Chapter 2 Literature Review
2.2 Bases of Taiwanese and Japanese Phonology
2.2.2 Japanese
Japanese vowels, especially high vowels, are devoiced in a particular
environment (Toda and Huang 2005, Tsujimura 2007). A high vowel will become voiceless when it appears between voiceless consonants, or when it is in the word-final position and is preceded by a voiceless consonant, as in (10).
(10) Devoicing of high vowels2 (Tsujimura 2007) V → V / C ___ C
[+high] [-voice] [-voice]
#
There are two kinds of codas in Japanese, nasal codas and obstruent codas. A coda nasal is placeless underlyingly in Japanese, which is represented by a capital ‘N’.
A word-medial coda nasal will acquire its place feature from the following onset consonant. A nasal sound is pronounced as the labial nasal [m] when it is followed by the labial consonants [p], [b], and [m]. It is pronounced as the alveolar nasal [n] when preceding the alveolar sounds [t], [d], [ʦ], [dz], [ʧ], [ʤ], [ɾ], and [n]. It is pronounced as the velar [ŋ] when it is followed by the velar sounds [k], [g], and [ŋ]. This
assimilation phenomenon is generalized in (11).
(11) Nasal assimilation3 (Tsujimura 2007) N → nasal / ___ C
[α place] [α place]
As for a coda nasal that is in the word-final position, it will be realized as a
uvular [ɴ]. In addition, a vowel that precedes the uvular [ɴ] is nasalized (Vance 2008).
A coda obstruent occurs only in the word-medial position in Japanese. Such kind of coda obstruent is the initial part of an extra-long obstruent (or a geminate
2 The circle under the capital V indicates that it is devoiced.
3 Tsujimura uses ‘n’ to represent the nasal underlyingly.
‧
obstruent). According to Vance (2008), a geminate obstruent is treated as “a syllable-final obstruent followed by a syllable-initial obstruent” in syllabification.
Segments including [ʧ], [ʤ], [ʃ], and [ç] are classified as palatal sounds in Japanese (e.g., Akamatsu 1997, Ito and Mester 2003, Tsujimura 2007, Yamaguchi 2007). A palatal sound is considered as [+high].Discussion of the [+high] property of these segments is found in Yoshida (1996, 2001), who conducted the analyses under the framework of element theory. Generally, the [+high] property of [ʧ], [ʤ], [ʃ], and [ç] results from feature spreading. Since [ʧ], [ʤ], [ʃ], and [ç] normally occur before the vowel [i], Yoshida suggested that “the spreading property of [i] must be [+high]”.
Moreover, Yoshida (2001) presented the liner representations of Japanese consonants in terms of elements. Certain of them that are related to the present study are listed in (12).
(12) Linear representation (Yoshida 2001) ɸ (h.U) ç (h.I)
ʧ (ʔ._); (h.I._) ʤ (L.ʔ._); (L.h.I._)
ʃ (h.I._)
Vowels are constructed from the elements I, U, and A, which are independently manifested as /i/, /u/, and /a/ in the nucleus, respectively. As shown in (12), [ɸ]
contains a U element in its representation and the other four segments all contain a I element in their representation. Based on the discussion, [ɸ] shall be [+labial] (the vowel [u] is a labial sound), and [ʧ], [ʃ], [ç], and [ʤ] are all [+high] in Japanese.
2.3. Previous Studies on Japanese Loanwords in Taiwanese
There is some research concerning segmental substitution of Japanese loanwords in Taiwanese.
‧
Chang (1993) in his paper examined segmental substitution of Japanese
loanwords. However, he did not analyze the substitution of every Japanese segment.
For example, there are five oral vowels in Japanese, but only the vowel [o] was analyzed. Moreover, a descriptive approach was utilized in his paper. For example, when discussing the substitution of the Japanese coda [s], he merely gave the
statement that “a Japanese syllabic non-nasal may not be realized in Taiwanese.” The motivation of such segment deletion was not provided in the study.
