreviation in Japarnese reveals a case of phonological opacity. Adopting the Weakly Parallel Model (Ito and Mester 2001b, 2003s), this study discussed the opaque case and argued mat CODA-Comm plays an active role in both lexical and postlexical phonology arnd applies to bound morphemes as well as prosodic words, whereas CODA-Cour)", plays an active role only in postlexical phonology and applies only to prosodic words. ThisideaissupportedbyverbstemsinJapanese. Third, this dissertation showed that the word-initial pitch of English loanwords in Kansai Japanese is predictable (Chapter 3). Pierrehumbert arnd Beckman (1988: 214) claim that the word-initial pitch of non-loanwords in Kansai Japanese is unpredictable. The present study revealed that the word-initial high pitch is more common and tlnat the word-initial low pitch of English loanwords in Kansai Japanese is triggered by the accent onthesecondmora. This studyalso showedthatKansaiJapaneseandTokyoJapanese arethesamewithrespecttoaccentassigmnentandthatthe Englishaccentispreservedin both dialects. Finally, this dissertation discussed the assimilation to the core part of the Japanese lexicon. Among the four phonological phenomena analyzed in this study, two phenomena, i.e. the realization of English lr/ (Chapter 4) and English compound abbreviation (Chapter 195 this as 6), showed the assimilation to the core part of the Japanese lexicon with regard to the constraint ‘[a:]]rw. Ito and Mester (1995) claim that recent loanwords violate more constraints and are less assimilated. The realization of English lr/ in Japanese, however, revealed that this is not always the case, since less recent English loanwords with the word-final coda /r/ violate this constraint but recent ones do not. 7.3. Future Research Regardingthefindingsofthis smdy,thcfollowingresearchneedstobedoneinthe future. meisdissatafiondiscmsedmeassimilatimmflrecmepmtofflrehpmese lexicon with regard to the constraints ’[azflm (Chapter 4 and 6) and ‘HL (Chapter 3). Butitisnotclearyetwhataspectofthewordtendstormdergotheassimilationin Japanese and cresslinguistically. The assimilation phenomena discussed in the present studysuggestthatprosodie feamresareassimilatedtothecorepartoftheJapanese lexicon more easily. But it is not clear whether this is the case crosslinguistically. This question needs to be further studied. Second, this study theoretically accounted for how the Japanese borrowers perceived English sounds, especially English lr/ (Chapter 4), and what is included in the input. But, the claims deveIOped in this dissertation lack empirical support through experimental studies. Experimental studies regarding the claims developed in the present study need to be conducted in the futme. 196 f'. ... APPENDIX The complete list of 1090 English loanwords in Kansai Japanese is given below, comparing accentuation in Kansai Japanese with that in Tokyo Japanese. The list is divided into three types, six 810098. and entries are listed in Japanese alphabetical order. Asterisks indicate words whose accentuation is difi‘erent in Kansai Japanese and Tokyo Japanese. English Type II (460Worda) Kansai Japanese a'acherii a'atisuto a'amondo a'itemu a'idoru a'iborli a'irisu a'kusidento a'kushon a'lcusesu a'kusento a'lnrtaa a'kutoresu a'ppuru adobe'nchaa a'tomu a'nimaru a'bereeji HLLLL HLLLL HLLLL HLLL HLLL HLLLL HLLL HLLLLL HLLL HLLL HLLLL HLLL HLLLL HLLL HHHLLL HLL HLLL HLLLL Tokyo Japanese a’acherii a'atisuto a'amondo a'itemu a'idoru a'iborii a’irisu a'kusidento a’lrushon a'kusesu a'kusento a'kutaa a'kutoresu a'ppuru adobe'nchaa a'tomu a'nimaru a'bereeji 197 HLLLL HLLLL HLLLL HLLL HLLLL HLLL HLLLLL HLLL HLLL HLLLL HLLL HLLLL LHHLLL HLL HLLL HLLLL “almond” 6‘ 1's. 1 \ est .I'l .- u ,. v I t. r I I! \. Ila} ‘. a O amari'risu a'nkaa 8 118“?“ i'iguru i’isutaa i'buningu imite'eshon i’yarinsu i'ngurando i'ngurisshu i'ntabyuu i'npakuto we'eruzu we'dingu ekijibi'jon e'kusasaizu ekusupu'resu e'ssee e'ssensu e'pisoodo e'puron e'raa e'ria e'remento e'njin e'nzeru entaate'imento e'ntorii e'nburemu o'iru HHHLL HLLL HLLL HLLLL HLLLL HHHLLL HLLLL HHHHLLL HLLLLL HLLLLL HLLLL HLLLL HLLL HHHHLL HLLLLL HHHHLL HLLL HLLLL HLLLL HLL HLLLL HLLL HLLL HHHHHLLLL HLLLL HLLLL a’nkaa a'nguru i’iguru i'isutaa i’buningu imite'eshon i’yarinsu i'ngurando i’ngurisshu i'ntabyuu i'npakuto we’eruzu we 'dingu ekijibi'jon e'kusasaizu ekusupu'resu e'ssee e'ssensu e’pisoodo e'puron e'raa e'ria e'remento e'njin e'nzeru entaate'imento e'ntorii e'nburemu o'iru 198 LHHLL HLLL HLLLL HLLLL LHHLLL HLLLL LHHHLLL HLLLLL l-ILLLLL HLLLL HLLLL HLLL HLLLLL LHHHLL HLLL HLLLL HLLLL HLLLL “amarylHS” “exhibition” -T o'zon o'nion o'fisu o'pushon oriente'eshon orijina'ritii o'regon ka'aten ka’atorijji ka'aton ka'anibaru kaane'eshon ka'apetto ka'abon ga'ariklcu ga'ido ka'kuteru ka'sutamu ka'ttaa ka'ppmu ka’baa ka'bareeji HLLLL HLLLL HLLLLL HLLLL HLLL HHHHLLL EEEEE HLLL HHHHHLLL HHHHLLL HLLL HLLLLL HLLLL HHHLLL HLLLL HLLL HLLLL HLLL HLLL HLLL HLLL HLLLL ooga'suta ooku'rando o'osutin o'odaa ootome’eshon o'onaa o’obun o zon o'nion o’fisu o'pushon oriente'eshon orijina'ritii o'regon ka'aten sa'aden ka'atorijji ka'aton ka'anibaru kaane'eshon ka'apetto ka'abon ga'ido ka’kuteru ka'sutamu ka'ttaa ka’ppuru ka'baa ka'bmeeji 199 HLLLL LHHLLL HLLLL HLLL LHHHLLL HLLL HLLL LHHHHLLL HLLL “orientation” “coverage” -C l b . : is ti’ 5 i I I I H t l I II J "1 o~r.. ka'husu ka'ruchaa ka'ruto ka'rejji ka'rotin sa'mn ka'nzasu ka'ntorii ka’npanii ki'ipaa ki'ttin gi'huto kya'sutaa kya'suto kya'cchaa kyn'dii kya'pimm kya’binetto age HLLL HLLL HLLLL HLLLL HLLL HLLL HLLL HLLL HLL HLLL HLLLL HLLL HLL HLLL HLLL HLLL HLLL HLLL HLLL HLLL ka'husu ka'ruchaa ka'ruto ka'refii ka'rotin sa'ron ka’nzasu ka'ntorii ka'npanii ki'ipaa ki'ttin gi'huto kya'wtaa kya'suw kya'cchaa kya'pitam kya'binetto kya’bin kya'pmhon kya'pmen kya'mban gya rarii sya'mppu kya'nm kya'ndii kya'ndom kya'nbasu kya'npasu 200 HLLLL “kitchen” “cabinet” “canvas” “may \ rhuyvrgjr D gya'nburu kyu'upiddo si'mdo ku'uraa ku'kkii ku'kkingu ku'sshon w'ppii kurarin'etto ge'suto kenta'kkii ko'iru ko’in ko'osutaa go'osuto koopore'eshon ko'orasu ko'suchuumu ko'suto ko'sumosu kosmnopo’ritan ko'tton ko'teeji ko'pii ko'bura ko'mikku ko'medii ko'ramu go'ruhu ko'ronii ko'nsaato HLLL I-ILLLL HLLL HLLL HLLLL HHHHLL HHHLLL HLLLL HLLL HHHHLLL HLLL HLLLL HLL HLLL HHHHLLL EEEE HLLL HLL HLLLL kenta'kkii ko'iru ko'in ko'osutaa sobsuto kmpore'eshon ko'orasu ko'suchmnnu ko'suto ko'stnnosu kowmopo'ritan ko'tton ko'teeji ko'pii ko'bura ko'mikku ko'medii ko'ramu go 'ruhu ko'ronii ko'nsaato 201 HLLLL HLLL HLLL HLLLL HLLL LHHHLL LHHLLL HLLLL HLLL LHHHLLL HLLL HLLLL HLLL LHHHLLL HLLL EEEEEEEE “gamble” guild” “cooler” “cookie” “cooking” “sappy” “clarinet” “Kentucky" I \ O I O l I 1 ' ‘ l I . ‘ ,[rb . . .