<%BANNER%>

New Typology of Japanese Compound Accents and an Analysis in Optimality Theory

Permanent Link: http://ufdc.ufl.edu/UFE0041790/00001

Material Information

Title: New Typology of Japanese Compound Accents and an Analysis in Optimality Theory
Physical Description: 1 online resource (91 p.)
Language: english
Creator: Chen, Si
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2010

Subjects

Subjects / Keywords: accent, compound, japanese, optimality
Linguistics -- Dissertations, Academic -- UF
Genre: Linguistics thesis, M.A.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: Abstract of Thesis Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Master of Arts NEW TYPOLOGY OF JAPANESE COMPOUND ACCENTS AND AN ANALYSIS IN OPTIMALITY THEORY By Si Chen August 2010 Chair: Caroline Wiltshire Major: Linguistics This paper explores a new typology of compound accents in Japanese, containing nine basic structures, and analyzes prefixes and complex compounds containing more than three words using these basic structures. Data of Japanese compound accents were collected from dictionaries and websites. Japanese native speakers were recorded reading all the data. Based on the recorded accent, further analysis has been done in assigning accent markers. After investigating the accent patterns through all different length categories and the various original accent patterns of the component words in isolation, common characteristics involved in word, monophrasal and biphrasal levels are proposed which form the basis of the new typology. In the later analysis, an OT (optimality theory) approach is provided to account for the assignment of the optimal accent pattern for a given input. Different levels of constraints such as MAXIO? (accent), MAXIO? (accent), ACCENT? and ACCENT ? correspond to distinct levels of the internal structure of compounds. More specifically, three constraints are proposed to account for the deaccentuation phenomenon. Constraints on the phonology-syntax interface are used to account for the difference between left-branching and right-branching compounds. My analysis accounts for the internal structure of compounds and redefines and associates different levels. More exceptions can be explained by this approach, and the OT analysis is more complete for deaccentuation phenomenon and for compounds with distinct structures.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by Si Chen.
Thesis: Thesis (M.A.)--University of Florida, 2010.
Local: Adviser: Wiltshire, Caroline R.
Electronic Access: RESTRICTED TO UF STUDENTS, STAFF, FACULTY, AND ON-CAMPUS USE UNTIL 2011-02-28

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2010
System ID: UFE0041790:00001

Permanent Link: http://ufdc.ufl.edu/UFE0041790/00001

Material Information

Title: New Typology of Japanese Compound Accents and an Analysis in Optimality Theory
Physical Description: 1 online resource (91 p.)
Language: english
Creator: Chen, Si
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2010

Subjects

Subjects / Keywords: accent, compound, japanese, optimality
Linguistics -- Dissertations, Academic -- UF
Genre: Linguistics thesis, M.A.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: Abstract of Thesis Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Master of Arts NEW TYPOLOGY OF JAPANESE COMPOUND ACCENTS AND AN ANALYSIS IN OPTIMALITY THEORY By Si Chen August 2010 Chair: Caroline Wiltshire Major: Linguistics This paper explores a new typology of compound accents in Japanese, containing nine basic structures, and analyzes prefixes and complex compounds containing more than three words using these basic structures. Data of Japanese compound accents were collected from dictionaries and websites. Japanese native speakers were recorded reading all the data. Based on the recorded accent, further analysis has been done in assigning accent markers. After investigating the accent patterns through all different length categories and the various original accent patterns of the component words in isolation, common characteristics involved in word, monophrasal and biphrasal levels are proposed which form the basis of the new typology. In the later analysis, an OT (optimality theory) approach is provided to account for the assignment of the optimal accent pattern for a given input. Different levels of constraints such as MAXIO? (accent), MAXIO? (accent), ACCENT? and ACCENT ? correspond to distinct levels of the internal structure of compounds. More specifically, three constraints are proposed to account for the deaccentuation phenomenon. Constraints on the phonology-syntax interface are used to account for the difference between left-branching and right-branching compounds. My analysis accounts for the internal structure of compounds and redefines and associates different levels. More exceptions can be explained by this approach, and the OT analysis is more complete for deaccentuation phenomenon and for compounds with distinct structures.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by Si Chen.
Thesis: Thesis (M.A.)--University of Florida, 2010.
Local: Adviser: Wiltshire, Caroline R.
Electronic Access: RESTRICTED TO UF STUDENTS, STAFF, FACULTY, AND ON-CAMPUS USE UNTIL 2011-02-28

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2010
System ID: UFE0041790:00001


This item has the following downloads:


Full Text





.NEW TYPOLOGY OF JAPANESE COMPOUND ACCENTS AND AN ANALYSIS IN
OPTIMALITY THEORY


















By

SI CHEN



















A THESIS PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
MASTER OF ARTS


UNIVERSITY OF FLORIDA

2010































2010 Si Chen









ACKNOWLEDGMENTS

First and foremost, I want to express my gratitude to my advisor, Caroline Wiltshire, and

to my committee members Ann Wehmeyer and Ratree Wayland for giving me valuable advice

and much help. Thank you also to my family for their love and support. Without my family, I

could not have successfully studied abroad. In data collection, many native speakers helped me

in determining the correct accent and I would really like to thank them: Yositaka Kawase, Mami

Tanaka, Yukari Nakamura and Miwa Horie. I would also like to express my appreciation to Dr.

Haruo Kubozono and Dr. Shin'ichi Tanaka for providing me resources to this problem and even

sending their own papers to me.










TABLE OF CONTENTS

page

A C K N O W L E D G M EN T S ......... ......... ......... ................................................................. 3

L IST O F T A B L E S ... .. .................. ..... ..................................................................... .............. ....... 6

LIST O F FIG U RE S ................................................................ 7

A B S T R A C T ....................................................................................... ......... ...... 9

CHAPTER

1 B A CK G R O U N D ............................................................................. ............... 11

Introduction ........................................ ..................... 11
B background ............... ............................................... 13
Basic Facts of Japanese Accent ......... ............................. 13
An Overview of Compound Accents ........................................ ................ 16
Early Studies on Word and Phrasal Compounding ..................................................20
D eaccentu ation ...................................................... 2 1
Analysis in an OT Approach.. ......... ................... ............... 24

2 NEW TYPOLOGY AND COMPLEX STRUCTURES ................................ 28

C current T ypology ............................ ... ... .... .......................... 28
The New Typology and the Application of Compound Accent Rules ............... ...............34
R ean aly sis of th e D ata ............................................................................................. 3 4
Definitions and Structures ................ ......................... ................... 36
A Review of the Nine Basic Types ....... ....................................................39
T y p e 1 ....................................... .............. ....... 3 9
T y p e 2 ..................... ...................................................................................................... 4 1
T y p e 3 ........................................................................................................................... 4 2
T y p e 4 ..................... ...................................................................................................... 4 4
T y p e 5 ........................................................................................................................... 4 5
T y p e 6 ..................... ...................................................................................................... 4 5
T y p e 7 ..................... ...................................................................................................... 4 6
T y p e 8 ..................... ...................................................................................................... 4 8
T y p e 9 ............... ................................................................................ ........... 4 9
C o m p lex Stru ctu res ........................... ............ ... ... ................................................. 5 1
T riphrasal Stru ctu re: ................................................................................................ 51
Right-branching com pounds: ........................................................................... 51
Biphrasal Structure ...................................................................................... 54
Complex Word (Monophrasal) Compound ................................................................ 57









3 AN ALY SIS IN AN O T APPR O A CH ............... .................................................................... 60

An OT Approach in Analyzing the N ine Types............................................. .....................60
Prefixes ......................... ...... ............... 71

4 CONCLUSION AND DISCUSSION ............. ................ .............. ................ ............... 76

The Advantages of the N ew Analysis.................... ......... ........................ ............... 76
The "2+2" Phenom enon in an O T A approach ................... ... ............................................ 76
Left-branching and Right-branching Differences in an OT Approach..............................78
Conclusions, Problems and Further Research ......... ...... ........... ..................... 79

APPENDIX

A D A T A O F N IN E B A SIC TY PE S ............................................. ........................................... 81

B THE TABLE FOR ALL ACCENT PATTERNS .............................................................. 85

R E F E R E N C E S .................. ................................................................................................................. 8 8

B IO G R A PH IC A L SK E T C H ........................................................ ............................................... 9 1
































5









LIST OF TABLES


Table page

2-1 The typology of compounds by Ito and Mester (2007) ............................ ............... 30

2-2 Possible accent patterns for N2 fewer than or equal to two morae................................. 35

2-3 Possible accent patterns for N2 of three or four morae long...............................................35

2-4 Possible accent patterns for N2 with more than five morae................................................36

2-5 The new typology for com pounds................................................. ............................ 37

3-1 Type 1 the constraint ranking for the word "sunabokori ta'isaku".................................. 61

3-2 Type 1 the constraint ranking for the word "nihonsi'gaku"........................ .................. 62

3-3 Type 2 the constraint ranking for the word "soosuya'kisoba"................ ................63

3-4 Type 3 the constraint ranking for the word "kaigai ryu'ugaku" ............... .....................63

3-5 Type 3 the constraint ranking for the word "inaka'ma" .......... ............. ............... 64

3-6 Type 3 the constraint ranking for the word "asa'giri"........................ .................64

3-7 Type 3 the constraint ranking for the word "kinuito".............. ..................... .. 65

3-8 Type 3 the constraint ranking for the word "garasudama"............................. ...............65

3-9 Type 3 the constraint ranking for the word "settyuuta'ue" .............................................66

3-10 Type 4 the constraint ranking for the word "amerikasan ore'nzi"............. .................. 66

3-11 Type 5 the constraint ranking for the word "genzi monoga'tari" ................................... 67

3-12 Type 6 the constraint ranking for the word "kita kariforunia"......................................... 68

3-13 Type 6 the constraint ranking for the word "singata infurue'nza"................ ................. 68

3-14 Type 7 the constraint ranking for the word "zimukyoku syoku'in" ................................ 69

3-15 Type 8 the constraint ranking for the word "infurue'nza kansenka'ku" ...........................70

3-16 Type 9 the constraint ranking for the word "ka'ngoku hoomon" ................................... 71









LIST OF FIGURES


Figure page

1-1 Bennett's analysis of accent markers and tone assignment.................... ............... 15

1-2 Oda's analysis of short and long N2 structures.......... ....... ............... ............... 19

1-3 Oda's analysis of the word"garasu dama"...................................................................... 22

1-4 O da's analysis of the w ord "kuro neko" ............................................................................... 23

1-5 O da's analysis of the w ord "hosi aw abi"............................................................................... 23

1-6 C onstraint ranking by Tanaka (2001) .......................................................................... ... 26

2 -1 Ju n ctu ral a ccen t ...................... .. .. ......... .. .. ........... .....................................2 8

2-2 Biphrasal structure and Rendaku distribution ..............................................................38

2-3 The structure of a word compound dominated by a phrasal node ......................................38

2-4 Type 1 Left-branching w ord com pounds ........................................ ......................... 39

2-5 Type 2 Right-branching word compounds .................. ........................................ 41

2-6 T ype 3 w ord com pounds............................................................... ............................. .. 42

2-7 Type 4 left-branching monophrasal compounds....... .................................................44

2-8 Type 5 right-branching monophrasal compounds..........................................................45

2-9 Type 6 m onophrasal com pounds.................................................. ............................. 46

2-10 Type 7 left-branching biphrasal compounds.....................................................................47

2-11 Type 8 right-branching biphrasal com pounds.............................................. ..................... 48

2-12 T ype 9 biphrasal com pounds...................................................................... ..................... 49

2-13 The complex structure for "dooso'ukaikozinzyo'ohoohogo ho'osin"..............................51

2-14 The complex structure for the word "sangaku kooryu'ukai" ........................................ 52

2-15 The complex structure for the word "hookeiga'kubukeizaigak'kaso'tu"...........................52

2-16 The complex structure for "syutoda'igakutookyoo soogookyoogi ta'ikai" ......................53

2-17 The complex structure for the word "tyo'o kantansetuyakure'sipi"............................... 53









2-18 The complex structure for "kitakariforunia asupara'gasu/ore'nzi" ............................... 53

2-19 The complex structure for the word "keizaikikaku'tyoo nyuutyoo" ............................... 54

2-20 The complex structure for the word "kaisya setumei'kai setuei" ................................... 54

2-21 The complex structure for the word "oosakabankoku hakura'nkai" ............................... 55

2-22 The complex structure for the word "nihon sei'hu daihyoo'bu" .................................... 55

2-23 The complex structure for the word "oosakaba'nkoku hakura'nkai"............................... 55

2-24 The complex structure for the word "minamioo'sawa hakubutu'kan"............................56

2-25 The complex structure for the word "syuusyoku'zi mensetusi'doo".............................56

2-26 The complex structure for the word "nihon si'gaku ko'oza"...................... ..................57

2-27 The complex structure for "minamioosawa kya'npasusangakukooryu'u kai"................... 57

2-28 The complex structure for the word "yo zakura ginzi zi'ken" ......................................... 58

2-29 The complex structure of the word "dokuiriorenzizi'ken"............ ....... ............... 58

2-30 The complex structure of the word "nihon sigaku ko'oza"............................................. 58

3-1 The structure of the w ord "koo bunsi"........................ .............. ................. ...............72

3-2 The structure of the word "koo enerugi" ......... ...... .......................................... 72

3-3 The structure of the word "ko'o byoogen sei" ............ ........................................72

3-4 The structure of the word "koo u'irusu".................... ............................................ ........... 74

3-5 The structure of the word "ko'o uirusu ya'ku" ........................................................74

3-6 The wrong structure of the word "kou u'irusu ya'ku"................................................... 74

3-7 The wrong structure of the word koo u'irusu ya'ku"......... ............................. 75

4 -1 T h e b alan ced stru ctu res............................................................................... ..................... 7 7

4-2 The structure of the word "ondankabo'osi" .............................................. ...... ......... 79









Abstract of Thesis Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Master of Arts

NEW TYPOLOGY OF JAPANESE COMPOUND ACCENTS AND AN ANALYSIS IN
OPTIMALITY THEORY

By

Si Chen

August 2010

Chair: Caroline Wiltshire
Major: Linguistics

This paper explores a new typology of compound accents in Japanese, containing nine

basic structures, and analyzes prefixes and complex compounds containing more than three

words using these basic structures. Data of Japanese compound accents were collected from

dictionaries and websites. Japanese native speakers were recorded reading all the data. Based on

the recorded accent, further analysis has been done in assigning accent markers. After

investigating the accent patterns through all different length categories and the various original

accent patterns of the component words in isolation, common characteristics involved in word,

monophrasal and biphrasal levels are proposed which form the basis of the new typology.

In the later analysis, an OT (optimality theory) approach is provided to account for the

assignment of the optimal accent pattern for a given input. Different levels of constraints such as

MAXIOco (accent), MAXIOQ (accent), ACCENTco and ACCENT D correspond to distinct

levels of the internal structure of compounds. More specifically, three constraints are proposed to

account for the deaccentuation phenomenon. Constraints on the phonology-syntax interface are

used to account for the difference between left-branching and right-branching compounds. My

analysis accounts for the internal structure of compounds and redefines and associates different









levels. More exceptions can be explained by this approach, and the OT analysis is more complete

for deaccentuation phenomenon and for compounds with distinct structures.









CHAPTER 1
BACKGROUND

Introduction

The accent patterns of standard Japanese have been an interesting topic since as early as

the 1960s. In Japanese, similar to tone languages, accent is assigned to almost every word, but it

is not found in tone languages. There are only level tones in Japanese, while tone languages can

also have contour tones.

The accents of compounds are not easily predicted by phonological rules because

compounds may have different structures influenced by semantic or pragmatic information1

(Vance, 1997). Previous research by McCawley (1977), Tsujimura (1987), Kubozono (2001) and

Oda (2006) dealt well with the accent patterns of compounds at the word level. They agree that

the length of N2 (the second member of a compound) will affect the accent location. Most of

them agree that when N2 is long (three morae and more), the accent will fall on the first mora of

N2 and when N2 is short (two morae or fewer), the accent will fall on the final mora ofNl (the

first member of a compound). Ito and Mester (2007) made use of this difference in the length of

N2 and proposed the concept of the junctural accent, which is a default position where

compound accents will fall. This position is at the boundary ofN1 and N2, either the final mora

ofN1 or the first mora of N2. Oda (2006) analyzed the structural difference between short and

long N2. Kubozono, Ito and Mester (1997) notice the different levels of compounds: word level

and phrasal level. Ito and Mester (2007) have further divided the levels into word, monophrasal

and biphrasal compounds, where junctural accent is the only standard to distinguish word and

phrasal levels. The length of N2 determines if the compound is in the word or phrasal level.


1 In Vance's (1997) review of Kubozono's (1993) book, biphrasal compounds are described as
more likely to be used by speakers in careful speech. If a speaker has a "pragmatic focus" on the
compound, then the whole compound or part of it are more likely to be biphrasal.









This thesis further investigates accent patterns in Japanese compounds and explores nine

basic structures of compounds which may affect accent patterns. A more complex standard is set

up based on characteristics of the three levels (word, monophrasal and biphrasal levels). As this

standard reflects the internal association between these levels and through the change of accent

location, the change across these levels is also demonstrated, such as the change from a

monophrasal compound to a word compound. This standard can account for the exceptions such

as kuda 'mono "fruit" which does not have ajunctural accent but is predicted to have one

according to the standard in Ito and Mester (2007). Also, compounds containing N2 of more than

four morae are predicted to have nojunctural accent by Ito and Mester (2007) while taimingu

"timing" and akusento "accent" are exceptions to this prediction, which can be handled using my

standard. My analysis also provides a new typology and accounts for the difference between left-

branching and right-branching compounds. More complex structures and prefixes which have

never been analyzed for different levels are analyzed based on basic structures within the

typology.

For the OT analysis, constraints from different levels are proposed to deal with cases of

junctural accent and accent retention. As for deaccentuation phenomenon, in which the whole

compound loses its compound, I propose a new constraint IDENT O-Ico (accent) which explains

why final accented N2 will trigger deaccentuation. Previous research on this topic will be

mentioned later in this chapter. I also proposed two constraints in Chapter 4 to account for the

structure of data in "2+2" phenomenon proposed by Oda (2006).

In Chapter 1, the basic facts of Japanese accent are introduced, with a summary of previous

analyses and an overview of the proposed analysis of compound accents. Chapter 1 also

summarizes previous research on the difference between word and phrasal accents and a special









phenomenon called deaccentuation, as well as previous analyses done in OT (Prince and

Smolensky, 1993).

Chapter 2 deals with the typology of compound accents and proposes nine basic structures

for word, phrasal and biphrasal structures. Data are presented after each type. Some of the data

were collected through the two dictionaries ",s\/uninelk'i Nihongo Akusento Ziten (2001)" and

"NHKNihongo Hatuon Akusento Ziten (1998)". Other compound data were collected through

the Internet and read by four Japanese native speakers. I confirmed some of the data from the

dictionary with the native speakers. They were asked to judge the appropriateness of the

compounds, and inappropriate compounds were deleted. I detected the location of the pitch-fall

throughout the data and confirmed some locations with the native speakers. The accent markers

on single words and some compounds are written in the dictionaries. Some of the Rendaku

(sequential voicing in compounds) changes can be found in the dictionary and others were

confirmed with the native speakers. Based on the nine basic structures, some complex structures

can be built to predict the placement of the accent marker. In the following chapters, data from

other researchers will be cited, and the rest of the data are all collected by the author and can be

founded in Appendix A.

Chapter 3 analyzes every typical structural pattern mentioned in Chapter 2 using an OT

approach, and Chapter 4 discusses the advantages of the new typology and some remaining

issues for further research.

Background

Basic Facts of Japanese Accent

While syllables can account for phonological structures in English, the notion of mora

plays an important role in analyzing Japanese phonological structure. A mora can be a vowel or a

CV. It can also be the first part of a long consonant, a geminate, and the syllable-final nasal /n/.









For example, in the word ansin "not worried", there are four morae a, n, si, n. A long vowel such

as oo contains two morae. (Tsujimura 2007).

Japanese is a pitch-accent language which bears a tone on each mora. Unlike tone

languages, the tonal pattern can be predicted given the location of an accent. The accent marker

"*" generally indicates a pitch fall in different analyses. In this paper, the will be used for

convenience to indicate the accent marker. For example, in the word so 'ra "sky", the accent is

located at the first syllable on the vowel o, and the word has a tonal pattern of high-low. If the

accent is located at the second mora, as is in koko 'ro "heart", the tonal pattern is low-high-low. If

the accent is on the third mora, the tonal pattern would be low-high-high. Moreover, if there is no

accent, the whole word would be low-high-high. All these types can be predicted by rules found

in the analyses of McCawley (1968, 1977), Haraguchi (1977) and Bennett (1981).

While tones are assigned to morae, a syllable bears an accent, so pitch-fall will not occur

after a coda: N(nasal),Q(a long consonant) or a long vowel, and any rule that tries to assign an

accent to the coda will fail. The accent will move to the vowel before the nasal or to the middle

of a long vowel (Iwasaki 2002). For example, in the word tenki "weather", the accent is assigned

as in te 'nki instead often 'ki.

McCawley, Haraguchi and Bennett have slightly different rules for the association of tones.

McCawley's pitch assignment rules initially make every mora high pitched. Then he assigns low

tones after the accented mora, and finally he assigns a low tone to the first mora if the second one

is high pitched (McCawley, 1977).

Haraguchi generated the placement of tone through four steps involving four rules. The

basic tone HL is associated with the CV tier according to the rules in order: Tone Associate Rule

(TA), Universal Tone Association Convention (UTAC), Initial Lowering (IL) and Tone









simplification (TS). Haraguchi argued that the first step is to associate the high tone to a vowel

with an accent mark or to the rightmost vowel in cases where there is no accent mark. After the

application ofUTAC, the initial lowering rule is applied to assign a low tone to the first mora

with no underlying accent. The last step is to eliminate contour tones by delinking the second

tone associated with the same vowel Haraguchi (1977).

Unlike Haraguchi, Bennett (1981) analyzed Japanese accents on two levels: accent foot

(@) in a phonological word and an accent marker (*) in one morpheme. As shown in Figure 1-1,

she uses a left-branching tree to derive accent patterns. She argued that within @, H is assigned

to the right node, otherwise L is associated with the right node.







i no ti ko ko ro a ta ma mi ya ko


H a L L

HL L L H L L H H L H H

I* *
i no ti ko ko ro a ta ma mi ya ko

Figure 1-1. Bennett's analysis of accent markers and tone assignment

Haraguchi (1991) analyzed the accent patterns on single words including nouns, verbs and

adjectives.The accent of nouns is usually assigned to the antepenultimate mora by rules.

Loanwords and compound nouns also have this tendency. Adjectives and verbs can either have

an accent or not. The accent will fall in "the final vowel of the stem". He also specified his

analysis based on the length of nouns. The accent of short nouns cannot be predicted while long

accented noun words tend to have the accent in the antepenultimate mora. When a high vowel is









devoiced because it appears between two voiceless consonants or word finally after a voiceless

consonant, the accent will be shifted to the left. For example, the word on gaku' kai "music

concert" will become onga 'kUkai; however, the accent on the initial mora will shift to the right

from hu 'kaku to hUka 'kU"deep".

An Overview of Compound Accents

The formation of compound nouns has been investigated intensively. In McCawley's

(1968) analysis, the second noun in a compound (N2) has a dominant role in determining the

accent of the whole compound regardless of the first noun (N1). Most later analyses also follow

this approach. Long compounds and short compounds are divided based on the length of N2. A

word is considered short if it is only made of one or two morae, and it is considered long when it

has three or more morae (McCawley, 1968).

Though analyses of compound rules vary, they basically accounted for unaccented

compounds and similar accent positions in compounds: the first syllable of N2, the last syllable

ofN1, the accent position retained in N2. Different analyses for these cases are shown by the

following.

I. The accent of the whole compound will be assigned on the first syllable of N2:

McCawley (1977) when:

N2 is long (having at least three morae)
N2 is a Sino-Japanese word
N2 is final accented or completely unaccented;

Tsujimura (1987) when:

A long N2 is unaccented or has an accent on the penultimate or final mora;

Oda (2006) when:

N2 is long.









II. The accent will be on the final syllable ofN1

Kubozono (2001) when:

If N2 is short and has an accent on the final syllable or is unaccented.

Oda (2006) when:

N2 is short.

III. The whole compound has the accent at the same position as N2:

McCawley (1977)

In other cases except for I and IV;

Kubozono (2001)

In other cases except for I and II.

IV. The whole compound would be unaccented:

McCawley (1977) when:

N2 is short and the final mora is accented

Oda (2006)

In other cases except for I and II.

A more detailed summary of the different analyses mentioned above is provided below.

McCawley (1968) proposed three cases of noun compound formation. In the first case, the

whole compound has the accent in the same position as in isolation, N2 such as in genzi "genzi

(name)" +monoga 'tari "story" genzi-monoga 'tari 'Tale of Genzi'. Kubozono proposed some

exceptions to this rule including certain morphemes such as hi 'me, which do not follow the rule

above: sirayuki "white snow"+hi 'me "princess"=sirayuki 'hime "Princess Snow White"

(Kubozono, 2001). In McCawley's second case, when long N2 is a Sino-Japanese word or has at

least three morae and is final accented or unaccented, the accent of the whole compound will be









assigned to the first syllable of N2. He gives some examples: no 'ogyoo "agriculture" +kumiai

"union"= noogyo-ku 'miai "agriculture union", inaka "country" +musume "daughter"=

inakamu 'sume "country girl". In McCawley' s third case, when N2 is short and final accented,

the whole compound loses its accent.

In order to explain some exceptions such as ka 'buto "helmet" +musi "bug"=

kabuto 'musi "a beetle", which bears an accent on the last syllable of N1 instead of being

unaccented, he proposes an underlying preaccented 'musi so that the accent of the whole

compound is dominated by the accent of N2 as in the first case (McCawley 1977). However, to

assume an underlying preaccented N2 may not explain other cases. Higurashi (1983) gave some

examples: sato "hometown"+ 'koko 'ro "heart"= *sato 'gokoro "homesickness", correct:

satogo 'koro; yude (ru) "boiled" +tama'go "eggs"= **yudetama'go 'boiled eggs' correct:

yudeta 'mago Higurashi (1983).

Tsujimura (1987) also analyzed long compound formation rules:

1. For a compound whose second member has an accent in the penultimate or the final mora,
the accent of the whole compound is assigned to the first mora of N2.

2. For a compound whose N2 is not assigned an accent, the accent of the whole compound is
assigned to the first mora of N2 such as ni "load"+kuruma "car"= ni gu 'ruma "cart".

Kubozono (2001) divided compounds into long and short ones. For long compounds, the

accent of the whole compound will be the same as the original accent of N2 unless N2 has its

accent in the final syllable. If the accent of N2 is in the final syllable or N2 is unaccented, the

whole compound will be assigned a new accent on the final syllable ofN1. Also in short

compounds, the accent will appear in the final syllable ofN1 such as ka '.bu.to "helmet"+mu.si

"bug"= ka.bu.to '.mu.si "a beetle".

Oda (2006) proposed that there are three patterns of compound accent placement. The first

is that N2 retains its accent. The second is that if N2 is long, the original accent will be at the left









edge of N2, otherwise, it will move to the end ofNl. The third possible case is for the compound

to be unaccented. He did not specify the condition for N2 to retain its accent or shift.

Oda distinguishes between word compounds and extended-word compounds to avoid the

arbitrary difference between short and long compounds. He based his analysis on the five

following assumptions and gave the internal structure of short and long N2: short N2 can only

form a single foot while long N2 will always form more than one foot.

Short N2 Long N2


F F F





Figure 1-2. Oda's analysis of short and long N2 structures

1. Assumption 1 Prosodic constituents should be parsed in the leftward direction and the unit
of parsing should be a mora.

2. Assumption 2 Parsing has to follow the Strict Layering Hypothesis (Selkirk, 1984), which
prohibits a prosodic level that is not dominated by a unit in the immediate level above. It
means that the node in the top level can only dominate the middle level which can only
dominate the bottom level. However, the top level node cannot directly dominate the
bottom level.

