Chapter 4 Optimality Theory Analysis
4.3 Substitution of Coda Consonants
Candidate (a) does not preserve the [+high] feature of [ʃ], which violates MAX[+high], and is ruled out. Candidate (c) satisfies the higher ranked constraint MAX[+high] but violates IDENT[MR], and is ruled out. Candidate (b) is chosen as the optimal output at the expense of violating DEP.
Finally, let us take a look at loanwords like [se:.ɸɯ] to [se.hu] ‘safe’ and [mẽ:.ʃ
i]
to [mẽ.si] ‘calling card’ substitutions. We assume that epenthesis occurs in all loanwords that contain the segments [ɸ] and [ʃ] (also [ʧ], [ʤ], and [ç]), but that the epenthesis segments [i] or [u] are then deleted if they are the same as their following vowels.
4.3. Substitution of Coda Consonants
The Japanese consonants [p], [t], [k], [m], [n], [ŋ], [s], [ʃ] and [ɴ] are observed in the coda position. While the stops [p], [t], and [k] and the nasals [m], [n], and [ŋ] can serve as coda in Taiwanese, the fricatives [s] and [ʃ] and the uvular [ɴ] can not. The following discussion focuses on the substitution of [s], [ʃ], and [ɴ].
4.3.1. Fricatives
First, consider the adjustment of the Japanese fricative [s] (also [ʃ]), which is not allowed in the coda position in Taiwanese. As can be observed, the consonants, [p], [t], [k], [m], [n], and [ŋ], that are allowed in the coda position in Taiwanese are all
[-continuant]. In addition, other than nasals, the coda consonants [p], [t], and [k] are
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
all [-voice]. Although the fricative [s] is [-voice], it is also [+continuant]. Therefore, it cannot be a coda consonant in Taiwanese. To account for the illegitimacy of the fricative [s], we propose CODACONDITION (CODACON), as in (41).
(41) CODACONDITION (CODACON): Assign one violation mark for every coda segment if it conforms to one of the following conditions.
a. The coda segment is [+continuant].
b. The coda segment is a voiced obstruent.
Segment deletion and segment substitution are both observed in the present corpus to fix the problematic coda [s]. The two strategies may be employed to the same word simultaneously. That is, a single Japanese word with the coda [s] will have two counterparts in Taiwanese. For example, [as.sa.ɾi] ‘frankly’ is borrowed as
[a.sa.liʔ] and [at.sa.liʔ].
Scholars (Orgun 1996, Anttila 1997, Inkelas and Zoll 2007) developed the cophonology approach to account for the diversity within a language. It is assumed that there are different phonological systems (results from different ranking of the constraints) that co-exist in a single language. The alternation between the use of two strategies in repairing the problematic coda [s] that mentioned above indicates two different rankings of MAX and IDENT[MR]. If MAX dominates IDENT[MR], we will get [at.sa.liʔ], as in (42). In contrast, [a.sa.liʔ] will be selected if MAX is dominated by IDENT[MR], as in (43).
‧
IDENT[nasal] is not decisive and is omitted in this tableau. All legitimate coda
segments are listed above in candidates (a) to (f). However, candidates (a) and (c) and candidates (d) to (f) are eliminated for the former two violate IDENT[coronal], and the later three violate IDENT[voice]. Although candidate (g) is the most faithful output, it is ruled out as its violation of CODACON is fatal. MAX dominates IDENT[MR] in this tableau, candidate (b) is therefore selected over candidate (h) by taking deletion as a strategy to fix the problem of the [s] in the coda position.
Next, consider another substitute of [assaɾi]. All possible substitutes in the coda position apart from [t] are omitted in the following tableau, since they will be
eliminated by higher ranked constraints, as in (42).
(43) Input: [assaɾi] ‘frankly’
CODACON IDENT[MR] MAX
a. at.sa.liʔ [cont]!
b. as.sa.liʔ *!