-.‘,“ J . .l hr ‘l' ‘ . U E - ‘ n‘ I . ‘ J. \ s 1" ,‘ r ‘ ‘\ ’.'a I ‘ I 3' I. Y , l . .‘ ' i I). l l' I“, . \ | ’. - . I ‘r ‘ I I — .' . ..l .- ' " - r t‘d' it " I “ I; ‘ I‘, : ( ' ‘tn’l"d ’ a I . . ' ( ‘ a l ‘9 , ‘° .‘ it 't . . ‘ (I, ' h ' I ' ‘s'i ' 5. . .l A ’ 1’7 ,.‘ ko'nseputo konse'nsasu ko'ntakuto ko'ntesuto kondomi'niamu ko'ntorasuto ko'npakuto sa'akasu sa'akitto sa'abisu sa'afaa sa'afin sa'amon sa'iensu sa'ikuringu sa'ikuru sa'ikuron sa'izu sa'ihon sa'iren sa'in saktu'ame'nto sa'supensu HLLLL HHHLLL HLLLL HLLLL HHHHLLL HLLLLL HLLL HLLLL HLLL HLLL HHHHLL HLLLL HLLL HLLL HLLL HLLLLL ke'nseputo konse'nsasu ko'ntakuto ko'ntesuto kondomi'niamu ko'ntorasuto ko'npakuto sa'akasu sa'aldtto sa'abisu sa'afaa sa’afin sa'amon sa’iensu sa'ikuringu sa'ikuru sa'ikuron sa izu sa'ihon sa iren as in sakurame'nto sa supensu sa'kkaa sa'maa sa'rada sa'rarii sa'ndee sa'npuru si'ataa si'ikuretto 202 HLLLL LHHLLL HLLLL LHHHLLL HLLLLL HLLLL HLLL HLLLL HLLL HLLL HLLLLL HLLL HLLLL HLLL HLLL HLL HLLLL HLLL HLL HLLL HLLL HLLL HLLLLL “circuit l, ‘. . 9 I} K l ". Q 0 . . O '1... . ll ,. ’ a . I a‘. ' . s r ‘3' r r J (l . E s j t. . . . t ‘ l v. . ‘ \ | " l 'V o .1" :,. ‘ J ’30 i . .I _ . J .'_-' \ :J. . ‘H 3 si'izun si'isoo' she'ebaa je'rasii she 'rutaa si’sutemu si'huto sha'abetto ja'saa ja'ketto sha'ttaa sha'waa ja'nsum jo'okaa jo'ojia sho 'kkingu she'ppinsu sho'rudaa si'riaru si'rikon si'ruku si'nguru si'nfonii si'nboru se'etaa se'orii se'kushon se'kutaa HLLL HLLL HLLL HLLL HLLLL HLLLL HLLL HLLL HLLL HLLLL HLLL HLLL HLLL HLLL si’izun si'isoo she 'ebaa je'suchaa je'rasii she'rutaa je’ndaa si'sutemu si’huto sha'abetto ja'saa ja'ketto sha'waa sha'beru sha'waa ja'nsuru jo'okaa jo'ojia sho’kkingu sho'ppingu sho'rudaa si’riaru si'rikon si'ruku si'nguru si’nfonii si'nboru se’orii se'kushon se'kutaa 203 HLLL HLLL HLLL HLLL HLLL HLLL HLLL HLLLL HLLL HLLL HLLL HLLL “shaver” “symbol” DVV' se'kkusu se'ssion se'rapii se’remonii se'rohan sense'eshon se nsaa se'ntaa se'ntensu so'kkusu so'faa ta'agetto ta'abin da'ietto ta'itoru da'ibingu ta’imaa ta'imu ta’iya ta'iru ta’kusii da'kuto da'suto ta'buretto ta'pesutorii da'rasu ta'waa ta'nkaa che'rii ti'kin cha'imu HLLL HLLL HLLL HLLLL HLLL HHHLLL HLLL HLLLL HLLL HLLLL HLLL HLLLL HLLL HLLLL sea: 5 LL EEE LL HLLLLL HLL HLLL HLL se'kkusu se'ssion se'rapii se'remonii se'rohan sense'eshon se'nsaa se'ntaa se'ntensu so'kkusu so'faa ta'agetto ta'abin da'ietto ta'itom da'ibingu ta'imaa ta'imu tsiya ta'iru ta'kusii da‘kuto ta’bmetto tape’sutorii da'rasu ta'waa ta'nkaa che'rii ti'kin cha'imu 204 HLLL HLLL HLLLL HLLL HLLL HLLLL HLLL HLLLL HLLL HLLLL HLLL HLL HLLL HLL HLLLL LHLLLL “smsation” “tile” ataxia “duct” “dust” “tablet” “Dallas” “tower” “chicken” “chime” MI— I . i r *‘I «'. t » ' .J " J2. - v .I . r . s 0‘ ‘ L . .. o) ”d cha'peru cha'ritii cha'npion chu'urippu tu'aa tu'urisuto di'suku te'kisasu te'kisuto de'suku te'smo te'nisu de'nimu te’rasu te'nshon de'nbaa to'osutaa to'onamento do'kyumento do'kutaa do 'naa to'nikku to'pikku do'mino toransi’ibaa toransure'eshon toronbo'on * to'nneru HHHHLLL HHHHHLLL HHHHLL HLLL cha'peru cha’ritii cha'npion chu'un'ppu tu'aa tu'urisuto di'suku di'supuree te'kisasu te'kisuto de'suku te'wto te'nisu de'nimu te'rasu te'nshon de’nbaa to’osutaa do’onatu to'onamento do’kyumento do'kutaa do'naa to'nikku to’pikku do'mino toransi'ibaa toransure'eshon toronbo'on tonneru 205 LHHHHLLL LHHHLL “display” “Texas” “text” “harrseiver” “translation” “trombone” “tunnel” na'iron na'nbaa nu’udoru ne'on ne'kutai ne'kkuresu no'zuru paaka'sshon ba'agen ba'ajin ba'asudee passe'nto paasona’ritii pa'atii pa'atonaa ba'anaa ba'abaa ha'amonii pa'am ha'ikingu ba'ison haibi'sukasu pa'irotto ha'usu ba’suto ha'zubando pa'zuru ba'taa pa'm HLLL HLLL HLLL HLL HLLL HLLLL HHHLLL HLLL HLLLL HHHLL HHHHLLL HLLL HLLLL HLLLL HLLLL HHHLLL na'iron na'pukin na'nbaa nu'udoru ne'on ne'kutai ne'kkuresu no'zuru W’sshon ba'agen ba’ajin ba'asudee paase’nto paasona'ritii pa’atii pa'atonaa ba'anaa ba'abaa ha'amonii pa'araa ha'ikingu ba'ison haibi'sukasu pairO'tto ha'usu ba'suto ha'zubando pa'zuru ba'taa pa'taa 206 HLLL HLLL HLLL HLLLL LHHLLL “nylon” “number” “harmony” “parlor” “hiking” “hibiscus” “pilot” “house” “powda'” “husband” “puzzle” 1‘, J.‘ ba'tahurai pa'kkeeji pa'neru ha'ntaa ha'ntingu ha'ndoru ba'npaa pa'nhmetto ha'nmaa hi'itaa Pi'inattu hi'iroo pi'kunikku pi'kurusu bi 'jinesu bi'jon pi 'suton bi 'deo hyu'usuton pi 'riodo bi'rudingu Esgssgégsggg HLLLL HLLL I-ILLLL HLLLLL HLLL HLLL HLLLL HLLL HLLLL HLLL HLLL HLLL HLLLL HLLLL ba’tahurai pa’kkeeji pa'sshon pa'nem ha'mminsu ba'buru pa'radaisu ba'ria pa'm pa'rodii pa was 0 ha'ntaa ‘ ha'ntingu ba'npaa pa'npuldn pa'nhuretto ha'nmaa hi'itaa pi'inattu hi'iroo pi'ktmikku pi'kurusu bi 'jinesu bi'jon pi'suton bi'deo hyu'usuton pi'riodo bi 'r'udingu 207 HLLLL HLLL HLLLL “peanut” “picnic” “pickles” “business” “vision” “piston” “video” “Huston” W0C!” .-. ii;.‘vd_l"""‘ “ " A . \ ' ts ,l, ., . r ‘* .', I v s' i \ ‘I‘I' l o a .‘I ‘ V 1 fa'itaa HLLL fa'inansu HLLLL fa'iru HLL fa'kutaa HLLL fa'sunaa HLLL fa'kkusu HLLL fa'sshon HLLL fa’mirii HLLL fa'ntajii HLLLL fi'ibaa HLLL fi’gyua HLL fi'kushon HLLL fi’ssingu HLLLL fe'sutibaru HLLLL fe'romon HLLL pensiruba'nia HHHHHLL fe'nsingu HLLLL fo’omatto HLLLL foome'eshon l-IHHLLL hurasutore'eshon HHHHHLLL purante'eshon HHHHLLL pmoje'kutaa HHHLLL puroda'kushon HHI-ILLL purodu'usaa HHHLLL puropo'oshon HHHLLL puromo’oshon HHHLLL be’ekarii HLLLL be'suto HLL pe'ndanto HLLLL ho'iru HLL fa'itaa HLLL fa'inansu HLLLL fa'iru HLL fa'kutaa HLLL fa'sunaa HLLL fa'kkusu HLLL fa'sshon HLLL fa'mirii HLLL fa'ntajii HLLLL fi'ibaa HLLL fi'gyua HLL fi'kushon HLLL fi'ssingu HLLLL fe'sutibaru HLLLL fe'rii HLL fe'romon HLLL pensiruba'nia LHHHHLL fe'nsingu HLLLL fo'omatto HLLLL foome'eshon LHHLLL hurasutore'eshon LHHHHLLL purante'eshon LHHHLLL puroje'kutaa LHHLLL puroda'kushon LHHLLL purodu'usaa LHHLLL puropo'oshon LHHLLL puromo’oshon LHHLLL be'ekarii l-ILLLL be’suto HLL pe'ndanto HLLLL ho'iru HLL 208 “fighter” “finance” “file” “factor” “fastener” “mas “fashion” “family” “flutes?” “fever” “figure” “fiction” “fishing” “featival” “fan” “pheromone” “Pennsylvania” “fencing” “format” “foil” 0‘ .... 