3. Assumption 3 Parsing should be maximally binary, not to have more than two branches.

4. Assumption 4 Parsing is also morphologically sensitive.

5. Assumption 5 Minimality requirement in Japanese It6 (1990)

a. Minimal Stem Requirement: Min (STEM) = F = [[Lm]
b. Minimal Word Requirement: Min (WORD) > a

His analysis focused on the different inner structures in short and long compounds based

on the above assumptions.









Early Studies on Word and Phrasal Compounding

Kubozono, Ito and Mester (1997) noticed five grammatical limitations on compounding.

Two members in a compound may retain their own accents if they do not combine

phonologically as a prosodic word, which is treated as compounding in the phrasal level by Ito

and Mester (2007). Kubozono et al. gave five scenarios where semantic relations may cause two

words not to combine as a word compound:

1. Certain prefixes such as ze 'n "before" and bo 'o "certain" 2. However, if each member has
no more than two morae, the compound will have one combined accent.

2. Two members are in an equal status. For example, ha'ku syu kassai "clapping hands and
acclaiming" retain their own accents.

3. Two members have a subj ect-obj ect relation. For example, ka'zi tetuda 'i "helping of
chores"

4. Names plus titles such as ku ri 'n ton daito 'oryoo "President Clinton"

5. Family names and first names such asyu 'kawa hi 'deki. However, if the internal structure
of a name is not considered, the whole name can bear one accent.

If two parts of a compound refer to the same content such as NI: content N2: label, N1:

last name N2: first name, N1: name N2: organization, or NI: title N2 name such as ka 'wase

ha 'kase "Dr. Kawase ", they tend to be biphrasal compounds which retain their accents on both

N1 and N2. The semantic information described here is also considered in my definition in

Chapter 2.

Kubozono et al. also argued that Japanese has the tendency to block combining of the

accents of two members in right branching compounds.They pointed out that N2 with more than

5 morae or three-morpheme Sino-Japanese words tend to keep their accent pattern, namely, to

remain unaccented or to have the accent at the same position as the words in isolation.


2 All the long vowel "ou" in the Japanese data are represented as oo for consistency, even if the
source used ou.









They believed that in a compound word, the accent of the two members are dependent,

even if the two do not bear accents, and the whole compound will have at most one. By contrast,

in a phrase, the accent of both members is independent and will not change. Some phrases may

undergo the process of becoming a compound, for example, eriza 'be su "Elizabethan"+ zyoo 'o

"queen"= eriza ble,' zyoo 'o or erizabe suzyoo 'o "Queen Elizabeth". The affix connecting two

members in a word may cause the accent on the first member to disappear, and this helps keep

the accent of N2. This is similar to compounds in the de-accentuation pattern.

In order to account for the similarity between 5 morae and three -morpheme Sino-Japanese

words, Kubozono introduces the standard of counting in feet. Two morae or a Sino-Japanese

morpheme can constitute one foot. For words of five morae and three-morphemes, there are at

least three feet.

When N2 is less than or equal to two feet, the whole compound is a word compound. For

three-foot N2, accent pattern of N2 will be retained. For N2 with more than three feet, the accent

of each of the two members will be retained (Kubozono et al. 1997). Oda (2006)'s analysis on

short and long compounds correspond to the word compound level in Kubozono et al. (1997)'s

research. Different length within word compounds may result in different accent locations.

Deaccentuation

Oda (2006) proposed three puzzles concerning deaccentuation phenomena in compounds:

First, finally accented N2 may cause the whole compound to be deaccented (McCawley

1968: 168).

Second, when the whole compound consists of four light syllables and N2 has two light

syllables, the whole compound tends to be deaccented.

Third, if N2 is long, the whole compound will never be deaccented.








He labeled the second puzzle as the "2+2 deaccentuation": with the same N2, a compound

will have an accent in the antepenultimate position when its N1 is long, but it is realized with no

accent when its N1 is bimoraic (with exceptions such as hi 'me "princess").

At the first glance, "2+2 deaccentuation" seems to be against the common assumption that

N1 is invisible. Internal structure mapping may explain the invisibility of N. Here is an example

ofgarasu-dama "glass ball" given by Oda to illustrate that N1 is mapped as a whole with no

concern about its internal structure (y means foot stem).



Metrical Tier

W --.. ----------- --w ------ .V
,,-------,
/X- Prwso'fic Tier


A. A


ga ra su + da ma
MDEP NHD

Figure 1-3. Oda's analysis of the word"garasu dama"
Oda proposed conditions on compounding that an accent must appear for compounds that

have a culminating point (the point at the top node of the structure) of extended prosodic word

and stated that the 2+2 phenomenon is not in contrast with the invisibility of Nl because the

structure ofN1 is visible but often ignored. The '2+2'structure does not have a culminating point

of an extended prosodic word as follows:









0)

-...--------- ----V Prosodic Tier


F F--""


R RL RI Lh Metrical Tier
ku ro + ne ko

Figure 1-4. Oda's analysis of the word "kuro neko"

It is different from a structure with a culminating point "co+" of an extended Prwd such as

the following structure:




CC
cor






ho si a ia bi

Figure 1-5. Oda's analysis of the word "hosi awabi"

Most of the research on compound accents above does not distinguish between word

compounds and phrasal compounds. Almost everyone agrees that the assignment of accent relies

on the length or the structure of N2. In the above overview of previous research, little attention

has been given to the original accent on N2 and its role in determining the accent pattern. The

generalization is missing that N2 with an original accent on the first mora will almost always

retain it regardless of the length of N2. Even if a N2 is considered to be long or short, it may not

behave as the rules predict because deeper structure is involved. Recent research, especially that

in an OT approach, paid more attention to the comparison of the accent input and the output. In









later chapters, I develop a typology considering both the length, structure and the original accent

of N2 as well as the structure of the whole compound.

Analysis in an OT Approach

Kubozono (1995) has done an OT analysis of Japanese compounds and has proposed five

constraints:

1. PARSE (accent): N2 should retain its accent

2. NON-FINALITY (t'o'):Accents should not be assigned in the final mora or syllable in a
prosodic word.

3. Non-finality (F'): No final accented foot in a prosodic word.

4. ALIGN-CA: The accent should be aligned either left or right to the N1 -N2 boundary.

5. Rightmostness: The accent should be assigned to the right edge of a prosodic word.

He gave the ranking as "NON-FINALITY (t'o')>>PARSE (accent) >> NON-FINALITY

(F'),ALIGN-CA>> RIGHTMOSTNESS".

Sino-Japanese morphemes are exceptions to compound rules such as inyo.ya.k

"reserved"+ se 'k "seats"= yo.ya.k '. se.k "reservation". Kubozono explained that SJ

words behave as if they are monosyllabic, and if the final "i" in seki is invisible to the above

rule, then SJ words will not be an exception. Kubozono hence proposed a constraint NON-

FINALITY (u'a') to solve this problem. This prohibits a final accent in the final foot so that the

problem of SJ exceptions can be explained.

After the first analysis in the OT approach by Kubozono, Tanaka (2001) has concluded,

based on previous research, that there are three primary characteristics of Japanese compounds.

First, the accent on N2 is usually retained or a new accent will occur on the first mora of N2.

Second, accent on the final syllable or mora is avoided. Finally, for short N2, namely one or two

morae, an accent will occur just before the N1-N2 boundary, otherwise it will occur immediately









after it. These three properties may be in conflict when accent needs to appear before the

boundary and N2 needs to retain its accent.

Tanaka revised Kubozono's generalization and pointed out that compound accent is

usually assigned to the penultimate foot unless preservation is needed in foreign, archaic native

and Sino-Japanese heads. Accordingly, Tanaka based his constraints on Kubozono's MAX

(accent), which requires the accent of the head root to be retained; ALIGN-L (o',root), which

requires the accented syllable to align to the left edge of a head root; and ALIGN-R (PrWd, 5'),

which requires the alignment between the right edge of a prosodic word and that of the accented

syllable.

Tanaka explained that MAX (accent) accounts for unaccented words better than PARSE

(accent) in the case of unaccented N2. ALIGN-L (o',root) emphasizes the head which can be N1

in mimetic words. ALIGN-R (PrWd, c') is better because of its gradient property, and it can be

applied among many languages.

For native words such as ningyo 'hime "doll princess", the accent cannot occur on the first

mora of"hi'me" because it violates NON-FINALITY (['o'F'), which ranks higher than MAX

(accent) according to Tanaka.

For Sino-Japanese words, some final accented words are still parsed but there are also

unparsed variations such as niho 'n "Japan" +zi 'n "person"=nihonzi 'n or niho 'nzin "Japanese

people" Tanaka argued that NON-FINALITY (C'o'F') and MAX (accent) are re-rankable. For

foreign compounds, the ranking is almost always MAX (accent) >>NON-FINALITY (a'o'F')

with some exceptions of nativized loanwords such as su 'upaa "super "+ma 'n

man "=suupa 'aman "superman ".









Tanaka explained all the data using the constraints mentioned above. For foreign (parsed),

MAX (accent) ranks higher than NON-FINALITY (ja'o'F') so that N2 can retain its accent. For

some variations in archaic native and Sino-Japanese words, MAX (accent) and NON-FINALITY

(ja'c'F') are rerankable. For general and nativized foreign words MAX (accent) ranks higher

thanNON-FINALITY (['C'F'). For quadrimoraic heads, MAX (accent) and ALIGN-L (c', root),

which requires the left edge of an accented syllable to align to the root, ensure that the winner

will have its accent on the left edge of N2 mora.

Tanaka gave three cases where the de-accentuation phenomenon occurs: final-accented

N2, quadrimoraic words and words with both conditions met.

For this phenomenon, Tanaka proposed a new constraint called NON-FINALITY (PrWd')

which forbids the accent to occur on the final prosodic word. So the constraint ranking is revised

as below (the dotted lines are an indication of rerankable constraints):


NON-FINALITY (p'o'F')
(a)

MAX (accent)
(b ) ...... .... (c )

Al IGN- L (',root) NON-FINALITY (PrWd')



ALIGN-R(PrWd, o')

Figure 1-6. Constraint ranking by Tanaka (2001)

In this ranking, there are three pairs that can be re-ranked because Tanaka's analysis does

not distinguish different compound structures. It can be improved by using different constraints

to analyze different structures, namely, the basic nine types presented in Chapter 2. In my

analysis in Chapter 3, I split the constraint "MAX (accent)" into different levels to deal with both









word and phrasal compounds separately. Tanaka proposes the constraint "NON-FINALITY

(PrWd')" in order to account for the deaccentuation phenomenon. However, this constraint is

violated by almost every candidate and thus is not so convincing. My OT analysis approaches

this problem by using the constraint IDENT O-Ico (accent) which requires the accent to stay in

N2.

Except for compounds undergoing deaccentuation, there is no constraint in this ranking to

reflect the fact that most compounds need at least one accent regardless of their structure. The

data show that a constraint is needed to account for this generalization, and Chapter 3 provides a

more specific explanation.

The above sections provide an overview of Japanese accents on single words and

compounds and the OT analyses proposed to date. Except for the level of word and phrasal

compounds mentioned in Kubozono et al. (1997), most previous research has focused more on

word level. In the following discussion, I introduce recent research on dividing the compounds

into three levels by Ito and Mester (2007). Following that, I redefine the three levels and provide

a new typology for further analysis on complex structures in Japanese compounding.








CHAPTER 2
NEW TYPOLOGY AND COMPLEX STRUCTURES

Current Typology

Ito and Mester (2007) proposed three basic categories for Japanese: intonation group (i),

phrase (0) and word (co). The internal structure of compounds may cause different accent

patterns. Apart from these three categories, relational projections are defined: "maximal

(Q/co)/minimal (P/co) projection is the highest/lowest element in a projection." It is argued that

Japanese can have only one accent in each phrase (P) because the phrase has only one head for

pitch accent (Ito and Mester 2007).

Ito and Mester classified compounds into word compounds and phrasal compounds. Only

in a word compound can ajunctural accent occur. The standard for distinguishing a word

compound from a phrasal compound lies in the length of the second member. If the second

member contains more than four morae, then the whole compound has to be parsed as a phrasal

compound (Ito and Mester 1997).

Monophrasal and biphrasal compounds do not allow any junctural accent to happen

according to Kubozono (1988). In a word compound, ajunctural accent is placed in Figure 2-1.






co co junctural accent

A Az

Figure 2-1. Junctural accent

However, the position of the junctural accent may coincide with the original N2 accent.

For example, in the biphrasal compoundyu 'kawa hi 'deki "Yukawa Hideki (name)", the original









accent of N2 is on the junctural position. So on the surface, finding an accent in the junctural

position may not be a perfect standard to determine whether a compound is a word or a phrasal

compound. Although Ito et al.(2007) argued that N2 under four mora should form a word

compound, the compound a 'ka "red"+ke "hair"=akage "red hair", does not have a junctural

accent and should be an exception. The "2+2" structures also contain a N2 that is less than four

mora such as tabi' "travel"+hito "person"=tabibito "traveler", without any junctural accent.

Moreover, since N2 with more than 5 morae should form a phrasal compound, nojunctural

accent should occur, as is given in Ito and Mester (2007). However, in the word hoogen

a 'kusento "dialect and accents", there is an accent in the juncture position which coincides with

the original accent on a 'kusento "accent". Another exception to this analysis is the word tenka

ta 'imingu "the timing of the sparks". The N2 of this compound has five morae, which should

stay unaccented according to Ito and Mester's analysis, however, the junctural accent occurs

since taimigu is originally unaccented.

The following table shows the typology Ito and Mester proposed. In this table, the square

L represents the minimal projection, and the round shape Q represents the maximal projection.

The symbol "r" represents the phenomenon of Rendaku. Rendaku is a phenomenon that happens

in compounding where the first consonant, if it is an obstruent, of the second word becomes

voiced after being combined into a compound. For example, iro "color" and kami "paper"

combines to form irogami "colored paper" where the consonant "k" becomes the voiced "g"

(Tsujimura 2007). Rendaku can only occur at the place of a "+r" but not "-r". The word on the

right branching usually has the "+r".









Table 2-1. The typology of compounds by Ito and Mester (2007)
Word Compounds


According to Kubozono, Ito and Mester (1997), the crucial factor that determines whether

a compound is a word compound or phrasal compound is the length of N2. However, it is

apparently not true for biphrasal compounds since even if N2 contains fewer than 4 morae, there

is nojunctural accent. For minami amerika asupara 'gasu "South American asparagus", the

derivation steps are:

Input Output Rules

Step 1 minami +amerika minami a'merika junctural accent

Step 2 minami a'merika+asupara'gasu minami amerika asupara'gasu deaccentuate N1

It is clear that N1 minami amerika "South America" is deaccented because it is originally

minamia'merika "South America". When attached to a 5 morae word asuparagasu, N1

undergoes deaccentuation, and the whole compound is thus a monophrase. For words with fewer









than 5morae, such as bi 'iru "beer", they usually have an original accent on the first mora. So in a

left branching compound minami amerika bi 'iru "South America beer", one cannot be certain if

it is a word compound with a newjunctural accent on N2 or if it is a monophrasal compound

which retains its original accent. I will treat it as a monophrasal compound in the OT approach

section, to make it parallel with the "South American asparagus" case, which must be a

monophrase.

Words exceeding 4 morae can be more easily identified as monophrases because they can

have an original accent on a position other than the first mora. Thus when N1 is added, although

N1 is deaccented, the compound accent will not occur in the default position as the new accent

for a compound word, while N2 retains its own accent as seen in asupara 'gasu "asparagus" and

minami amerika asupara 'gasu "South American asparagus".

Words of fewer than four morae can also have an accent on a position other than the first

mora. In compounding, those words will also retain their accents, such as kuda 'mono "fruits"and

ore 'nzi "orange" where no junctural accent will appear.

(1) mu' "without"+nooyaku "pesticides"+kuda'mono "fruits"=muno'oyaku+kuda'mono

=munooyakukuda'mono "organic fruits"

(2) amerika+sa'n+ore'nzi=amerikasan+ore'nzi=amerikasan ore'nzi "American oranges"

Since the junctural accent does not occur as in *munooyaku ku 'damono "organic fruits", it

is a phrase and not a word compound. The derivation steps are as follows:

Input Output Rules

Step 1 mu+ nooyaku muno'oyaku junctural accent

Step 2 muno'oyaku+kuda'mono munooyaku kuda'mono deaccentuation ofNl









If it were a biphrasal compound, then it should be muno 'oyaku kuda 'mono "organic

fruits". This is because the two lexical items in the input of step 2 should have retained their

accents. So we can say that it is a monophrase with an N2 of only four morae.

The phenomenon of Rendaku (sequential voicing) has something to do with compound

structure as shown in the typology table above. Rendaku usually happens in word compounds.

However, the occurrence of Rendaku cannot guarantee the identity of a word compound. There

are variations to word compounds and monophrasal compounds (Kubozono 1995, 1997).

Monophrasal compounds can also have Rendaku.

Rendaku does not happen in a predicate-argument relation such as sakana turi "fishing".

Biphrasal compounds usually happen in words having a predicate-argument relation as in ka'zi

"chores"+tetu'dai "helping"= ka'zitetu'dai "helping with chores". Rendaku usually occurs in N2

of word compounds, but not always. For example, the compound sakana "fish"+turi "catch"

=sakana'turi "fishing", (data from Sugioka (1986)) has the characteristics of a word compound

because the accent falls on the last mora ofN1 which is a "junctural accent". However, if it is a

word compound, then Rendaku should appear according to the prediction of Ito & Mester

(2007). On the contrary, if a compound shows the characteristics of a biphrasal compound,

Rendaku will not occur between two phrases.

Tanaka points out that the version of the accent on N2 being retained in compounds is

archaic, which means that the second type in the output such as densyo ba'to and koomori ga'sa

is archaic as follows (Tanaka 2001):

(3) densyo "transferring messages"+hato "pigeon"= densyo'bato/densyo ba'to "homing pigeon"

(4) ko'omori "bat"+ka'sa "umbrella"= koomori'gasa/koomori ga'sa "black umbrella"

(5) hidari "left"+uti'wa "Japanese fan"= hidariu'tiwa /hidari uti'wa "comfortable life"









He did not treat these variations with different compound structures. But they can actually

be analyzed as having different compound structures: word compounds or monophrasal

compounds, so there is a trend for these words to change from a monophrasal compound to a

word compound. In English, a similar phenomenon called lexicalization can occur, where a

phrase may lose its internal phonological structure. For example, the phrase ice 'cream may lose

its stress on the word "cream" and the stress is assigned on "ice" because the whole phrase enters

into the lexicon and becomes 'ice cream". In Old English, lexicalization may happen so that

the whole compound may be fused into a word such as "earwig< OE eare 'ear'+wicga 'one that

moves' Brinton (2005). In French, syntactic phrases can be lexicalized with a pattern of"A+N"

and "N+A", such as saint-bernard"St Bernard Dog" and table ronde "round table meeting"

(Mathieu-Colas 1996). In Dutch, phrases become compounds frequently as well, such as

sneltrein "fast train" (Schlucker 2008).

In Japanese, there can be variations as in mono "things" +hosi'"drying" =monoho 'si

monohosi' "frame for drying clothes". The compound mono hosi' is a biphrasal compound which

retains its accent pattern on N1 and N2. If it is a word compound, N2 final accent should cause

the whole compound to be deaccented. Thus, the fact that the accent occurs on monoho 'si is still

unexplained.

No Rendaku happens in the monohosi "dried clothes because of the predicate-argument

relation, while we do see Rendaku happen in kagebosi "dry in the shadow". The accent is ka 'ge

"shadow"+hosi' "drying" kagebosi 'kagebosi "dry in the shadow". The forms kage 'bosi and

kagebosi reflect a word compound structure while kagebosi' is a monophrase.

Kubozono mentioned the variation: so 'n "Son (last name) "+goku 'wu "gokuwu (first

name)" so 'n goku 'wu/son go 'kuwu "Son Gokuwu (name)" which reflects the variation of a









biphrasal and word compound. Usually, for biphrasal compounds, Rendaku does not happen, as

for example between the first and last family name.

The New Typology and the Application of Compound Accent Rules

Reanalysis of the Data

It has been mentioned in Chapter 1 that there are exceptions to the standard of the length

of N2. In Ito & Mester (2007)'s data dai+sakusika "great lyric-person" is unaccented even

though sakusika has only four morae. They explain that the structure is (saku)(si)+(ka) instead of

binary (saku)(si+ka), so the word violates MAXBIN because of the superbinary second member.

Thus, the whole compound should be a monophrasal one. However, the four-morae word

danraku "paragraph" remains unaccented in the compound hukusu 'u "plural "+danraku

"paragraph" hukusuudanraku "more than one paragraph" although this word danraku

observes MAXBIN (two bimoraic feet). Moreover, the binary word hanayome has two

morphemes hana "flower" andyome "bride", which is binary, but the whole compound retains

its accent in nihonzin hana 'yome "Japanese bride". Using the current data, I explore the accent

pattern according to the length of N2 as follows (Please refer to Appendix B for more data):

It is interesting to note that if the length of N2 is greater than five morae, then final

accented examples are hard to find. The accent on N2 of more than five morae is the most stable

because the whole compound can keep its accent or remain unaccented if N2 is originally so. It is

relatively hard to remain unaccented if N2 contains fewer than 5 morae and does not have an

accent originally. On the contrary, the whole compound tends to keep the accent on the first

syllable if N2 has an original accent there, but N2 with 2 morae or fewer can be so active that

even the accent on the first syllable can shift. Moreover, finally accented N2 usually triggers

deaccentuation of the whole compound, but finally accented N2 with fewer than 5 morae may

still have a junctural accent when forming a compound.









Table 2-2. Possible accent patterns for N2 fewer than or equal to two morae
Short N2 Length<=2 morae
N2 Accented Retain the original accent of densyo "carrier"+ha'to "pigeon"=
on the first N2 on first syllable or shift to densyoba'to or desyo'bato "carrier pigeon"
syllable the final syllable ofN1 From Tanaka (2001)
(junctural accent)
Only shift the original accent ni'ngyo "mermaid"+hi'me
of N2 to the final syllable of "princess"=nigyo'hime "mermaid princess"
N1 From Tanaka (2001)
Only retain the original ka'fe "cafe"+ ba'a "bar"=kafeba'a "cafe bar"
accent of N2 From Tanaka (2001)
N2 A new junctural accent may sasa "bamboo leaf'+ame"candy"=sasa'ame
Unaccented appear "candy wrapped in bamboo leaves"
The lack of accent may eda "branch"+ke "hair"=edage "split hair"
remain
N2 Final The accent may shift to the suidoo "waterworks"+hasi' "bridge"=suidoo
accented final syllable ofN1. 'basi "Suido Bridge (place name)"
The whole compound may be akane "madder"+iro' "color"=akaneiro
deaccented "madder red"


Table 2-3. Possible accent patterns for N2 of three or four morae long
Short N2 Length=3 or 4 morae
N2 Accented Retain accent on the first so'osa "investigation"+ka'igi
on the first syllable: coincides with the "conference"=soosaka'igi "investigation
syllable junctural accent meetings"
N2 New junctural accent may si'n "New"+yokohama "Yokohama (place
Unaccented appear name)"= Sinyo'kohama "New Yokohama"
From Tanaka (2001)
The whole compound may hukusu'u "plural"+danraku
remain unaccented "paragraph"=hukusuu danraku "more than
one paragraph"
N2 Final Accent may shift to junctural deza'ato "dessert" +azuki' "red azuki
accented position bean"=dezaatoa'zuki "dessert made of red
azuki bean"
The whole compound may be hasan "bankruptcy" +tuutisyo'
deaccented "notification"= hasan tuutisyo "the
notification of bankruptcy"
N2 Accented The accent may be retained sa'ga "saga (name)"+teno'o
in the middle "emperor"=sagatenno'u
"Emperor Saga"
The accent may be shifted to the genkin "cash"+huriko'mi
first mora of N2 "deposit"=genkin hu'rikomi "cash
deposit" From Tanaka (2001)









Table 2-4. Possible accent patterns for N2 with more than five morae
Long N2 Length>=5 morae
N2 Accented Retain accent on the first sokuseki "on the spot" +da'ietto "diet"=
on the first syllable: coincides with the sokuseki da'ietto "diet that can be effective
syllable junctural accent immediately"

N2 New junctural accent may kyo'oka "strengthen"+taimingu
Unaccented appear "timing"=kyooka ta'imingu "fine tune
timing"

The lack of an accent may kooon "high temperature"+taikyuusei
remain "durability"= kooon taikyuusei "durability of
high temperature"

N2 Accented The accent may be retained sya'nai "in a company"+danketu'ryoku
in the middle "power of unity"= syanai danketu'ryoku "the
power of unity in a company"



Looking through all of the tables, the length of N2 is related to the accent pattern of the

whole compound, but it is not decisive. The inner structure of the compounds concerning its

semantics and other factors might have led to the accent pattern. There are some common

characteristics in the accent patterns across the tables such as the junctural accent and the

retention of the accent. Based on those characristics, I offered new definitions of word,

monophrasal and biphrasal compound in the next section.

Definitions and Structures

A word compound has a junctural accent either on the first syllable of N2 or the last

syllable ofN1 except for deaccentuation, which happens in the "2+2 phenomenon"and word

compounds with final accented N2. A monophrasal compound retains the accent on N2, and N1

will be deaccented if it has an accent. No junctural accent will occur even if neither Nlor N2 has

an accent. A biphrasal compound retains both N1's accent and N2's accent, and nojunctural

accent will occur, even if neither N or N2 has an accent. By "retain", I mean to retain its accent









position on N2 or keep unaccented if N2 is originally so. Rendaku can only happen in word and

monophrasal compounds, but some word compounds may not have Rendaku. So if there is a

case in which N1 (unaccented)+N2 (unaccented)=Compound (unaccented), then this compound

can either be a monophrasal or biphrasal compound. However, ifRendaku occurs, it is definitely

not a biphrasal compound.

Table 2-5. The new typology for compounds
Word Compounds
0) O CO)




-r +r +r -r -r +r -r +r
sunabokori ta'isaku denkia'misori minami a'merika
dust treatment "electric shaver" "'south America"
Mono-phiasal Compounds
4) '> ,





-r +r +r -r -r +r -r +r
amerikasan ore'nzi kaigai ryuygakuse'ido Siro asupara'gasu
American oranges" overseaea) study abroad system" "White asparagus
Biphrasal Compounds








-r +r -r -r -r +r -r -r
sunabo'kori boosi infulule'nza kansen ka'kadai "the tyu'ugoku hoomon "Visits
:'dust treatmentJ spreading of influenza" to China"









I believe that there should be no potential Rendaku position occurring in the right

branching under a phrase containing two phrases as in






-r -r

Figure 2-2. Biphrasal structure and Rendaku distribution

It can happen in the branching under co dominated by P as in







w (L

-r +r

Figure 2-3. The structure of a word compound dominated by a phrasal node

So within a Japanese name such asyamaguti sakura, Rendaku can happen in guti but not sakura.

The insertion rule of [+voice] does not happen between two phrases.

Ito and Mester (2003, 71-99) divide the accentuation patterns of reduplicated words into

two kinds. Intensive/pluralizing reduplication may have junctural accents. Mimetic reduplication

has its accent on the first syllable of the mimetic while dvanda compounds (compounds with two

co-ordinated lexical items) retain the accent on N1 (Shibatani 2003).

They propose an analysis of Rendaku by treating it as a morpheme "R" with a feature

[+voice]. Coordinating compounds and object-verb compounds (OV compounds) with a

predicate-argument relation will not have Rendaku. Unger (1975) proposes a hypothesis that the

origin of Rendaku is a nasal sound. Rendaku requires N2 to be able to add the suffix no or ni.

OV compounds cannot meet this requirement because the suffix no and ni cannot be added to N2.









Thus, Rendaku does not happen accordingly. The following section includes data for nine basic

types; more data can be found in Appendix A.

A Review of the Nine Basic Types

Type 1

0)








-r +r +r

Figure 2-4. Type 1 Left-branching word compounds

Thejunctural accent rule happens at the last mora ofNl if N2 has one or two morae.