Certain of Chang’s analyses of loanword adjustments are based on step by step rule applications. He suggested that like Japanese, a coda consonant will assimilate with its following onset in the place of articulation in Taiwanese. After assimilation rule, the phonotatic rules (which were not explained in the paper) will be applied to the forms that meet the requirements, as in (13). Here we consider only the
substitution of the nasal coda.
(13) Input: /bok.sĩŋ.ɡu/
/bok.siŋ.ɡu/
Output: /bok.sin.ɡu/
In the present corpus, the [kãm.p,o:] to [k*an.pjə] ‘dried gourd’ change is observed, which contradicts Chang’s analysis. Based on the rule application in (13), the assimilation rule applies to [kãm.p,o:], resulting in the wrong output [k*am.pjə].
The phonotatic rule is inactive since [am] is a legitimate rime in Taiwanese. As will be discussed in chapter 4, the change of [am] to [an] can be properly accounted for by adopting other universal constraints.
assimilation rule
phonotatic rule
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
17
2.3.2. Tsao’s (2000) Analysis
Tsao (2000) proposed a rule for glottal insertion when discussing the assignment of tones to Japanese loanwords in Taiwanese. He classified syllables into short and long ones. Syllables with structures such as VC, V:, and VV are considered as heavy syllables. The first part of these structures is assigned a letter N, and the second part is assigned a letter A. Accordingly, a long syllable can be represented as having an NA structure, while a short syllable can be represented as N. Regarding glottal insertion, it will not occur when the penultimate syllable has an NA structure. Compare two loanwords in (14).
(14) Glottal insertion
(a) su pa na ‘spanner’ (b) bok sin ku ‘boxing’
N N N NA NA N
M H Lh M HH LL
As shown in (14), the glottal h is added to the loanword supana ‘spanner’ when both the final and the penultimate syllables are open syllables. In contrast, glottal insertion will not take place when the penultimate syllable contains an NA structure, even though the final syllable is an open one.
This generalization of the role of glottal insertion may be too rigid in that counter-examples are likely to be found. For example, loanwords like tsiok.kih
‘scissors’ and ji.sia.khuh ‘magnet’ are observed in the present corpus. Therefore, when examining the phenomenon of glide insertion, we have to distinguish the VC syllable from the V: syllable and the VV syllable, rather than grouping them together.
‧
Hung (2002) gave a short discussion of the modification of the onset nasal in an NVN syllable in Japanese loanwords in Taiwanese when examining the phenomenon of the nasal spreading. His analysis was conducted under the framework of Optimality Theory. Constraints that relate to the modification of such nasal are presented below.
(15) *-ṽC: No nasalized nucleus in a close syllable.
(16) SURVIVENAS: The underlying nasal has to be presented in surface.
Variations in the modification of a NVN syllable can be explained by different rankings of the same constraint set, as in tableaux (17) and (18).
(17) mãŋ31ga11 is obtained with the opposite ranking hierarchy. The different results of the output choice are accounted for by the deployment of constraints and ranking.
The grammar in (17) obeys the phonological restrictions that are imposed by Taiwanese. However, the grammar in (18) takes the similarity between sources and loans as the primary concern. Moreover, Hung suggested that generations who do not receive education under Japanese rule will adopt the ranking in (17). On the contrary, those who received education under Japanese rule will adopt the ranking in (18).
‧
However, he also pointed out that Taiwanese speakers who can not speak Japanese may also adopt the ranking in (18) when pronouncing Japanese loanwords.
There are some problems in Hung’s analysis. First, other than the adjustment of the nasal feature, place assimilation also occurs in substituting the Japanese word
man-ga ‘cartoon’. The coda consonant will assimilate with its following onset in both
Japanese and Taiwanese. The candidates in (17) and (18) show the phenomenon of such assimilation, but it was not discussed at all.
Second, only the -VN syllable is allowed in Taiwanese in terms of the constraint
*-ṽC given in (15). It prohibits the possibility of *[sãp] (-ṽC) and *[sĩãŋ] (-ṽN)
correctly. However, the constraint *-ṽC will wrongly excludes the possibility of forms like [mãʔ]. Therefore, the constraint is better defined as *-ṽN. The lack of the nasality of a vowel mainly results from the existence of the following nasal coda, but not the obstruent coda.