1.. ,4? y. r 5 I 0" l \h ’l ‘I 7;. J I 1 l I';] v ’i s po'emu po'okaa bo'onasu bo'kusingu ho'sutesu ho'suto po'suto ho'raa po'risii ho'ridee ma'agarin ma'aketto ma’gajin ma'suku ma'sutaa ma'ttoresu ma'nia ma'huraa ma'nshon ma'nmosu mi’kisaa mi'sigan misisi’ppii mi’suterii mi'neram myu'ujiamu myu'ujikaru myu'ujikku mi’raa po'emu po'okna bo'onasu bo'kusingu ho'sutesu ho'suto pa’suto bo'suton ho'raa po'r'isii ho'ridee ma’agarin ma’aketto ma'gajin ma'suku ma’sutaa ma'ttoresu ma'nia ma'huraa ma'nshon ma'nmosu mi'kisaa mi'sigan misisi'ppii mi'suterii mi 'neraru myu'ujiamu myu'ujikaru myu'ujikku mi 'raa 209 HLLLL HLLL HLLLL HLLLL HLLLL “mansion” sl'.‘ 1 s "a a“4~ - ll... ’ ‘ \ ' . HA 3 C ll 1"]?' *‘H t l .p . | V A." I 4 .. . Q ( ‘h 1 0 '0 _' I . I r t t ' x ‘ s r v i ,1 t . ro,.--il "-"‘ ' it I mi'rakuru mi 'rion miruwo'okii mi'ruku mu'ubii mu'ubumento me'ekaa me'gahon me’jaa * me 'daru me'sseeji me'dia me'doree me’nyuu me'ritto me'rodii me'ntenansu mo'oshon mo'otaa mo'deru mo'nitaa mo 'nyumento morato'riamu mo 'raru mo 'nkii mo'nsutaa yu'uzaa yu'umoa yu'nion yu'nifoomu * ra'ion HHHLLL HLLL HLLLLL HLLL HLLL HLLL HLL HLLLL HHHLLL HLLL HLLLL HLLL HLLL HLLLL mi'rakuru mi 'rion miruwo'okii mi'ruku mu'ubii mu'ubumento me'ekaa me’gahon me'jaa medam me'sseeji me'dia me'doree me 'nyuu me 'ritto me'rodii me 'ntaransu mo'oshon mo 'otaa mo'deru mo 'nitaa mo’nyurnento morato'rlamu mo 'raru mo nlm mo’nsutaa w'um yu'umoa yu'nion yu'nifoomu 210 HLLL HLLL LHHLLL HLL HLLL HLLLLL HLLL HLLL HLL LHH HLLLL HLL HLLL HLL HLLL HLLL HLLLLL HLLL HLLL HLL HLLL HLLL LHHLLL HLL HLLL HLLLL HLLL HLLLL “miracle” “million” “Milwaukee” “milk” “movie” “niovement” “maker” “megaphone” “meastue” “medal” “message” i “media” “uniform” “lion” [e.d rhr' a ‘e All ‘I ‘ tx ra'isensu ra'itaa ra’idaa ra'ito ra'ihu ra'iburarii ra'kkaa ra'beru ra'ndorii ri'idaa rikue'suto ri'suku ri'bon ru aa ru'bii re'esaa re'edaa re'kuchaa re zaa re'sipi re’jaa re suraa re'suringu re'tasu re'dii re’baa re'beru ro'oshon roote'eshon ro oraa ro'jikku HLLL HHHLLL HLLL HLLL ra'isensu ra’itaa ra’idaa ra’ito ra'ihu ra'iburarii ra'ldcaa ra'beru ra'ndorii ri'idaa rikue’suto ri'suku ri’bon m'aa ru'bii re'esaa re'kuchaa re'zaa re'sipi re'jaa re'suraa re'suringu re'tasu re'dii re'baa re'beru ro'oshon roote'eshon 211 HLLLL HLLL HLLL HLLLLL HLLL HLLLL “license” “lighter” “rider” “light” “life” “library” “label” “WW “leader” “MW” “mks “ribbon” “lure “mbY” “lecture” “leather” “leisure” “wrestler” “wrestling” elm» “lady” “level” “lotion” “roller” ’0 0" -.. I a: 5 .J \l s. - oh. l I 212 ro'kkaa HLLL ro'kkaa HLLL “locker” ro'bii HLL ro’bii HLL “lobby” wa’ipaa HLLL wa'ipaa HLLL “wiper” wa'ihu HLL wa'ihu HLL “wife” wa'iya HLL wa'iya HLL “wire” wa'in HLL wa’in HLL “wine” wa'kkusu HLLL wa'kkusu HLLL “wax” wa'hhuru HLLL wa'hhuru HLLL “waffle” English Type L (272Worda) Kansai Japanese Tokyo Japanese Gloss * aide'a LLHL a'idea HLLL “idea” aide ntrtn LLHLLLL aide 'ntitii Ll-IHLLLL “identity” asi'sutanto LHLLLL asi'sutanto LHLLLL “assistant” ase'smnento LHLLLL ase'sumento LHLLLL “assessment” ata'kku LHLL ata’kku LHLL “attack” ada'putaa LHLLL ada'putaa LHLLL “adopter” ada'ruto LHLL ada'ruto LHLL “adult” atora’kushon LLHLLL atora’kushon LHHLLL “attraction” atora'nta LLHLL atora'nta LHHLL “Atlanta” anime'eshon LLHLLL anime'eshon LHHLLL “animation” ahutanu'un LLLHLL ahutanu'un LHl-ll-ILL “afternoon” apurike'eshon LLLHLLL apurike'eshon LHHHLLL “application” apuro'oti LLHLL apuro'oti LHHLL “approach” amyu’uzumento LHLLLLL amyu'uzumento LHLLLLL “amusement” ame'nitii LHLLL ame’nitii LHLLL “amenity” ara'amu LHLL ara'amu LHLL “alarm” ari’ina LHLL ari 'ina LHLL “arena” ie’roo LHLL ie'roo LHLL “yellow” imajine'eshon LLLHLLL imaj ine'eshon LHHHLLL “imgination” ‘7 a' ‘- I I s .I ‘1 but OI“ . I 'i ~.' 0- ‘A ‘ 0.1 ff} ‘ insutora'kutaa insupire'eshon intone'eshon infome’eshon inhure'eshon inpure'sshon inbe'edaa ui'sukii wisuko'nsin ui'nku ui'nti ue'itaa ue'itoresu ue'suto eko’nomii eko'rojii ooso'n'tii oha'io sme'eii kyapa'sitii kyanpe'en kui’in kuo'ritii gura'idaa LLLLHLLL LLLLHLLL LLLHLLL LLLHLLL LLLHLLL LLLHLLL LLHLLL LHLLL LLHLLL LHLL LHLL LHLL LHLLLL LHLL LHLLL LHLLL LLHLLL LHLL LLHLLL LLHLLL LHLLLL LHL LHLL LLHLL LHLL LHLLL LLHLL LHLL LHLLL LHLLL LHLLLL insutora'kutaa insupire'eshon intone’eshon infome'eshon inhure'eshon inpure'sshon inbe'edaa wisuko'nsin ui’nku ui'nti ue'itaa uc'itoresu eko'nomii eko'rojii ooso'ritii oha'io obuza’abaa anime'eshon ltau'nseraa ka'kao kase'tto kafete'ria sme'eji kyapa'sitii kyanpe'm kui'in kuo'ritii gm'a'idaa 213 LHHHHLLL LHHHHLLL LHHHLLL LHHHLLL LHHHLLL LHHHLLL LHHLLL LHLL LHLLLL LHLLL LHLLL LI-IHLLL LHHLLL LHHLLL LHLLLL LHHLL LHLLL LHHLL LHLLL LHLLL LHLLLL “intonation” r.- “I it I a. ' ..I.4 s 3" r 's i, ‘ . 3 ,1‘4’0 kura’kkaa LHLLL kura'tti LHLL kura'bu LHL gura'maa LHLL kuri'iku LHLL kuri'imu LHLL guri'in LHLL kurie'etaa LLHLLL kuri'sutaru LHLLLL kuri'suchan LHLLL kuri‘ppu LHLL auri'ppu LHLL kuri'nti LHLL km‘u'u LHL amu'urm LHLL kme'e LHL gure'e LHL kure'en LHLL kme'jitto LHLLL kure'nzaa LHLLL lmro'ozetto LHLLLL kuro'obaa LHLLL guro'obu LHLL kuro'oru LHLL knro'on LHLL koka'in LHLL koma'ndo LHLL komi'sshon LHLLL komyrmike'eshon LLLHLLL komyu'nitii LHLLL kome'dian LHLLL kura'kkaa LHLLL km'a'tti LHLL ku'rabu HLL gura'maa LHLL kuri'iku LHLL kuri'imu LHLL guri'in LHLL kurie'etaa LhHLLL kuri'sutaru LHLLLL kuri'suchan LHLLL kuri'ppu LHLL auri'ppu LHLL kin-i'nti LHLL kuru'u LHL sunl'upu LHLL kure'e LHL gure'e LHL kure'en LHLL kure'jitto LHLLL kure'nzaa LHLLL kuro’ozetto LHLLLL kuro'obaa LHLLL guro'obu LHLL kuro'oru LHLL kuro'on LHLL koka'in LHLL komando LHHH komi'sshon LHLLL komyunike'eshon LHHHLLL komyu'nitii LHLLL kome'dian LHLLL 214 “cracker” “clutch” “club” “clinch” kore'kushon ‘ kore'kutaa konsa'rutanto konsa'rutingu konda'kutaa kondi'shon konte'na konde’nsaa kontoro'oru konpa'nion konbine'eshon konpyu'utaa konpure'ssaa konbe'yaa saba'ibaru LHLLL LHLLL LLHLLLL LLHLLLL LLHLLL LLHLL LLHL LLHLLL LLLHLL LLHLLL LLLHLLL LLHLLL LLLHLLL LLHL LHLLL LHLL LHLLL LLHLLL LLHLLL LHLL LHLLL LHLL LHLL LHLL LHLL LHLL LHLL LHLL kore 'kushon korekutaa konsa'rutanto konsa'rutingu konda'kutaa kondi'shon ko’ntena konde'nsaa kontoro'oru konpa'nion konbine'eshon konpyu'utaa 215 LHLLL LHHLLLL LHHLLLL LHHLLL LHHLL LHHLLL LHHHLL LHHLLL LHHHLLL LHHLLL LHHHLLL LHLLL LHLLL LHHLLL LHLL LHLL LHLL LHLL LHLL “collection” “collector” “simulation” “school” 'i .3 i, (‘1- r 1 l \ \ a . . ‘ ‘ ( t . . -,,.n'. ' M e v.... . .. ' ‘li ' I I l 5 Dr 5 i T ‘1‘: tu- t t . I e “l ‘u'; .' - _'il,' . fl: fl ' “'1"“J) l“. I 7. i.