However, when the original accent on N2 is on the first syllable, two powers are competing with

each other: either to retain the accent or to move it to the junctural location. If N2 has three

morae or more, then the junctural accent will appear on the first mora of N2. Unlike in the case

when N2 has only one or two morae, there will be no conflicting power to shift the accent to

word boundary because the default junctural position coincides with the first mora of N2.

The compound rule first applies to col+co2, then col' will then combine with co3 where the

compound rule applies again. Rendaku may happen in co2 and co3 as in Ito and Mester (2007).

For example, the word suna "sand" and hokori "dust" form a compound col'sunabo 'kori "sand

and dust". When the third word taisaku "treatment" is added, the whole compound sunabokori

ta 'isaku "dust treatment" has an accent on the junctural place, namely on the first mora of co3.

Problems are that some examples can be categorized either as a compound or a

monophrase such as kikoo "climate" +hendoo "change"+ sa 'mitto "summit" kikoohe 'ndoo









sa 'mitto= kikoohendoo sa 'mitto "climate change summit". This is because the original accent

ofsa 'mitto is on its first mora, which coincides with the junctural position.

It is clear that if o3 has an accent shift, then there is no ambiguity of categorization such as

iti' "one"+ 'zi "time"+hara 'i "payment" iti 'zi+hara 'i=itiziba 'rai "one-time payment". It is not

the case that every o3 will shift its accent in this type such as kudamono "fruit" and orenzi

"orange", which is considered to be another type of compound. However, for words more than

three morae that already have an original accent on their first mora, they cannot "shift" to form

the junctural accent since the original accent coincides with the junctural position.

Data of this type are as follows:

(6) tabi' "travel"+hito "person"+ne'tto "net"=tabibito+ ne'tto=tabibitone'tto

"traveler's net"

(7) suna "sand"+hokori "dust"+taisaku "treatment"= sunabo'kori+taisaku

(unaccented)= sunabokori ta'isaku

"dust treatment"

The following can be categorized into type 1 or 4 since N2 in the compound has its accent

on the first mora and the compound accent still shows on the first mora of N2, so it is hard to tell

if the accent on N2 is retained or a newjunctural accent has occurred.

(8) yo' "night"+sakura "cherry blossoms"+gi'nzi "ginzi (name)"= yoza'kura+ gin'zi=yo zakura

gi'n zi "Yozakura Ginzi (name)"

(9) maki'zusi "hand-rolled sushi"+wa'arudo "world"= makizusi wa'arudo "hand-rolled sushi

world"

(10)nyu'u "new"+sa'maa "summer"+ore'nzi "orange"= nyuusa'maa+ ore'nzi= nyuusamaa

o'renzi "new summer orange"









Type 2


CO
20





32'
w2 w3

-r -r +r


Figure 2-5. Type 2 Right-branching word compounds

The compound rule first applies to co2+o3, then col' will combine with col where the

compound rule applies again. Rendaku may occur only in o3 as in Ito and Mester (2007). This

type has been explored extensively in Ito and Mester (2007). Examples given by them are:

ta 'nuki tani no 'bori "valley climbing by badgers", genkinfu 'rikomi "cash deposit", aka

ta 'manegi "red (round)-onion", nootopa 'sokon "notebook PC". I have found a word that used to

be in another type, then changes its accent to be categorized as type 2: muhuu "without

wind"+ti '"earth" +ta 'i "zone" muhuuti'tai (it used to be muhuutita'i). "the region of calms"

Data of this type are as follows:

(11)ka'igai "oversea"+ryuugaku "study abroad"+se'ido "system"= ka'igai+ryuugakuse'ido

=kaigai ryuugakuse'ido overseaa) study abroad system"

(12)niho'n "Japan"+si' "history"+gaku "subject"=niho'n+si'gaku=nihon si'gaku

"Japanese History Studies"

Some data from Tanaka (2001) fall into this type

(13)de'nki "electric"+kami' "hair"+so'ri "shave"=de'nki+kamiso'ri= denkika'misori

"electric shaver"









(14)so'osu"sauce"+yaki"fried"+so'ba"noodle"=so'osu+yakiso'ba=soosuya'kisoba "chow mein

with sauce"

Type 3

(0





wlI w2

-r +r

Figure 2-6. Type 3 word compounds

This type basically has two possible accentuations: one is to have a junctural accent, the

other is to be unaccented. The compound rule ofjunctural accent applies here too. Unaccented

types must contain a word that loses its original accent. If the words are originally unaccented,

they are considered to have retained their accent patterns. Unaccented pattern usually happens on

"2+2" and for final-accented co2, but there are exceptions as follows:

(15)ma'kura "pillow"+ki' "wood"=makura'gi crosstiess"

(16)iro "color"+kami' "paper"=iro'gami "colored paper"

(17)a'sa "morning"+kiri "mist"=asa'giri "morning mist"

(18)hosi "star"+so'ra "sky"=hosizo'ra "starry sky"

One problem is that since 4 morae N2 can also have a junctural accent, the mora length

does not determine which compound can have junctural accents. If the boundary of word length

is abandoned, and instead used to categorize monophrasal and word compounds according to

their characteristics, some of the words may fall into two categories. Consider the example

so 'osa "investigation" +ka 'igi "meeting" soosaka 'igi "the meeting of investigation"; it has a









junctural accent which is a characteristic of a word compound. However, the position of the

junctural accent coincides with the original accent on N2 so that the same compound can be

understood as to retain N2 and deaccent N1, which is a typical characteristic of the monophrasal

compounds. It can be predicted that Rendaku can happen on N2 if it is a monophrasal or a word

compound. In later types, it can be seen that Rendaku will not happen on N2 in biphrasal

compounds. N2 here refers to the item directly dominated by the top node.

There are some similar difficulties for distinguishing word, monophrasal and biphrasal

compounds. IfN1 is originally unaccented and N2 has an original accent on the first mora such

as hosi "star"+so 'ra "sky" hosizo 'ra "starry sky", this will make the combination look like a

biphrasal compound where two accents ofNl and N2 are retained. However, Rendaku occurs in

this case, which suggests that it should be a word or a monophrasal compound.

Word compound data are as follows:

The juncture type:

(19)a'sa "morning"+kiri "mist"=asa'giri "morning mist"

(20)hana' "flower"+kotoba' "languages" =hanako'toba "the language of flowers"

(21)a'ka "red"+tonbo "dragonfly"= aka to'nbo "red dragonfly"

The following show ajunctural accent, but usually compounds with a final-accented N2 will be

deaccented:

(22)suido "suido (name)"+hasi' "bridge"=suidoo 'basi "Suido Bridge (place name)"

(23)to'kati "tokati (place name)"+hasi' "bridge"= tokati'basi "Tokati Bridge (place name)"

The unaccented type:

(24)ma'e "before"+asi' "leg"=maeasi "fore-leg"

(25)ma'e "before"+kami' "hair"=maegami "forelock"











Type 4


a>







W1 w2

-r +r +r

Figure 2-7. Type 4 left-branching monophrasal compounds

The compound rule applies to col and co2 first. Ajunctural accent may appear in ol' or it

may remain unaccented. When combined with o3, the accent on o3 will be retained and ol' will

be deaccented. If there is no accent on col' or o3, they will still remain unaccented. Problems as

illustrated in type 1, are that when o3 has an accent on its first mora, it is hard to tell whether it is

a word compound or a monophrasal compound. For the case of the unaccented ol' and co3, it

may look like biphrasal compounds which retain their accent patterns, but biphrasal compounds

do not have Rendaku while monophrasal compounds can.

Data are as follows:

(26)amerika "America"+sa'n "product"+ore'nzi "orange"=amerikasan+ore'nzi=amerikasan

ore'nzi

"American oranges"

(27) minami "south"+amerika "America"+asupara'gasu "asparagus"

=minmia'merika+asupara'gasu= minmiamerika asupara'gasu "South American asparagus"









Type 5


-T -r +r

Figure 2-8. Type 5 right-branching monophrasal compounds

The compound rule applies in the combination of co2 and o3. Alternatively, col' can

remain unaccented if co2+o3 are unaccented. Rendaku can happen on co3. The problems are

similar to previous types that when the accent of col' falls on the first mora, and col is

deaccented, the whole can be treated both as a word compound or a monophrasal right-branching

compound such as niho 'n "Japan" +si' "history" +gaku "study" =niho 'n+si 'gaku nihon si 'gaku

"Japanese history study". Ito & Mester (2007) have data which fall into this type: genzi

monoga 'tari "Genzi story-telling", sinkokuritu ge 'kizoo "new national theater", daisakusika

"great lyric person". I have also found a word:

(28)ka'igai "abroad"+ryuugaku "study"+se 'ido "system"= ka'igai+ryuugakuse'ido= kaigai

ryuugakuse 'ido

overseaea) study abroad system" and muhuu (unaccented)+(ti'+ta'i)=muhuutita'i "region of

calms"

Type 6

In this type, co2 retains its accent pattern: either to be unaccented or to be accented in the

original position while col should be deaccented. However, if ol does not have an accent

originally, it may look like a biphrasal compound in which both accent patterns are retained.

Rendaku can indicate how to categorize the data. Consider the two words monohosi' "dried
















-r +T

Figure 2-9. Type 6 monophrasal compounds

things" and kagebosi' "drying in the shade"; the accent pattern looks alike, but Rendaku occurs

only in the second one, which does not have a predicate-argument as the first one. The accent in

kagebosi' can be explained as to retain the accent on hosi' "dry" and deaccent ka 'ge "shadow"

while the accent in monohosi' can be said to have retained both accent types of col and o02. Thus,

kagebosi' belongs to type 6 while monohosi' belongs to type 9 (biphrasal compound).

Another problem has been accounted for type 3 when co2 has an accent on the first mora:

(29)so'osa "investigation"+ka'igi "conference"=soosaka'igi "investigation meetings"

(30)ryuugaku (unaccented)+se'ido=ryuugakuse'ido "study abroad system"

(31)a'ka "red"+ke "hair"=akage "red hair"

(32)kanzi'kaki "writing characters"+zyun "order"=kanzi kaki zyun "the stroke order of

characters"

However, since co2 with one or two morae should have the accent on the final mora of col,

or be deaccented in "2+2", so the word hosi "star"+so 'ra "sky"= hosizo 'ra "starry sky" is a

monophrasal compound.

Type 7

In this type, compound rules first apply to col and co2 to form a prosodic ol' which is

dominated by O1. Then D2 dominating co3 will combine with D1. Both (1 and D2 tend to

preserve their own accents. No accent is shifted and no new accent emerges if they are originally












cl (I2
01 '2



wl w2

-r +r -r

Figure 2-10. Type 7 left-branching biphrasal compounds

unaccented. Rendaku can happen in the level of prosodic words but not between phrases. Names

are good examples of phrasal compounds. Some family names are made of two morphemes that

may stand alone and conform to compound formation rules such asyama' "mountain"+kuti

"mouth" +sa 'kura "cherry blossom"= yama 'guti+sa'kura "Sakura Yamaguti (name) ".

Kubozono (1994) has mentioned one type of biphrasal compounds "organization

name+position". It allows the internal structure such as ziti 'kai kaityo- "the president of the self-

governed region" in which ziti 'kai "self-governed regions" is a word compound which is further

combined with kaityo- "president", the name of the position.

Data of this type are as follows:

(33)zi'mu "clerical work"+'kyoku "department"+syoku'in "staff'=zimu'kyoku+syoku'in=

zimu'kyokusyoku'in "staff in Secretariat"

(34)kezai "Economics"+'kyoku "department"+syoku'in "staff'=kezai'kyoku+syoku'in=

kezai'kyokusyoku'in

'staff in Economic Bureau'

(35)a'ka "red"+matu "pine" + ma'sato "masato (name)" =aka'matu +ma'sato=aka'matu ma'sato

"Masato Akamatu (name)"









This word ko 'o uirusu ma 'suku "anti-virus mask"can be treated as a monophrasal

compound referring to a special mask as a whole. It can be also treated as a biphrasal left-

branching compounds where the accents on ko 'o "anti"and ma 'suku "mask" are retained but

uirusu "virus"is deaccented. However, the combination ofkoou 'irusu has a junctural accent and

when the third word is attached to it, the accent on koo appears again. Similarly, one speaker

treatedyozakura ginzi zi 'ken "The event ofYozakura Ginzi (name)" as a biphrasal compound

with the accentyoza 'kura ginzi zi 'ken, though she deaccentedyoza 'kura in the combination of

yozakura gi 'nzi. However, the accent on yoza 'kura reappears after combining with zi 'ken while

gi 'nzi is instead deaccented.

Similar examples are as follows:

(36)ko'o "anti"+u'irusu "virus"+tiryoo "treatment"=koouirus+tiryoo (unaccented)=ko'ouirusu

ti'ryoo "anti-virus treatment"

(37)ko'o "anti"+uirusu "virus"+ya'ku "medicine"=koo uirusu+ya'ku=ko'ouirusuya'ku

"anti-virus medicine"

(38)ko'o "anti"+uirusu "virus"+zai "dose"=koouirusu+zai=ko'o uirusu zai "anti-virus dose"

Type 8



02'
T 1 ---,02




w2 w3
-r -r +r

Figure 2-11. Type 8 right-branching biphrasal compounds

In this type, co2 and o3 combine as a word compound where compound rules apply. The

accent co2' will be retained as a phrasal accent and so (1 retains its accent.









Although the "name+title" is considered to be a biphrasal compound, if the title has an

internal structure, the accent will be retained on a higher level such as ka 'wase meiyo kyo 'ozyu

"Honored Professor Kawase". If there is no meiyo "honored", the accent will be ka 'wase kyoozyu

"Professor Kawase", in which kyoozyu remains unaccented. However, meiyokyo 'o zyu "honored

professor" is combined as a compound word at the prosodic level before being further combined

as a phrasal compound. Another possibility is to treat it as a tri-phrasal compound ka 'wase meiyo

kyoozyu.

Data from this type are as follows:

(39)yo'ru "night"+yo' "night"+na'ka "middle"=yo'ru+yonaka'=yo'ru yonaka "midnight"

(40)infurue'nza "influenze"+kansen "spreading"+kakudai "spreading"=infurue'nza+

kansenka'kudai= infurue'nza kansen ka'kudai "the spreading of influenza"

(41) nyuugaku"admission"+si'kan "application"+hyoo "forms"= nyuugaku+sikanhyoo=

nyuugaku sikanhyoo "application forms for universities"

Type 9










-r -r

Figure 2-12. Type 9 biphrasal compounds

In this type, the accent patterns on (1 and D2 will be retained. If there is no accent on (1

and D2, no new accent should appear and there should be no accent shifts from any position. The

problem is that some words may have the same accent pattern as the monophrasal compounds

since "unaccented+unaccented=unaccented" can be explained as to retain the accent of the last









word and deaccent the first word or combined word, such as the pair monohosi and kagebosi

explained in type 6.

As Kubozono (1994) mentioned, compounds with a predicate-argument relation are

usually biphrasal compounds.

(42) tyu'ugoku "China"+hoomon "visit"=tyu'ugoku hoomon "visits to China"

(43) ka'ngoku "Korea"+hoomon "visits"=ka'ngoku hoomon "visits to Korea"

(44) kita "North"+tyoose'n "challenge"+ hoomon "visits"= kitatyoose'n+

hoomon=kitatyoose'n hoomon "visits to Northern Korea"

However, with the same structure and the same verbal noun hoomon "visits", the

compound katei ho 'omon "home visits"is treated as a word compound referring to home visits as

a fixed expression. (katei "home "+hoomon "visits" katei ho'umon "home visits ")

Although names usually fall into the biphrasal compound types, some names show the

accent pattern of a word compound consistently such as "X-taroo".

X-taroo compounds:

There are three kinds of accents in X-taroo compounds Kubozono (2001):

1. Word compounds.

When N1 has only one syllable, the whole compound will be unaccented.

2. Word compounds.

When N1 has two syllables and two morae, the compound will be assigned an accent to the final
syllable ofN1.

3. Monophrasal compounds.

When N1 has three moras or more, N2 will retain the accent.

Data of the biphrasal compounds (type 9) are as follows:

(45)ka'wase "kawase (name)"+senpai "senior colleague"=ka'wase senpai "Senior Colleague









Kawase"

(46)yo'ru "night"+ him' "day" = yo'ru him "night and day"

(47)ma'e "before"+usiro "after"=ma'eusiro "before and behind"

Some data from Kubozono (1994) also fall into this type:

(48)i'ppu "one husband"+tasai "many wives"=i'ppu tasai polygynyy"

(49)ha'kusyu "clapping"+kassai "cheering"=ha'kusyukassai "clapping and cheering"

(50)kaisya "company"+syatyoo "president"=kaisyasyatyoo "president of the company"

(51)niho'n "Japan"+ze'ngoku "the whole nation"=niho'nze'ngoku "the whole nation of Japan"

(52)syoosoku "correspondence"+humei "lost"= syoosokuhumei "lost in contact"

(53)i'siki "conscious"+humei "lost"=i'sikihumei "losing one's consciousness"

Complex Structures

Some compounds have complex structures based on the nine basic structures analyzed

above.

Triphrasal Structure:

Right-branching compounds:

(54) doosoo"members of the alumni"+ka'i "organization"+ko'ozin "private"+zyoohoo

"information"+ho'go "protection"+hoosin "policy"=dooso'ukai+

kozinzyo'uhoo+hogoho'osin= dooso'okai+ kozinzyo'ohoohogoho'osin=

dooso'okaikozinzyo' ohoohogoho' osin

"protection policy of private information in the alumni"




Figure 2-13. The complex structure for "dooso'ukaikozinzyo'ohoohogo ho'osin"



Figure 2-13. The complex structure for "dooso'ukaikozinzyo'ohoohogo ho' osin"









(55) minami "south"+oosawa "oosawa (name)"+kya'npasu

"campus"=minamio'osawa+kya'npasu=minamioosawa kya'npasu "South 6sawa Campus"

(56) sa'ngaku "industry and education"+kooryuu "communication"+'kai "organization"=

sa'ngaku+kooryuu'kai= sangaku kooryu'ukai

"Industry and education forum of South 6sawa campus"



(I







Figure 2-14. The complex structure for the word "sangaku kooryu'ukai"

Left branching compounds:

(57) hookei "law and economics"+ga'kubu "department"+ke'izai "economics"+gakka

"subj ect"+so'tu "graduation"=hookeiga'kubu+keizaigak'ka+so'tu=

hookeiga'kubukeizaigak'ka+so'tu=hookeiga'kubukeizaigak'kaso'tu "economics maj or

from the department of law and economics"







(0 (0 (0


0 0) (0 (0

Figure 2-15. The complex structure for the word "hookeiga'kubukeizaigak'kaso'tu"

(58) syuto "capital"+daigaku "university"+tookyoo "Tokyo"+soogoo

"comprehensive"+kyo'ogi "sports"+taikai "meeting"=

syuto+da'igaku+tookyoo+soogookyo'ogi+taikai=









syutoda'igakutookyoo+soogookyoogita'ikai= syutoda'igakutookyoo soogookyoogita'ikai

"sports meeting in Tokyo Metropolitan University"









03 0 03 O CO

Figure 2-16. The complex structure for "syutoda'igakutookyoo soogookyoogi ta'ikai"

(59) tyo'o "super"+kantan "simple"+setuyaku "saving"+re'sipi "recipe"=tyo'o

kantan+setuyakure'sipi= tyo'o kantansetuyakure'sipi "very simple time-saving recipe"





/I I



(0 CO

Figure 2-17. The complex structure for the word "tyo'o kantansetuyakure'sipi"

(60)kita "North"+kariforunia "California"+asuparagasu "asparagus"=kitakariforunia +

asupara'gasu= kitakariforunia asupara'gasu "North California asparagus"

(61)kita "North"+kariforunia "Califomia"+ore'nzi "orange"=kitakariforunia + ore'nzi=
kitakariforunia ore'nzi "North California orange"


/Fm"s


( co


(0 (0

Figure 2-18. The complex structure for "kitakariforunia asupara'gasu/ore'nzi"









Biphrasal Structure

(62) keizai "economics"+kikaku "plan"+tyo'o "agency"+nyutyoo

"joining"=keizaikikaku+tyo'o+nyuutyoo (unaccented)= keizaikikaku'tyoo+nyuutyoo=

keizaikikaku'tyoo nyuutyoo "joining the planning agency"












Figure 2-19. The complex structure for the word "keizaikikaku'tyoo nyuutyoo"

(63) kaisya "company"+setumei "explanation"+'kai "meeting"+ setuei=kaisya setumei'kai

setuei

"setting up the venue for company presentations"










C o (O



Figure 2-20. The complex structure for the word "kaisya setumei'kai setuei"

(64) oosaka "oosaka (name)"+ ba'nkoku "world"+hakuran "exposition"+'kai

"meeting"=oosaka+ ba'nkoku +kakura'nkai=oosaka+bankoku hakura'nkai=

oosakabankoku hakura'nkai

"osaka World Exposition"























Figure 2-21. The complex structure for the word "oosakabankoku hakura'nkai"

(65) niho'n "Japan"+se'ihu "government"+daihyoo "representitives"+'bu "department"=nihon

sei'hu daihyo'obu "Delegation of Japanese government"










Figure 2-22. The complex structure for the word "nihon sei'hu daihyoo'bu"

(66) oosaka "oosaka (name)"+ ba'nkoku "world"+hakuran "exposition"+'kai

"meeting"=oosaka+ ba'nkoku +kakura'nkai=oosaka+ba'nkoku hakura'nkai=

oosakaba'nkoku hakura'nkai

"osaka World Exposition" (This one has a different structure from the same word in the
Figure 2-21)






23 (0


C 2 (0T CO

Figure 2-23. The complex structure for the word "oosakaba'nkoku hakura'nkai"










(67) minami "south"+ oosawa "oosawa (name)"+ haku'butu "natural history"+'kan

"building"=minamioo'sawa+ hakubutu'kan= minamioo'sawa hakubutu'kan

"South Osawa museum"










CO CO


) 0 ) CO CO


Figure 2-24. The complex structure for the word "minamioo'sawa hakubutu'kan"

(68) syuusyoku "employment"+zi "time"+mensetu "interview"+sidoo "instruction"=

syuusyoku'zimensetusi'doo= syuusyoku'zi mensetusi'doo

"the instruction of employment interviews"












(0D 0




Figure 2-25. The complex structure for the word "syuusyoku'zi mensetusi'doo"

(69) niho'n "Japan"+ si'gaku "history studies"+kooza "lectures"=nihon si'gaku+ ko'oza= nihon

si'gaku ko'oza "lectures of Japanese history studies"





















(0 0)

Figure 2-26. The complex structure for the word "nihon si'gaku ko'oza"

(70) minami "south"+oosawa "oosawa (name)"+kya'npasu "campus"+ sa'ngaku "industry and

education"+kooryu "communication"+'kai "meeting" =minamio'osawa+kya'npasu+

sa'ngaku+kooryuu'kai =minamioosawa kya'npasu sangaku kooryu'ukai

"Industry and education forum of South Osawa campus"



"..


I I
CO (0



(0 C CO (0


Figure 2-27. The complex structure for "minamioosawa kya'npasusangakukooryu'u kai"

Complex Word (Monophrasal) Compound

(71) yo' "night"+sakura "cherry blossoms"+gin'zi "ginzi (name)"+zi'ken "event"= yoza'kura+

gin'zi+ zi'ken =yo zakura gin' zi+ zi'ken= yo zakura ginzi zi'ken

"Yozakura Ginzi (name) event"

The word "yozakura ginzi" has a word structure but it is a name










(o(cD)






(0 CO

Figure 2-28. The complex structure for the word "yo zakura ginzi zi'ken"

(72) doku' "poison"+iri "added"+ore'nzi "orange"+zi'ken "incident"=

dokuiri+ore'nzi+zi'ken=dokuiriorenzizi'ken "poisonous orange incident"



o()


(0 (0




Figure 2-29. The complex structure of the word "dokuiriorenzizi'ken"

(73) niho'n "Japan"+si' "history"+gaku "subject"+kooza "lectures"= nihonsi+gaku+kooza=

nihonsi'gaku+kooza= nihon sigaku ko'oza

"lectures of Japanese history studies"




( (0CO







Figure 2-30. The complex structure of the word "nihon sigaku ko'oza"

This chapter introduced a new typology and further analyzes the nine basic structures. The

internal structure of more complex compounds may also be constructed based on those basic









structures predicting the location of accent in the complex structures based on the generalization

found in the nine basic types. In the following chapter, I analyze the data using the OT approach

and propose constraints for compounds of word and phrasal levels. In the last part of the next

chapter, I also explore the internal structures of prefixes.









CHAPTER 3
ANALYSIS IN AN OT APPROACH

An OT Approach in Analyzing the Nine Types

Chapter 3 analyzes all the data collected under the optimality theory framework.

Constraints proposed can account for the internal structures of compounds in two levels and for

the deaccentuation phenomenon. At the end of this chapter, prefixes are also analyzed in those

two levels. In this analysis, I will use constraints proposed by Kubozono (1995) and Tanaka

(2001): NON-FINALITY (t',a'), NON-FINALITY (F'), MAX (accent), ALIGN-L(o',root) and

ALIGN-R (PrWd,o'). The constraint NON-FINALITY (t',a') and NON-FINALITY (F') tend to

avoid final accents on N2. The constraint MAX (accent) requires the accent on the head of a

compound to be retained. The alignment constraint ALIGN-L (a', root) requires that the

accented syllable align with the root to the left while ALIGN-R (PrWd,o') requires the accented

syllable to align with the prosodic word to the right.

I have restricted the constraints of NON-FINALITY (t',a'), NON-FINALITY (F'), MAX

(accent) only to word compounds and rewrote them as NON-FINALITYco (t',a'), NON-

FINALITYco (F') and MAXIOco (accent).

The definitions for some new constraints used are as follows:

1. MAXIOo (accent) requires an accent corresponding to the head in a prosodic word

2. MAXIOO (accent) requires an accent corresponding to the head in a phrasal compound

3. and forbid more than one accent

4. ACCENTco requires at least one accent in the highest level of a prosodic compound

5. ACCENT P requires at least one accent on the head in a phrasal compound

6. IDENT O-Io (accent) requires the accent to stay in N2 (The second prosodic word of the
first branch) without movement.









This IDENT constraint above is devoted only to the retention of the accent position. If an

accent is deleted, then it violates MAXIOco (accent) but not IDENT O-Ico (accent).

In type 1 compounds, since the second candidate in Table 3-1 "{ { sunabokori}

co{(tai')(saku)} co} co" should lose to the winner, ALIGN-L (o',root) must rank higher than

ALIGN-R (PrWd,o'). Similarly, since the fourth candidate "{ { sunabokori} co{(tai)(saku)} co}

co" loses, ACCENT co has to rank higher than ALIGN-R (PrWd,a'). The third candidate

"{ sunabokori} co{(tai)(sa'ku) } co co" has the accent on the final syllable, so it violates the

constraints NON-FINALITYco ([t',a') and NON-FINALITY (F').

Table 3-1. Type 1 the constraint ranking for the word "sunabokori ta'isaku"
/{ { {suna co+{hokori}co}
co+{taisaku}co}co/


U 3 C


r {{sunabokori}
co{(ta'i)(saku)} co co
{{sunabokori} *! **
co{(tai')(saku)} co} co
{{sunabokori} *! *
co{(tai)(sa'ku)} co} co
{{sunabokori} *! *
co{(tai)(saku)} co} co



In Table 3-1, since the constraint ALIGN-L(o' root) requires aligning the prosodic word

with the accented syllable, all candidates except for the winner violate this constraint. The last

candidate does not have an accent which violates ACCENTco. The winner violates ALIGN-

R(PrWd,o') which is a gradient constraint requiring the accented syllable to align with the

prosodic word from the right. So violations are counted from the right in the unit of a syllable.