As for the absence of forms like *[sãp], it can be excluded by the constraint C-V N.H, which was also provided in Hung’s paper, and is represented in (19).
(19) C-V N.H.: Assign one violation mark for every output that its onset and nucleus do not agree in [nasal].
This constraint concerns the nasality of an onset consonant and its following nucleus.
However, it wrongly excludes the forms like [tã] ‘carry things (on shoulder)’ in Taiwanese. As can be observed, the onset segment in [tã] is a voiceless obstruent [t].
Hung also noticed this situation, and he proposed a constraint *c̃-, saying that the [nasal] feature will not spread to a voiceless onset obstruent. By ranking *c̃- higher than C-V N.H, forms like [tã] can be correctly predicted.
Finally, the constraint SURVIVENAS can be replaced by IDENT[nasal], since they both require the input and the output to be the same in nasality.
‧
Studies of loanwords from a phonological aspect in other languages have been made by several researchers.
Cheng (2001) in his thesis examined English loans in Japanese under the framework of Optimality Theory. His analysis was based on a comparison of the similarity of the features of a segment in the donor language and the recipient one.
The constraint IDENT[F] is adopted to account for the process of segmental
substitutions. He concluded that segmental substitutions can be seen as a result of the interaction between markedness constraints concerning segmental substitutions and faithfulness constraints. These constraints are part of Universal Grammar, and they form a specific ranking that belongs to Japanese.
Kenstowicz (2003) used the term ‘Reconfiguration’ to describe the phenomenon of segmental substitution when reviewing Flavien Gbéto’s study of the loanword phonology of Fon. The process of reconfiguration refers to the argument is that segments will be split off when they enter other languages. For example, the French /R/ is realized as [ʁl] in Fon. Since [ʁ] is an obstruent but /R/ is [+sonorant], [l] is added to fulfill the feature identity of [+sonorant]. Japanese loanwords also reveal similar phenomenon of reconfiguration when they are borrowed into Taiwanese, which will be discussed in the following chapters.
Some scholars suggested that both perception and grammar play roles in the word-loaning process (Steriade 2001, Yip 2006, Kim 2008). People first perceive non-native segments from speakers of the donor language. They then find similar segments that correspond to the one that they heard within their native sound inventory. After excluding or modifying the patterns that do not comply with phonological constraints in the recipient language, the adapted loanwords appear (Kenstowicz 2005). Such intermediate view stands between a purely perceptual modal
‧
and a purely grammatical modal. This intermediate view of loanword adjustments is adopted in this thesis.
2.5. Previous Studies with Loanword Corpus
Much research has been analyzed loanword adjustments based on loanword corpus. Researchers built a corpus, with an attempt to find out the dominant substitute tendency. With the objective numerical basis obtained from corpus, researchers can distinguish the major tendencies from minor ones, and forming generalizations of loanword substitutions.
Guo (1999) examined the modification of the consonant clusters of English loanwords in Mandarin. He found that onset clusters are all parsed in Mandarin with vowel epentheses to fix the problematic CCV structure. In contrast, about three quarters of coda clusters are parsed faithfully with vowel insertions or segmental substitutions, but about a quarter of coda clusters are fixed by segment deletions.
Lü (2005) also examined English loanwords in Mandarin, but he focused on only coda consonants. He found that different categories of coda consonants show different adjustment strategies. While stop, fricative, affricate, and nasal segments that in the coda position are more likely to be preserved, retroflex codas tend to be deleted. The tendency of the preservation or deletion of a coda consonant is consistent whether it stands alone or is part of a consonant cluster.