☞ c. a.sa.liʔ *
The most faithful output candidate (b) is ruled out by CODACON. Candidate (a) incurs
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
one violation of IDENT[MR]. Candidate (c) is selected over candidate (a) as the optimal output by taking segment substitution as the repairing strategy.
From a cophonology point of view, Taiwanese has two ranking hierarchies to decide substitutes of the Japanese coda [s].
(44) …>> MAX >> IDENT[MR] >>…
(45) …>> IDENT[MR] >> MAX >>…
The [s] to [t] substitution occurs with the ranking in (44), and the [s] deletion takes place with the ranking in (45).
4.3.2. Nasals
This subsection discusses the substitution of the Japanese word-final uvular [ɴ]
and the word-medial nasals [m], [n], and [ŋ].
For the word-final uvular [ɴ], it is illicit in Taiwanese, and is substituted by either an alveolar nasal [n] or a velar nasal [ŋ], as in (46). The patterns that are illustrated are the dominant substitutes.
(46) Japanese sources Taiwanese loans [ɯ̃ɴ], [ĩɴ], [ẽɴ] → [un], [in], [en]
[ãɴ], [õɴ] → [#ŋ], [oŋ]
As for the substitution of the word-medial coda nasals [m], [n], and [ŋ], they will assimilate with its following onset consonant in the place of articulation in both Japanese and Taiwanese. Since the two languages show the same assimilation phenomenon, there will not be any adjustment. However, substitutions are still observed in the present corpus. Consider (47) and (48). The patterns that are illustrated are the dominant substitutes.
‧
(47) Substitution of word-medial coda nasals following [u], [i], and [e]
Japanese sources Taiwanese loans [ɯ̃n], [ĩn], [ẽn] → [un], [in], [en]
[ɯ̃m], [ĩm], [ẽm] → [un], [in], [en]
[ĩŋ], [ẽŋ] → [in], [en]
(48) Substitution of word-medial coda nasals following [o] and [a] / [#]18. Japanese sources Taiwanese loans
[ãn], [õn] → [an], [oŋ] / [on]
[ãm], [õm] → [an], [oŋ]
[ãŋ], [õŋ] → [aŋ], [oŋ]
Generally, a Japanese coda nasal, whether it is [ɴ], [m], [n], or [ŋ], is substituted by [n] when following the front vowels [i] and [e], or the back vowel [u].
In contrast, the substitutes of the coda nasals that follow the vowels [a] / [#] and [o] is not so consistent. First, while the coda nasals [ɴ] and [ŋ] that follow the vowels [a] / [#] is substituted by the velar [ŋ], the coda nasals [m] and [n] that follow the same vowels is substituted by the alveolar [n]. Second, the coda nasals [ɴ], [m], and [ŋ] that follow the back vowel [o] are substituted by the velar [ŋ]. Finally, the coda nasal [n] may either become [ŋ] or remain intact when following a back vowel [o].
Since [n] occurs after front vowels mostly and [ŋ] occurs after back vowels only, we assume that a word-final nasal has to be the same with its preceding vowel in backness, as in (49).
(49) SHARE[back]: Assign one violation mark for every nucleus and nasal coda sequence that are not linked to the same token of [back]. (McCarthy 2008b)
If SHARE[back] is adopted, [un] will wrongly be excluded and [uŋ] will be
18 [a] and [#] are allophones of /a/.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
mistaken, since [un] is a [+high] [-high] sequence, and [uŋ] is a [+high] [+high]
sequence. Therefore, we need another constraint that dominates SHARE[back] to rule out [uŋ]. According to Chung (1996), the two segments in a rime cannot both be [+high] in Taiwanese. The constraint can be formulated as *V-C [+high], as in (50).
(50) *V-C [+high]: Assign one violation mark for every nucleus and coda that are both [+high].
Moreover, although the labial [m] is legitimate in the coda position in Taiwanese, it is never chosen as the substitute, as in (47) and (48). It seems that a labial nasal is not preferred in substituting the coda nasals in the loanword adjustments. The Place markedness constraints (Prince and Smolensky 1993, Lombardi 2001) are adopted to account for this situation.