In the Table 3-2, since the second candidate "{ {nihonsi}co{(ga'ku)} co}co" loses to the

winner, NON-FINALITYo (t',a') and NON-FINALITYco (F') should rank higher than ALIGN-

L (o',root). So it can be concluded that NON-FINALITYo (t',a') NON-FINALITYo (F')>>

ALIGN-L (o',root), ACCENT o>> ALIGN-R (PrWd,o').

Table 3-2. Type 1 the constraint ranking for the word "nihonsi'gaku"
/{{ {niho'n} o+{si' }o} co+{gaku}
co}co/









{ {nihonsi' }o{(gaku)} o}o,
Snihonsi (ga'ku) *







{{nihonsi}co{(gaku')} o} *!* o*

{{nihonsi}co{(gaku)} oco* *****
nihonsi(gaku) *1 *


In Table 3 -3, NON-FINALITYco (t',a') and NON-FINALITYco (F') should rank higher

than MAXIOco (accent) to ensure that the second candidate" { { soosu}Co{(yaki)(so'ba)}o} co"

loses.

A compound such as hosi (unaccented) +so 'rahosizo 'ra may look similar to a word

compound but can only be a monophrase because if it is a prosodic word, NON-FINALITYco

(j',a') will move the accent on so 'ra. Moreover, the occurrence of Rendaku precludes the

structure from being biphrasal. In Table 3-4, the constraint ALIGN-L(a',root) rules out the third

and last candidates.









Table 3-3. Type 2 the constraint ranking for the word "soosuya'kisoba"

{so'osu}o+{ {yaki} s "
o+{ so'ba} o)}c/ s
i co *s; I







soosu co (yaki)(so'
ba)}Ico}co
{ i* **
{soosu} co{(yaki')(so
ba)co}co
S*I *******
{soosu} co{(yaki)(so
ba)}co}co
{ *I *** ******
{so'osu}co{(ya'ki)(s
oba)}}co co

Table 3-4. Type 3 the constraint ranking for the word "kaigai ryu'ugaku"












kaigai(ryu'u)(gaku)
{kaigai(ryuu)(ga'ku)}* *








{kaigai'(ryuu)(gaku) } co '*! ***
Skaigai(ryuu)(gaku) **
3 *







Ikaigai'(ryuu)(gaku)}co i *! ***
{kaigai(ryuu)(gaku) }o *! i *****









In Table 3-5, the last constraint does not have an accent and violates ACCENT co.

Table 3-5. Type 3 the constraint ranking for the word "inaka'ma"
/{{inaka}om+{ma}lo}(o/ ^




S{ {inaka'ma} o
{ {ina'kama }o *
{inakama' }o ***
{inaka ma}o *! ****

In Table 3-6, the constraint ACCENT co rules out candidate without an accent. The

constraints NON-FINALITYo(t(',c') and NON-FINALITYco(F') rule out candidates

with a final accent.

Table 3-6. Type 3 the constraint ranking for the word "asa'giri"
/{ {a'sa}co+{kiri}o}o/ | co/








f asa(giri)} I c
SH H asa(gi'ri) c

{asa(giri)} o *1 *
{asa(gi'ri)}o, *

{asa(giri')}co *! *


Even for N2 with final accent, the whole compound will be deaccented only when N2

contains fewer than 2 morae. For N2 with three morae or more, a junctural accent will appear. In

Table 3-7, a new constraint IDENT O-Ico (accent) is used which won't affect other tableaus

because candidates in other tableaus do not violate this constraint. To ensure the second

candidate "{kinu'(ito)}co" loses, IDENT O-Ico (accent) should rank higher than ACCENT co and









MAXIOco (accent). So the ranking turns to NON-FINALITYo (t',a') NON-FINALITYCo

(F')>> IDENT O-Ico (accent)>>MAXIOco (accent), ALIGN-L (o',root) ACCENT o>>

ALIGN-R (PrWd,o'). The second candidate {kinu'(ito)}co in Table 3-7 violates IDENT O-Ico

(accent) because the accent of N2 moves out to the N1-N2 boundary.

Table 3-7. Type 3 the constraint ranking for the word "kinuito"
{{ ki'nu}co+{i'to} p _


H 3



S* ****
{kinu(ito)} co


{kinu'(ito)} o *! **

{kinu(i'to)}o *!


Table 3-8. Type 3 the constraint ranking for the word "garasudama"
/{ {garasu}o +{tama'}








f garasu(dama)
>co/ S 5









{garasu'(dama) }o **

{garasu(da'ma) }co i*! *
garasu(dama') }o i i i
{ garasu(dama') }wo *









Table 3-9. Type 3 the constraint ranking for the word "settyuuta'ue"

settyuu}co+{taue' co





{ settyuu(ta')(ue))
CO 3 3



}oJ


{ settyuu(ta)(ue) } *****

{settyuu'(ta)(ue) }co *
{ settyuu'(ta)(ue) }c *!** ***

{settyuu(ta)(u'e) }co *! *

Table 3-10. Type 4 the constraint ranking for the word "amerikasan ore'nzi"
/{{{amerika}co+{sa'n}
coo o +{ ore'nzi} o /





*
S{ {amerikasan} co
(ore'n)(zi)} co} 0

{{amerikasan}co{(o're *! **
n)(zi)} co}

{{amerikasa'n}co{(ore *! ***
n)(zi)} co}

S{{amerikasan}co (oren *! *
)(zi)} o)} O ___

In the tableaus for type 4, 5, 6, 7, 8 and 9, two constraints MAXIOO (accent) and

ACCENT P are used, which will not affect previous tableaus because candidates in previous

tableaus do not violate this constraint. MAXIOO (accent) has to rank higher than NON-








FINALITYco ([t',o') because the second, third and fourth candidate have to lose to the winner.

The candidate {{amerikasa'n}co{(oren)(zi)} co} D in Table 3-10 does not have an accent on the

head N2 and violates ACCENT D. The candidate in Table 3-10 {{amerikasan} co{(o'ren)(zi)} co}

P violates MAXIOQ (accent) because the original accent on the head of the monophrase is

changed. Examples of N2 with other accent patterns and different internal structures are given

from Table 3-11 to Table 3-16.

In Table 3-11, the second candidate loses because it has a final accented syllable. The third

candidate does not have a corresponding accent on the head of the phrasal compound and

violates ACCENT P. The forth candidate has two accents in one phrasal compound and should

not win.

Table 3-11. Type 5 the constraint ranking for the word "genzi monoga'tari"
/
{genzi+ {{mono} co+{g
a'tari}co}co} /



U P






ta'ri)} } c

(tari)} c} D
{ {ge'nzi}o{mono(ga' *! : ****


ge'nzi}comono(ga' **
)(tari)} ***





































Table 3 -13. Type 6 the constraint ranking for the word "singata infurue'nza"
/ {{sin gata}
+{infurue'nza}c} I
ca o
SC 8
o H
^ < M e s i
O H H S _



*
"{ sin gata
(in)(furu)(e'nza)} _
{sin gata ****
(i'n)(furu)(enza)} P
{sin gata' *1 *****
(in)(furu)(enza)} I
{sin gata *****
(in)(furu)(enza)} O ***








Table 3-14. Type 7 the constraint ranking for the word "zimukyoku syoku'in"
/{ { { {zi'mu}m +{ f
'kyoku}}co}co}
D+ {syoku'in} 0
}

w Z o u x S
ooU







Szimu(kyoku)
"'^"{{{zimu'(k
yoku)}c} 0


{syoku'in} I }
{{{zimu(kyoku) *



}CO}
{syoku'in} 0 }
{ { {zimu(kyo'ku) ** ** ,*
}(D} O
{syoku'in} P }


{{{zimu'(kyoku) ** ***

{'syokuin} P }





In Tables 3-12 and 3-13, other candidates all violate MAXIO] (accent) and therefore lose.

In Table 3-14, the second and fourth candidates do not have a corresponding accent to the

original accent on the head of either the first or second phrase. The third candidate has a final

accented syllable and should lose to the winner.



69









Table 3-15. Type 8 the constraint ranking for the word "infurue'nza kansenka'ku"
/{{infurue'nza}
co}D++{ {kansen
}om+{kakudai} >



}} /{ { infume









mo}H{ {kansen(
l Il v




{ {in infuruenza
nzao}{ {kansenka
nsen(ka'ku)(dai



{ {infuruenza} *I ****
co}{ {kansen(
ka'ku)(dai)} m}


{{infurue'nza} ***
co}({ {kansen(
kaku)(da'i)} m}


{{infurue'nza} ***
oI(D}{{kansen( **
kaku)(dai)} co}



{{infurue'nza} **
co}>{ {kansen'(
kaku)(dai)} co}



In Table 3-15, the fourth candidate does not have an accent on kansenkakudai and violates

ACCENT co. The last candidate does not retain its accent on the head and violates ACCENT O.

The other candidates violate the constraint MAXIO Q(accent) and cannot win.









Table 3-16. Type 9 the constraint ranking for the word "ka'ngoku hoomon"
{{ka'ngoku}co}Q
+{{hoomon}o}Q(D




**
"{ {ka'ngoku}co

+{ {(hoo)(mon)} co}

{{ka'ngoku}Io}D *! ***
+{ {(ho'u)(mon)}co}

{ {kangoku}o}D *! *** **
+{ {(hoo)(mon)} co}



In the OT section above, I proposed five new constraints to deal with the internal structures

of compounds. Those constraints are also divided into the word and phrasal level which

correspond to the levels of compounds. MAXIO P (accent) and ACCENT P can deal with the

retention of accent in phrasal compounds. ACCENT co and MAXIOco (accent) operate at the

word level. IDENT O-Ico (accent) accounts for the deaccentuation phenomenon. This analysis

needs no rerankable constraints as in Tanaka (2001)

Prefixes


For prefixes, if they are in a biphrasal construction, they will retain their accent. If they

appear in a word structure, it means that the phrasal structure of the prefix is somewhat restricted

from being active in a whole word compound, indicated by '(D)'. I explore the following

examples with the prefix "koo" meaning high:










(74) A. I-# koo bu'nsi "high polymer"

(75) A -4x kooeneru'gi "high energy"

(76) AF. ; fJl ko'o byoogensei "highly pathogenic"

(77) A 'ft ko'o taikyuusei "high perduable"

Each represents the word, monophrasal or biphrasal compound.

Co

(4) o

* bu'nsi
koo

Figure 3-1. The structure of the word "koo bunsi"




I I

koo enern'gi

Figure 3-2. The structure of the word "koo enerugi"










ko'o byoogen sei

Figure 3-3. The structure of the word "ko'o byoogen sei"

Other prefixes such as tyo 'o and hi' can show either word or biphrasal compound structure.

Left branching tri-phrasal compounding is also possible such as:









(78)tyo'o "super"+ eiga "film"+ hi'han "criticism"=tyo'oeiga+hi'han=tyo'oeigahi'han

"super film criticism"

(79)tyo'o "super"+kookyuu "luxurious"+oni'giri "rice ball"= tyo'okookyuu

+oni'giri=tyo'okookyuu onigiri "super luxurious rice ball"

(80)tyo'o "super"+kookyuu "luxurious"+hure'nti "French"= tyo'okookyuu

(unaccented)+hure'nti= tyo'okookyuuhure'nti "super luxurious French dishes"

(81)tyo'o "super"+koosoo "high-rise"+bi'ru "buildings"= tyo'okoosoo+bi'ru=tyo'okoosoobi'ru

"super high-rise buildings"

(82)tyo'o "super"+kandan "easy"+ me'nyuu "menu"= tyo'okandan+ me'nyuu= tyo'okandan

me'nyuu "super easy menu"

(83)tyo'o "super"+manin "crowded"+de'nsya "train"= tyo'omanin+de'nsya=

tyo'omaninde'nsya

"super crowded trains"

(84)ko'o "anti"+u'irusu "virus"+ma'suku "mask"= koouirusu+ma'suku=

ko'ouirusuma'suku/koo uirusu masuku "anti-virus mask"

It is interesting that some combinations of a prefix and a root can restrict the biphrasal

structure of the prefix in a word compound such as koou 'irusu. When this word compound is put

under a phrasal compound ko 'ouirusu ya 'ku, the restricted biphrasal structure of the prefix gains

its biphrasal status again.

If in a phrasal compound, N1 has a phrasal structure while N2 only has a word structure,

then we expect that N2 will be deaccented because of Max ( (accent) which requires only the

phrasal head to retain its accent in a phrasal compound.

So the word compound koou'irusu has the structure:


1












(,)




koo u'irusu

Figure 3-4. The structure of the word "koo u'irusu"

When this word compound needs to be combined with another phrasal structure, a '0'

which dominates this current structure and the ko 'o retians its original accent and u 'irusu is

deaccented:








ya'ku


i ,

ko'o uirusu

Figure 3-5. The structure of the word "ko'o uirusu ya'ku"

Other structures such as the following are not possible:










o a ya'ku
koo ui iusu

Figure 3-6. The wrong structure of the word "kou u'irusu ya'ku"

















A ya'ku



koo u'irusu

Figure 3-7. The wrong structure of the word" koo u'irusu ya'ku"

The accent will be retained on "u'irusu" for the two structures in Figure 3-6 and 3-7. As

described above, prefixes can also have different internal structures at different levels.

Constraints from word and phrasal levels are more suitable for analyzing internal structures of

different levels. In the following chapter, two constraints are proposed to deal with the

deaccentuation phenomenon and two constraints are to account for the difference between left-

branching and right-branching compounds. The advantages and problems of my analysis are

summarized as well.









CHAPTER 4
CONCLUSION AND DISCUSSION

In this chapter, the advantages of the new typology and that of the new tableaux are

discussed. The chapter also explores an OT analysis for the "2+2" phenomenon and explains the

difference between left-branching and right-branching compounds using an OT approach.

Finally, I explain the limitations of my analysis and make suggestions on further research.

The Advantages of the New Analysis

The new typology is based on the redefinition of word, monophrasal and biphrasal

compounds. This new definition helps to categorize different levels of compounds by comparing

the input and the output of N2 rather than only focusing on the occurrence of the junctural accent.

The new definition helps compare and contrast the characteristics of word, monophrasal and

biphrasal compounds. Since junctural accent cannot be used to define biphrasal compounds, the

new definition is especially helpful in including biphrasal compounds in the new typology. This

way of categorization distinguishes each level well by considering the characteristics of the input

and output. Each level of compounds may have the internal structure of left-branching, right

branching or flat structure (N1+N2). The phenomenon of Rendaku can be also predicted in

correspondence to different structures. Based on the nine basic structures, more complex

structures can be analyzed successfully. The prefix, which has not been well analyzed in

previous research, can be analyzed using the new typology as well. Since compounds can be

divided into different levels, constraints of the corresponding levels are needed as well such as

MAXIOo (accent), MAXIOP (accent), ACCENTro and ACCENT D.

The "2+2" Phenomenon in an OT Approach


As for the deaccentuation problem, IDENT O-Ico (accent) can explain why N2 with a final

accent tends to be deaccented.Another phenomenon related to the deaccentuation is the "2+2"









phenomenon, mentioned in Chapter 1; however, not all "2+2" structure will be deaccented, only

balanced structures can be deaccented. The following data are from Tanaka (2001):

(85)ma'n "full"+ge'tu "moon"=ma'ngetu "full moon"

(86)tyo'o "punishment"+ba'tu "infliction"= tyo'obatu "punishment infliction"

These structures are weaker than the "2+2" structure, because the nasal and one part of a

long vowel cannot bear an accent. So the structure can be demonstrated as 4p(t)pja which is

different from the structure of unaccented kuro neko which has no nasals or long vowels with a

structure of pppp. Other "2+2" data are as follows Tanaka (2001):

(87)on "sound"+in "rhythm"=onin "phonology"

(88)usu "light"+azi "taste"=usuazi "light taste"

(89)ro'o "labor"+do'o "work"=roodoo "manual labor"

(90)ka'n "see"+ko'o "light"=kankoo "sightseeing"

For the data above, onin "phonology", roodoo "manual labor"and kankoo "sightseeing" all

have the structure p(u)u(u) while usuazi "light taste"has the structure of pppp. For "2+2" data,

balanced "2+2" structures do not need an accent such as


AA A A



Figure 4-1. The balanced structures

In the cases where N2 does not have a final accent, if two morphemes in a compound word

(<=4 morae) have a balanced structure "[[ [][ [ []" or [L ([)] [L ([)], no accent should appear

(Constraint A). On the contrary, if two morphemes fail to have the same structure, then an accent

has to occur (Constraint B). These are descriptive constraints whose motivation I leave for

further research.









Left-branching and Right-branching Differences in an OT Approach

In the following section, I examine the reason that left-branching compounds tend to be a

whole word compound. Truckenbrodt (1999) has investigated the relation between phonological

phrases and syntactic XP. He proposes the constraint WRAP-XP which requires that every XP

has to be contained in a phonological phrase. Selkirk (1995) modified the ALIGN constraint

proposed by McCarthy and Prince (1993) and created the constraint ALIGN-XP,R which was

defined as "For each XP there is a P such that the left edge of XP coincides with the left edge of

P."

In the case of Japanese, the constraint for biphrasal compounds requires that the left edge

of every accented prosodic word be aligned with an XP.The constraint ALIGN-NP,L can be used

to analyze Japanese compounds.

Left-branching:

[(N1+N2)NP2+N3]NP1

( )p

Right-branching

[N1+(N2+N3)NP2]NP1

(p( )P

In left-branching compounds, there is no conflict between ALIGN-NP,L and WRAP-NP.

In right-branching compounds, the left edge is required by ALIGN-NP, L which is in conflict

with WRAP-NP since the whole NP1 is not included in a prosodic word. So left-branching

compounds have the tendency to be treated as a whole prosodic word-word compound.

For Right-branching compounds, ifWRAP-NP>>ALIGN-NP,L, then the compound is

treated as a whole prosodic word---word compound. If ALIGN-NP,L, >>WRAP-NP, then it is









treated as a phrasal compound. This analysis in the OT approach can provide an explanation for

the difference of left-branching and right-branching compounds.

Conclusions, Problems and Further Research

My analysis of the new typology has some limitations which remain to be solved in the

future. According to the definition, it is difficult to distinguish a word compound from a

monophrasal one when a long N2 has its accent on the first mora, which coincides with the

default junctural accent position. The previous research by Ito and Mester (2007) relies only on

thejunctural accent to determine if it is a word or monophrasal compound and by my definition,

all "monophrasal compounds" which have an original accent on the first syllable of N2 will be

word compounds in Ito and Mester's standard. There is also one exception that Rendaku does not

show up in the predicted structure, which is the word sakanaturi "fishing" mentioned in chapter

two. This cannot be explained either by my analysis or previous analyses. In some compounds, a

word level can dominate the monophrasal level such as in:

(91) eda "branch"+ke "hair"+taisaku "treatment"=edage+taisaku= edageta'isaku

"split hair treatment"

(92)ondan "warming"+ka "change"+boosi "prevention" =ondanka+boosi=ondankabo'osi

"(global) warming prevention"

They have the following structure:






Figure 4-2. The structure of the word "ondankabo'osi"



Figure 4-2. The structure of the word "ondankabo'osi"









So a monophrasal compound is made first and then combined with another prosodic word to

form a word compound which dominates the monophrasal compound. It may be argued that the

word and phrasal level should have some order since the phrasal level dominates the word level

in most cases.

In the OT approach section, the input is already labeled with phrasal or word structures.

The constraints that determine the input's compound structure need to be found, which exist not

only in the phonological level but also in the semantic and pragmatic levels as mentioned by Ito

& Mester (2007). The constraint IDENT O-Ico (accent) explains the deaccentuation phenomenon,

but according to the data in Appendix B, final accented N2 may not always trigger

deaccentuation. IDENT O-Ico (accent) alone cannot explain this. Moreover, the constraint A and

constraint B mentioned in this chapter need further investigations in order to reflect deeper

reasons for the requirement of balanced structures. The competition of the power between

syllables and feet may be one of the reasons.

This thesis has succeeded in accounting for a variety of compounds using a new typology

and a set of OT constraints. It also provides further data for research to resolve the remaining

questions about the location of accents in compound structures.









APPENDIX A
DATA OF NINE BASIC TYPES

Type 1
(93)hosi "star"+so'ra "sky"+nyuumon "introduction"= hosizo'ra+nyuumon=hosizora nyu'umon

"Introduction to Astro observation"

(94)niho'n "Japan"+si' "history"+gaku "subject"= nihonsi+gaku= nihonsi'gaku

"The subject of Japanese history"

(95)sin gata "new-type"+infurue'nza "influenza"+zyoohoo "information"= sin gata

infurue'nza+zyoohoo= sin gata infuruenzazyo'uhoo

"New-type influenza information"

The following can be categorized into type 1 or 4:

(96)si'ro "white"+ asupara'gasu "asparagus"+ryo'uri "cooking"=siroasupara'gasu+ryo'uri

"white asparagus dish"

(97)minami "south"+oosawa "oosawa (place name)"+kya'npasu "campus" =minamio'osawa

+kya'npasu=minamioosawa kya'npasu "South Oosawa (place name) campus"

(98)minami "south"+oosawa "oosawa (place name)"+me'ibutu "special local product"

=minamioosa'wa+ me'ibutu=minamioosawa me'ibutu

"South Oosawa (place name) special local product"

(99)minami "south"+amerika "America"+bi'iru "beer" =minmia'merika+ bi'iru =

minamimerika bi'iru "South American beer"

(100) nyuusen "breast"+ga'n "cancer"+ se'nta "center"=nyuuse'ngan+ se'nta=nyuusengan

se'nta

"breast cancer center"

(101) hoppoo "northern"+kankei "relationship"+si'ryoo "resources"=hoppoo kankei si'ryoo

"northern studies collections"









Type 3

The junctural accent type

(102) inaka "country" +ma "space"=inaka'ma "tatami size(176cm*88cm) or a unit of measure,

1.8m"

(103) a'ka "red"+sato'o "sugar"=akaza'too "brown sugar"

(104) a'o "green"+kaeru "frog"=aoga'eru "green frog"

(105) ha'nabi "fireworks"+taikai "festival"=hanabita'ikai "fireworks festival"

(106) ha'nabi "fireworks"+kansyo "appreciation" =hanabika'nsyo

"fireworks appreciation"

(107) ha'nabi "firework"+tokusyu "specials"=hanabi to'kusyu "firework specials"

(108) a'o "blue"+tatami "tatami mat"=aoda'tami "new tatami mat"

(109) hude "pen"+tukai "usage" =hudedu'kai "brush work"

(110) kuti "mouth"+kuruma "car"=kutigu'ruma "cajolery"

(111) minami "south"+amerika "America"=minamia'merika 'South America'

(112) kikoo "climate"+hendoo "change"= kikoohe'ndoo "climate change"

(113) ti' "earth"+ta'i "zone"=ti'tai "area"

(114) kansen "inflection"+kakudai "spread"= kansen ka'kudai "spread of infection"

(115) ka'igai "overseas"+ryuugaku "study abroad"=kaigai ryu'ugaku "study abroad"

(116) nyuusen "breast"+ga'n "cancer"=nyuuse'ngan "breast cancer"

(117) suna "sand"+hokori "dust"=sunabo'kori "dust"

(118) hoppoo "northern"+kankei "relation"=hoppoo ka'nkei "northern relation"

(119) mu' "without"+nooyaku "pesticides"=muno'uyaku "no pesticides"

(120) hosi "star"+so'ra "sky"=hosizo'ra "starry sky"









(121) so'osa "investigation"+ka'igi "conference"=soosaka'igi "investigation meetings"

(122) ryuugaku "studying abroad"+se'ido "system"=ryuugakuse'ido

"the system of studying abroad"

The unaccented pattern

(123) tabi' "travel"+hito "person"=tabibito "traveler"

(124) niho'n "Japan"+hasi' "bridge"= nihon basi' "Bridge of Japan (place name)'

(125) musi "worm"+ha' "tooth"=musiba "decayed teeth"

(126) a'ka "red"+me' "eye"=akame "red eye"

(127) kuti "mouth"+kuse' "habit"=kutiguse "pet phrase"

(128) hosi "star"+so'ra "sky"=hosizora "starry sky"

(129) niho'n "Japan"+si' "history"=nihon si "Japanese history"

Type 4

(130) minami "South"+amerika "America"+ore'nzi "orange" =minmia'merika+ ore'nzi =

minmiamerika ore'nzi "South American orange"

(131) minami "South"+oosutora'ria "Australia"+kanga'ruu "kangaroo"=

minamio'osutoraria+kanga'ruu =minami oosutoraria kanga'ruu

"South Australia Kangaroo"

(132) hoppoo "north"+kankei "relationship"+si'ryoo "resources"= hoppoo ka'nkei +si'ryoo

=hoppoo kankei si'ryoo "Northern studies collection"

(133) mu "without"+nooyaku "pesticide"+kuda'mono "fruit"=

muno'uyaku+kuda'mono=munooyaku kuda'mono "organic fruit"

Type 6

(134) eda "branch"+ke "hair"=edage "split hair"









(135) a'ka "red"+mi "meat"=akami unaccented "lean meat"

(136) kita "North"+kariforunia "California"=kitakariforunia "North California"

(137) sin gata "new type"+infurue'nza "influenza"= sin gata infurue'nza "New type influenza"

(138) ti' "earth"+ta'i "zone"=tita'i "area"

(139) sa'ga "saga (name)"+tenno'o "emperor"= sagatenno'o "Emperor Saga"

(140) dokuiri "poisonous"+ore'nzi "orange"=dokuiriore'nzi "poisonous oranges"

Type 7

(141) suna"sand"+hokori "dust"+boosi "prevention"=sunabo'kori+boosi (unaccented)=

sunabo'kori boosi "prevention of dust"

(142) yama' "moutain"+kuti "mouth"+senpai "senior colleague"= yama'guti+senpai

"senior colleague Yamaguti (name)"

(143) syu'to "capital"+daigaku "university"+tookyoo "Tokyo"=syutoda'igaku tookyoo

"Tokyo Metropolitan University"

Type 9

(144) bo'n "bon festival"+mukae "welcome"+'bi "fire"=bo'n+mukae'bi=bo'n mukae'bi

"fireworks for Bon festival"

(145) ka'wase "kawase (name)"+sa'kura "cherry blossoms (name)"=ka'wase sa'kura

"Sakura Kawase (name)"

(146) a'ka "red"+tonbo "dragonfly"=a'katonbo (now: aka to'nbo) "red dragonfly"

(147) sa'ga "saga (name)"+tenno'o "emperor"=it was "sa'gatennoo" now sagatenno'o

"Emperor Saga"









APPENDIX B
THE TABLE FOR ALL ACCENT PATTERNS

Possible accent patterns for N2 less than or equal to two morae


Short N2 Length<=2 morae


Retain the
original accent
of N2 on first
syllable or shift
to the final
syllable ofNl
(junctural
accent)
Only shift the
original accent
of N2 to the
final syllable of
N1

Only retain the
original accent
of N2


densyo "carrier"+ha'to "pigeon"= densyoba'to or
desyo'bato "carrier pigeon"
ko'omori "Western"+ka'sa "umbrella"= koomori'gasa
"Western umbrella" From Tanaka (2001)


ni'ngyo "mermaid"+hi'me "princess"=nigyo'hime
"mermaid princess"
kansou "dry"+ha'da "skin"=kanso'uhada "dry skin"
kansoo "dry"+ne'gi "leek"= kanso'onegi "dry leek"
From Tanaka (2001)

ka'fe "cafe"+ ba'a "bar"=kafeba'a "cafe bar"
eiga "movie"+fa'n "fan"=eigafa'n "movie fan"
be'suto "best"+te'n "ten"=besutote'n "best ten"
From Tanaka (2001)


N2 A new junctural Sasa "bamboo leaf'+ame"candy"= sasa'ame "candy
Unaccented accent may wrapped in bamboo leaves"
appear a'sa "morning"+kiri "mist"=asa'giri "morning mist"
inaka "country" +ma "space"=inaka'ma "tatami
size(176cm*88cm) or a unit of measure, 1.8m"
The lack of eda "branch"+ke "hair"=edage "split hair"
accent may a'ka "red"+mi "meat"=akami unaccented "lean meat"
remain kanzi'kaki+zyun (unaccented)=kanzi kaki zyun
N2 Final The accent may Suido "waterworks"+hasi' "bridge"=suidoo 'basi "Suido
accented shift to the final Bridge (place name)"
syllable of N. to'kati "tokati (place name)"+hasi' "bridge"= tokati'basi
"Tokati Bridge (place name)"
abura "oil"+kami' "paper"= abura'gami "oil-absorbent
paper"
The whole akane "madder"+iro' "color"=akaneiro "madder red"
compound may
be deaccented