Miao (2005) examined loanwords form three languages (English, German, and Italian) in Mandarin Chinese. Her findings about the tendencies of preservation or deletion of coda segments are similar with Lü (2005). Besides, Miao also discussed the phenomena of the consonant substitutions. She pointed out that for obstruents and nasals, the faithful outputs constitute the majority of the observed outputs, and these
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
outputs usually agree with their sources in manner features. The deviant outputs make up the minor proportion of the observed outputs, and the deviant features are usually voicing/aspiration and/or place features. As for laterals, retroflex, and glides, the faithful outputs are still the majority, but certain manner features, such as
approximancy and sonorancy, are likely to be changed.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
23
Chapter 3
Corpus and Generalization of Loanword Substitutions
This chapter introduces a corpus of Japanese loanwords in Taiwanese. Segmental substitutions between the two languages are presented. Based on the corpus data, the numbers and percentages of each segmental substitution are calculated. The dominant substitutes of the segments can be pointed out in terms of the evidence from these statistics.
3.1. The Corpus
The Japanese loanwords are gathered from a dictionary edited by Chang (2005).
About 1003 loanwords are collected, which contains only phonologically oriented loanwords. The data includes all of the listed Taiwanese variations of a single Japanese word. ‘Microsoft Excel’ is the software used for the corpus, as it is inexpensive and is widely used. It is also easy for researchers to use in computing, sorting, and filtering data.
All loanwords in the corpus are serially numbered. The serial number of the data is composed of the syllabary (K for katakana and H for hiragana4), syllable count, and ordinal number. Take (1) for example, the capital “H” means hiragana, a native Japanese or Sino-Japanese vocabulary item; the number “3” indicates three syllables;
4 In Japanese, Katakana is used in transcribing foreign (mainly Western language) borrowings, while Hiragana is used in transcribing native (Yamato) or Sino-Japanese (Chinese borrowings) vocabulary items.
‧
the number “004” indicates the fourth three-syllable loanword in the corpus.
(1) Serial# H-3-004
Each loanword datum in the corpus contains three parts: the original Japanese vocabulary, the Taiwanese correspondent, and the English translation. For the original Japanese word, the syllabary, kanji, and its romanization5 are provided in the corpus.
For the Taiwanese loanword, the romanization and its meaning are provided. Example (2) shows detailed information on the parts coded.
The key information of each loanword is the phonetic transcription. The phonetic pronunciation of each segment of a word is coded in separated columns. This
enhances the clarity of vision when making a comparison between Japanese and Taiwanese. The division of words into segments is based on the syllable structures of the two languages. Japanese allows maximally a CVC structure, while Taiwanese allows maximally a CGVX structure.
There is one additional column ‘E’ (stands for entering tone) in Taiwanese. The glottal stop [ʔ] is used to represent a syllable with an entering tone (or check tone)
5 The Romanization system used in transcribing Japanese sources is the Hepburn system, while the one in transcribing Taiwanese loans is the system made by National Languages Committee (2007).
6 Abbreviations: ROM = Romanization, C = Consonant, V = Vowel (the small numeral indicates the ordinal of the syllable in question), G = Glide, X = Glide or Consonant, E = Entering tone, S = Syllable, TRS = Translation.
‧
when the syllable does not end with a consonant. This segment has to be placed in another column in order to correctly transcribe the vowels (especially the diphthongs) that precede it, as shown in (2). However, if the segment in ‘X’ is a consonant, there will never be a glottal stop [ʔ] in ‘E’.
Suprasegmental information, the pitch accent of each Japanese word and the tone of each Taiwanese syllable, are also included in the corpus.7 For example, the
Japanese word tsuyu ‘sauce’ has two moras. The numeral ‘1’ in Peak indicates that the first mora carries a high pitch, and the pitch pattern is thus HL. For the Taiwanese loanword, a five level scheme is used to represent the tone.
3.2. Segmental Substitutions
In this section, we present the substitution of each segment with loanword data.
Statistics are provided to allow for identification of the major tendencies.
3.2.1. Vowels
There are five vowels [a], [i], [e], [o], and [ɯ] in the Japanese vowel system, and they show the diversity in substitution. Generally, a Japanese vowel will remain intact, be substituted for by other vowels, or be omitted, when it is borrowed in Taiwanese.