- It I. o-u e v . s - I s s I v a \ 7 I ‘ l s ‘ a O 'l r . . . . .re a r . r.' .. .‘l .' l .. r 17 sukru'a'ppu sukm'i’in suke'eto suke'eru suke'tti suko'a suta'a suta'ato suta'iru suta'jiamu suta'hhu suta'nsu suta'npu suti'imu sute'eki sute'eji sute'eshon sute'etasu sute’kki suto'a suto'obu suto'orii suto'kku sutora'iki sutora'ipu sutori'ito suna'kku supfi supa'isu sum'na LLHLL LLHLL LHLL LHLL LHLL LHLL LHLLL suke'tti suko'a suta'a suta'ato suta'iru suta'jiamu suta'hhu suta'nsu suta'npu suti'imu sute'eki sute'eji sute'eshon sute'etasu sute'kki We PP“ suto'a suto'obu suto’orii suto'kku sutora'ipu sutori'ito suna’kku m'i supa'isu 811]” 08 216 LHHLL LHHLL LHLL LHLL LHLLL LHLL LHLL LHLL LHHLL LHLL LHLL LHL um” “screen some, “scale” “sketch” -Q ,I a-‘o. r, "?O ,sLLOI' m'n supi'ikaa supi'iti supi'ido supi'ritto m'tm supuri’ntaa supe’a sm'esu snpe'rinsu unpo'oku amo'otu mpo'nsaa suma'iru sumo'oku sumo 88“ sura'ngu smu'npu suri'ppa suro'ogan suro'opu sere’kushon turi'i disuka'unto disuka'sshon tekuno'rojii dekore'eshon deza’ato deza'inaa LHLLL LHLL LHLL LHLLL LHLL LLHLLLL LLHLLL LLHLLL LHLLL LLHLLL LLHLLL LHLLL m'n supi'ikaa supi'iti supi'ido supi'ritto m'lm supuri'nkuraa supuri'ntaa supe a supe'esu supe'rinsu amo'olm suro'nsaa suma'iru sumo'oku sumo 88“ sum'ngu suta'npu su'rippa sum 0830 suro'opu sere'kushon trn'i'i disuka'unto disuka'sshon tekuno'rojii dekore'eshon deza'ato deza'inaa 217 LHLLL LHLLL LHLL LHHLLL LHLLL LHHLLL LHLLL LHHLLL LHHLLL LHLL LHLLL . .U r - I I c. . Q t ‘ I] I I l V ' , . . . . r . l I as. .t |.. 0 ll - . .. fl . r .l’ tr - m I t .l Yr. ' a t 0 .ll . II V . . - a r .s 1‘] al.. I '0 ... a O r . r. A (u . l I Isl f a ea. deza'in detoro'ito deme'ritto toma'to dora'ibaa dora'ibu tori'itomento dori'imu tori'ppu dori'ppu dori'buru dori'ru dori'nku tore'enaa toreeningu tore'eraa tore'ndo toro'oti nare'eshon nebura'suka LHLL LLHLL LHLLL LHLLL LHLLL LHLLL LHLL LHLL LHLL LHLLLL LHLLLLL LHLL LHLL LHLL LHLL LHLL LHLLL LHLLLL LHLLL LHLL LHLLL LLHLL detoro'ito deme'ritto to'mato doraibaa dora'ibu domiyaa toqu don ssn toradi'shon tora'buru tora'nku tora’nporin tori'itomento dori'imu dori'nku tore'esu tore'enaa tore'eningu tore'eraa tore'ndo toro'oti nare’eshon nebura'suka 218 LHLLLL ,H 4*: i‘ l 's . l. .1“- I l: '»v(,f. o as 9 ‘04 V" baibure'eshon bairi'ngaru bake'eshon patoro'oru pafo'omansu barie'eshon baru'un pare'edo bigi'naa hui'rumu pura'isu pura'ibasii bura'usu bura'un hura'sshu huramrngo hura'waa puri'nsesu buru'u bura'usu hum 'utu huru'uto pmu'un LLLHLLL LLHLLL LLHLL LHLLLL LHLL LHLL LHLL LHLLL LHLL LHLL LLHLL LHLL LHLL LHLLLL LHLLL LHLL LHLL LHLLLL LHLLL EEEEE baibure'eshon bairi'ngaru bake 'eshon ba’nana pafo'omansu barie'eshon baru'un m'edo ha'wai bi 'ginaa hui'rumu pura'isu pura'ibasii bura'usu bura'un hura'sshu hurami'ngo hura'waa bura’nku puranku'ton huri'izaa buri'jii puri'nsu M'nsesu buru’u buru'usu huru'utu hum 'uto puru'un 219 LHHHLLL LHHLLL LHLLL LHHLL LHLLLL LHHLLL HLLL LHLL LHLLL LHLL LHLL “flamingo” “flower” “blank” ‘plankton” “bridge” “Princeton” “blue” I glt 1;. i a ii I a fle . 6.‘, I r , 1. .l r ' ' '0 I) .3 ' s uric! C - 'I}.- 2' U A U bure'in m’e bure'eki pure'eyaa bure'zaa pure'jidento bure'suretto purezente’eshon pure'zento pure'sshaa huro'a buro'iraa buro'oti me'jdmto puro'sesu buro'kku pote'to masi'in maji'shan massa'aji myuuji'shan moza'iku motibe'eshon ria'ritii risa'ati rim'oto riha'asaru rifo'omu rekm'ie'eshon LHLL LHL LHLL LHLLL LHLL LHLLLL LHLLLL LLLLHLLL LHLLL LHLLL LHL LHLLL LHLL LHLLL LHLL LHLL LHL LHLL LHLL LLHLL LHLL LHLL LLHLL LHLL LLHLLL LHLLL LHLL LHLL LHLLL LHLL LLLHLLL bure'in pure’e bure'eki pm'eyaa bure'zaa pure'jidento bme'suretto purezente'eshon pure'zento huro a buro'iraa buro'oti puro'jekuto m'sesu buro'kku pa'teto masi'in maji'shan massa'aji mizu uri myuuji'shan moza’iku motibeeshon ria'ritii risa'ati rizo'oto riha'asaru rifo’omu rekurie'eshon 220 LL LHLL LHLLLL LHLLLL LHHHHLLL LHLLL LHLLL LHLL LHLL HLL LHLL LHLL LHLL LHHLL LHLL LHLLL LHLL LHLL LHLLL LHLL LHHHLLL n“ ‘ .0 ...\ .I ' l t . . l 4 J . . . a I . Y . 1 t ‘ .‘ I l ‘- .’ . ob ". .. ! -('1, VI! resi'ito rese'pushon mpo'oto reri’ihu roke'eshon LHLL LHLLL LHLL LHLL LHLLL resi'ito rese'pushon repo'oto reri'ihu roke'eshon Non-Eaglbh Type B (150 Words) Kansai Japanese aaka'nsoo a'akeedo aari'nton a’kusesarii akuroba'tto ajite'etaa W’sasu asufa'ruto a'dobaisu adobante'eji anau’nsaa anari'suto apuriko'tto arige'etaa arumini'umu anka'rejji anpa'ia inisia'tibu i’nisharu irasutore’etaa indianapo'risu inde'kkusu HHHLLL HLLLL HHHLLL HLLLLL HHHHLL HHHLLL HHHHLL HHHLL HLLLL HHHHHLL HHHLLL HHHLL HHHHLL HHHLLL HHHHLL HHHLLL HHHLL HHHHLL HLLL HHHHHLLL HHHHHHLL HHHLLL Tokyo Japanese aaka'nsoo a'akeedo aari’nton a'kusesarii akuroba'tto ajite'etaa asupmsm asufa'ruto a'dobaisu adobante'eji anau'nsaa anari'suto apuriko'tto arige'ctaa arumini’umu anka'rejji anpa'ia inisia'tibu i'nisharu irasutore’etaa indianapo'risu inde'kkusu 221 LHLL LHLLL LHLLL LHHLLL HLLLL LHHLLL LHHLLL LHHLL HLLLL LHHLLL LHHHLL LHHLLL LHHLLL LHHLL LHHHLL LHHHHLLL LHHLLL “reception” “relief” “location” n». ..o!