N2
Accented
on the first
syllable









Possible accent patterns for N2 equal to three or four morae


Short N2 Length=3 or 4 morae


N2 Accented Retain accent on so'osa "investigation"+ka'igi "conference"=soosaka'igi
on the first the first syllable "investigation meetings"
syllable coincides with the yama' "mountain"+kuti "mouth"+sa'kura "cherry
junctural accent blossom"= yama'guti+sa'kura 'Sakura Yamaguti (name)'
minami "south"+oosawa "oosawa (place
name)"+kya'npasu "campus" =minamio'osawa
+kya'npasu=minami oosawa kya'npasu "South Osawa
(place name) campus"
N2 New junctural si'n "New"+yokohama "Yokohama (place name)"=
Unaccented accent may Sinyo'kohama "New Yokohama"
appear kuti "mouth"+yakusoku "promise"=kutiya'kusoku
"verbal promise" From Tanaka (2001)
a'ka "red"+tonbo "dragonfly"= a'katonbo(before) =aka
to'nbo (now) "red dragonfly"
a'o "green"+kaeru "frog"=aoga'eru "green frog"
ha'nabi "fireworks"+taikai"festival"= hanabita'ikai
"fireworks festival"
ha'nabi "fireworks"+kansyo "appreciation"
=hanabika'nsyo "fireworks appreciation"
ha'nabi "firework"+tokusyu "specials"= hanabi to'kusyu
"firework specials"
The whole hukusu'u "plural"+danraku "paragraph"=hukusuu
compound may danraku "more than one paragraph"
remain ori'zinaru "original"+tezome "hand
unaccented dyeing"=orizinarutezome "original hand dyeing"
kookyuu "luxurious"+tyanomi "tea-drinking"= kookyuu
tyanomi "luxurious tea-drinking"
N2 Final Accent may shift deza'ato "dessert" +azuki' "red azuki
accented tojunctural bean"=dezaatoa'zuki "dessert made of red azuki bean"
position syuumatu "weekend"+danziki "fast"' =syuumatu da'n
ziki "fast in weekends"
hana "flower"+tubomi' "bud"=hanatu'bomi "the bud of
flowers"
hue'ru "expand"+wakame' "seaweed"= hueruwa'kame
"seaweed that can expand"
The whole hasan "bankruptcy" +tuutisyo' "notification"= hasan
compound may tuutisyo "the notification of bankruptcy"
be deaccented naitei "unofficial decision"+ tuutisyo' "notification"=
naitei tuutisyo "the notification of an unofficial decision"
N2 Accented The accent may mura'saki "purple"
in the middle be retained sa'ga "saga (name)"+tenno'o "emperor"=sagatenno'u
"Emperor Saga"













The accent may
be shifted to the
first mora of N2


mu "without"+nooyaku "pesticide"+kuda'mono "fruit"=
muno'uyaku+kuda'mono=munooyaku kuda'mono
"organic fruit"
genkin "cash"+huriko'mi "deposit"=genkin hu'rikomi
"cash deposit"
onna' "woman"+ kokoro "heart"=onnago'koro "woman's
heart"
yude' "boiled"+ tama'go "egg"= yudeta'mago "boiled
egg"
kami' "paper" + omu'tu "diaper"= kamio'mutu "paper
diaper"
From Tanaka (2001)


Possible accent patterns for N2 more than five morae


Long N2 Length>=5 morae


N2 Retain accent sokuseki "on the spot" +da'ietto "diet"= sokuseki da'ietto
Accented on the first "diet that can be effective immediately"
on the first syllable kantan "simple"+da'ietto "diet"= kantan da'ietto "simple
syllable coincides with diet"
the junctural sukyu'uba "scuba"+da'ibingu "diving"=sukyuuba da'ibingu
accent "scuba diving"
bo'oto "boat"+da'ibingu "diving"= bootoda'ibingu "boat
diving"
N2 New junctural kyo'oka "strengthen"+taimingu "timing"=kyooka ta'imingu
Unaccented accent may "fine tune timing"
appear tenka "sparks"+ta'imingu "timing"=tenka ta'imingu "the
timing of sparks"
The lack of an kooon "high temperature"+taikyuusei "durability"= kooon
accent may taikyuusei "durability of high temperature"
remain zyuunan "softness"+taikyuusei "durability"=zyuunan
taikyuusei "durability of softness"
kyu'udo "archery"+aikooka "lover"= kyu'udo aikooka "a
lover of archery"
N2 The accent sya'nai "in a company"+danketu'ryoku "power of unity"=
Accented in may be syanai danketu'ryoku "the power of unity in a company"
the middle retained kaisansei tansai'boo=kaisansei tansai'boo
tyoobun "long article"+dokkai'ryoku
"comprehensibility"=tyoobun dokkai'ryoku "the
comprehensibility of long articles"
zi'ntai "human body"+kaiboo'zu "atlas of anatomy"=
zintaikaiboo'zu "the atlas of human anatomy"









REFERENCES


Bennett, Diana Mary (1981). Pitch accent in Japanese: a metrical analysis. University of Texas at
Austin Thesis.

Brinton, Laurel & Traugott, Elizabeth (2005). Lexicalization and Language change. Cambridge:
Cambridge University Press.

Giegerich, Heinz (2004). Compound or phrase? English noun-plus-noun constructions and the
stress criterion. English Language and Linguistics. Cambridge: Cambridge University Press.
8.1.1-24.

Haraguchi, Shosuke (1977). The Tone Pattern ofJapanese: An Autosegmental Theory of
Tonology. Tokyo: Kaitakusha.

Haraguchi, Shosuke (1991). A Theory of Stress and Accent.Dordrecht-Holland. Providence RI-
U.S.A.: Foris Publications.

Higurashi, Yoshiko (1983). The Accent of Extended Word Structures in Tokyo Standard
Japanese. Tokyo, EDUCA Inc.

Ito, Junko (1990). Prosodic minimality in Japanese. Chicago Linguistics Society 26-11: Papers
from the Parasession on the Syllable in Phonetics and Phonology. 213-239.

Ito, Junko & Armin Mester (1997). Correspondence and compositionality: The ga-gyo variation
in Japanese phonology. In Derivations and Constraints in Phonology, ed. Iggy Roca(ed.), 419-
62. Derivations and Constraints in Phonology. Oxford: Oxford University Press.

Ito, Junko & Armin Mester (2003). Japanese Morphophonemics Markedness and Word
Structure. Cambridge, Massachusetts: The MIT Press.

Ito, Junko & Armin Mester (2007). Prosodic Adjunction in Japanese Compounds. In Yoichi
Miyamoto and Masao Ochi (eds.) Formal Approaches to Japanese Linguistics 4, MIT
Working Papers in Linguistics 55. 97-112.

Iwasaki, Shoichi (2002). Japanese. Amsterdam/ Philadelphia: John Benjamins Publishing
Company.

Kindaichi, Kyosuke & Kindaichi, Haruhiko (2001). Meikai kogojiten [Explanatory dictionary of
Japanese archaisms].Tokyo: Sanseido.

Kubozono, Haruo (1988). The organization of Japanese Prosody. Doctoral Dissertation,
University of Edinburgh. Tokyo: Kurosio Publishers.

Kubozono, Haruo (1994). Word Formation and Phonological Structure. (Gokei sei to Onin
Kouzoo). Tokyo: Kurosio Publishers.









Kubozono, Haruo (1995). Constraint interaction in Japanese phonology: Evidence from
Compound Accent. In Rachel Walker, Ove Lorentz, and Haruo Kubozono (eds.) Phonology at
Santa Cruz 4. 21-38.

Kubozono, Haruo (2001). Epenthetic Vowels and Accent in Japanese: Facts and Paradoxes. In
Jeroen Van Deweijer and Tetsuo Nishihara (eds.) Issues in Japanese Phonology and
Morphology. 111-140. Berlin: Mouton de Gruyter.

Kubozono, Haruo, Ito, Junko & Arimin Mester (1997). On'inkoozoo-kara mita go-to ku-no
kookai: fukugoo-meishi akusento-no bunseki[The word/phrase boundary from the perspective
of phonological structure: the analysis of nominal compound accent]. In Spoken Language
Research Group (eds.) Bunpoo to onsei. page number[Speech and Grammar], 147-166. Tokyo:
Kurosio Publishers.

Mathieu-Colas, Michel (1996). Essai de typologic des norms composes francais. Ca hiers de
lexicologie 69. 71-125.

McCarthy, John & Alan Prince (1993). Genralized Alignment. In Geert Booij and Jaap van
Marle (eds.), 79-153. Yearbook of morphology. Dordrencht: Kluwer.

McCawley, James (1968). The Phonological Component ofa Grammar ofJapanese. The Hague:
Mouton.

McCawley, James (1977). Accent in Japanese. In Hyman,L.M.(1977) (ed.). Studies in Stress and
Accent. Southern California Occasional Papers in Linguistics 4: 262-272.

NHK (1998). Nihongo Hatuon Akusento Ziten [Dictionary of Japanese Pronunciation and
Accent]. Tokyo: NHK (Nippon Hoosoo Kyookai) Syuppan.

Oda, Kenji (2006). The accentuation patterns of nominal compounds in Japanese. Toronto
Working Papers in Linguistics 26: 23-64.

Prince, Alan; & Smolensky, Paul. (1993). Optimality Theory: Constraint interaction in
generative grammar. RuCCS Technical Report 2, Rutgers University. 1-262 Piscateway, NJ:
Rutgers University Center for Cognitive Science. (Revised version published 2004). (Online:
roa.rutgers.edu/view.php3?id=845).

Schlucker, Barbara (2008). Naming strategies in West Germanic Lanugages. Paper presented at
the ICLC5-conference. Leuven.

Selkirk, Elizabeth (1984). Phonology and Syntax: the Relation between Sound and Structure.
Massachusetts: MIT Press.

Selkirk, Elisabeth (1995). The prosodic structure of function words. In University of
Massachusetts occasionalpapers 18: Papers in Optimality Theory, 439-469. GLSA,
University of Massachusetts, Amherst.









Shibatani, Masayoshi (2003). Directional verbs in Japanese. Amsterdan: John Benjamins
Publishing Co.

Tanaka, Shinichi (2001). The Emergence of the 'Unaccented': Possible Patterns and Variations
in Japanese Compound Accentuation. In Jeroen Van Deweij er and Tetsuo Nishihara (eds.)
Issues in Japanese Phonology and Morphology, 163-192. New York: Mouton de Gruyter.

Truckenbrodt, H. (1999). On the relation between syntactic phrases and phonological phrases.
Linguistic Inquiry 30. 219-255.

Tsujimura, Natsuko. & Davis, Stuart (1987). The accent of long nominal compounding in Tokyo
Japanese. Studies in Language 11. 199-206.

Tsujimura, Natsuko. (2008). An Introduction to Japanese Linguistics. Malden: Blackwell
Publishing.

Unger, James Marshall (1975). Studies in early Japanese morphophonemics. Doctoral
dissertation, Yale University. Oxford: Oxford University Press.

Vance, Timothy J. (1997) Review of The Organization ofJapanese Prosody, by Haruo
Kubozono. Tokyo: Kurosio Publishers, 1993. The Journal of the Association of Teachers of
Japanese 28: 93-103.









BIOGRAPHICAL SKETCH

Si Chen was born in 1986 in Nanjing, China. She is the only child in her family. Since her

father and mother are musicians, Chen has received musical training from an early age and

gained the high proficiency certificate of piano skills in China. Her interests in languages were

developed since middle school. After graduation from Jinling High School in Nanjing, she chose

English linguistics and literature as her major and received her bachelor's degree from Beijing

International Studies University. Upon graduation, she pursued her master's degree in linguistics

in University of Florida and began to focus her research on phonology and phonetics in Chinese

and Japanese.





PAGE 1

1 NEW TYPOLOGY OF JAPANESE COMPOUND ACCENTS AND AN ANALYSIS IN OPTIMALITY THEORY By SI CHEN A THESIS PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF ARTS UNIVERSITY OF FLORIDA 2010

PAGE 2

2 2010 Si Chen

PAGE 3

3 ACKNOWLEDGMENTS First and foremost, I want to express my gratitude to my advisor, Caroline Wiltshire, and to my committee members Ann Wehmeyer and Ratree Wayland for giving me valuable advice and much help. Thank you also to my family for their love and support. Without my family, I could not have successfully studied abroad. In data collection, many native speakers helped me in determining the correct accent and I would reallly like to thank t hem: Yositaka Kawase, Mami Tanaka, Yukari Nakamura and Miwa Horie. I would also like to express my appreciation to Dr. Haruo Kubozono and Dr. Shinichi Tanaka for providing me resources to this problem and even sending their own papers to me.

PAGE 4

4 TABLE OF CON TENTS page ACKNOWLEDGMENTS .................................................................................................................... 3 LIST OF TABLES ................................................................................................................................ 6 LIST OF FIGURES .............................................................................................................................. 7 ABSTRACT .......................................................................................................................................... 9 CHAPTER 1 BACKGROUND ......................................................................................................................... 11 Introduction ................................................................................................................................. 11 Background .................................................................................................................................. 13 Basic Facts of Japanese Accent .......................................................................................... 13 An Overview of Compound Accents ................................................................................. 16 Early Studies on Word and Phrasal Compounding ........................................................... 20 Deaccentuation ..................................................................................................................... 21 Analysis in an OT Approach ............................................................................................... 24 2 NEW TYPOLOGY AND COMPLEX STRUCTURES .......................................................... 28 Current Typology ........................................................................................................................ 28 The N ew Typology and the Application of Compound Accent Rules .................................... 34 Reanalysis of the Data ......................................................................................................... 34 Definitions and Structures ................................................................................................... 36 A Review of the Nine Basic Types .................................................................................... 39 Type 1 ........................................................................................................................... 39 Type 2 ........................................................................................................................... 41 Type 3 ........................................................................................................................... 42 Type 4 ........................................................................................................................... 44 Type 5 ........................................................................................................................... 45 Type 6 ........................................................................................................................... 45 Type 7 ........................................................................................................................... 46 Type 8 ........................................................................................................................... 48 Type 9 ........................................................................................................................... 49 Complex Structures ..................................................................................................................... 51 Triphrasal Structure: ............................................................................................................ 51 Right -branching compounds: ...................................................................................... 51 Biphrasal Structure ....................................................................................................... 54 Complex Word (Monophrasal) Compound ....................................................................... 57

PAGE 5

5 3 ANA LYSIS IN AN OT APPROACH ....................................................................................... 60 An OT Approach in Analyzing the Nine Types ........................................................................ 60 Prefixes ........................................................................................................................................ 71 4 CONCLUSION AND DISCUSSION ........................................................................................ 76 The Advantages of the New Analysis ........................................................................................ 76 The 2+2 Phenomenon in an OT Approach ............................................................................ 76 Left branching and Right branching Differences in an OT Approach .................................... 78 Conclusions, Problems and Further Research ........................................................................... 79 APPENDIX A DATA OF NINE BASIC TYPES .............................................................................................. 81 B THE TABLE FOR ALL ACCENT PATTERNS ..................................................................... 85 REFERENCES ................................................................................................................................... 88 BIOGRAPHICAL SKETCH ............................................................................................................. 91

PAGE 6

6 LIST OF TABLES Table page 2 1 The typology of compounds by Ito and Mester (2007) ....................................................... 30 2 2 Possible accent patterns for N2 fewer than or equal to two morae ..................................... 35 2 3 Po ssible accent patterns for N2 of three or four morae long ............................................... 35 2 4 Possible accent patterns for N2 with more than five morae ................................................ 36 2 5 The new typology for compounds ......................................................................................... 37 3 1 Type 1 the constraint ranking for the word sunabokori taisaku ..................................... 61 3 2 Type 1 the constraint ranking for the word nihonsigaku ................................................ 62 3 3 Type 2 the constraint ranking for the word soosuyakisoba ............................................ 63 3 4 Type 3 the constraint ranking for the word kaigai ryuugaku ......................................... 63 3 5 Type 3 the constraint ranking for the word inakama ...................................................... 64 3 6 Type 3 the constraint ranking for the word asagiri ......................................................... 64 3 7 Type 3 the constraint ranking for the word kinuito .......................................................... 65 3 8 Type 3 the constraint ranking for the word garasudama .................................................. 65 3 9 Type 3 the constraint ranking for the word settyuutaue ................................................. 66 3 10 Type 4 the constraint ranking for the word amerikasan orenzi ...................................... 66 3 11 Type 5 the constraint ranking fo r the word genzi monogatari ....................................... 67 3 12 Type 6 the constraint ranking for the word kita kariforunia ............................................ 68 3 13 Type 6 the constraint ranking for the word singata infuruenza ...................................... 68 3 14 Type 7 the constraint ranking for the word zimukyoku syokuin ................................... 69 3 15 Type 8 the constraint ranking for the word infuruenza kansenkaku ............................ 70 3 16 Type 9 the constraint ranking for the word kangoku hoomon ....................................... 71

PAGE 7

7 LIST OF FIGURES Figure page 1 1 Bennetts analysis of accent markers and tone assignment ................................................. 15 1 2 Odas analysis of short and long N2 structures .................................................................... 19 1 3 Odas analysis of the wordgarasu dama ............................................................................ 22 1 4 Odas analysis of the word kuro neko ............................................................................... 23 1 5 Odas analysis of the word hosi awabi .............................................................................. 23 1 6 Constrai nt ranking by Tanaka (2001) ................................................................................... 26 2 1 Junctural accent ...................................................................................................................... 28 2 2 Biphrasal structure and Rendaku distribution ...................................................................... 38 2 3 The structure of a word compound dominated by a phrasal node ...................................... 38 2 4 Type 1 Left -branching word compounds ............................................................................. 39 2 5 Type 2 Right -branching word compounds ........................................................................... 41 2 6 Type 3 word compounds ........................................................................................................ 42 2 7 Type 4 left -branching monophrasal compounds .................................................................. 44 2 8 Type 5 right -branching monophrasal compounds ................................................................ 45 2 9 Type 6 monophrasal compounds ........................................................................................... 46 2 10 Type 7 left -branching biphrasal compounds ........................................................................ 47 2 11 Type 8 right -branching biphrasal compounds ...................................................................... 48 2 12 Type 9 biphrasal compounds ................................................................................................. 49 2 13 The c omplex st ructure for doosoukaikozinzyoohoohogo hoosin ................................ 51 2 14 The complex structure for the word sangaku kooryuukai .............................................. 52 2 15 The complex structure for the word hookeigakubukeizaigakkasotu ........................... 52 2 16 The complex structure for syutodaigakutookyoo soogookyoogi taikai ....................... 53 2 17 The complex structure for the word tyoo kantansetuyakuresipi ................................... 53

PAGE 8

8 2 18 The complex structure for kitakariforunia asuparaga su / orenzi .................................... 53 2 19 The complex structure for the word keizaikikakutyoo nyuutyoo .................................. 54 2 20 The complex struct ure for the word kaisya setumeikai setuei ....................................... 54 2 2 1 The complex structure for the word oosakabankoku hakurankai .................................. 55 2 22 The complex structure for the word nihon seihu daihyoobu ........................................ 55 2 2 3 The complex structure for the word oosakabankoku hakurankai ................................. 55 2 24 The complex structure for the word minamioosawa hakubutukan ............................... 56 2 2 5 The complex structure for the word syuusyokuzi mensetusidoo .................................. 56 2 26 The complex structure for the word nihon sigaku kooza .............................................. 57 2 2 7 The complex structure for minamioosaw a kyanpasusangakukooryuu kai ................... 57 2 28 The complex structure for the word y o zakura ginzi ziken ............................................ 58 2 29 Th e complex structure of the word dokuiriorenziziken .................................................. 58 2 30 The complex structure of the word nihon sigaku kooza ................................................. 58 3 1 The structure of the word koo bunsi .................................................................................. 72 3 2 The structure of the word koo enerugi .............................................................................. 72 3 3 The structure of the word koo byoogen sei ..................................................................... 72 3 4 The structure of the word koo uirusu ............................................................................... 74 3 5 The structure of the word koo uirusu yaku .................................................................... 74 3 6 The wrong structure of the word kou uirusu yaku ......................................................... 74 3 7 The wrong st ructure of the word koo uirusu yaku ........................................................ 75 4 1 The balanced structures .......................................................................................................... 77 4 2 The structure of the word onda nkaboosi ......................................................................... 79

PAGE 9

9 Abstract of Thesis Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Master of Arts NEW TYPOLOGY OF JAPANESE COMPOUND ACCENTS AND AN ANALYSIS IN OPTIMALITY THEORY By Si Chen August 2010 Chair: Caroline Wiltshire Major: Linguistics This paper explores a new typology of compound accents in Japanese, containing nine basic structures, and analyzes prefixes and complex c ompounds containing more than three words using these basic structures. Data of Japanese compound accents were collected from dictionaries and websites. Japanese native speakers were recorded reading all the data. Based on the recorded accent, further anal ysis has been done in assigning accent markers. After investigating the accent patterns through all different length categories and the various original accent patterns of the component words in isolation, common characteristics involved in word, monophras al and biphrasal levels are proposed which form the basis of the new typology. In the later analysis, an OT (optimality theory) approach is provided to account for the assignment of the optimal accent pattern for a given input. Different levels of constra ints such as correspond to distinct levels of the internal structure of compounds. More specifically, three constraints are proposed to account for the deaccentuation phenomenon. Constraints on the phonology-syntax interface are used to account for the difference between left -branching and right -branching compounds. My analysis accounts for the internal structure of compounds and redefines and associates different

PAGE 10

10 levels. More exceptions can be explained by this approach, and the OT analysis is more complete for deaccentuation phenomenon and for compounds with distinct structures.

PAGE 11

11 CHAPTER 1 BACKGROUND Introduction The accent patterns of standard Japanese have been an interesting topic since as early a s the 1960s. In Japanese, similar to tone languages, accent is assigned to almost every word, but it is not found in tone languages. There are only level tones in Japanese, while tone languages can also have contour tones. The accents of compounds are not easily predicted by phonological rules because compounds may have different structures influenced by semantic or pragmatic information1 1 In Vances (1997) review of Kubozonos (1993) book, biphrasal compounds are described as more likely to be used by speakers in careful speech. If a speaker has a pragmatic focus on the compound, then the whole compound or part of it are more likely to be biphrasal. (Vance, 1997). Previous research by McCawley (1977), Tsujimura (1987), Kubozono (2001) and Oda (2006) dealt well with th e accent patterns of compounds at the word level. They agree that the length of N2 (the second member of a compound) will affect the accent location. Most of them agree that when N2 is long (three morae and more), the accent will fall on the first mora of N2 and when N2 is short (two morae or fewer), the accent will fall on the final mora of N1 (the first member of a compound). Ito and Mester (2007) made use of this difference in the length of N2 and proposed the concept of the junctural accent, which is a default position where compound accents will fall. This position is at the boundary of N1 and N2, either the final mora of N1 or the first mora of N2. Oda (2006) analyzed the structural difference between short and long N2. Kubozono, Ito and Mester (1997) notice the different levels of compounds: word level and phrasal level. Ito and Mester (2007) have further divided the levels into word, monophrasal and biphrasal compounds, where junctural accent is the only standard to distinguish word and phrasal levels The length of N2 determines if the compound is in the word or phrasal level.

PAGE 12

12 This thesis further investigates accent patterns in Japanese compounds and explores nine basic structures of compounds which may affect accent patterns. A more complex standard is set up based on characteristics of the three levels (word, monophrasal and biphrasal levels). As this standard reflects the internal association between these levels and through the change of accent location, the change across these levels is also demon strated, such as the change from a monophrasal compound to a word compound. This standard can account for the exceptions such as kudamono fruit which does not have a junctural accent but is predicted to have one according to the standard in Ito and Mest er (2007). Also, compounds containing N2 of more than four morae are predicted to have no junctural accent by Ito and Mester (2007) while taimingu timing and akusento accent are exceptions to this prediction, which can be handled using my standard. My analysis also provides a new typology and accounts for the difference between left branching and right -branching compounds. More complex structures and prefixes which have never been analyzed for different levels are analyzed based on basic structures with in the typology. For the OT analysis, constraints from different levels are proposed to deal with cases of junctural accent and accent retention. As for deaccentuation phenomenon, in which the whole compound loses its compound, I propose a new constraint IDENT O which explains why final accented N2 will trigger deaccentuation. Previous research on this topic will be mentioned later in this chapter. I also proposed two constraints in Chapter 4 to account for the structure of data in 2+2 phenome non proposed by Oda (2006). In Chapter 1, the basic facts of Japanese accent are introduced, with a summary of previous analyses and an overview of the proposed analysis of compound accents. Chapter 1 also summarizes previous research on the difference bet ween word and phrasal accents and a special

PAGE 13

13 phenomenon called deaccentuation, as well as previous analyses done in OT (Prince and Smolensky, 1993). Chapter 2 deals with the typology of compound accents and proposes nine basic structures for word, phras al and biphrasal structures. Data are presented after each type. Some of the data were collected through the two dictionaries Shinmeikai Nihongo Akusento Ziten (2001) and NHK Nihongo Hatuon Akusento Ziten (1998). Other compound data were collected thro ugh the Internet and read by four Japanese native speakers. I confirmed some of the data from the dictionary with the native speakers. They were asked to judge the appropriateness of the compounds, and inappropriate compounds were deleted. I detected the l ocation of the pitch -fall throughout the data and confirmed some locations with the native speakers. The accent markers on single words and some compounds are written in the dictionaries. Some of the Rendaku (sequential voicing in compounds) changes can be found in the dictionary and others were confirmed with the native speakers. Based on the nine basic structures, some complex structures can be built to predict the placement of the accent marker. In the following chapters, data from other researchers will be cited, and the rest of the data are all collected by the author and can be founded in Appendix A. Chapter 3 analyzes every typical structural pattern mentioned in Chapter 2 using an OT approach, and Chapter 4 discusses the advantage s of the new typolog y and some remaining issues for further research. Background Basic Facts of Japanese Accent While syllables can account for phonological structures in English, the notion of mora plays an important role in analyzing Japanese phonological structure. A mor a can be a vowel or a CV. It can also be the first part of a long consonant, a geminate, and the syllable -final nasal /n/.