(51) Place Markedness:
a. *Lab: Assign one violation mark for every labial consonant.
b. *Dor: Assign one violation mark for every dorsal consonant.
c. *Cor: Assign one violation mark for every coronal consonant.
d. Ranking hierarchy: *Lab, *Dor >> *Cor
Since the preference of the non-labial nasals are observed only in the coda position, we assume that the Place markedness imposes only on the coda nasals in the loanword adjustments. Moreover, when evaluating candidates in the following tableaux, we consider only the place features of consonants, but not vowels.
With the new proposed constraints, consider first the word-final uvular [ɴ] to alveolar [n] change that follows the high back vowel [u].
‧
All legitimate coda segments are listed in this tableau.19 Input information is maximally retained in the candidate (g), but it is still ruled out for having an illicit consonant [ɴ]. Candidates (c) and (f) fatally violate *V-C [+high], and are ruled out.
Candidates (b) and (e) win over candidates (a) and (d), as the former two violate only SHARE[back], but the later two violate both SHARE[back] and *Lab. Candidate (b) is further selected over candidate (e) as the optimal outputs, since candidate (b) violates only *Cor, but candidate (e) violates both *Cor and IDENT[nasal]. As shown in the tableau, *V-C [+high] has to ranked higher than SHARE[back], or candidate (b) will not be the optimal output.
As shown in (52), candidates that end with [p], [t], and [k] are harmonically bounded by candidates that end with [m], [n], and [ŋ], respectively. Since these candidates cannot win under any constraint ranking, they will not be taken into consideration in the following tableaux. In addition, a candidate that contains [ɴ] will also be ignored. Although it is the most faithful output, it is never chosen as the optimal one, since [ɴ] is illicit in Taiwanese. Finally, if the candidates that mentioned above are omitted, constraint IDENT[nasal] that is violated by [p], [t], and [k] and constraint *ɴ will no longer have to be listed, for they are inactive.
19 Please refer to section 4.2.2 for a description of the adjustment of the nasality of segments.
‧
SHARE[back] is omitted, since candidates that violate SHARE[back] will also violate
*V-C [+high] in this tableau. Candidates (c) incurs one violation of *V-C [+high], as [i]
and [ŋ] are both [+high]. Candidate (a) violates *Lab, and is ruled out. Candidate (b) is chosen as the optimal output at the expense of *Cor.
Next, consider the [ɴ] to [n] substitution that follows the front vowel [e].
(54) Input: [ke:ɾẽɴ] ‘convulsion’
SHARE[back] *Lab *Dor *Cor
a. k*elem *!
☞ b. k*elen *
c. k*eleŋ *! *
*V-C [+high] is inactive and is left out. Candidate (c) is ruled out by SHARE[back], since [e] is [-back] but [ŋ] is [+back]. Candidate (b) is selected over candidate (a), as candidate (a) violates *Lab but candidate (b) satisfies it.
Next, consider the [ɴ] to [ŋ] change after the low back vowel [#].
‧
*V-C [+high] and SHARE[back] are inactive and are left out. Candidates (a) and (b) are eliminated as they violate IDENT[dorsal]. Candidate (c) is chosen as the optimal output at the expense of violating *Dor. This suggests that IDENT[dorsal] ranks higher than
*Dor.
Next, consider the [ɴ] to [ŋ] change after the back vowel [o].
(56) Input: [ka:tõɴ] ‘carton’
SHARE[back] *Lab *Dor *Cor
a. k*atom *! *
b. k*aton *! *
☞ c. k*atoŋ *
*V-C [+high] is inactive and IDENT[anterior] results in the same violating situation as SHARE[back]. The two constraints are omitted here. Candidates (a) and (b) are
excluded by SHARE[back]. Candidate (c) is selected as the optimal output. This tableau shows that SHARE[back] has to rank higher than *Dor, or candidate (c) will not be chosen as the optimal one.
Next, consider the [m] to [n] change after the low front vowel [a].