_ I I 0 ‘. ,. l I. 1 ‘0 r . l'. ‘ i d x’ ' 1 I d r r \- d ‘ s I winbu’rudon earobi’kusu ekisu’tora ekisupa'ato egoi'suto esukare'etaa epito'osu emera'rudo erekutoroni'kusu erebe'etaa o'asisu ookesu'tora okura'homa okkusufo’odo omuni'basu kaaba'ido kasuta'ado katapa'ruto ka'nuu karikyu'ramu karihura'waa 'ria kuraima'kkusu ku'rasu su'msu gu'rahu kurike'tto kurisu'masu gu'riru HHHHLLL HHHLL HHHHLLL HHHLL HHHHHHLL HHHLLL HHHHLL HHHLL HHHHHLL HHHLL HHHLL HHHLL HHHHLLL HLL HLL HHHLL HHHLL winbu'rudon earobi'kusu ekisu'tora ekisupa'ato egoi suto esukare'etaa epim'osu emera'rudo erekutmoni'kusu erebe'etaa o'asisu ooke'sutora okura'homa okkusufo'odo kaaba'ido kasuta'ado katapa'ruto ka'nuu ka'rikyuramu karihura'waa kya'ria kuraima'kkusu ku'rasu su’msu gu'rahu su'mbu kurike'tto kurisu’masu guriru 222 LHHHLLL LHHHLL LHHLL LHHI-ILL LHHHLLL LHHLL LHHHHHLL LHHLLL HLLL LHHLLL LHHHLLL LHHLL LHHLL “Wimbledon” “aerobics” nexus, “expert” “egoist” “epilogue” “emerald” “electronics” “elevator” “oasis . . . ' ‘ t- . a r I “” V . t ‘ ”e . . - I a. , l a , .’.l 'l. I. l '- I ’0 u‘ l . v ‘ . . ‘ ... . < . l a l . t . I . u . o l . 1 l I ‘ t""."“l '.‘ ku'rosu koodinc'etaa konctikn'tto komsutero'om ko'mna konko'osu ko'nsento kondo'omu suro'in saikuroto'ron saibo'ogu satem'ito sa'fari sandoi'tti jaanm'i'stno jaanari'zumu shande’ria si'atoru sinsi'nati sinme'torii soose'eji dainama'ito daiynmo'ndo ta'nbarin tinpa'njii HLL HHHHLLL HHHHLL HHHHHLL HHHLL HLLLL HHHLL HHHHHLL HHHLL HHHHLL HHHHLL HHHLL HLLL HHHLL HHHLLL HHHHLL HLLL HHHHLL HHHHLL HLLLL HHHLLL ku'row koodine'etaa konetikn'tto koresutero'oru ko'rona konko'osu ko'nsento kondo’omu sam'in saikumto’mn saibo'ogu satm'ito sa'fari sandoi’ni jaanari'suto jaanati'mu shande'fin aha npuu sia'toru sinsi'nati sinme'torii sukottom'ndo sutaMa'ado su titan se'dan so'oseeji dainama’ito daiyamo'ndo ta'nbarin tinpa'njii 223 LHHHLLL LHHHLL LHHHHLL HLLLL LHHLL LHHLL LHHHHLL LHHLL LHHLL LHHHLL LHHHLL LHHLLL LHHHHLL LHHHLL LHHHLL HLLLL LHHHLL HLLLL LHHLLL . 9. v '. o. t. . . n I . . I. ‘ .c I . u a . s A .I It «I. r .0! \I u ’I . ll . t . . , 1 a ‘47 - . . q \o h ( ' u . u \’ I I, . .0. I O I, . h .7 \ u . 0 .n c E I» . I . 1 b . o . ' ‘ ‘ I ‘ I. AD: (l. \.‘ .11 All \ A ' o ’ j . . . I A} f, , l. . I I, o . v Q. . 4 . ‘ . . t . ~ ~ lIal' te'klmikku te'nesii derawc'a terori'suto term-i'zumu do’mmn do'ramu toranpe'tto do'mu to'rofii nashonari'zumu msi'suto hanemu'un pm-ashu'mo bafike'edo harike'en baruko'nii hanba'agaa hanmo’kku I-ILLLL tc'lmnikku HLLLL HLLL te'ncsii HLLL demo-samba. HHHHHHLLL W'uhon LHHHHHLLL HHHL dcmwe'a LI-IHL HHHLL temri’smo LHHLL HHHLL terofi'zumu LHHLL HLL do 'rama HLL HLL do'mmu HLL HHHHLL tomnpe'tto - LHHHLL HLL do'resu HLL HLLL to'rofii HLLL HHHHLL nashonari'zumu LHHHLL HHHLL namsi'suto LHHLL HLL ne'bada HLL HHHl-ILLL paakom'ctaa LHHHLLL HHHLLL baate'ndaa LHHLLL HHHLL baabc’kyuu LI-IHLL HLLLL baamo'nto LHHLL HHHLL haija'kku LHHLL HHHLL basuke'tto LHI-ILL HHHLL pasupo'oto LHHLL HLL baton LHH HLL ba'nim HLL HHHLL hanemu'un LHHLL HHHLL panshu'mo LHHLL HHHLL barike'edo LHHLL HHHLL harike'en LHHLL HHHLL bamko'nii LHHLL HHHLLL hanba'agaa LHHLLL HHHLL hanmo'kku LHHLL HHHLL piani'suto LHHLL 224 “Tennessee” “Delaware” “nationalism” I. ‘BIQ. ‘7' I ‘ ‘ ' i. : .3‘HH‘ :- ‘ ' ‘i‘,"', ‘t , \‘1 , 3 . ‘ . ) 9 ’ “l ’4 l . 4«.2 _. I ~ ‘ O ' l i E . .- ' » ‘ E 4 . c ‘ - I o I r 4 n ‘ t . 1 I . ‘ k 1 l 0 \ w v ' u v bisuke'tto hittiha'iku pittuba'asu hiyasi'nsu firadem'fia pirami'ddo biriya'ado femini'zumu pmaka'ado purasuti'kku pm'attoho'omu hurancha'izu purosu'ramaa pmsu'mmu pmoro'ogu heriko'putaa hemme'tto bomte'eji ma'agaretto maamare'edo maame'ido magme'tto masachuuse'ttu masuka'tto masuko'tto masuta'ado masshm'u'umu mekani'zumu medari'suto meriigo'orando meriira'ndo HHHHLL HHHHLL HHHLL HHHHLL HHHLL HHHLL HHHLL HHHHLL HHHHHLL HHHHLL HI-IHLLL HHHLL HHHLL HHHLLL HHHLL HLLLLL HHHLL HHHLL HHHLL HHHLL HHHLL HHHHLL HHHLL HHHLL HHHHLLLL HHHHLL bisuke'tto hittiha'iku pittuba'agu hiyasi'nsu firademfi'a biriya'ado femini'zumu purasuti’kku purattoho'omu hurancba'izu W'mmaa W'ramu puma'osu heriko'putaa he'mmetto bomte'eji ma'agaretto maamare edo maame'ido magune'tto masachuuse'ttu masuka'tto masuko'tto masuta'ado masshm'u'umu mekani'zumu medari'suto meriigo’orando meriira'ndo 225 LHHLL LHHHLL LHHLL LHHLL LHHLL LHHLLL LHHLL LHHLLL HLLLL HLLLLL LHHHLL LHHLL LHHHHLL LHHLL LHHLL LHHHLL LHI-ILL LHHLL LHHHLLLL LHHHLL “hitchhike” “hyacinth” “Philadelphia” Wmid” “billiards” “feminism” “placard” “plastic” “platform” “mermaid” “magnet” “Massachusetts” “mnscat” “mustard” “mushroom” “mechanism” I.- I. ,'. o. Al . Po‘ 9 | y} l a. T13: . .4.. s I o I n’ 1 . . . t t. v ‘F. 3‘ ll r .l I. . l l n A I . 0. ‘l . . .. A t I. av ' .~ I V 312‘ J: I. I. . a . 4 p . O A l i . o I c Q . a. . u . . . < . ... 4 ‘I 1 o‘ t . v . u w v L. . . . . I J“ . I I . 3 4 . a. i - - . I i ;' l ' . ‘ -, . .o. monokuro'omu HHHHLL monoto'on HHHLL re'ferii HLLL ro'mansu HLLL Non-English Type L Kansai Japanese * ai'su LHL * akoodi'on LLLHLL anbu'rera LLHLL ime'eji LHLL * insuta'nto LLLHLL inta'abaru LLHLLL inta'an LLHLL inte’ria LLHLL edi'nbara LHLLL enji 'nia LLHLL opere'etaa LLHLLL ori'ibu LHLL ore'nji LHLL * gau'n LHL " gaso'rin LHLL * kame'ra LHL kamere'on LLHLL kare'ndaa LHLLL * kyara'kutaa LHLLL * kui'zu LHL * kmiini'ngu LLLHLL ' koko'a LHL konkuri'ito LLLHLL (65 Words) monokuro'omu LHHHLL monoto'on re'ferii ro'mansu Tokyo Japanese a'isu ako'odion anbu’rera ime'eji i'nsutanto inta'abaru inta’an inte'ria edi'nbara enji'nia om'etaa ori'ibu one'nji sa'un gasorin ka'mera kamere'on kane'ndaa kya'rakutaa ku'izu kuri'iningu ko'koa konkuri'ito 226 LHHLL HLLL HLLL LHLLLL LHLL HLLLLL LHHLL LHHLL LHLLL LHHLL LHHLLL LHLL LHLL LHHLL LHLLL HLLLL LHLLLL LHHHLL “monochrome” “monotone” “referee” “romance” .a I; 4 o, . I e a. ' a. a ,‘ cl 4 I . I . I . u l ‘| . . 1 its. I - I o . J u ,1 rl‘ . - .. . I n . m . 1‘ .’ a I I ‘ . .... e f . ‘ a D - I‘ l 1 ‘ I IA l I . s ; 'II‘ 1. .\u I . 4' ‘ .‘ III I r u .I I . . . n .a . I a ... . I o I . v v . f )- ' e . .n ‘9 pt I C ' - - ... 