PAGE 14

14 For example, in the word ansin not worried, there are four morae a, n, si, n. A long vowel such as oo contains two morae. (Tsujimur a 2007). Japanese is a pitch accent language which bears a tone on each mora. Unlike tone languages, the tonal pattern can be predicted given the location of an accent. The accent marker generally indicates a pitch fall in different analyses. In this p aper, the will be used for convenience to indicate the accent marker. For example, in the word sora sky, the accent is located at the first syllable on the vowel o and the word has a tonal pattern of high -low. If the accent is located at the seco nd mora, as is in kokoro heart, the tonal pattern is low -highlow. If the accent is on the third mora, the tonal pattern would be low -high-high. Moreover, if there is no accent, the whole word would be low -high -high. All these types can be predicted by rules found in the analyses of McCawley (1968, 1977), Haraguchi (1977) and Bennett (1981). While tones are assigned to morae, a syllable bears an accent s o pitch -fall will not occur after a coda: N (nasal) ,Q (a long consonant) or a long vowel and any ru le that trie s to assign an accent to the coda will fail T he accent will move to the vowel before the nasal or to the middle of a long vowel (Iwasaki 2002). For example, in the word tenki weather, the accent is assigned as in tenki instead of tenki. Mc Cawley, Haraguchi and Bennett have slightly different rules for the association of tones. McCawleys pitch assignment rules initially make every mora high pitched. Then he assigns low tones after the accented mora, and finally he assigns a low tone to the first mora if the second one is high pitched (McCawley, 1977). Haraguchi generated the placement of tone through four steps involving four rules. The basic tone HL is associated with the CV tier according to the rules in order: Tone Associate Rule (TA), U niversal Tone Association Convention (UTAC), Initial Lowering (IL) and Tone

PAGE 15

15 simplification (TS). Haraguchi argued that the first step is to associate the high tone to a vowel with an accent mark or to the rightmost vowel in cases where there is no accent m ark. After the application of UTAC, the initial lowering rule is applied to assign a low tone to the first mora with no underlying accent. The last step is to eliminate contour tones by delinking the second tone associated with the same vowel Haraguchi (19 77). Unlike Haraguchi, Bennett (1981) analyzed Japanese accent s on two levels: accent foot (@) in a phonological word and an accent marker (*) in one morpheme. As shown in Figure 1 1, s he uses a left -branching tree to derive accent patterns. She argued that within @, H is assigned to the right node otherwise L is associated with the right node Figure 1 1. Bennetts analysis of accent markers and tone assignment Haraguchi (1991) analyzed the accent patterns on single words including nouns, verbs and a djectives.T he accent of nouns is usually assigned to the antepenultimate mora by rules. L oanwords and compound nouns also have this tendency. Adjectives and verbs can either have an accent or not. The accent will fall in the final vowel of the stem. He a lso specified his analysis based on the length of nouns. T he accent of short nouns cannot be predicted while long accented noun words tend to have the accent in the antepenultimate mora. When a high vowel is

PAGE 16

16 devoiced because it appears between two voiceles s consonants or word finally after a voiceless consonant the accent will be shifted to the left. For example, the word on gaku kai music concert will become ongakUkai ; however, the accent on the initial mora will shift to the right from hukaku to h U k a k U deep. An Overview of Compound Accents The formation of compound nouns has been investigated intensively. In McCawleys (1968) analysis the second noun in a compound ( N2 ) has a dominant role in determining the accent of the whole compound regardless of the first noun (N1) Most later analyses also follow this approach. Long compounds and short compounds are divided based on the length of N2. A word is considered short if it is only made of one or two morae, and it is considered long when it has three or more morae (McCawley, 1968). Though analyses of compound rules vary, they basically accounted for unaccented compounds and similar accent positions in compounds: the first syllable of N2, the last syllable of N1, the accent position retained in N2. Di fferent analyses for these cases are shown by the following. I. The accent of the whole compound will be assigned on the first syllable of N2: McCawley (1977) when: N2 is long (having at least three morae) N2 is a Sino Japanese word N2 is final accented or completely unaccented; Tsujimura (1987) when: A long N2 is unaccented or has an accent on the penultimate or final mora; Oda (2006) when: N2 is long.

PAGE 17

17 II. The accent will be on the final syllable of N1 Kubozono (2001) when: If N2 is short and has an accent o n the final syllable or is unaccented. Oda (2006) when: N2 is short. III. The whole compound has the accent at the same position as N2: McCawley (1977) In other cases except for I and IV; Kubozono (2001) In other cases except for I and II. IV The whole compou nd would be unaccented: McCawley (1977) when: N2 is short and the final mora is accented Oda (2006) In other cases except for I and II. A more detailed summary of the different analyses mentioned above is provided below. McCawley (1968) proposed three c ases of noun compound formation. In the first case, the whole compound has the accent in the same position as in isolation, N2 such as in genzi genzi (name) +monogatari story = genzi -monogatari Tale of Genzi Kubozono proposed some exceptions to this rule inc luding certain morpheme s such as hime which do not follow the rule above: sirayuki white snow+ hime princess= sirayukihime Princess Snow White (Kubozono, 2001). In McCawleys second case, when long N2 is a Sino Japanese word or ha s at leas t three morae and is final accented or unaccented, the accent of the whole compound will be

PAGE 18

18 assigned to the first syllable of N2 He gives some examples: noogyoo agriculture +kumiai union = noogyo-kumiai agriculture union inaka country +musume da ughter = inakamusume country gir l In McCawleys third case, when N2 is short and final accented, the whole compound loses its accent. In order to explain some exceptions such as kabuto helmet +musi bug = kabutomusi a beetle which bears an accent on the last syllable of N1 i nstead of being unaccented he proposes an underlying preaccented musi so that the accent of the whole compound is dominated by the accent of N2 as in the first case ( McCawley 1977) However, to assume an underlying preaccente d N2 may not explain other cases. Higurashi (1983) g ave some examples: sato hometown +kokoro heart=* satogokoro homesickness, correct: satogokoro; yude (ru ) boiled +tamago eggs = **yudetamago boiled eggs correct: yudetamago Higurashi (1983). Tsujimura (1987) also analyzed long compound formation rules: 1 For a compound whose second member has an accent in the penultimate or the final mora, the accent of the whole compound is assigned to the first mora of N2 2 For a compound whose N2 is not assigned an accent, the accent of the whole compound is assigned to the first mora of N2 such as ni load + kuruma car = ni guruma cart Kubozono (2001) divided compounds into long and short ones. For long compounds, the accent of the whole compound wi ll be the same as the original accent of N2 unless N2 has its accent in the final syllable. If the accent of N2 is in the final syllable or N2 is unaccented, the whole compound will be assigned a new accent on the final syllable of N1 Also in short compounds, the accent will appear in the final syllable of N1 such as ka.bu.to helmet +mu.si bug = ka.bu.to.mu.si a beetle Oda (2006) proposed that there are three patterns of compound accent placement The first is that N2 retains its accent. The second is that if N2 is long, the original accent will be at the left

PAGE 19

19 edge of N2, otherwise, it will move to the end of N1. The third possible case is for the compound to be unaccented. He did not specify the condition for N2 to retain its accent or shift. Oda di stinguishes between word compounds and extended-word compounds to avoid the arbitrary difference between short and long compounds. He based his analysis on the five following assumptions and gave the internal structure of short and long N2: s hort N2 can on ly form a single f oot while long N2 will always form more than one foot. Figure 1 2. Odas analysis of short and long N2 structures 1 Assumption 1 P rosodic constituents should be parsed in the leftward direction and the unit of parsing should be a mora. 2 Assumption 2 P arsing has to follow the Strict Layering Hypothesis (Selkirk, 1984), which prohibits a prosodic level that is not dominated by a unit in the immediate level above It means that the node in the top level can only dominate the middle level wh ich can only dominate the bottom level. However, the top level node cannot directly dominate the bottom level. 3 Assumption 3 P arsing should be maximally binary, not to have more than two branches. 4 Assumption 4 Parsing is also morphologically sensitive. 5 Ass umption 5 M inimality requirement in J apanese It (1990) a b His analysis focused on the different inner structures in short and long compounds based on the above assumptions.

PAGE 20

20 Early Studies on Word and Phrasal Compounding Kub ozono, Ito and Mester (1997) noticed five grammatical limitations on compounding. Two members in a compound may retain their own accents if they do not combine phonologically as a prosodic word, which is treated as compounding in the phrasal level by Ito a nd Mester (2007). Kubozono et al. gave five scenario s where semantic relations may cause two words not to combine as a word compound: 1 Certain prefix es such as zen before and bo o certain 22 Two members are in an equal status. For example haku syu kassai c lapping hands and a cclaiming retain their own accents. However, if each member has no more than two morae, the compo und will have one combined accent. 3 Two members have a subject object relation. For example kazi tetudai helping of chores 4 Names plus titles such as ku rin ton daito o ryoo President Cli n ton 5 Family name s and first name s such as yukawa hideki However, if the internal structure of a name is not considered, the whole name can bear one accent. If two parts of a compound refer to the same content such as N1: content N2: label, N1: last name N2: first name, N1: name N2: organization, or N1: title N2 name such as kawase hakase Dr. Kawase they tend to be biphrasal compounds which retain their accents on both N1 and N2. The semant ic information described here is also considered in my definition in Chapter 2. Kubozono et al. also argued that Japanese has the tendency to block combining of the accents of two members in right branching compounds. They pointed out that N2 with more than 5 morae or three -morpheme Sino Japanese words tend to keep their accent pattern, namely, to remain unaccented or to have the accent at the same position as the words in isolation 2 All the long vowel ou in the Japanese data are represe nted as oo for consistency, even if the source used ou.

PAGE 21

21 They believed that i n a compound word, the accent of the two members are d ependent, even if the two do not bear accents, and the whole compound will have at most one. By contrast, i n a phrase, the accent of both members is independent and will not change. Some phrase s may undergo the process of becoming a compound f or example, erizabe su Elizabethan + zyoo o queen = er i zabesu zyoo o or erizabe suzy o o o Queen Elizabeth The affix connecting two members in a word may cause the accent on the first member to disappear and this helps keep the accent of N2. This is similar to c ompounds in the de accentuation pattern. In order to account for the similarity between 5 morae and three -morpheme SinoJapanese words, Kubozono introduces the standard of counting in f ee t. Two morae or a Sino Japanese morpheme can constitute one foot. F or words of five morae and three -morphemes, there are at least three feet. When N2 is less than or equal to two fee t, the whole compound is a word compound. For three -foot N2, accent pattern of N2 will be retained. For N2 with more than three feet the accent of each of the two members will be retained ( Kubozono et al. 1997). Oda (2006)s analysis on short and long compounds correspond to the word compound level in Kubozono et al. (1997)s research. Different length within word compounds may result in diff erent accent locations. Deaccentuation Oda (2006) proposed three puzzles concerning deaccentuation phenomena in compounds : First, finally accented N2 may cause the whole compound to be deaccented (McCawley 1968: 168) Second, when the whole compound cons ists of four light syllables and N2 has two light syllables, the whole compound tends to be deaccented. Third, if N2 is long, the whole compound will never be deaccented.

PAGE 22

22 He labeled the second puzzle as the 2+2 deaccentuation : with the same N2 a compou nd will have an accent in the antepenultimate position when its N1 is long, but it is realized with no accent when its N1 is bimoraic ( with exceptions such as hime princess) At the first glance, 2+2 deaccentuation seems to be against the common assum ption that N1 is invisible. Internal structure mapping may explain the invisibility of N1. Here is an example of garasu-dama glass ball given by Oda to illustrate that N1 is mapped as a whole with no concern about its internal structure ( ). Figure 1 3. Odas analysis of the wordgarasu dama Oda proposed conditions on compounding that an accent must appear for compounds that have a culminating point (the point at the top node of the structure) of extended prosodic word and stated that the 2+2 phenomenon is not in contrast with the invisibility of N1 because the structure of N1 is visible but often ignored. The 2+2structure does not have a culminating point of an extended prosodic word as follows:

PAGE 23

23 Figure 1 4. Odas analysis of the word kuro neko the following structure: Figure 1 5. Odas analysis of the word hosi awabi Most of the research on compound accents above does not distinguish between word comp ounds and phrasal compounds. Almost everyone agrees that the assignment of accent relies on the length or the structure of N2. In the above overview of previous research, little attention has been given to the original accent on N2 and its role in determining the accent pattern. The generalization is missing that N2 with an original accent on the first mora will almost always retain it regardless of the length of N2. Even if a N2 is considered to be long or short, it may not behave as the rules predict beca use deeper structure is involved. R ecent research, especially that in an OT approach, paid more attention to the comparison of the accent input and the output. In

PAGE 24

24 later chapters, I develop a typology considering both the length, structure and the original accent of N2 as well as the structure of the whole compound. Analysis in an OT Approach Kubozono (1995) has done an OT analysis of Japanese compounds and has proposed five constraints: 1 PARSE (accent): N2 should retain its accent 2 NONprosodic word. 3 Non -finality (F): No final accented foot in a prosodic word. 4 ALIGN CA: The accent should be aligned either left or right to the N1 N2 boundary. 5 Rightmostn ess: The accent should be assigned to the right edge of a prosodic word. He gave the ranking as NON >> NON-FINALITY (F),ALIGN CA>> RIGHTMOSTNESS Sino Japanese morphemes are exceptions to compound rules such as in yo.ya.k reserved + sek seats = yo.ya.k. se.k reservation. Kubozono explained that SJ words behave as if they are monosyllabic and if the final i in seki is invisible to the above rule, then SJ words will not be an exception. Kubozono hence pr oposed a constraint NON ) to solve this problem. This prohibits a final accent in the final foot so that the problem of SJ exceptions can be explained. After the first analysis in the OT approach by Kubozono, Tanaka (2001) has concluded, base d on previous research, that there are three primary characteristics of Japanese compound s First, the accent on N2 is usually retain ed or a new accent will occur on the first mora of N2. Second, accent on the final syllable or mora is avoided. Finally, for short N2, namely one or two morae, an accent will occur just before the N1 -N2 boundary, otherwise it will occur immediately

PAGE 25

25 after it. These three properties may be in conflict when accent needs to appear before the boundary and N2 needs to retain its acc ent. Tanaka revised K ubozonos generalization and pointed out that compound accent is usually assigned to the penultimate foot unless preservation is needed in foreign, archaic native and Sino Japanese heads. Accordingly, T anaka based his constraints on Ku bozonos MAX (accent) which requires the accent of the head root to be retain ed ; ALIGN which requires the accented syllable to align to the left edge of a head root; and ALIGN-R which requires the alignment between the right edge of a prosodic word and that of the accented syllable. Tanaka explained that MAX (accent) accounts for unaccented words better than PARSE (accent) in the case of unaccented N2. ALIGN in mimetic words. ALIGN because of its gradient property and it can be applied among many languages. For native words such as ningyohime doll princess the accent cannot occur on the first mora of hime because it violates NON which ranks higher than MAX (accent) according to Tanaka. For Sino Japanese words, some final accented words are still parsed but there are also unparsed variation s such as nihon Japan +zin person= nihonzin or nihonzin Japanese people Tanaka argued that NON rankable. For foreign compounds, th e ranking is almost always MAX (accent) >>NON with some exceptions of nativized loanwords such as suupaa super + man man = suupaaman superman

PAGE 26

26 Tanaka explained all the data using the constraints mentioned above. For foreign (parsed), MAX (accent) ranks higher than NON some variations in archaic native and Sino Japanese words, MAX (accent) and NON FINALITY t) ranks higher thanNON root) which requires the left edge of an accented syllable to align to the root, ensure that the winner will have its accent o n the left edge of N2 mora. Tana ka g ave three cases where the de accentuation phenomenon occurs: finalaccented N2, quadrimoraic words and words with both conditions met. For this phenomenon, Tanaka proposed a new constraint called NON FINALITY (PrWd) which forbids the accent to occur o n the final prosodic word. So the constraint ranking is revised as below (t he dotted lines are an indication of rerankable constraints ): Figure 1 6. Constraint ranking by Tanaka (2001) In this ranking, there are three pairs that can be re ranked becau se Tanakas analysis does not distinguish different compound structures. It can be improved by using different constraints to analyze different structures, namely, the basic nine types presented in Chapter 2. In my analysis in Chapter 3, I split the constr aint MAX (accent) into different levels to deal with both

PAGE 27

27 word and phrasal compounds separately. Tanaka proposes the constraint NON-FINALITY (PrWd) in order to account for the deaccentuation phenomenon. However, this constraint is violated by almost e very candidate and thus is not so convincing. My OT analysis approaches this problem by using the constraint IDENT O N2. Except for compounds undergoing deaccentuation, there is no constraint in this ranking to reflect the fact that most compounds need at least one accent regardless of their structure. The data show that a constraint is needed to account for this generalization, and Chapter 3 provides a more specific explanation. The above sections provide a n overview of Japanese accents on single words and compounds and the OT analyses proposed to date. Except for the level of word and phrasal compounds mentioned in Kubozono et al. (1997), most previous research has focused more on word level. In the followi ng discussion, I introduce recent research on dividing the compounds into three levels by Ito and Mester (2007). Following that, I redefine the three levels and provide a new typology for further analysis on complex structures in Japanese compounding.

PAGE 28

28 CH APTER 2 NEW TYPOLOGY AND COM PLEX STRUCTURES Current Typology Ito and Mester (2007) proposed t hree basic categor ies T he internal structure of compounds may cause different accent patterns. Apart from these three categories, relational projections are defined: maximal est/lowest element in a projection. It is argued that Japanese can have only one accent in each or pitch accent ( Ito and Mester 2007) Ito and Mester classified compounds into word compounds and phrasal com pounds. Only in a word compound can a junctural accent occur. The standard for distinguishing a word compound from a phras al compound lies in the length of the second member. If the second member contains more than four morae, then the whole compound has t o be parsed as a phrasal compound (Ito and Mester 1997). M onophrasal and biphrasal compounds do not allow any junctural accent to happen according to Kubozono (1988). In a word compound, a junctural accent is placed in Figure 2 1. Figure 2 1. Junctura l accent Howeve r, the position of the junctural accent may coincide with the original N2 accent. For example, in the biphrasal compound yukawa hideki Yukawa Hideki (name), the original

PAGE 29

29 accent of N2 is on the junctural position. So on the surface, findi ng an accent in the junctural position may not be a perfect standard to determine whether a compound is a word or a phrasal compound Although Ito et al.(2007) argued that N2 under four mora should form a word compound the compound aka red + ke hair = ak age red hair, does no t have a junctural accent and should be an exception. The 2+2 structures also contain a N2 that is less than four mora such as tabi travel + hito person = tabibito traveler without any junctural accent. Moreover, since N2 with more than 5 morae should form a phrasal compound, no junctural accent should occur as is given in Ito and Mester (2007). However, in the word hoogen akusento dialect and accents there is a n accent in the juncture position which coincides with the orig inal accent on akusento accent Another exception to this analysis is the word tenka taimingu the timing of the sparks. The N2 of this compound has five morae, which should stay unaccented according to Ito and Mesters analysis, however, the junctur al accent occurs since taimigu is originally unaccented. The following table shows the typology Ito and Mester proposed. In this table, the square represents the minimal projection, and the round shape represents the maximal projection. The symbol r represents the phenomenon of Rendaku. Rendaku is a phenomenon that happens in compounding where the first consonant, if it is an obstruent, of the second word becomes voiced after being combined into a compound. For example, iro color and kami paper c ombines to form irogami colored paper where the consonant k becomes the voiced g (Tsujimura 2007). Rendaku can only occur at the place of a +r but not r. The word on the right branching usually has the +r.

PAGE 30

30 Table 2 1. The typology of compo unds by Ito and Mester (2007) According to Kubozono, Ito and Mester (1997), t he crucial factor that determines wh ether a compound is a word compound or phrasal compound is the length of N2. However, it is apparently not true for biphrasal compounds since even if N2 containts fewer than 4 morae there is no junctural accent. For minami amerika asuparagasu South American asparagus, the derivation steps are: Input Output Rules Step 1 minami +amerika minami amerika junctural accent Step 2 minami amerika+asuparagasu minami amerika asuparagasu deaccentuate N1 It is clear that N1 minami amerika South America is deaccented because it is originally minamiamerika South America When attached to a 5 morae word asuparagasu, N1 undergoes deaccentuation, and the whole compound is thus a monophrase. For words with fewer

PAGE 31

31 than 5morae, such as biiru beer they usually have an original accent on the first mora. So in a left branching compound minami amerika biiru South America beer one cannot be certain if it is a word compound with a new junctural accent on N2 or if it is a monophrasal compound which retains its original accent. I will treat it as a monophrasal compound in the OT approa ch section, to make it parallel with the South American asparagus case, which must be a monophrase. W ords exceeding 4 morae can be more easily identified as monophrase s because they can have an original accent on a position other than the first mora Thu s when N1 is added, although N1 is deaccented, the compound accent will not occur in the default position as the new accent for a compound word, while N2 retains its own accent as seen in asuparagasu asparagus and minami amerika asuparagasu South Amer ican asparagus W ords of fewer than four morae can also have an accent on a position other than the first mora In compounding, those words will also retain their accents, such as kudamono fruitsand orenzi orange where no junctural accent will appe ar (1 ) mu without +no o yaku pesticides+kudamono fruits =muno o yaku+kudamono =munooyakukudamono organic fruits (2 ) amerika+san+orenz i =amerikasan+orenzi=amerikasan orenzi American oranges S ince the junctur al accent does not occur as in munooyaku kudamono organic fruits it is a phrase and not a word compound The derivation steps are as follows: Input Output Rules Step 1 mu+ nooyaku munooyaku junctural accent Step 2 munooyaku+kudamono munooyaku kudamono deaccentuation of N1

PAGE 32

32 If it were a biphrasal compound, then it should be muno o yaku kudamono organic fruits. This is because the two lexical items in the input of step 2 should have retained their accents. S o we can say that it is a monophrase with a n N2 of only four morae. The phenomenon of Rendaku (sequential voicing) has something to do with compound structure as shown in the typology table above. Rendaku usually happens in word compounds. However, t he oc currence of Rendaku cannot guarantee the identity of a word compound. There are variations to word compounds and monophrasal compound s (Kubozono 1995, 1997). Monophrasal compounds can also have Rendaku. Rendaku does not happen in a predicate argument relat ion such as sakana turi fishing. Biphrasal compounds usually happen in words having a predicate argument relation as in kazi chores+tetudai helping= kazitetudai helping with chores. Rendaku usually occurs in N2 of word compounds, but not always For example, the compound sakana fish+turi catch =sakanaturi fishing, (data from Sugioka (1986)) has the characteristics of a word compound because the accent falls on the last mora of N1 which is a junctural accent. However, if it is a word com pound, then Rendaku should appear according to the prediction of Ito & Mester (2007). On the contrary, if a compound shows the characteristics of a biphrasal compound, Rendaku will not occur between two phrases. Tanaka points out that the version of the ac cent on N2 being retained in compounds is archaic, which means that the second type in the output such as densyo bato and koomori gasa is archaic as follows (Tanaka 2001): (3 ) densyo transferring messages +hato pigeon = densyobato/densyo bato homing pig eon (4 ) koomori bat +kasa umbrella= koomorigasa/koomori gasa black umbrella (5 ) hidari left +utiwa Japanese fan= hidariutiwa /hidari utiwa comfortable life

PAGE 33

33 He did not treat these variations with different compound structures. But they can actuall y be analyzed as having different compound structures: word compounds or monophrasal compounds, s o there is a trend for these words to change from a monophrasal compound to a word compound. In English, a similar phenomenon called lexicalization can occur, where a phrase may lose its internal phonological structure. For example, the phrase ice cream may lose its stress on the word cream and the stress is assigned on ice because the whole phrase enters into the lexicon and becomes ice cream. In Old English, lexicalization may happen so that the whole compound may be fused into a word such as earwig< OE eare ear+wicga one that moves Brinton (2005). In French, syntactic phrases can be lexicalized with a pattern of A+N and N+A, such as saint -bernard St Bernard Dog and table ronde round table meeting (Mathieu Colas 1996). In Dutch, phrases become compounds frequently as well, such as sneltrein fast train (Schlucker 2008). In Japanese, t here can be variations as in mono things +hosi dryingmonohosi/ monohosi frame for drying clothes The compound mono hosi is a biphrasal compound which retains its accent pattern on N1 and N2. If it is a word compound, N2 final accent should cause the whole compound to be deaccented Thus, the fact that the accent occur s on monohosi is still unexplained. No Rendaku happens in the monohosi dried clothes because of the predicate argument relation while we do see Rendaku happen in kagebosi dry in the shadow The accent is kage shadow + h osi dr ying =kagebosi/ kagebosi dry in the shadow The forms kagebosi and kagebosi reflect a word compound structure while kagebosi is a monophrase. Kubozono mentioned the variation: son Son (last name) +gokuwu gokuwu (first name) =son gokuwu/ son go kuwu Son Gokuwu (name) which reflects the variation of a

PAGE 34

34 biphrasal and word compound. Usually, for biphrasal compounds, Rendaku does not happen as for example between the first and last family name. The New Typology and the Application of Compound Accent Rules Reanalysis of the Data It has been mentioned in Chapter 1 that there are exceptions to the standard of the length of N2. In Ito & Mester (2007)s data dai+sakusika great lyric -person is unaccented even though sakusika has only four morae. They ex plain that the structure is (saku)(si)+(ka) instead of binary (saku)(si+ka) so the word violates MAXBIN because of the superbinary second member. Thus, the whole compound should be a monophrasal one. However, the four -morae word danraku paragraph remain s unaccented in the compound hukusuu plural +danraku paragraph= hukusuudanraku more than one paragraph although this word danraku observes MAXBIN (two bimoraic feet). Moreover, the binary word hanayome has two morphemes hana flower and yome brid e, which is binary, but the whole compound retains its accent in nihonzin hanayome Japanese bride Using the current data, I explore the accent pattern according to the length of N2 as follows (Please refer to Appendix B for more data): It is interest ing to note that if the length of N2 is greater than five morae, then final accented examples are hard to find. The accent on N2 of more than five morae is the most stable because the whole compound can keep its accent or remain unaccented if N2 is origina lly so. It is relatively hard to remain unaccented if N2 contains fewer than 5 morae and does not have an accent originally. On the contrary, the whole compound tends to keep the accent on the first syllable if N2 has an original accent there, but N2 with 2 morae or fewer can be so active that even the accent on the first syllable can shift. Moreover, finally accented N2 usually triggers deaccentuation of the whole compound, but finally accented N2 with fewer than 5 morae may still have a junctural accent w hen forming a compound.

PAGE 35

35 Table 2 2. Possible accent patterns for N2 fewer than or equal to two morae Short N2 Length<=2 morae N2 Accented on the first syllable Retain the original accent of N2 on first syllable or shift to the final syllable of N1 (junct ural accent) densyo carrier+hato pigeon= densyobato or desyobato carrier pigeon From Tanaka (2001) Only shift the original accent of N2 to the final syllable of N1 ningyo mermaid+hime princess=nigyohime mermaid princess From Tanaka (200 1) Only retain the original accent of N2 kafe cafe+ baa bar=kafebaa caf bar From Tanaka (2001) N2 Unaccented A new junctural accent may appear sasa bamboo leaf+amecandy=sasaame candy wrapped in bamboo leaves The lack of accent may r emain eda branch+ke hair=edage split hair N2 Final accented The accent may shift to the final syllable of N1. suidoo waterworks+hasi bridge=suidoo basi Suido Bridge (place name) The whole compound may be deaccented akane madder+iro co lor=akaneiro madder red Table 2 3. Possible accent patterns for N2 of three or four morae long Short N2 Length=3 or 4 morae N2 Accented on the first syllable Retain accent on the first syllable: coincides with the junctural accent soosa investigat ion+kaigi conference=soosakaigi investigation meetings N2 Unaccented New junctural accent may appear sin New+yokohama Yokohama (place name)= Sinyokohama New Yokohama From Tanaka (2001) The whole compound may remain unaccented hukusuu pl ural+danraku paragraph=hukusuu danraku more than one paragraph N2 Final accented Accent may shift to junctural position dezaato dessert +azuki red azuki bean=dezaatoazuki dessert made of red azuki bean The whole compound may be deaccented hasan bankruptcy +tuutisyo notification= hasan tuutisyo the notification of bankruptcy N2 Accented in the middle The accent may be retained saga saga (name)+tenoo emperor=sagatennou Emperor Saga The accent may be shifted to the first mor a of N2 genkin cash+hurikomi deposit=genkin hurikomi cash deposit From Tanaka (2001)

PAGE 36

36 Table 2 4. Possible accent patterns for N2 with more than five morae Long N2 Length>=5 morae N2 Accented on the first syllable Retain accent on the first syllab le: coincides with the junctural accent sokuseki on the spot +daietto diet= sokuseki daietto diet that can be effective immediately N2 Unaccented New junctural accent may appear kyooka strengthen+taimingu timing=kyooka taimingu fine tune timing The lack of an accent may remain kooon high temperature+taikyuusei durability= kooon taikyuusei durability of high temperature N2 Accented in the middle The accent may be retained sya nai in a company+ danketuryoku power of unity= syanai danketuryoku the power of unity in a company Looking through all of the tables, the length of N2 is related to the accent pattern of the whole compound, but it is not decisive. The inner structure of the compounds concerning its semantics and other factors might have led to the accent pattern. There are some common characteristics in the accent patterns across the tables such as the junctural accent and the retention of the accent. Based on those characristics, I offered new definitions of word, mon ophrasal and biphrasal compound in the next section. Definitions and Structures A word compound has a junctural accent either on the first syllable of N2 or the last syllable of N1 except for deaccentuation, which happens in the 2+2 phenomenonand word co mpounds with final accented N2. A monophrasal compound retains the accent on N2, and N1 will be deaccented if it has an accent. No junctural accent will occur even if neither N1or N2 has an accent. A biphrasal compound retains both N1s accent and N2s accent, and no junctural accent will occur, even if neither N1or N2 has an accent. By retain, I mean to retain its accent

PAGE 37

37 position on N2 or keep unaccented if N2 is originally so. Rendaku can only happen in word and monophrasal compounds, but some word comp ounds may not have Rendaku. So if there is a case in which N1 (unaccented)+N2 (unaccented)=Compound (unaccented), then this compound can either be a monophrasal or biphrasal compound. However, if Rendaku occurs, it is definitely not a biphrasal compound. Table 2 5. The new typology for compounds

PAGE 38

38 I believe that there should be no potential Rendaku position occurring in the right branching under a phrase containing two phrases as in Figure 2 2. Biphrasal structure and Rendaku distribution It can happ en in the branching under dominated by as in Figure 2 3. The structure of a word compound dominated by a phrasal node So within a Japanese name such as yamaguti sakura, Rendaku can happen in guti but not sakura T he insertion rule of [+voice] does not happen between t w o phrase s Ito and Mester (2003, 7199) divide the accentuation patterns of reduplicat ed words into two kinds. Intensive/pluralizing reduplication may have junctural accents. Mimetic reduplication ha s its accent on the first syllable o f the mimetic wh ile dvanda compounds (compounds with two co -ordinated lexical items) retain the accent on N1 (Shibatani 2003). They propose an analysis of Rendaku by treating it as a morpheme R with a feature [+voice] Coordinating compounds and object -verb compounds (OV compounds) with a predicate argument relation will not have Rendaku. Unger (1975) proposes a hypothesis that the origin of Rendaku is a nasal sound. Rendaku requires N2 to be able to add the suffix no or n i OV compounds cannot meet this requirement because the suffix no and n i cannot be added to N2.