(57) Input: [kãmpai] ‘toast’
IDENT[dorsal] *Lab *Dor IDENT[coronal] *Cor
☞ a. k*an.pai * *
b. k*am.pai *!
c. k*ɑ
ŋ.pai
*! **V-C [+high] and SHARE[back] are inactive and are left out. Candidate (c) is ruled out for it fatally incurs one violation of IDENT[dorsal]. Candidate (b) violates *Lab for the substitute [m] is a labial nasal, and is eliminated. Candidate (a) is chosen as the optimal output at the expense of violating *Cor. This tableau shows that both
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
IDENT[coronal] and *Cor are outranked by IDENT[dorsal], *Lab, and *Dor, or candidate (a) could not be selected as the optimal output.
Finally, consider the adjustment of the word-medial coda alveolar nasal [n] that follows the back vowel [o]. It may be substituted by the velar nasal [ŋ] or remain intact. The two substitutes occur in the same phonological environment. For example, [põn.do] → [p*oŋ.lo] ‘pound’ and [bõn.do] → [bon.lo] ‘bond’. In the two
examples, the alveolar [n] appears between the back vowel [o] and the alveolar stop [d]. To account for this substituted variations, we suggest that the [on] and the [oŋ]
variation can be explained by different rankings of SHARE[back] and IDENT[anterior]. Let us first consider the [on] to [oŋ] change in the word-medial position.
(58) Input: [põndo] ‘pound’
SHARE[back] IDENT[dorsal] *Lab *Dor IDENT[coronal] *Cor
a. p*on.lo *! *
b. p*om.lo *! * *
☞ c. p*oŋ.lo * * *
*V-C [+high] is inactive and is omitted. In this tableau, SHARE[back] is ranked higher than IDENT[dorsal]. Candidates (a) and (b) are ruled out immediately by SHARE[back]. Candidate (c) is chosen as the optimal output. In this case, the speakers’ perception of Japanese sounds is shaped by the Taiwanese syllable constraints, since [on] is not a legitimate VC syllable.
Next, consider the unchanged situation of the word-medial coda [n] that follows the back vowel [o].
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
91
(59) Input: [bõndo] ‘bond’
IDENT[dorsal] SHARE[back] *Lab *Dor IDENT[coronal] *Cor
☞ a. bon.lo * *
b. bom.lo * *! *
c. boŋ.lo *! * *
In this tableau, SHARE[back] is ranked lower than IDENT[dorsal]. Candidate (c) is ruled out immediately by the top ranked IDENT[dorsal]. Candidates (a) and (b) both violate SHARE[back], but candidate (a) is selected over candidate (b) as the optimal output, as candidate (b) further violates *Lab. In this case, Taiwanese speakers take the
similarity between the loanword and its sources as the primary concern.
In short, Taiwanese has two ranking hierarchies to decide the substitutes of the Japanese word-medial coda nasal [n] that follows a vowel [o], as in (60) and (61).
(60) …>> SHARE[back] >> IDENT[dorsal] >>…
(61) …>> IDENT[dorsal] >> SHARE[back] >>…
The [n] to [ŋ] substitution takes place with the ranking in (60), and the [n] remains intact under the ranking in (61).
4.4. Syllable Combination
Two syllables may join together to make a single syllable. The phonological environments where such modification happens are presented in (62). The onset consonant in the first syllable is not necessary. In addition, forms such as [ao], [eo], and [ie] that are illegitimate in Taiwanese will not be taken into consideration. Here C stands for certain arbitrary consonants and N for certain arbitrary nasals.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
(62) Loanwords with and without syllable combinations
Japanese sources Taiwanese loans Combine Ca.i, Ca.u, Cu.i, Ci.u → Cai, Cau, Cui, Ciu Not combine Ca.iN, Ca.uN, Cu.iN → Ca.iN, Ca.uN, Cu.iN
As shown in (62), two syllables combine to one when the second syllable is only made up of a nucleus. It seems that an onsetless syllable is less preferred by
Taiwanese speakers. To prevent the onsetless syllable in the surface, ONSET is employed, as in (63).