First, consider the substitution of the Japanese short vowel [a] and the
lengthened one [a:], as given in (3). The short vowel [a] is chosen as the substitute for both segments. No exceptions are found in the present corpus, as in (4).
(3) Vowel [a] / [a:] substitution
a. kaabu [ka:bɯ] → khabu [k*abu] ‘curve’
b. arei [aɾe:] → ale [ale] ‘dumb-bell’
7 The pitch accents of Japanese words are based on the NHK Dictionary of Japanese Accent, and the tone information of Taiwanese loanwords is included in the source dictionary.
‧
(4) Statistics of [a] substitution Japanese
In Japanese, a short vowel contrasts with a long one (Iwasaki 2002, Vance 2008).
Vowel length is phonemic. However, the difference between a short vowel and its lengthened counterpart is viewed as a phonetic variation in Taiwanese. They are treated as the same segment. Thus, vowel length has no effect on the word meaning.
In addition, since short and lengthened vowels do not contrast in Taiwanese, the numbers and percentages of the two types of the same vowel are combined together when providing statistics.
Next, consider the substitution of the Japanese vowel [i].
(5) Dominant substitute of the vowel [i]
a. nori [nõɾi] → noolih [nõliʔ] ‘dried laver’
b. kimono [miso] → khimoonooh [k*imonoʔ] ‘kimono’
As in (5), most of the vowel [i] will remain intact. The statistics in (7) show that this substitution accounts for 99.21% of all occurrences.
The vowel [i] can also be substituted by the vowels [a], [u], or [e], as in (6).
(6) Minor substitutes of the vowel [i]
a. ribidoo [ɾibido:] → libalooh [libaloʔ] ‘libido’
b. rakkii [ɾakki:] → lakkhuh [lakk*uʔ] ‘lucky’
c. pisutoɴ [pisɯtõɴ] → phesutong [p*esutoŋ] ‘piston’
However, the amount of the three substitutes only makes up about 1% of all occurrences, as shown in (7). Since there is no clear phonological effect, the
substitution of the vowels [a], [u], and [e] for the Japanese vowel [i] may be a result
‧
of the borrower’s mis-production. Such instances are considered as exceptions.
(7) Statistics of [i] substitution Japanese
sources Taiwanese
loans Number Total Percentage
i / i: i 504 508 99.21%
a 2 0.39%
u 1 0.20%
e 1 0.20%
Next, consider the substitution of the Japanese vowel [ɯ]. This vowel is illicit in the Taiwanese sound inventory, and will be substituted by the most similar vowel [u], as in (8). The two sounds differ only in roundness. The vowel [ɯ] is unrounded, but the vowel [u] is a rounded one. The vowel [u] is the dominant substitute of the Japanese vowel [ɯ], accounts for 97.62% of all occurrences.
(8) Dominant substitute of the vowel [ɯ]
a. maiku [mãikɯ] → maikhu [mãj̃k*u] ‘microphone’
b. kuro [kɯɾo] → khuloohʔ [k*uloʔ] ‘black’
In addition, the vowel [ɯ] may be deleted when it appears in the word-final position with an onset labial [m] in the same syllable, as in (9a). After deleting, the labial [m] will further combine with the penultimate syllable and become a coda. As shown in (11), there are seven cases of the [ɯ] deletion, and this kind of deletion occurs in four cases.
(9) Deletion of the vowel [ɯ]
a. araamu [aɾa:mɯ̃] → alam [a.lam]8 ‘alarm’
b. bokushingu [bo.kɯ.ʃĩŋ.ɡɯ] → boksingu [bok.sin.ɡu] ‘boxing’
As shown in (9b), the [ɯ] deletion will also take place when following a
8 The dot between two segments stands for the syllable boundary.
‧
voiceless velar [k] and preceding a voiceless consonant. As shown in (11), there are seven cases of the [ɯ] deletion, and three of them are in this environment. In fact, about nineteen loanwords in the present corpus meet the requirement, but only three of them are deleted. Therefore, this [ɯ] deletion is better treated as the result of perception, other than the effect of Taiwanese phonotactic constraints. In Japanese, a short high vowel such as [i] and [ɯ] will be devocalized when it occurs between two voiceless consonants. Such vowels typically, though not consistently, may not be perceived by Taiwanese speakers since they are not sounded in pronunciation.