PAGE 39

39 Thus, Rendaku does not happen accordingly. The following section includes data for nine basic types; more data can be found in Appendix A. A Review of the Nine Basic Types Type 1 Figure 2 4. Type 1 Left -branching word compounds The junctural accent rule happens at the last mora of N1 if N2 has one or two morae. However, when the original accent on N2 is on the first syllable, two powers are competing with each other: either to retain the accent or to move it to the junctural location. If N2 has three morae or more, then the junctural accent will appear on the first mora of N2. Unlike in the case when N2 has only one or two morae, there will be no conflicting power to shift the accent to wo rd boundary because the default junctural position coincides with the first mora of N2. r (2007). For example, the word suna sand and hokori dust form a compound sunabokori sand and dust. When the third word taisaku treatment is added, the whole compound sunabokori taisaku dust treatment has an accent on the junctural place, na mely on the first mora of Problems are that some examples can be categorized either as a compound or a monophrase such as kikoo climate +hendoo change + samitto summit =kikoohendoo

PAGE 40

40 +samitto= kikoohendoo samitto climate change summit This is b ecause the original accent of samitto is on its first mora, which coincides with the junctural position. iti one +zi time +harai payment =itizi+harai=itizibarai one -time payment It is not kudamono fruit and orenzi orange which is considered to be another type of compound However, for words more than three morae that already hav e an original accent on their first mora they cannot shift to form the junctural accent since the original accent coincides with the junctural position. Data of this type are as follows: (6 ) tabi travel +hito person +ne tto net =tabibito+ ne tto tabibitone tto travelers net (7 ) suna sand +hokori dust +taisaku treatment =sunabokori+taisaku (unaccented)=sunabokori taisaku dust treatment The following can be categorized into type 1 or 4 since N2 in the compound has its accent on the first mora and the compound accent still shows on the first mora of N2, so it is hard to tell if the accent on N2 is retained or a new junctural accent has occurred. (8 ) yo night +sakura cherry blossoms +gi nzi ginzi (name) = yozakura+ ginzi=yo zakura gin zi Y o zakura G inzi (name) (9 ) makizusi hand rolled sushi +waarudo world = makizusi waarudo hand rolled sushi world (10) nyuu new +sama a summer +orenzi orange = nyuusa ma a + orenzi= nyuusama a orenzi new summer orange

PAGE 41

41 Type 2 Figure 2 5. Type 2 Right br anching word compounds the type has been explored extensively in Ito and Mester (2007) Examples given by them are: tanuki tani nobori valley climbing by badgers genkin furikomi cash deposit aka tamanegi red (round) -onion nooto pasokon notebook PC I have found a word that used to be in another type, then changes its accent to be categorized as type 2: muhuu without wind +ti earth +ta i zone = muhuutitai (it used to be muhuutitai). the region of calms Data of this type are as follows: (11) kaigai oversea +ryuugaku study abroad +seido system = kaigai+ryuugakuseido =ka igai ryuugakuseido (oversea) study abroad system (12) nihon Japan +si history +gaku subject =nihon+sigaku=nihon sigaku Japanese History Studies Some data from Tanaka (2001) fall into this type (13) denki electric+kami hair+sori shave=denki+k amisori=denkikamisori electric shaver

PAGE 42

42 (14) soosusauce+yakifried+sobanoodle=soosu+yakisoba=soosuyakisoba chow mein with sauce Type 3 Figure 2 6. Type 3 word compounds This type basically has two possible accentuations: one is to have a junc tural accent, the other is to be unaccented. The compound rule of junctural accent applies here too. Unaccented types must contain a word that loses its original accent. If the words are originally unaccented, they are considered to have retained their acc ent patterns. Unaccente d pattern usually happens on 2+2 and for final as follows: (15) makura pillow +ki wood =makuragi crossties (16) iro color +kami paper =irogami colored paper (17) asa morning +kiri mist =asagiri morning mist (18) hosi star+sora sky=hosizora starry sky One p roblem is that since 4 morae N2 can also have a junctural accent, the mora length does not determine which compound can have junctural accents. If the boundary of word length is abandoned, and instead used to categorize monophrasal and word compounds according to their characteristics, some of the words may fall into two categories. Consider the example so o sa investigation +kaigi meeting =soosakaigi the meeting of investigation; it has a

PAGE 43

43 junctural accent which is a characteristic of a word compound. However, the position of the junctural accent coincide s with the original accent on N2 so that the same compound can be understood as to retain N2 and deaccent N1, which is a typical characteristic of the monophrasal compounds. It can be predicted that Rendaku can happen on N2 if it is a monophrasal or a word compound In later types, it can be seen that Rendaku will not happen on N2 in biphrasal compounds. N2 here refers to the item directly dominated by the t op node. There are some similar difficulties for distinguishing word, monophrasal and biphrasal compounds. If N1 is originally unaccented and N2 has an original accent on the first mora such as hosi star +sora sky =hosizora starry sky this will ma ke the combination look like a biphrasal compound where two accents of N1 and N2 are retain ed However, Rendaku occurs in this case, which suggests that it should be a word or a monophrasal compound. Word compound data are as follows: The juncture type: (19) asa morning +kiri mist =asagiri morning mist (20) hana flower +kotoba languages =hanakotoba the language of flowers (21) aka red +tonbo dragonfly = aka tonbo red dragonfly The following show a junctural accent, but usually compounds with a final -accented N2 will be deaccented: (22) suido suido (name) +hasi bridge = suidoo basi Suido Bridge (place name) (23) tokati tokati (place name) +hasi bridge = tokati basi Tokati Bridge (place name) The unaccented type: (24) mae before+asi leg=maeasi fore -leg (25) mae before+kami hair=maegami forelock

PAGE 44

44 Type 4 Figure 2 7. Type 4 left -branching monophrasal compounds a word may look like biphrasal compounds which retain their accent patterns, but biphrasal compounds do not have Rendaku while monophrasal compounds can. Data are as follows: (26) ameri ka America +san product +orenz i orange =amerikasan+orenzi=amerikasan orenzi American oranges (27) minami south+amerika America+asuparagasu asparagus =minmiamerika+asuparagasu= minmiamerika asuparagasu South American asparagus

PAGE 45

45 Type 5 Figure 2 8. Type 5 right branching monophrasal compounds remain unaccented if 2+ 3 are unaccented The problems are deaccented, the whole can be treated both as a word compound or a monophrasal right -branching compound such as nihon Japan +si history +gaku study =nihon+sigaku=nihon sigaku Japanese history study Ito & Mester (2007) ha ve data which fall into this type: gen z i monogatari Genzi story -telling sinkokuritu gekizoo new national theater daisakusika great lyric person I have also found a word: (28) kaigai abroad +ryuugaku study +seido system =kaigai+ryuugakuseido= kaigai ryuugakuseido (o versea) study abroad system and muhuu (unaccented)+(ti+tai)= muhuutitai region of calms Type 6 1 does not have an accent originally, it may look like a biphrasal compound in which both accent patterns are retained. Rendaku can indicat e how to categor ize the data. Consider the two words monohosi dried

PAGE 46

46 Figure 2 9. Type 6 monophrasal compounds things and kagebosi d rying in the shade ; the accent pattern looks alike, but Rendaku occur s only in the second one which does not have a predicate argument as the first one. The accent in kagebosi can be explained as to retain the accent on hosi dr y and deaccent kage shadow while the accent in monohosi can be said to have retain ed Thus, kagebosi belongs to type 6 while monohosi belongs to type 9 (biphrasal compound). Another problem has been accounted for type 3 when (29) soosa investigation+kaigi conference=soosakaigi i nvestigation meetings (30) ryuugaku (unaccented)+seido=ryuugakuseido study abroad system (31) aka red +ke hair =akage red hair (32) kanzikaki writing characters+zyun order= kanzi kaki zyun the stroke order of characters or be deaccented in 2+2, so the word hosi star +sora sky = hosizora starry sky is a monophrasa l compound. Type 7 preserve their own accents. No accent is shifted and no new accent emerges if they are originally

PAGE 47

47 Figure 2 10. Type 7 left branching biphrasal compounds unaccented. Rendaku can happen in the level of prosodic words but not between phrases. Names are good examples of phrasal compounds. Some family names are made of tw o morphemes that may stand alone and conform to compound formation rules such as yama mountain +kuti mouth +sakura cherry blossom = yamaguti+sakura Sakura Yamaguti (name) Kubozono (1994) has mentioned one type of biphrasal compound s organizatio n name+position. It allows the internal structure such as zitikai kaityo the president of the self governed region in which zitikai self -governed regions is a word compound which is further combined with kaityo president the name of the positio n. Data of this type are as follows: (33) zimu clerical work + kyok u department+syokuin staff=zimukyoku+syokuin= zimukyokusyokuin staff in Secretariat (34) kezai Economics+kyoku department+syokuin staff=kezaikyoku+syokuin= kezaikyokusyokuin staff in Economic Bureau (35) aka red+matu pine + masato masato (name) =akamatu +masato=akamatu masato Masato Akamatu (name)

PAGE 48

48 This word ko o uirusu ma suku anti vi rus mask can be treated as a monophrasal compound referring to a special mask as a whole. It can be also treated as a biphrasal left branching compounds where the accents on ko o anti and masuku mask are retained but uirusu virus is deaccented. Howeve r, the combination of koo u irusu has a junctural accent and when the third word is attached to it, the accent on koo appears again. Similarly, one speaker treated yozakura ginzi ziken The event of Yozakura Ginzi (name) as a biphrasal compound with the a ccent yozakura ginzi ziken, though she deaccented yozakura in the combination of yozakura ginzi However, the accent on yozakura reappears after combining with ziken while ginzi is instead deaccented. Similar examples are as follows: (36) koo anti+ui rusu virus+tiryoo treatment=koouirus+tiryoo (unaccented)=koouirusu tiryoo anti -virus treatment (37) koo anti+uirusu virus+yaku medicine=koo uirusu+yaku=koouirusuyaku anti -virus medicine (38) ko o anti+uirusu virus+zai dose=koouirusu+zai=ko o uirusu zai anti -virus dose Type 8 Figure 2 11. Type 8 right -branching biphrasal compounds In this type, 2 and 3 combine as a word compound where compound rules apply. The accent 2 will be retained as a phrasal accent and so 1 retain s its accent.

PAGE 49

49 Although the name+title is considered to be a biphrasal compound, if the title has an internal structure, the accent will be retained on a higher level such as kawase meiyo kyo o zyu H onored P rofessor K awase If there is no meiyo honored the accent will be kawase kyoozyu Professor Kawase in which kyoozyu remain s unaccented. However, meiyokyo o zyu honored professor is combined as a compound word at the prosodic level before being further combined as a phrasal compound. Another possibility is to treat it as a tri -phrasal compound kawase meiyo kyoozyu Data from this type are as follows: (39) yoru night + yo night +naka middle =yoru+yonaka=yoru yonaka midnight (40) infu ruenza influenze +kansen spreading +kakudai spreading =infu ruenza+ kansenkakudai= infuruenza kansen kakudai the spreading of influenza ( 41) nyuugaku admission +sikan application +h yoo forms = nyuugaku+sikanhyoo= nyuugaku sikanhyoo application forms for universities Type 9 Figure 2 12. Type 9 biphrasal compounds and o new accent should appear and there should be no accent shifts from any position. The problem is that some words may have the same accent pattern as the monophrasal compounds since unaccented+unaccented=unaccented can be explained as to retain the accen t of the last

PAGE 50

50 word and deaccent the first word or combined word such as the pair monohosi and kagebosi explained in type 6. As Kubozono (1994) mentioned, compounds with a predicate argument relation are usually biphrasal compounds. (42) tyuugoku China +hoomon visit =tyuugoku hoomon visits to China (43) kangoku Korea +hoomon visits =kangoku hoomon visits to Korea (44) kita North +tyoosen challenge+ hoomon visits = kitatyoosen+ hoomon=kitatyoosen hoomon v isits to Northern Korea However, with the same s tructure and the same verbal noun hoomon visits the compound katei ho o mon home visits is treated as a word compound referring to home visits as a fixed expression. (katei home +hoomon visits =katei houmon home visits ) Although names usually fall into the biphrasal compound types, some names show the accent pattern of a word compound consistently such as X -taroo. X taroo compounds: There are three kinds of accents in X -taroo compounds Kubozono (2001): 1 Word compounds. When N1 has only one syllable the whole compound will be unaccented. 2 Word compounds. When N1 has two syllables and two morae, the compound will be assigned an accent to the final syllable of N1. 3 Monophrasal compounds. When N1 has three moras or more, N2 will retain the accent. Data o f the biphrasal compounds (type 9) are as follows: (45) kawase kawase (name) +senpai senior colleague =kawase sen pai Senior Colleague

PAGE 51

51 Kawase (46) yoru night + hiru day = yoru hiru night and day (47) mae before +usiro after =maeusiro before and behind Some data from K ubozono (1994) also fall into this type: (48) ippu one husband +tasai many wives=ippu tasai polygyny (49) hakusyu clapping +kassai cheering =hakusyukassai clapping and cheering (50) kaisya company +syatyoo president =kaisyasyatyoo preside nt of the company (51) nihon Japan +zengoku the whole nation =nihonzengoku the whole nation of Japan (52) syoosoku correspondence +humei lost = syoosokuhumei lost in contact (53) isiki conscious +humei lost =isikihumei losing ones consciousness Comple x Structures Some compounds have complex structures based on the nine basic structures analyzed above. Triphrasal Structure: Right -branching compounds: (54) doosoomembers of the alumni+kai organization +koozin private +zyoohoo information +hogo protec tion +hoosin policy =doosoukai+ kozinzyouhoo+hogohoosin= doosookai+ kozinzyoohoohogohoosin= doosookaikozinzyoohoohogohoosin protection policy of private information in the alumni Figure 2 13. The c omplex structure for doosoukaikozinzyoohoohogo hoosin

PAGE 52

52 (55) minami south+oosawa oosawa (name)+kyanpasu (56) sangaku industry and education +kooryuu communication+kai organization = sangaku+kooryuukai= sangaku kooryuukai Industry Figure 2 14. The complex structure for the word sangaku kooryuukai Left branching compounds: (57) hookei law and economics +gakubu department +keizai economics +gakka subject+sotu graduation =hookeiga kubu+keizaigakka+sotu= hookeigakubukeizaigakka+sotu=hookeigakubukeizaigakkasotu e conomics major from the department of law and economics Figure 2 15. The complex structure for the word hookeigakubukeizaigakkasotu (58) syuto capital +daigaku university +tookyoo Tokyo +soogoo comprehensive +kyoogi sports +taikai meeting = syuto+ daigaku+tookyoo+soogookyoogi+taikai=

PAGE 53

53 syutodaigakutookyoo+soogookyoogitaikai= syutodaigakutookyoo soogookyoogitaikai sports meeting in Tokyo Metropolitan Uni versity Figure 2 16. The complex structure for syutodaigakutookyoo soogookyoogi taikai (59) tyoo super +kantan simple +setuyaku saving +resipi recipe =tyoo kantan+setuyakuresipi= tyoo kantansetuyakuresipi very simple time -saving recipe F igure 2 17. The complex structure for the word tyoo kantansetuyakuresipi (60) kita North+kariforunia California+asuparagasu asparagus=kitakariforunia + asuparagasu= kitakariforunia asuparagasu North California asparagus (61) kita North+kariforunia California+ orenzi orange=kitakariforunia + orenzi = kitakariforunia orenzi North California orange Figure 2 18. The complex structure for kitakariforunia asuparagasu / orenzi

PAGE 54

54 Biphrasal Structure (62) keizai economics +kikaku plan +tyoo agency +nyutyoo joining =keizaikikaku+tyoo+nyuutyoo (unaccented)= keizaikikakutyoo+nyuutyoo= keizaikikakutyoo nyuutyoo joining the planning agency Figure 2 19. The complex structure for the word keizaikikakutyoo nyuutyoo (63) kaisya company+setumei ex planation+kai meeting+ setuei=kaisya setumeikai setuei setting up the venue for company presentations Figure 2 20. The complex structure for the word kaisya setumeikai setuei (64) oosaka oosaka (name)+ bankoku world+hakuran exposition+kai meeting=oosaka+ bankoku +kakurankai=oosaka+bankoku hakurankai= oosakabankoku hakurankai

PAGE 55

55 Figure 2 2 1. The complex structure for the word oosakabankoku hakurankai (65) nihon Japan +seihu government +daihyoo representitives+bu department =nihon seihu daihyoobu Delegation of Japanese government Figure 2 22. The complex structure for the word nihon seihu daihyoobu (66) oosaka oosaka (name)+ bankoku world+hakuran exposition+kai meeting=oosaka+ bankoku +kakurankai=oosaka+bankoku hakurankai= oosakabankoku hakurankai his one has a different structure from the same word in the Figure 2 21) Figure 2 2 3. The complex structure for the word oosakabankoku hakurankai

PAGE 56

56 (67) minami south+ oosawa oosawa (name)+ hakubutu natural history+kan building=minamioosawa+ hak ubutukan= minamioosawa hakubutukan South sawa museum Figure 2 24. The complex structure for the word minamioosawa hakubutukan (68) syuusyoku employment+zi time+mensetu interview+sidoo instruction= syuusyokuzimensetusidoo= syuusyokuzi mensetusidoo the instruction of employment interviews Figure 2 2 5. The complex structure for the word syuusyokuzi mensetusidoo (69) nihon Japan+ sigaku history studies+kooza lectures=nihon sigaku+ kooza= nihon sigaku kooza lectures of Japanese history studies

PAGE 57

57 Figure 2 26. The complex structure for the word nihon sigaku kooza (70) minami south+oosawa oosawa (name)+kyanpasu campus+ sangaku industry and education+kooryu communication+kai meeting =minamioosawa+kyanpasu+ sangaku+kooryuukai =minamioosawa kyanpasu s angaku kooryuukai Industry and education forum of South sawa campus Figure 2 2 7. The complex structure for minamioosawa kyanpasusangakukooryuu kai Complex Word (Monophrasal) Compound (71) y o night +sakura cherry blossoms +ginzi ginzi (name) +zi ken event = yozakura + ginzi+ ziken =y o zakura gin zi+ ziken = y o zakura ginzi ziken Yozakura Ginzi (name) event The word yozakura ginzi has a word structure but it is a name

PAGE 58

58 Figure 2 28. The complex structure for the word y o zakura ginzi zi ken (72) doku poison +iri added +orenzi orange +ziken incident = dokuiri+orenzi+ziken=dokuiriorenziziken poisonous orange incident Figure 2 29. The complex structure of the word dokuiriorenziziken (73) nihon Japan +si history +gaku subject+kooza lectures= nihonsi+gaku+kooza= nihonsigaku+kooza= nihon sigaku kooza lectures of Japanese history studies Figure 2 30. The complex structure of the word nihon sigaku kooza This chapter introduced a new typology and further analyzes the nine basic structures. The internal structure of more complex compounds may also be constructed based on those basic

PAGE 59

59 structures predicting the location of accent in the complex structures based on the generalization found in the nine basic types. In the fol lowing chapter, I analyze the data using the OT approach and propose constraints for compounds of word and phrasal levels. In the last part of the next chapter, I also explore the internal structures of prefixes.

PAGE 60

60 CHAPTER 3 ANALYSIS IN AN OT AP PROACH An O T Approach in Analyzing the Nine Types Chapter 3 analyzes all the data collected under the optimality theory framework. Constraints proposed can account for the internal structures of compounds in two levels and for the deaccentuation phenomenon. At the end of this chapter, prefixes are also analyzed in those two levels. In this analysis, I will use constraints proposed by Kubozono (1995) and Tanaka (2001): NON, NON -FINALITY (F), MAX (accent), ALIGN ALIGN nstraint NON and NON -FINALITY (F) tend to avoid final accents on N2. The constraint MAX (accent) requires the accent on the head of a compound to be retained. The alignment constraint ALIGN accented syllable align with the root to the left while ALIGN syllable to align with the prosodic word to the right. I have restricted the constraints of NON -, NON FINALITY (F), MAX (accent) only to word compounds and rewr ote them as NON The definitions for some new constraints used are as follows: 1 2 esponding to the head in a phrasal compound 3 and forbid more than one accent 4 5 6 IDENT Oires the accent to stay in N2 (The second prosodic word of the first branch) without movement.

PAGE 61

61 This IDENT constraint above is devoted only to the retention of the accent position. If an I In type 1 compounds, since the second candidate in Table 3 1 {{sunabokori} { (tai)(saku) } should lose to the winner, ALIGN must rank higher than ALIGN Similarly, since the fourth candidate { (ta i)(saku) } ALIGN The third candidate constraints NON FINALITY (F). Table 3 1. Type 1 the constraint ranking for the word sunabokori taisaku NONNON(F) ALIGN-L ALIGN-R {{sunabokori} { (tai)(saku)} *** {{ sunabokori} { (tai)(saku) } *! ** {{sunabokori} { (tai)(saku) } * {{sunabokori} { (tai)(saku) } *! In Table 3 1, since the constraint ALIGN with the accented syllable, al l candidates except for the winner violate this constraint. The last prosodic word from the right. So violations are counted from the right in the unit of a syllable.

PAGE 62

62 In the Table 3 2, since the second candidate { (gaku) } loses to the winner, NON and NON should rank higher than ALIGN L So it can be concluded that NON NON ->> ALIGN ALIGN Table 3 2. Type 1 the constraint ranking for the word nihonsigaku NONNON(F) ALIGN-L ALIGN-R { (gaku) } ** { (gaku) } { (gaku) } { (gaku) } *! ***** In Table 3 3, NON FINAL NON (F) should rank higher { { soosu } loses. A compound such as hosi (unaccented)+sora=hosizora may look similar to a word compound but can only be a monophrase because if it is a prosodic word, NONwill move the accent on sora Moreover, the occurrence of Rendaku precludes the structure from being biphrasal. In Table 3 4, the constraint ALIGN and las t candidates.

PAGE 63

63 Table 3 3. Type 2 the constraint ranking for the word soosuyakisoba / { { {yaki } { NONNON) t) ALIGN-L ALIGN-R { { soosu } *** { { soosu } * { { soosu } *! ** { { soosu } *! ******* { { soosu } ki)(s *! ***,****** Table 3 4. Type 3 the constraint ranking for the word kaigai ryuuga ku NONNON(F) ALIGN-L ALIGN-R { kaigai(ryuu)(gaku) ** { * *** {kaigai(ryuu)(gaku ) *! *****

PAGE 64

64 In Table 3 T able 3 5. Type 3 the constraint ranking for the word inakama NONFINALIT NONFINALIT (F) ACCENT (accent) AL IGN -L ALIGN-R { { ** { { * *** *! **** In Table 3 constraints NON and N ON with a final accent. Table 3 6. Type 3 the constraint ranking for the word asagiri /{{asa} +{kiri} } / NONNON(F) ALIGN-L ALIGN-R {asa(gi ri)} ** {asa(giri)} *! {asa(giri)} *! {asa(giri)} *! Even for N2 with final accent, the whole compound will be deaccented only when N2 contains fewer than 2 morae. For N2 with three morae or more, a junctural accent will appear. In Table 3 7, a new constraint IDENT O because candidates in other tableaus do not violate this constraint. To ensure the second candidate {kinu(ito)} loses, IDENT O

PAGE 65

65 NON NON(F)>> IDENT O ALIGN ALIGN The second candidate {kinu(ito)} in Table 3 7 violate s IDENT O (accent) because the accent of N2 moves out to the N1 -N2 boundary. Table 3 7. Type 3 the constraint ranking for the word kinuito {{ kinu ito } } NON-FINALITY NON-FINALITY IDENT O ) ALIGN-L ALIGN-R {kinu(ito)} **** {kinu(ito)} *! ** {kinu(ito)} *! Table 3 8. Type 3 the constraint ranking for the word garasudama /{{ garasu tama } NONNON(F) IDENT O I ALIGN-L ALIGN-R { garasu(dama) ***** { garasu(dama) *! ** { garasu(dama) *! { garasu(dama) *!

PAGE 66

66 Table 3 9. Type 3 the constraint ranking for the word se ttyuutaue /{{ settyuu taue NONNON(F) IDENT O ALIGN-L ALIGN-R { settyuu(ta)(ue)) ** { settyuu(ta)(ue) *! ***** { settyuu(ta)(ue) *!* *** { settyuu(ta)(ue) *! Table 3 10. Type 4 the constraint ranking for the word amerikasan orenzi In the tableaus for type 4, 5, 6, 7, 8 and 9, two constraints and tableaus do not violate this constraint. has to rank higher than NON +{ orenz i} M AXIO NONNON(F) ALIG N -L ALIGN-R { (oren)(zi)} { (ore n)(zi) } ** { (ore n)(zi) } * *** { (oren )(zi)}

PAGE 67

67 because the sec ond, third and fourth candidate have to lose to the winner. The candidate { (oren)(zi) } in Table 3 10 does not have an accent on the 10 { (oren)(zi)} violates MAX changed. Examples of N2 with other accent patterns and different internal structures are given from Table 3 11 to Table 3 16. In Table 3 11, the second candidate loses because it has a final accented syllable. The third candidate does not have a corresponding accent on the head of the phrasal compound and violates ACCENT The forth candidate has two accents in one phrasal compound and should not win. Table 3 11. Type 5 the constra int ranking for the word genzi monogatari / {genzi + {{mono} +{g atari} } / M AXIO NONNON(F) ALIGN-L ALIGN-R { { genzi } {mon o(ga)(tari)} } ** { { g enzi } {mono(ga)( tari)} } * { { genzi } {mono(ga) (tari)} } *! **** { { genzi } {mono(ga )(tari)} } **,*** ***

PAGE 68

68 Table 3 12. Type 6 the constraint ranking for the word kita kariforunia / {{kita} +{kariforunia} } / M AXIO NONNON(F) IDENT O ent) ALIGN-L ALIGN-R {kita(kari)(foru)(ni a) } ** ***** *** {kita(kari)(foru)(nia) } {kita(kari)(foru)(nia) } * {kita(kari)(foru)(nia) } * ***** Table 3 13. Type 6 the constraint ranking for the word singata infuruenza / {{sin gata} +{infuruenza} } / M AXIO NONNON(F) IDENT O ) ALIGN-L ALIGN-R {sin gata (in)(furu)(enza)} * {sin gata (in)(furu)(enza)} **** {sin gata (in)(furu)(enza)} * ***** {sin gata (in)(furu)(enza)} ***** ***

PAGE 69

69 Table 3 14. Type 7 the constraint ranking for the word zimukyoku syokuin + M AXIO NONNON-FINALITY (F) IDENT O ALIGN-L ALIGN-R {{{zimu(k ** ** ,* {{{zimu(kyoku) * {{{zimu(kyoku) * ,* {{{zimu(kyoku) ** ** ,* In Tabl es 3 12 and 3 In Table 3 14, the second and fourth candidates do not have a corresponding accent to the original accent on the head of either the first or second phrase. The third candida te has a final accented syllable and should lose to the winner.