(63) ONSET: Assign one violation mark for every onsetless syllable.
ONSET predicts, for example, Japanese [an.na.i] ‘guide’ and [kɯ.ɾi.a] ‘clear’, which will be borrowed as [an.nai] and [k*u.lja], respectively in Taiwanese. However, the phenomenon of syllable combination does not take place in [kɯ.ɾi.a] based on the corpus data. It shall be borrowed as [k*u.li.a]. To account for the substituted
difference between [an.na.i] and [kɯ.ɾi.a], IDENT[μ] is adopted, as in (64).
(64) IDENT[μ]: Assign one violation mark for every output that differs from its input in the numbers of mora.
This paragraph briefly introduces the way in counting the numbers of moras.
Japanese allows maximally a CVX syllable structure (X can be a consonant, a nasal, or the second part of a lengthened vowel). Both V and X can bear mora. Thus, a CV syllable contains one mora, and a CVC (or CVN) syllable contains two moras. In contrast, Taiwanese allows maximally a CGVX syllable structure (X can be a
consonant, a nasal, or the second part of a diphthong). The glide (G) is considered as a part of onset (Duanmu 1993). Segments in the onset position do not affect the syllable
‧
weight. In other words, only segments in rime, that is, V and X, are assigned mora.
We adopt this point of view when counting numbers of the moras in a syllable. Thus, while a CGV syllable contains one mora, a CVG syllable contains two moras.
Let us now consider loanwords that go through syllable combination.
(65) Input: [an.na.i] ‘guide’
IDENT[μ] ONSET
a. an.na.i * !
☞ b. an.nai
The input and candidate (a) is syllabified as VC.CV.V; candidate (b) is syllabified as VC.CVG. They each contain four moras. Thus, both candidates (a) and (b) satisfy IDENT[μ]. But candidate (a) is then ruled out by ONSET, as it has an onsetless syllable.
Candidate (b) satisfies all the listed constraints, and it is chosen as the optimal output.
Next, consider loanwords that do not go through syllable combination.
(66) Input: [kɯ.ɾi.a] ‘clear’
IDENT[μ] ONSET
☞ a. k*u.li.a *
b. k*u.lja *!
The input and candidate (a) is syllabified as CV.CV.V, which contains three moras.
Candidate (b) is syllabified as CV.CGV, which contains two moras. Thus, candidate (b) incurs one violation of IDENT[μ], as it has fewer moras than the input. Candidate (a) is the optimal output, though it violates the lower ranked ONSET.
As shown in (62), syllable combination also does not happen if the second
syllable ends with a coda nasal. IDENT[μ] accounts for the absence of forms like Cja or Cwe in the loanwords, but it is unable to rule out forms such as *CaiN, *CauN,
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
*CuiN. Consider tableau (67).
(67) Input: [ka.ɯ̃n.to] ‘count’
IDENT[μ] ONSET
a. k*a.un.to *!
☟ b. k*aun.to
The input and candidates (a) and (b) all contain the same number of moras. Thus, IDENT[μ] is not violated by any candidates. Candidate (a) is further ruled out by ONSET. Candidate (b) is chosen as the optimal output with this constraint set and ranking. However, the optimal output shall be candidate (a) rather than candidate (b) based on the present corpus. It seems that syllables that are ‘too big’ are
comparatively not preferred by Taiwanese speakers. According to Duanmu (1993), a Taiwanese syllable contains no more than two moras. Therefore, super-heavy
syllables like *CauN will violate constraint *[μμμ].
(68) *[μμμ]: Assign one violation mark for every trimoraic syllable. (McCarthy 2008a)
(69) Input: [ka.ɯ̃n.to] ‘count’
*[μμμ] ONSET
☞ a. k*a.un.to *
b. k*aun.to *!
IDENT[μ] is inactive and is omitted in this tableau. To correctly chose the optimal output, *[μμμ] has to dominate ONSET. Candidate (b) is ruled out for having a trimoraic syllable [k*aun]. Candidate (b) is chosen as the optimal output at the expense of violating ONSET.