0 r v _ .ta I . D . . ,1: t p D o 1 t t .. F e. . V. ' e | I. OI. . s a a o '5‘ a ‘J 1 l .l u 3 o a f a l O - c o A l 227 konba'in LLHLL konba'in LHHLL “combine” konpa'undo LLHLLL konpa'undo LHHLLL “compound” konpme'kkusu LLLHLLL konpure'kkusu LHHHLLL “complex” safa'ia LHLL safa'ia LHLL “sapp ' ” siri'izu LHLL si'riizu HLLL “series” suku'ramu LHLL suku'ramu LHLL “scrum” suku'ryuu LHLL suku'ryuu LHLL “screw” suchuwa'aadesu LLHLLL suchuwa'aadesu LHHLLL “stewardess” suchuwa’ado LLHLL suchuwa'ado LHHLL “steward” suto'resu LHLL suto'resu LHLL “stress” suto'roo LHLL suto'roo LHLL “straw” sutorobe'rii LLLHLL sutorobe’rii LHHHLL “strawberry” supu'ree LHLL supu'ree LHLL “spray” supeshari'suto LLLHLL supeshari 'suto LHHHLL “specialist” sero'ri LHL se'rori HLL “celery” dame 'eji LHLL dame'eji LHLL “damage” chokore'eto LLHLL chokore'eto LHHLL “chocolate” teku'nishan LHLLL tekuni'shan LHHLL “technician” dcbem'ppaa LLHLLL debcro'ppaa LI-IHLLL “developer” dcri'kasii LHLLL deri'kasii LHLLL “delicacy” topa'azu LHLL topa'azu HLLL “topaz” bake 'tu LHL baketu LHH “bucket” pase'ri LHL pa'seri HLL “persley” pata'an LHLL pata'an LHLL “pattern” hiro'ine LHLL hiro'ine LHLL “heroine” pena'rutii LHLLL pena'rutii LHLLL “penalty” hero'in LHLL hero’in LHLL “heroin” poke 'tto LHLL poke 'tto LHLL “pocket” bora'ntia LHLLL bora'ntia LHLLL “voltmteer” * maji'kku LHLL ma'jikku HLLL “magic” mane'ejaa LHLLL mane’ejaa LHLLL “manager” ‘l I . I v ... 9 l r ‘ 4 ~ a,‘ 4.01 '.\:.J l i 1 ’ at l ‘r ' \ “J a... . . . 0’ , ,3 . -~ . " a: " 4’ t\ r . - ‘0'). J. J ' _ . ,7 4 I 4. \‘ ..: s V Q ( “ ' I 1;." '5' - I I ‘I-~v0 misa’iru LHLL misa'iru moju'uru LHLL moju'uru yoogu'ruto LLHLL yoogu’ruto rake 'tto LHLL rake'tto rajie'etaa LLHLLL rajie'etaa * raji’o LHL ra'jio reko'odo LHLL reko’odo repu'rika LHLL repu’rika roke'tto LHLL roke'tto * robo'tto LHLL ro'botto wasi'nton LHLLL wasi'nton Unaccented Type B (134 Words) Kansai Japanese Tokyo Japanese aiowa HHHH aiowa aikon HHHH aikon aidaho I-IHHH aidaho airon HHHH airon * adobaizaa HHHHHH adoba'izaa * adoresu HHHH a'doresu anarcgu HHHH anamgu * aparatia HHHHH apara'tia * apareru HHHH a'pareru abogado mam abogado amachua HHHH amachua smerika HHHH emerika arasuka HHHH arasuka arabama HHHH arabama arizona HHHH arizona aribai HHHH aribai 228 LHLL LHHLL LHLL LHHLLL LHLL LHLL HLLL LHLLL “missile” “module” ‘Voshm” Gloss “Iowa “icon” “Idaho” “avocado” antena iguana ibento irinoi inshurin * indiana karifonmia * hmufinsu * kureetaa gureedo kureernu HHHH HHHH km'uiinsu kine'etsa kureemu 229 LHLLL LHLLLL LHLLL “antenna” 'l‘ 1" *keeburu komento kororado *jaanaru sutajio sutando taimingu daiyam dabinsu * memo * tiketto * chuunaa *diiraa toppinsu torooringu HHHHHH ke'eburu kootingu komento sa'almru ja'anaru smajio 230 all ‘ a ‘ "Mp Q > VI 1. O ' t'. l . ‘l' '0‘ 1 I O 4 I 1. M. *o l naiagara *baajinia *bmjon fiin'ngu fiirudo hurorida Pa'akhsu baq'i’nia ba’ajon hure'ezu hurorida 231 LHLLL LHLLL “Florida” “plate” f .' I s e I l i J t I I . f 0 32- I l 0 A0 I hureemu HHHI-l purobaidaa HHHHHH * worm HHHH buronzu HHHH burondo HHHH bearingu HHHHH poem HHHH perikan HHHH beruto HHH pensin HHHH pointo HHHH boorlngu HHHHH botoru HHH * maaketingu HHHHHH manikyua HHHH miitingu HHHHH * misshon HHHH minichua HHHH minesota HHHH * memorii I-IHI-IH * raunji HHHH * rankingu HHHHH retaringu HHHHH * waiomingu HHHHHH Unaccented Type L (9 Words) Kansai Japanese okura LLB karee LLH hureemu pointo botoru maake'tingu miitingu mi'sshon minichua minesota me morii ra'unji ra'nkingu retains“ waiomi'ngu Tokyo Japanese okura 232 LHHH LHHHHH HLLL HLLL HLLLL “lounge” “Louisiana” “lettering” “Wyoming” 46“” ——~ * sutookaa dokkinsu baipasu manekin * rekoodingu LLLLH LLLH LLLLH LLLH LLLH LLLH LLLLLH marason reko'odingu 233 LHLLLL “stalker” “gelatin” “docking” BIBLIOGRAPHY Akahori, Kanji (ed). 1999. Hyoojyun Pasokon Yougo Jiten [A Standard Dictionary of Computer-related Words]. Tokyo: Syuwa System. Akinaga, Kazue. 1960. Tokyo Akusento Hosoku ni Tsuite [0n the Laws of Tokyo Accent]. InI-Iaruhiko Kindaichi, ed., Meikai Nihongo Akusento Jiten [Japanese Accent Dictionary]. 1-68. Tokyo: Kuroshio Sanseido. Asano,Makiko. 1999. GairaigoniokeruAkusentoKinshiRyooikitoAkusentogatano Henka [A Constraint-based Analysis of Loanword Accentuation in Japanese]. In Y. A. Sasaki, ed., Gengogaku to Nihongo Kyoiku [Linguistics and Japanese Pedagogy]. 65- 79. Tokyo: Kuroshio Publishas. Beckman, Jill. 1998. Positional Faithfirlness. Ph.D. dissertation, Amherst: University of Massachusetts. [RDA-234, http:/lmamtgers.edu]. Benua, Laura. 1995. Identity Effects in Morphological Truncation. University of Massachusetts Occasional Papers in Linguistics 18, ed. by Beckman, Walsh Dickey, and Urbanczyk, 77-136. Amherst, MA: GLSA. [RCA-74, http://roarutgersedu/l Bloch, Bernard. 1950. Studies in Colloquial Japanese IV : Phonemics. Language 26. 86- 125. Downes, William. 1998. Language and Society: Second Edition. Cambridge: Cambridge University Press. FukauwaflmnandMafilyuKimhmmappear.RmkingParadoxinConsonant Voicing in Japanese. Voicing in Japanese. Berlin: Mouton de Gruyter. Fukuzawa, Haruka, Mafuyu Kitahara, and Mitsuhiko Ota. 1998. Lexical Stratification andRankingInvarianceinConshahrt-basedGmmmars.PmceedingsofflreClS 34. [RCA-267, http://roamtgersedu]. Fukazawa, Haruka, Mafuyu Kitahara, and Mitsuhiko Ota. 2002. Constraint-based Modeling of Split Phonological Systems. On’in Kenkyuu [Phonological Studies] 5. 115-120. Hamans, Camiel. 1997. Clipping in Modern French, English, German and Dutch. In R. Hickey and S. Puppel, eds., Language History and Linguistc Modeling: A Festschrift for Jacck Fisiak on his 60th Birthday. 1733-1741. Berlin: Mouton de Gruyter. Hayashi, Oold (ed). 1982. Zusetsu Nihongo [Illustrated Japanese]. Tokyo: Kadokawa Shoten. 234 ..J Hayes, Bruce. 1980. A Metrical Theory of Stress Rules. Ph.D. dissertation, MIT. Published by Garland Press, New York, 1985. Horiuchi, Katsuaki, ed. 1996. Kataltana Gairaigo Ryakugo Jiten [Abbreviated Loanword Dictionary]. Tokyo: J iyuu Kokuminsya. Inaba, Seiichiro. 2001 . Ninchiteki-hyoso-3renhaku-firtto-to Rizumu-shido [Cognitive Surface Ternary Feet and Rhythm Instruction]. In Masahiko Minsmi and Yukiko Alam Sasaki, eds., Linguistics and Japanese Language Education 11. 293-295. Tokyo: Kusoshio Publisher. Ito, Jtmko. 1986. Syllable Theory in Prosodic Phonology. Ph.D. dissertation, Amherst: University of Massachusetts. Published by Garland Press, New York, 1988. Ito, erko. 1989. A Prosodic Theory of Epenthesis. Natural Language and Linguistic Theory 7. 217-259. Ito, Junko. 1990. Prosodic Minimality in Japanese. In K. Deaton et al., eds., Chicago Linguistic Society 26: Parasession on the Syllable in Phonetics and Phonology. 213- 239. Chicago: Chicago Linguistic Society, University of Chicago. Ito, Junko, and Armin Mester. 1992. Week Layering and Word Binarity. LRC-92-09, Linguistics Research Center, UCSC, Santa Cruz, CA. Ito, Junko, and Armin Mester. 