PAGE 70

70 T able 3 15. Type 8 the constraint ranking for the word infuruenza kansenkaku / {{ infuruenza } + {{kanse n kakudai } / M AXIO NON-FINALITY NON-FINALITY IDENT O ALIGN-L ALIGN-R {{infurue nza ka nse n( ka ku)(dai *,*** {{ infuruenza } kanse n( ka ku)(dai *! *** {{ infuruenza } kanse n( kaku)(da i *! *,* {{ infuruenza } kanse n( kaku)(dai *! **** ** {{ infuruenza } kanse n( kaku)(dai *! **** In Table 3 15, the fourth candidate does not have an accent on kansenkakudai and violates

PAGE 71

71 Table 3 16. Type 9 the constraint ranking for the word kangoku hoomon {{ kangoku + {{hoomon M AXIO NONFINALITY NONFINALITY ent) ALIGN-L ALIGN-R {{kangoku + {{( hoo)( mon ** {{ kangoku + {{( hou )( mon *! **,* {{ kangoku + {{( hoo)( mon *! ***,** In the OT secti on above, I proposed five new constraints to deal with the internal structures of compounds. Those constraints are also divided into the word and phrasal level which ret word level. IDENT O needs no rerankable constraints as in Tanaka (2001) Prefixes For prefixes, if they are in a biphrasal construction, they will retain their accent. I f they appear in a word structure, it means that the phrasal structure of the prefix is somewhat restricted from being active in a whole word compound, indicated by I explore the follo wing examples with the prefix koo meaning high :

PAGE 72

72 (74) koo bunsi high polymer (75) kooenerugi high energy (76) koo byoogensei highly pathogenic (77) koo taikyuusei high perduable Each represents the word, monophrasal or biphrasal compound. Figure 3 1. The structure of the word koo bunsi Figure 3 2. The structure of the word koo enerugi Figure 3 3. The structure of the word koo byoogen sei Other prefix es such as tyoo and hi can show either word or biphrasal compo u nd struc ture. Left branching tri -phrasal compound ing is also possible such as:

PAGE 73

73 (78) tyoo super+ eiga film + hihan criticism =tyooeiga+hihan=tyooeigahihan super film criticism (79) tyoo super+kookyuu luxurious +onigiri rice ball = tyookookyuu +onigiri=t yookookyuu onigiri super luxurious rice ball (80) tyoo super+kookyuu luxurious +hurenti French= tyookookyuu (unaccented)+hurenti= tyookookyuuhurenti super luxurious French dishes (81) tyoo super+koosoo highrise +biru buildings = tyookoosoo+bi ru=tyookoosoobiru super highrise buildings (82) tyoo super+kandan easy + menyuu menu = tyookandan+ menyuu= tyookandan menyuu super easy menu (83) tyoo super+manin crowded +de nsya train= tyoomanin+de nsya= tyoomanindensya super crowded tr ains (84) koo anti+uirusu virus+masuku mask= koouirusu+masuku= koouirusumasuku/koo uirusu masuku anti -virus mask It is interesting that some combination s of a prefix and a root can restrict the biphrasal structure of the prefix in a word compound such as koouirusu. When this word compound is put under a phrasal compound koouirusu yaku, the restricted biphrasal structure of the prefix gains its biphrasal status again. If in a phrasal compound, N1 has a phrasal structure while N2 only has a word structure, phrasal head to retain its accent in a phrasal compound. So the word compound koou irusu has the structure:

PAGE 74

74 Figure 3 4. The structure of the word koo uirusu which dominates this current structure and the koo retians its original accent and uirusu is deaccented: Figure 3 5. The structure of the word koo uirusu yaku Other structures such as t he following are not possible: Figure 3 6. The wrong structure of the word kou uirusu yaku

PAGE 75

75 Figure 3 7. The wrong structure of the word koo uirusu yaku The accent will be retained on uirusu for the two s tructures in Figure 3 6 and 3 7. As described above, prefixes can also have different internal structures at different levels. Constraints from word and phrasal levels are more suitable for analyzing internal structures of different levels. In the followin g chapter, two constraints are proposed to deal with the deaccentuation phenomenon and two constraints are to account for the difference between left branching and right -branching compounds. The advantages and problems of my analysis are summarized as well

PAGE 76

76 CHAPTER 4 CONCLUSION AND DISCU SSION In this chapter, the advantages of the new typology and that of the new tableaux are discussed. The chapter also explores an OT analysis for the 2+2 phenomenon and explains the difference between left -branching an d right -branching compounds using an OT approach. Finally, I explain the limitations of my analysis and make suggestions on further research. The Advantages of the New Analysis The new typology is based on the redefinition of word, monophrasal and biphrasa l compounds. This new definition helps to categorize different levels of compounds by comparing the input and the output of N2 rather than only focusing on the occurrence of the junctural accent. The new definition helps compare and constrast the character istics of word, monophrasal and biphrasal compounds. Since junctural accent cannot be used to define biphrasal compounds, the new definition is especially helpful in including biphrasal compounds in the new typology. This way of categorization distinguishe s each level well by considering the characteristics of the input and output. Each level of compounds may have the internal structure of left -branching, right branching or flat structure (N1+N2). The phenomenon of Rendaku can be also predicted in correspon dence to different structures. Based on the nine basic structures, more complex structures can be analyzed successfully. The prefix, which has not been well analyzed in previous research, can be analyzed using the new typology as well. Since compounds can be divided into different levels, constraints of the corresponding levels are needed as well such as The 2+2 Phenomenon in an OT Approach As for the deaccentuation problem, IDENT O accent tends to be deaccented.Another phenomenon related to the deaccent uation is the 2+2

PAGE 77

77 phenomenon, metioned in Chapter 1; however, not all 2+2 structure will be deaccented, only balanced structures can be deaccented. The following data are from Tanaka (2001): (85) man full+getu moon=mangetu full moon (86) tyoo punishme nt+batu infliction= tyoobatu punishment infliction These structures are weaker than the 2+2 structure, because the nasal and one part of a different from the structure of unaccented kuro neko which has no nasals or long vowels with a (87) on sound+in rhythm=onin phonology (88) usu light+azi taste=usuazi light taste (89) roo labor+doo work=roodoo manual labor (90) kan see+koo light=kankoo sightse eing For the data above, onin phonology, roodoo manual laborand kankoo sightseeing all usuazi balanced 2+2 structures do not need an accent such as Figure 4 1. The balanced structures In the cases where N2 does not have a final accent, if two morphemes in a compound word (Constraint A). On the contrary, if two m orphemes fail to have the same structure, then an accent has to occur (Constraint B). These are descriptive constraints whose motivation I leave for further research.

PAGE 78

78 Left-branching and Right -branching Differences in an OT Approach In the following sectio n, I examine the reason that left -branching compounds tend to be a whole word compound. Truckenbrodt (1999) has investigated the relation between phonological phrases and syntactic XP. He proposes the constraint WRAP -XP which requires that every XP has to be contained in a phonological phrase. Selkirk (1995) modified the ALIGN constraint proposed by McCarthy and Prince (1993) and created the constraint ALIGN -XP,R which was defined as For each XP there is a P such that the left edge of XP coincides with the left edge of P. In the case of Japanese, the constraint for biphrasal compounds requires that the left edge of every accented prosodic word be aligned with an XP.The constraint ALIGN NP,L can be used to analyze Japanese compounds. Left branching: [(N1+N2 )NP2+N3]NP1 ( )p Right -branching [N1+(N2+N3)NP2]NP1 ( )p( )P In left branching compounds, there is no conflict between ALIGN NP,L and WRAP NP. In right -branching compounds, the left edge is required by ALIGN -NP, L which is i n conflict with WRAP NP since the whole NP1 is not included in a prosodic word.So left -branching compounds have the tendency to be treated as a whole prosodic word word compound. For Right branching compounds, if WRAP NP>>ALIGN NP,L, then the compound is t reated as a whole prosodic word ---word compound. If ALIGN NP,L, >>WRAP NP, then it is

PAGE 79

79 treated as a phrasal compound. This analysis in the OT approach can provide an explanation for the difference of left -branching and right -branching compounds. Conclusions Problems and Further Research My analysis of the new typology has some limitations which remain to be solved in the future. According to the definition, it is difficult to distinguish a word compound from a monophrasal one when a long N2 has its accent o n the first mora, which coincides with the default junctural accent position. The previous research by Ito and Mester (2007) relies only on the junctural accent to determine if it is a word or monophrasal compound and by my definition, all monophrasal com pounds which have an original accent on the first syllable of N2 will be word compounds in Ito and Mesters standard. There is also one exception that Rendaku does not show up in the predicted structure, which is the word sakanaturi fishing mentioned in chapter two. This cannot be explained either by my analysis or previous analyses. In some compounds, a word level can dominate the monophrasal level such as in: (91) eda branch +ke hair +taisaku treatment=edage+taisaku= edagetaisaku split hair tr eatment (92) ondan warming +ka change+boosi prevention =ondanka+boosi=ondankaboosi (global) warming prevention They have the following structure: Figure 4 2 The structure of the word ondankaboosi

PAGE 80

80 So a monophrasal compound is made first and then combined with another prosodic word to form a word compound which dominates the monophrasal compound. It may be argued that the word and phrasal level should have some order since the phrasal level dominates the word level in most cases. In the OT a pproach section, the input is already labeled with phrasal or word structures. The constraints that determine the inputs compound structure need to be found, which exist not only in the phonological level but also in the semantic and pragmatic levels as m entioned by Ito & Mester (2007). The constraint IDENT O (accent) explains the deaccentuation phenomenon, but according to the data in Appendix B, final accented N2 may not always trigger deaccentuation. IDENT O (accent) alone cannot explain this. Mor eover, the constraint A and constraint B mentioned in this chapter need further investigations in order to reflect deeper reasons for the requirement of balanced structures. The competition of the power between syllables and feet may be one of the reasons. This thesis has succeeded in accounting for a variety of compounds using a new typology and a set of OT constraints. It also provides further data for research to resolve the remaining questions about the location of accents in compound structures.

PAGE 81

81 APPE NDIX A DATA OF NINE BASIC T YPES Type 1 (93) hosi star +sora sky +nyuumon introduction = hosizora+nyuumon=hosizora nyuumon I ntroduction to Astro observation (94) nihon Japan +si history +gaku subject = nihonsi+gaku= nihonsigaku The subject of J apanese history (95) sin gata new type +infuruenza influenza +zyoohoo information = sin gata infuruenza+zyoohoo=sin gata infuruenzazyouhoo New type influenza information The following can be categorized into type 1 or 4: (96) siro white + asuparagasu asparagus+ryouri cooking =siroasuparagasu+ryouri white asparagus dish (97) minami south +oosawa oosawa (place name) +kyanpasu campus =minamioosawa +kyanpasu=minamioosawa kyanpasu South Oosawa (place name) campus (98) minami south +oosawa oosawa ( place name) +meibutu special local product =minamioosawa+ meibutu=minamioosawa meibutu South Oosawa (place name) special local product (99) minami south+amerika America+biiru beer =minmiamerika+ biiru = minamimerika biiru South American beer (100) nyuusen breast +gan cancer + senta center =nyuusengan+ senta=nyuusengan senta breast cancer center (101) hoppoo northern +kankei relationship +siryoo resources=hoppoo kankei siryoo northern studies collections

PAGE 82

82 Type 3 The junctural accent type (102) inaka country +ma space =inakama tatami size(176cm*88cm) or a unit of measure, 1.8m (103) aka red +satoo sugar =akazatoo brown sugar (104) ao green +kaeru frog =aogaeru green frog (105) hanabi fireworks+taikai festival= hanabitaikai fireworks festi val (106) ha nabi fireworks+ kansyo appreciation = hanabikansyo fireworks appreciation (107) hanabi firework+tokusyu specials= hanabi tokusyu firework specials (108) ao blue +tatami tatami mat =aodatami new tatami mat (109) hude pen +tukai usage =hudedukai brush work (110) kuti mouth +kuruma car =kutiguruma cajolery (111) minami south+amerika America=minamiamerika South America (112) kikoo climate+hendoo change= kikoohendoo climate change (113) ti earth+tai zone=titai area (114) kansen inflection+kakudai spread= kansen kakudai spread of infection (115) kaigai overseas+ryuugaku study abroad=kaigai ryuugaku study abroad (116) nyuusen breast+gan cancer=nyuusengan breast cancer (117) suna sand+hokori dust=sunabokori dust (118) hoppoo northern+kankei rel ation=hoppoo kankei northern relation (119) mu without+nooyaku pesticides=munouyaku no pesticides (120) hosi star+sora sky=hosizora starry sky

PAGE 83

83 (121) soosa investigation+kaigi conference=soosakaigi investigation meetings (122) ryuugaku studying abroad+seido system=ryuugakuseido the system of studying abroad The unaccented pattern (123) tabi travel+hito person=tabibito traveler (124) nihon Japan+hasi bridge= nihon basi Bridge of Japan (place name) (125) musi worm+ha tooth=musiba decayed tee th (126) aka red+me eye=akame red eye (127) kuti mouth+kuse habit=kutiguse pet phrase (128) hosi star+sora sky=hosizora starry sky (129) nihon Japan+si history=nihon si Japanese history Type 4 (130) minami South+amerika America+ orenzi orange =minm iamerika+ orenzi = minmiamerika orenzi South American orange (131) minami South+oosutoraria Australia+kangaruu kangaroo= minamioosutoraria+kangaruu =minami oosutoraria kangaruu South Australia Kangaroo (132) hoppo o north +kankei relationship +si ryo o resources= hoppoo ka nkei + siryoo = hoppoo kankei siryoo Northern studies collection (133) mu without +nooyaku pesticide +kudamono fruit = munouyaku+kudamono=munooyaku kudamono organic fruit Type 6 (134) eda branch +ke hair =edage split hair

PAGE 84

84 (135) ak a red +mi meat =akami unaccented lean meat (136) kita North+kariforunia California=kitakariforunia North California (137) sin gata new type +infuruenza influenza = sin gata infuruenza New type influenza (138) ti earth+tai zone =titai area (139) saga saga (name) +ten n o o emperor = sagatenno o Emperor Saga (140) dokuiri poisonous +orenzi orange =dokuiriorenzi poisonous oranges Type 7 (141) sunasand+hokori dust+boosi prevention=sunabokori+boosi (unaccented)= sunabokori boosi prevention of dust (142) yama moutain+kuti mouth+senpai senior colleague= yamaguti+senpai senior colleague Yamaguti (name) (143) syuto capital+daigaku university+tookyoo Tokyo=syutodaigaku tookyoo Tokyo Metropolitan University Type 9 (144) bon bon festival +mukae welcome +b i fire =bon+mukaebi=bon mukaebi f ireworks for Bon festival (145) kawase kawase (name)+sakura cherry blossoms (name)=kawase sakura Sakura Kawase (name) (146) aka red+tonbo dragonfly=akatonbo (now: aka tonbo) red dragonfly (147) saga saga (name)+ ten n o o emperor=it was sagatennoo now sagatenno o Emperor Saga

PAGE 85

85 APPENDIX B THE TABLE FOR ALL AC CENT PATTERNS Possible accent patterns for N2 less than or equal to two morae Short N2 Length<=2 morae N2 Accented on the first syllable Retain the or iginal accent of N2 on first syllable or shift to the final syllable of N1 (junctural accent) densyo carrier+hato pigeon= densyobato or desyobato carrier pigeon koomori Western+kasa umbrella= koomorigasa Western umbrella From Tanaka (200 1) Only shift the original accent of N2 to the final syllable of N1 ningyo mermaid+hime princess=nigyohime mermaid princess kansou dry+hada skin=kansouhada dry skin kansoo dry+negi leek= kansoonegi dry leek From Tanaka (2001) Only retain the original accent of N2 kafe cafe+ baa bar=kafebaa caf bar eiga movie+fan fan=eigafan movie fan besuto best+ten ten=besutoten best ten From Tanaka (2001) N2 Unaccented A new junctural accent may appear Sasa bambo o leaf+amecandy=sasaame candy wrapped in bamboo leaves asa morning+kiri mist=asagiri morning mist inaka country +ma space=inakama tatami size(176cm*88cm) or a unit of measure, 1.8m The lack of accent may remain eda branch+ke hai r=edage split hair aka red+mi meat=akami unaccented lean meat kanzikaki+zyun (unaccented)=kanzi kaki zyun N2 Final accented The accent may shift to the final syllable of N1. Suido waterworks+hasi bridge=suidoo basi Suido Bridge (place n ame) tokati tokati (place name)+hasi bridge= tokatibasi Tokati Bridge (place name) abura oil+kami paper= aburagami oil absorbent paper The whole compound may be deaccented akane madder+iro color=akaneiro madder red

PAGE 86

86 Possible accent patterns for N2 equal to three or four morae Short N2 Length=3 or 4 morae N2 Accented on the first syllable Retain accent on the first syllable coincides with the junctural accent soosa investigation+kaigi conference=soosakaigi investigation meetings yama mountain+kuti mouth+sakura cherry blossom= yamaguti+sakura Sakura Yamaguti (name) minami south+oosawa oosawa (place name)+kyanpasu campus =minamioosawa +kyanpasu=minamioosawa kyanpasu South sawa (place nam e) campus N2 Unaccented New junctural accent may appear sin New+yokohama Yokohama (place name)= Sinyokohama New Yokohama kuti mouth+yakusoku promise=kutiyakusoku verbal promise From Tanaka (2001) aka red+tonbo dragonfly= akatonbo(before) =aka t onbo (now) red dragonfly ao green+kaeru frog=aogaeru green frog hanabi fireworks+taikaifestival= hanabitaikai fireworks festival hanabi fireworks+kansyo appreciation =hanabikansyo fireworks appreciation hanabi firework+tokusyu specials= hanabi tokusyu firework specials The whole compound may remain unaccented hukusuu plural+danraku paragraph=hukusuu danraku more than one paragraph orizinaru original+tezome hand dyeing=orizinarutezome original hand dyeing ko okyuu luxurious+tyanomi tea drinking= kookyuu t yanomi luxurious tea drinking N2 Final accented Accent may shift to junctural position dezaato dessert +azuki red azuki bean=dezaatoazuki dessert made of red azuki bean syuumatu weekend+ danzi ki fast = syuumatu dan ziki fast in weekends hana flower+ t u bomi bud = hanatubomi the bud of flowers hueru expand+wakame seaweed= hueruwakame seaweed that can expand The whole compound may be deaccented hasan bankruptcy +tuutisyo n otification= hasan tuutisyo the notification of bankruptcy naitei unofficial decision+ tuutisyo notification= naitei tuutisyo the notification of an unofficial decision N2 Accented in the middle The accent may be retained murasaki purple sag a saga (name)+tennoo emperor=sagatennou Emperor Saga

PAGE 87

87 mu without +nooyaku pesticide +kudamono fruit = munouyaku+kudamono=munooyaku kudamono organic fruit The accent may be shifted to the first mora of N2 genkin cash+hurikomi deposit= genkin hurikomi cash deposit onna woman+ kokoro heart=onnagokoro womans heart yude boiled+ tamago egg= yudetamago boiled egg kami paper + omutu diaper= kamiomutu paper diaper From Tanaka (2001) Possible accent patterns for N 2 more than five morae Long N2 Length>=5 morae N2 Accented on the first syllable Retain accent on the first syllable coincides with the junctural accent sokuseki on the spot +daietto diet= sokuseki daietto diet that can be effective immediately ka ntan simple+daietto diet= kantan daietto simple diet sukyuuba scuba+daibingu diving=sukyuuba daibingu scuba diving booto boat+daibingu diving= bootodaibingu boat diving N2 Unaccented New junctural accent may appear kyooka stren gthen+taimingu timing=kyooka taimingu fine tune timing tenka sparks+taimingu timing=tenka t aimingu the timing of sparks The lack of an accent may remain kooon high temperature+taikyuusei durability= kooon taikyuusei durability of high temperature zyuunan softness+taikyuusei durability=zyuunan taikyuusei durability of softness kyuudo archery+aikooka lover= kyuudo aikooka a lover of archery N2 Accented in the middle The accent may be retained sya nai in a company+ danke turyoku power of unity= syanai danketuryoku the power of unity in a company kaisansei tansaiboo = kaisansei tansaiboo tyoobun long article+dokkairyoku comprehensibility=tyoobun dokkairyoku the comprehensibility of long articles zintai human body+ka i boozu atlas of anatomy= zintaika i boozu the atlas of human anatomy

PAGE 88

88 REFERENCES Bennett, Diana Mary ( 1981). Pitch accent in Japanese: a metrical analysis University of Texas at Austin Thesis Brinton, Laurel & Traugott, Elizabeth (2005) Lexicalization and Language change. Cambridge: Cambridge University Press. Giegerich, Heinz (2004). Compound or phrase? English noun-plus noun constructions and the stress criterion. English Language and Linguistics Cambridge: Cambridge University Pres s. 8.1 .1 24. Haraguchi, Shosuke (1977). The Tone Pattern of Japanese: An Autosegmental Theory of Tonology Tokyo: Kaitakusha. Haraguchi, Shosuke ( 1991). A Theory of Stress and Accent.Dordrecht -Holland Providence RI U.S.A. : Foris Publications Higurash i, Yoshiko ( 1983). The Accent of Extended Word Structures in Tokyo Standard Japanese. Tokyo, EDUCA Inc. Ito, Junko ( 1990). Prosodic minimality in Japanese. Chicago Linguistics Society 26-II: Papers from the Parasession on the Syllable in Phonetics and Phonology 213 239. Ito, Junko & Armin Mester (1997). Correspondence and compositionality: The ga -gyo variation in Japanese phonology. In Derivations and Constraints in Phonology ed. Iggy Roca (ed.), 41962. Derivations and Constraints in Phonology Oxford: Oxford University Press. Ito, Junko & Armin Mester (2003). Japanese Morphophonemics Markedness and Word Structure Cambridge, Massachusetts : The MIT Press. Ito, Junko & Armin Mester (2007). Prosodic Adjunction in Japanese Compounds. In Yoichi Miyamoto a nd Masao Ochi (ed s .) Formal Approaches to Japanese Linguistics 4, MIT Working Papers in Linguistics 55. 97 112. Iwasaki, Shoichi ( 2002). Japanese Amsterdam/ Philadelphia : John Benjamins Publishing Company Kindaichi, Kyosuke & Kindaichi, Haruhiko (2001). Meikai kogo jiten [Explanatory dictionary of Japanese archaisms].Tokyo: Sanseido. Kubozono, Haruo (1988). The organization of Japanese Prosody. Doctoral Dissertation, University of Edinburgh. Tokyo: Kurosio Publishers. Kubozono, Haruo ( 1994). Word For mation and Phonological Structure (Gokeisei to Onin Kouzo o ). Tokyo: Kurosio Publisher s.

PAGE 89

89 Kubozono, Haruo ( 1995). Constraint interaction in Japanese phonology: Evidence from Compound Accent. In Rachel Walker, Ove Lorentz, and Haruo Kubozono ( eds. ) Phonology at Santa Cruz 4 21 38. Kubozono, Haruo ( 2001). Epenthetic Vowels and Accent in Japanese: Facts and Paradoxes. In Jeroen Van Deweijer and Tetsuo Nishihara ( ed s ) Issues in Japanese Phonology and Morphology 111 140. Berlin: Mouton de Gruyter. Kubozono, Haruo, Ito, Junko & Arimin Mester (1997). Oninkoozoo kara mita go to ku no kookai: fukugoo-meishi akusento -no bunseki[The word/phrase boundary from the perspective of phonological structure: the analysis of nominal compound accent]. In Spoken Language Re search Group ( ed s ) Bunpoo to onsei. page number[ Speech and Grammar ], 147166. Tokyo: Kurosio Publi shers Mathieu Colas, Michel (1996). Essai de typologic des norms composes francais. Ca hiers de lexicologie 69. 71 125. McCarthy, John & Alan Prince (1993). Genralized Alignment. In Geert Booij and Jaap van Marle (eds.), 79 153. Yearbook of morphology. Dordrencht: Kluwer McCawley, James ( 1968). The Phonological Component of a Grammar of Japanese. The Hague : Mouton McCawley, James (1977). Accent in Japanese. In Hyman,L.M. (1977) (ed.) Studies in Stress and Accent. Southern California Occasional Papers in Linguistics 4 : 262272. NHK (1998). Nihongo Hatuon Akusento Ziten [Dictionary of Japanese Pronunciation and Accent]. Tokyo: NHK (Nippon Hoosoo Kyookai) Syuppan. Oda, Kenji (2006). The accentuation patterns of nominal compounds in Japanese Toronto Working Papers in Linguistics 26: 23 64. Prince, Alan; & Smolensky, Paul. (1993). Optimality Theory: Constraint interaction in generative grammar. RuCCS Technical Report 2 Rutgers University. 1 262 Piscateway, NJ: Rutgers University Center for Cognitive Science. (Revised version published 2004). (Online: roa.rutgers.edu/view.php3?id=845 ). Schlucker, Barbar a (2008). Naming strategies in West Germanic Lanugages. Paper presented at the ICLC5 -conference. Leuven. Selkirk, Elizabeth ( 1984). Phonology and Syntax: the Relation between Sound and Structure Massachusetts : MIT Press. Selkirk, Elisabeth (1995). The prosodic structure of function words. In University of Massachusetts occasional papers 18: Papers in Optimality Theory 439469. GLSA, University of Massachusetts, Amherst.

PAGE 90

90 Shibatani, Masayoshi (2003). Directional verbs in Ja panese. Amsterdan: John Benjamins Publishing Co. Tanaka, Shinichi ( 2001). The Emergence of the Unaccented: P ossible Patterns and Variations in Japanese Compound Accentuation In Jeroen Van Deweijer and Tetsuo Nishihara (eds.) Issues in Japanese Phonology and Morphology 163-192. New York : Mouton de Gruyter Truckenbrodt, H. (1999). On the relation between syntactic phrases and phonological phrases. Linguistic Inquiry 30. 219255. Tsujimura, Natsuko. & Davis, Stuart ( 1987). The accent of long nomina l compounding in Tokyo Japanese. Studies in Language 11. 199 206. Tsujimura N atsuko (2008). An Introduction to Japanese Linguistics Malden: Blackwell Publishing. Unger James Marshall ( 1975). Studies in early Japanese morphophonemics. Doctoral dissert ation, Yale University. Oxford: Oxford University Press. Vance, Timothy J. (1997) Review of The Organization of Japanese Prosody by Haruo Kubozono. Tokyo: Kurosio Publishers, 1993. The Journal of the Association of Teachers of Japanese 28: 93103.

PAGE 91

91 BIO GRAPHICAL SKETCH Si Chen was born in 1986 in Nanjing, China. She is the only child in her family. Since her father and mother are musicians, Chen has received musical training from an early age and gained the high proficiency certificate of piano skills in China. Her interests in languages were developed since middle school. After graduation from Jinling High School in Nanjing, she chose English linguistics and literature as her major and received her bachelors degree from Beijing International Studies Uni versity. Upon graduation, she pursued her masters degree in linguistics in University of Florida and began to focus her research on phonology and phonetics in Chinese and Japanese.