‧
We have developed an OT analysis of the adjustments of Japanese loanwords in Taiwanese. Analysis shows that adjustments of loanwords mainly result from the deployment of faithfulness and markedness constraints and rankings.
The Hasse diagrams for all the mentioned constraints are presented in (70).
The constraint hierarchy concerning vowel substitution is presented in (70a). The hierarchy is made up of three levels, and the constraints at each level are mutually unranked, respectively.
The constraint hierarchy concerning syllable modification is shown in (70b).
Both IDENT[μ] and *[μμμ] dominate ONSET, but they are mutually unranked.
The constraint hierarchy concerning the consonant adjustments is given in (70c).
First, consonantal markendess (which is simplified as *C in the diagrams), *[NVN]σ, CODACON, and V-C[+high] are at the highest ranking. The former three immediately dominate MAX, IDENT[voice], and IDENT[dorsal]. *[NVN]σ is the only constraint that dominates C-V N.H.. V-C[+high] immediately dominates SHARE[back], and SHARE[back]
further dominates IDENT[dorsal]. MAX, IDENT[voice], and IDENT[dorsal] all dominate IDENT[coronal], but only IDENT[dorsal] dominates *Dor and *Lab. The two constraints further dominate IDENT[coronal] as well as *Cor. IDENT[nasal] is dominated by
IDENT[coronal], and it together with MAX[+high] and MAX[+labial] dominate IDENT[MR]. Finally, DEP is the lowest ranked constraint in this ranking hierarchy, which is
dominated by IDENT[MR].
There are four cophonologies in the Japanese loanwords in Taiwanese. Other than cophonology A that is illustrated in (70), cophonologies B, C, and D are given form (71) to (73), respectively. The four cophonologies differ in the rankings of SHARE[back] and IDENT[dorsal] as well as MAX and IDENT[MR]. In cophonology A, SHARE[back] dominants IDENT[dorsal] and MAX dominates IDENT[MR]. In
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
cophonology B, IDENT[dorsal] dominants SHARE[back] and MAX dominates IDENT[MR]. In cophonology C, SHARE[back] dominants IDENT[dorsal] and IDENT[MR] dominates MAX. In cophonology D, IDENT[dorsal] dominants SHARE[back] and IDENT[MR]
dominates MAX.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
97
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
99
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
101
Chapter 5 Conclusion
5.1. Thesis Summary
This thesis built a corpus of Japanese loanwords in Taiwanese, comprised of only phonologically oriented loanwords. About 1003 Japanese loanwords are collected.
Numbers and percentages of each pair of substitution are calculated. Based on the statistics obtained from the corpus data, this thesis has revealed several patterns of substitutions of the Japanese segments. Most of the Japanese segments are in one-to-one mapping with the Taiwanese segments. Some Japanese segments may have one-to-two mapping due to Taiwanese phonotatics, such as the modification of the NVN syllable; feature preservation between the sources and the loans may also occur, such as the phenomenon of reconfiguration.
Other than segmental substitution, we also examined the phenomenon of glottal insertion in the word-final position. Generally, glottal insertion will take place when the penultimate syllable in a loanword ends with a vowel or a non-sonorant consonant, and the last syllable is an open syllable.
Moreover, we examined a phenomenon of syllable combination in loanwords.
Two adjacent syllables will combine into one if the moras are not deleted. In addition, the combination will not happen if the second syllable ends with a nasal coda, which would result in an illegitimate super-heavy syllable.
This thesis has analyzed the dominant and systematic substitutions of Japanese
‧
segments in the loanwords under the framework of OT.
Regarding vowel substitution, only Japanese [ɯ] changes to another segment when borrowing into Taiwanese. A set of IDENT constraints is adopted to show the similarity between the input and output. The rankings of these IDENT constraints require feature identity between the corresponding segments. Since this vowel only differs with [u] in roundness, IDENT[round] is in the lowest position in the ranking hierarchy. The [o] to [ə] change and the vowel shortening were also discussed. Both are considered as the result of the Taiwanese phonotatics.