1995. Japanese Phonology. The Handbook of Phonological Theory, ed. by John Goldsmith. 817-838. Oxford: Blackwell Publisher. Ito, erko, and Armin Mester. 1997a. Correspondence and Cornpositionality: The Ga-gyo Variation in Japanese Phonology. Derivations and Constraints in Phonology, ed. by Iggy Roca, 419-462. Oxford: Clarendon Press. Ito, erko, and Armin Mester. 1997b. Sympathy Theory and German Tnmcations. University of California, Santa Cruz ms. [RCA-211, http:/Imamtgers.edu/] Ito, Junko, and Armin Mester. 1997c. Featursl Sympathy. Phonology at Santa Cruz 45. 29-36. Ito, Junko, and Armin Mester. 1998. Markedness and Word Structure: OCP Effects in Japanese. Santa Cruz: UCSC ms. [RCA-255, http://roarutgerseduf]. Ito, Junko, and Armin Mester. 1999. The Phonological Lexicon. In Natsuko Tsujimura, ed., The Handbook of Japanese Linguistics. 62-100. Oxford: Blackwell Publisher. Ito, Junko, and Armin Mester. 2001a. Covert Generalizations in Optimality Theory: the Role of Stratal Faithfulness Constraints. Studies in Phonetics, Phonology and Morphology 7.2. 273-299. 235 "h Ito, erko, and Armin Mester. 2001b. Structure Preservation and Stratal Opacity in German. Segmental Phonology in Optimality Theory, ed. by Linda Lombardi. 261-295. Cambridge: Cambridge University Press. Ito, Junko, and Armin Mester. 2003a. Lexical and Postlexical Phonology in Optimality Theory: Evidence from Japmese. Ms., University of California, Santa Cruz. rm, Junko, and Armin Mester. 2003b. Japanese Morphophonemics: Markedness and Word Structure. Cambridge: The MIT Press. Jorden, Elenor H. and Hamako I. Chaplin. 1976. Reading Japanese. New Haven and London: Yale University Press. Jun, Jongho. 1995. Perceptual and Articulatory Factors in Place Assimilation: An Optimality Theoretic Approach. Ph.D. dissertation, UCLA. Kager, Rene. 1999. Optimality Theory. Cambridge: Cambridge University Press. Kanno, Ken. 1971. Gairaigo Akusento (3) [Loanword Accentuation (3)]. Bunkyu Geppou [Literature Research Journal] 3. 47-55. Katayama, Motoko. 1995. Loanword Accent and Minimal Reranking in Japanese. Phonology at Santa Cruz 4. 1-12. Kay, Gillian S. 1995. English Lomwords in Japanese. World Englishes. 14(1). 67-76. Kenstowicz, Michael, and Atiwong Suchato. 2004. Issues in Loanword Adaptation: A Case Study from Thai. Ms., MIT. Kenyon, John Samuel, and Thomas Albert Knott. 1944. A Pronouncing Dictionary of American English. Springfield: G. & C. Merriam Company. Kindaichi, Haruhiko, and Kazue Akinaga, ed. 2001. Shimmeikai Nihongo Akusento Jiten [Japanese Accent Dictionary]. Tokyo: Sanseido. Kobayashi, Mina, Hiroko Quackenbush, and Atsushi Fukuda. 1991. Gairaigo-m Mirareru Nihongokakisoku-no Syuutoku [The Acquisition of English Loanwords in Japanese]. Nihongo Kyouiku [Japanese-language Education] 74. 48-59. Kubozono, Haruo. 1988. The Organization of Japanese Prosody. Ph.D. dissertation, Edinburgh University. [Published by Kuroshio Publishers, Tokyo, 1993.] Kubozono, Haruo. 1994. Syllable and Accent in Japanese: Evidence from Inanword Phonology, Ms., University of California, Santa Cruz. 236 .1". 31.1' 9 ll ’. D" It Kubozono, Haruo. 1995a. Constraint Interaction in Japanese Phonology: Evidence from Compound Accent. Phonology at Santa Cruz 4. 21-38. Kubozono, Haruo. 1995b. Gookeisei to Onin-koozoo [Word Formation and Phonological Structure]. Tokyo: Kuroshio Publishers. Kubozono, Haruo. 1999. Mora and Syllable. Tsujimura, Natsuko ed., The Handbook of Japanese Linguistics. 31-61. Oxford: Blackwell Publisher. Kubozono, Haruo. 2003. The Syllable as a Unit of Prosodic Organization in Japanwe. The Syllable in Optimality Theory, ed. by Caroline Féry and Ruben van de Vijver, 271-303. Cambridge, UK: Cambridge University Press. Kubozono, Haruo and Satoshi Ohta. 1998. Onin-koozoo to Akusento [Phonological Structure and Accent]. Tokyo: Kenkyusha. Labrune, Iamence. 2002. The Prosodic Structure of Simple Abbreviated Loanwords in Japanese: a Constraint-based Account. Journal of Phonetic Society of Japan 6. 98-120. [ROA-532, http://roa.nrtgers.edu]. LaCharite, Darlene, and Carole Paradis. 2005. Category Preservation and Proximity versus Phonetic Approximation in Loanword Adaptation. Linguistic Inquiry 36. 223- 258. Ledefoged, Peter. 1993. A Course in Phonetics. 3rd edition. Fort Worth: Harcourt Brace Jovanovich. Lombardi, Linda. 1999. Positional Faithfulness and Voicing Assimilation in Optimality Theory. Natural Language and Linguistic Theory 17. 267-302. McCarthy, John. 1986. OCP Efl’ects: Gemination and Antigemination. Linguistic Inquiry 17. 207-263. McCarthy, John. 1999. Sympathy and Phonological Opacity. Phonology 16. 331-399. McCarthy, John, and Alan Prince. 1990. Foot and Word in Prosodic Morphology: The Arabic Broken Plurals. Natural Language and Linguistic Theory 8. 209-282. McCarthy, John, and Alan Prince. 1993a. Generalized Alignment. In G. E. Booji and J. van Marle, ed., Yearbook of Morphology 1993. 79-153. Dordrecht: Kluwer. McCarthy, John, and Alan Prince. 1993b. Prosodic Morphology I: Constraint Interaction and Satisfaction. Technical Report No.3, Rutgers University Center for Cognitive Science. Cambridge, Massachusetts: MIT Press. 237 z o x a . , f r x O n .a s . I . ‘ a. I all s n t 5, . I ~ ‘ . a r 1.. ... .1. r l‘ . . r l ... s n I- \ Io .- rI . .0. 1 t u I .J.‘ r. s . : .... .Ir '1 l t t n. 3 1 v . v . .y t. ‘, l. . ' a t .I ‘ I . I. . l " I: r v .. . o McCarthy, John, and Alan Prince. 1995. Faithfirlness and Reduplicative Identity. University of Massachusetts Occasional Papers in Linguistics 18, ed. by Beckrnan, Walsh Dickey, and Urbanczyk, 249-384. Amherst, MA: GLSA. McCawley, James D. 1968. The Phonological Component of a Grammar of Japanese. The Hague: Mouton & Co. Mutsukawa, Masahiko. 2004. The Nativization and the Realization of English Word-final [r] in Japanese. University of Washington Working Papers in Linguistics 23: Proceedings of the 20th Northwest Linguistics Conference. 161-174. Seattle: University of Washington. Mutsukawa, Masahiko. 2005a. Lomword Accentuation in Japanese. Penn Working Papers in Linguistics 11.1: Proceedings of the 28th Annual Penn Linguistics Colloquium. 199-212. Pennsylvania: Penn Linguistics Club. Mutsukawa, Masahiko. 