Other than being substituted by the vowel [u] and being deleted, the vowel [ɯ] can also change to the vowels [i], [o], and [e], as in (10).
(10) Minor substitutes of the vowel [ɯ]
a. kezuri [kezɯɾi] → khejilih [k*eziliʔ] ‘zelkova’
b. kurabu [kɯɾabɯ] → khulabooh [k*ulaboʔ] ‘club’
c. marutoosu [mãɾɯto:sɯ] → malutooje [mãlutoze] ‘maltose’
However, each of these accounts for less than 1% of all occurrences, as in (11), and they will be viewed as exceptions.
(11) Statistics of [ɯ] substitution Japanese
sources Taiwanese
loans Number Total Percentage
ɯ / ɯ: u 616 631 97.62%
ψ9 7 1.11%
i 5 0.79%
o 2 0.32%
e 1 0.16%
Next, consider the substitution of the Japanese vowel [e]. It will remain intact or become [ja] in Taiwanese. The two substitutes are in complementary distribution.
9 The symbol stands for segments that have been deleted.
‧
(12) Vowel [e] substitution
a. kakeru [kakeɾɯ] → khakheluh [k*ak*eluʔ] ‘bet’
b. tare [taɾe] → thaleh [t*aleʔ] ‘sauce’
c. enjiɴ [ẽnʤĩɴ] → ianjin [janzin] ‘engine’
d. setto [setto] → siattooh [sjattoʔ] ‘to set’
As shown in (12a) and (12b), the Japanese vowel [e] will not go through any adjustment when it is in an open syllable or when it is followed by a glottal [ʔ] in Taiwanese. However, this vowel will change to a diphthong [ja] when it is followed by a nasal [n] or a stop [t] in the coda position, as in (12c) and (12d)10, respectively.
Kenstowicz (2003) named such segment splitting phenomenon as “Reconfiguration”.
As provided in (13), about 28.38% of the occurrences of the Japanese vowel [e]
change to [ja], but most of them still remain unchanged in Taiwanese, making up 71.62% of all occurrences. Since CV syllables comprise the majority in Japanese, it is normal that the percentage of the [e] to [ja] substitution is comparatively low.
(13) Statistics of [e] substitution Japanese
Finally, consider the substitution of the Japanese vowel [o]. The majority of the occurrences of this vowel will remain intact in Taiwanese, as in (14). As provided in (16), this accounts for 92.65% of all occurrences.
(14) Dominant substitution of vowel [o]
a. furo [ɸɯɾo] → hulooh [huloʔ] ‘bath tub’
b. koshi [koʃi] → khoosih [k*osiʔ] ‘dado’
10 Here is [jan] is in fact pronounced as [en] in Taiwanese. However, [jan] instead of [en] is adopted to represent the sound, in order to show variations of substitutions of the Japanese [e].
‧
The vowel [o] can also be substituted by [ə]. As shown in (15a), this vowel is a monophthong in Japanese, but it becomes part of a diphthong in Taiwanese loans. The [o] to [ə] substitution takes place since the diphthong io is typically pronounced as [jə]
instead of [jo] in Taiwanese. Although the [o] to [ə] change only accounts for 5.73%
of all occurrences, all of the cases show phonological consistency. Thus, this kind of substitution is treated as a systematic adjustment.
(15) Minor substitutes of vowel [o]
a. chousei [ʧ
o:se:] → tsiose [
ʦjəse] ‘adjustment’b. to [to:] → too [to] ‘grade’
c. kopii [kopi:] → khappih [k*appiʔ] ‘copy’
The vowels [u] and [a] can also be the substitutes of the Japanese vowel [o], as in (15b) and (15c). However, they will be disregarded here due to the low percentage of
The vowels [u] and [a] can also be the substitutes of the Japanese vowel [o], as in (15b) and (15c). However, they will be disregarded here due to the low percentage of