As for consonant substitution, the situation is relatively complex. The ranking hierarchy proposed for the substitutions of the illicit onset consonants shows that the identities of voice and place features are much more important, while the identity of manner features is comparatively less important.
When looking at the [m] to [b] change and the [n] to [l] change in the NVN syllable and the adjustment of nasality of vowels, the constraints *[NVN]σ and C-V N.H. were proposed. Both constraints forbid two nasals in a syllable, and require no nasalized vowel in a close syllable.
Japanese segments [ɸ], [ç], [ʃ], [ʧ], and [ʤ] show the phenomenon of
reconfiguration. When dealing with this phenomenon, MAX[labial] and MAX[+high] are adopted to account for the preservation of the [labial] feature of [ɸ] and for the
preservation of the [+high] feature of the other consonants.
As for the Japanese coda fricatives, which are not allowed in the coda position in Taiwanese, are subject to segment substitution or deletion. We suggested that this results from two cophonologies that differ in the ranking between MAX and
IDENT[MR]. When MAX dominates IDENT[MR], [t] substitution takes place. In contrast, deletion happens when IDENT[MR] dominates MAX.
The substitutions of the coda nasals are accounted for by *V-C[+high],
‧
SHARE[back], and the Place markedness constraints (*Lab, *Dor >> *Cor), either in the word-final or word-medial position. However, we suggested that the rankings between SHARE[back] and IDENT[dorsal], show that two cophonologies are needed. For example, there are two possible substitutes, [on] and [oŋ], for the same source [õn].
When SHARE[back] dominates IDENT[dorsal], the substitute [oŋ] is obtained. In contrast, when IDENT[dorsal] dominates SHARE[back], the substitute [on] is chosen.
Finally, three constraints are proposed to account for syllable combination in loanwords, they are, IDENT[μ], *[μμμ], and ONSET. This constraint set accounts well for the absence of forms like [k*aun.ta] ‘counter’ and [k*u.lja] ‘clear’. The former violates *[μμμ] by having three moras in the rime, while the latter violates IDENT[μ]
by changing the vowel [i] in the source form to the glide [j] in the borrowed form, yielding the deletion of the mora.
To conclude, the substitutions of Japanese segments in Taiwanese are accounted for by the high ranking of the markedness constraints. Alternative ranking of the same constraint set then explain certain variations of the corresponding segments.
5.2. Further Issues
There are some issues which were not dealt with in this thesis.
First, glottal insertion and voiceless stop aspiration are discussed in the present study with the statistics from corpus data. However, the two adjustment phenomena are not analyzed under the framework of OT.
Second, we have shown that there are variations in the substitutions of Japanese segments. However, the present study only analyzes the dominant substitute and the substitute that reveal a systematic adjustment of each Japanese segment when
developing an OT analysis. For further research, one may analyze all of the variations
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
of every Japanese segment by adopting the concept of floating constraints.
Finally, while Japanese words contrast each other by high and low pitches, as in ha-si (HL) ‘chopsticks’ and ha-si (LH) ‘bridge’, Taiwanese use tones to distinguish words, as in tong ‘east’ and tōng ‘hole’. The adjustment of pitches in Japanese loanwords in Taiwanese has been of interest to researchers. However, it is still controversial whether there are correlations between Japanese and Taiwanese
supersegmentally. The complex mapping of tones of Japanese loanwords and pitches of the original sources may be better accounted for by investigating the statistics obtained from corpus data.
‧
Akamatsu, Tsutomu. 1997. Japanese phonetics: Theory and practice. München, Newcastle: Lincom Europa.
Anttila, Arto. 1997. Deriving variation from grammar. In Frans Hinskens, Roeland Van Hout & W. Leo Wetzels (eds.), Variation, change and phonological theory, 35-68. Amsterdam: John Benjamins.
Chang, Yu-Hung. 1993. The assimilation of Japanese loanwords in Taiwanese
Chang, Yu-Hung. 1993. The assimilation of Japanese loanwords in Taiwanese