2005b. The Realintion of the English Plural Morpheme in Japanese. Proceedings of the 3rd Seoul International Conference on Phonology. 159- 183. Seoul: The Phonology- Morphology Circle of Korea. Mutsukawa, Masahiko. to appear. Loanword Accentuation in Tokyo Japanese. CLS 40-1: The Main Session. Chicago: Chicago Linguistic Society. Nakajo, Osamu. 1989. Nihongo-no Onin-to Akusento [Phonology and Accent of Japanese]. Tokyo: Keisou Shobou. National Language Research Institute. 1990. Gairaigo-no Keisei-to sono Kyouiku [A Study of Loanword Formation]. Tokyo: Ministry of Finance Printing Bureau. Nihon Hoso Kyokai, ed. 1998. Nihongo Hatsuon Akusento Jiten [Japanese Accent Dictionary]. Tokyo: Nihon Hoso Syuppan Kyokai. Niikura, Shunichi. 1996. Furansugo Handobukku [A Handbook of French]. Tokyo: Hakusuisha. Nishihara, Tetsuo, Jeroen van de Weijer, and Kensuke Nanjo. 2001. Against Headedness in Compormd Truncation: English Compounds in Japanese. In Jeroen van de Weijer and Tetsuo Nishihara, eds., Issues in Japanese Phonology and Morphology. 299-324. New York: Mouton de Gruyter. Nishimura, Kohei. 2001. Lyman’s La in Japanese Loanwords. Handout of Talk Given at the Meeting of the Phonological Association in Kansai (PAIK), Kobe University, October 2001. Ohso, Mieko. 1991 . Eitango-no Nihongoka [Japanization of English Loanwords]. Nihongo Kyouiku [Japanwe Language Education] 74. 34-47. 238 . r r r . , : u 4 . . . . , . I I .. , - l , a r - . t' . r I ' , l . r. \ r .' l as. " -"4 . s a \ "I . r f , I” ‘ r .t‘. I .'r ’ n I o ,r».- s A O ‘ u o C I‘ r: . l 14' ‘ . 4. . ' N I t I. ,A a- - I . i l J! t 1 . .‘. '11 ‘- ‘I O .' r-f' Ono, Koji. 1991. Accentuation of Loan Words in Japanese. Gengo Kyoiku [Language Inquiry] 100. 9-20. Paradis, Carole, md Darlene LaCharité. 1997. Preservation and Minimality in Loanword Adaptation. Journal of Linguistics 33, 379-430. Peperkamp, Sharon, and Emmanuel Dupoux. 2003. Reinterpreting Loanword Adaptations: The Role of Perception Proceedings of the 15th International Congress of Phonetic Sciences, 367-370. Pierrehumbat, Janet, and Mary Beckman. 1988. Japanese Tone Structure. Cambridge, Massachusetts: MIT Press. Poser, William J. 1984. The Phonetics and Phonology of Tone and Intonation in Japanese. Ph.D. dissertation, MIT. Poser, William J. 1990. Evidence for Foot Structure in Japanese. Language 66. 78-105. Prince, Alan. 1980. A Metrical Theory for Estonian Quantity. Linguistic Inquiry 11. 511- 562. Prince, Alan, and Paul Smolensky. 1993. Optimality Theory: Constraint Interaction in Generative Grammar. Technical Report No.2, Rutgers University Center for Cognitive Science. Cambridge, Massachusetts: MIT Press. Shibatani, Masayosdri. 1990. The languages of Japan. Cambridge, England: Cambridge University Press. Shimmura, Izrn'u (ed). 1955. Kojien: dai 1 pan [Comprehensive Japanese Dictionary. lst edition]. Tokyo: Iwamura Shoten. Shimmura, Izuru (ed.). 1969. Kojien: dai 2 han [Comprehensive Japanese Dictionary. 2nd edition[. Tokyo: Iwamura Shoten. Shimmura, Izuru (ed.). 1983. Kojien: dai 3 pan [Comprehensive Japanese Dictionary. 3rd edition]. Tokyo: Iwamura Shoten. Shimmura, Izuru (ed.). 1991. Kojien: dai 4 han [Comprehensive Japanese Dictionary. 4th edition]. Tokyo: Iwamura Shoten. Silverman, Daniel. 1992. Multiple Scansions in Loanword Phonology: Evidence from Cantonese. Phonology 9. 289-328. Smith, Jermifer L. 1998. Noun Faithfulness: Evidence from Accent in Japanese Dialects. Japanese Korean Linguistics 7. 611-627. 239 .’ 3‘29 I 1" t. lwr(.,' I ,I . r ‘7 oA . I] I I? | v , 1' r n! Mu Smolensky, Paul. 1993. Harmony, Markedness, and Phonological Activity. Handout to Talk Presented at Rutgers Optimality Workshop 1, New Brunswick, N. L. October 23. [RCA-87, http://roa.rutgers.edu]. Smolensky, Paul. 1996. The Initial State and ‘Richness of the Base’ in Optimality Theory. Technical Report JHU-CogSci-96-4, Cognitive Science Department, Johns Hopkins University. Steriadc, Donca. 2000. Padadigrn Uniformity and the Phonetics-phonology Boundary. In Michael B. Broe and Janet B. Pierrehumbert, eds., Papers in Laboratory Phonology V. 313-334. Cambridge: Cambridge University Press. Sugito, Miyoko. 1982. Nihongo Akusento no Kenkyu [Study of Japanese Accentuation]. Tokyo: Sanseido. Szymanek, Bogdan. 1989. Introduction to Morphological Analysis. Warszawa: Wydawnictwo Naukowe. Tanaka, Shin-ichi. 1992. Accentuation and Prosodic Constituenthood in Japanese. Tokyo Linguistic Forum 5. 195-216. Tashiro, Koji. 1953. Hyoojungo no Akusento Kyoohon [Accent Textbook for Standard Japanese]. Osaka: Sogensha Tateishi, Koichi. 1989. Theoretical Irnplications of the Japanese Musicians’ Language. WCCFL 8. 384-398. Tateishi, Koichi. 2001. On’irr Jisho Kurasu Seeyaku-no Bunpu-nitsuite [On the Distribution of Constraints for Phonological Sub-lexical. Paper Presented at the 26‘” Meeting of the Kansai Linguistics Society, Ryukoku University, Kyoto. Tateishi, Koichi. 2003. Are Borrowed Morphernes Always Foreign? In Takeru Honma, Masao Okazaki, Toshiyuki Tabata, and Shin-[chi Tanaka, eds., A New Century of Phonology and Phonological Theory: A Festschrifr for Professor Shosuke Haraguchi on the Occasion of His Sixtieth Birthday. 258-267. Tokyo: Kaitakusha. Tsuchida, Ayako. 1995. English Loans in Japanese: Constraints in Loanword Phonology. Working papers of the Cornell Phonetics Laboratory 10. 145-164. Tsujimura, Natsuko. 1996. An Introduction to Japanese Linguistics. Oxford: Blackwell Publisher. Tsujimura, Natsuko (ed.). 1999. The Handbook of Japanese Linguistics. Oxford: Blackwell Publisher. 240 .l' \j. Vance, Timothy J. 1987. An Introduction to Japanese Phonology. New York: SUNY Press. Wiese, Richard. 1996. The Phonology of German. Oxford: Oxford University Press. Yip, Moira. 1991. Coronals, Consonant Clusters and the Coda Condition. In Carole Paradis and Jean-Francois Prunet, eds., The Special Status of Coronals. 61-78. San Diego, CA: Academic Press. Yip, Moira. 2002. Tone. Cambridge: Cambridge University Press. Yoshida, Yirko Z., and Hideki Zamma. 2001. The Accent System ofthe Kyoto Dialect of Japanese: A Study of Phrasal Patterns and Paradigms. In Jeroen van de Weijer and Tetsuo Nishihara, eds., Issues in Japanese Phonology and Morphology. 215-241. New York: Mouton de Gruyter. 241