Chapter 4 Optimality Theory Analysis
4.2 Substitution of Onset Consonants
4.2.1 Illicit Consonants
國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
67
with the deployment of markedness as well as faithfulness constraints and rankings.
The faithfulness constraints concerning the feature identities are adopted since the output is assumed to be as close to the input as possible. The other markedness constraints are employed to prevent the output from having illegitimate segments or structures. The overall rankings of all of the constraints related to the vowel
substitutions are provided in (14).
(14) *ɯ, *V:, *jo >> MAX, IDENT[Place] >> IDENT[round], IDENT[long]
4.2. Substitution of Onset Consonants
This section discusses the substitution of Japanese consonants in the onset position. As shown in the last chapter, each Japanese consonant may have more than one substitute in Taiwanese. When developing an OT analysis, we consider only the adaptation of the dominant substitute of the Japanese consonants that are illicit in Taiwanese sound system, or those that reveal systematic adjustments.
4.2.1. Illicit Consonants
There are eight Japanese onset consonants [d], [ɸ], [ç], [ʃ], [ʧ], [ʤ], [dz], and [ɾ]
that are not legitimate in the Taiwanese sound system. The segmental markedness constraints for consonants, as in (15), are employed to prohibit the occurrence of those consonants.
(15) Consonantal markedness constraints: Assign one violation mark for every consonant [d], [ɸ], [ç], [ʃ], [ʧ], [ʤ], [dz], and [ɾ].
Constraints in (15) cannot be violated, since these consonants are not found in the loanwords.
‧
The constraint family IDENT[F] is employed to decide which Taiwanese
consonants shall be the substitutes of the corresponding Japanese ones. Consider the feature specifications of the two languages, as illustrated in (16).
(16) Feature specifications of onset consonants of Japanese and Taiwanese
d l ɾ ɸ h ç ʃ s ʧ ʦ ʤ z dz
The value of [continuant] of the Taiwanese [l] is controversial in the literature. It is normally viewed as a [+continuant] sound in languages. However, since [l] is grouped with stop to form a natural class in Taiwanese, some researchers suggest that its [+continuant] value is not obvious (e.g. Luo 1993, Chung 1996, Hung 2005). In the present study, we follow the universal principle and treat [l] as a [+continuant] sound.
The affricates are neither [+continuant] nor [-continuant]. Thus, to distinguish the affricates from other segments, the feature [delayed release] is adopted. Only the affricates have positive value of this feature.
As shown in (16), all Japanese consonants and their substitutes are identical in the [voice] feature. This suggests that any violation of this feature is fatal. Besides, almost all the substitutions have the same value of [labial], [dorsal], and [coronal]
features. The glottal [h] is a laryngeal sound, it is considered as placeless here.
Accordingly, [h] incurs no violations of IDENT[dorsal], IDENT[labial], and IDENT[coronal].
‧
Moreover, although every pair in a consonant substitution in (16) is the same in [nasal], the identity of this feature is not so important. This is evident in the [m] to [b]
and [n] to [l] substitution, which will be discussed in the next section. Finally, the identity of the manner features, including [continuant], [delayed release], [sonorant], and [lateral], is relatively not significant, which is evident in the [d] to [l] change, or the [ʤ] and [dz] to [z] substitutions.
The relevant constraints and their rankings are given below.
(17) Constraints
a. IDENT[voice]: Assign one violation mark for every output that differs from its input correspondent in the feature [voice].
b. IDENT[dorsal]: Assign one violation mark for every output that differs from its input correspondent in the feature [dorsal].
c. IDENT[coronal]: Assign one violation mark for every output that differs from its input correspondent in the feature [coronal].
d. IDENT[nasal]: Assign one violation mark for every output that differs from its input correspondent in the feature [nasal].
e. IDENT[MR]: Assign one violation mark for every output that differs from its input correspondent in the features [continuant], [delay release], [sonorant], and [lateral].
>> IDENT[coronal] >> IDENT[nasal] >> IDENT[MR]
‧
With the above constraint set, let us first consider the substitution of the Japanese [d]. The dominant substitute of this segment is [l].
(19) Input: [ɯ:do] ‘hood’
IDENT[dorsal] is inactive and is left out in this tableau. Candidate (a) incurs one violation of IDENT[coronal] by changing a labial [b] to a coronal [d]. Candidate (b) is eliminated due to the violation of IDENT[voice]. Candidate (c) is ruled out for violating IDENT[nasal]. Although candidate (e) is most faithful to the input, it is ruled out for having an illicit segment [d] in the word. Candidate (g) violates MAX by deleting the target segment [d], and is excluded. Candidate (d) incurs three violations of
IDENT[MR], but candidate (f) incurs only one violation of the same constraint. Thus, candidate (f) is chosen as the optimal output. However, the optimal output shall be candidate (d) rather than candidate (f) based on the evidence from the present corpus.
In Taiwanese, only vowels [i] and [u] and glides are allowed to follow the fricative [z]. As can be observed, the sequence [zo] in candidate (f) violates this Taiwanese phonotatic. To rule out the candidate (f), constraint *zV[-hi] is employed.
(20) * zV[-hi]: Assign one violation mark for every sequence where [z] is followed by a non-high vowel.
‧
* zV[-hi] IDENT[nasal] IDENT[MR]
☞ a. hulo [lat] [son] [cont]
b. huzo *! [cont]
c. hunõ *! [son]
The other higher ranked constraints are omitted in this tableau, as they are inactive.
Candidate (b) is eliminated by *zV[-hi] in that [z] is followed by a mid vowel [o]. The optimal output is candidate (a) at the expense of IDENT[MR]. Moreover, IDENT[MR]
has to be ranked lower than IDENT[nasal], or candidate (c) will be mistaken as the optimal output.
Next, consider the flap [ɾ], another Japanese consonant that becomes [l] in Taiwanese. As shown in (19), candidates with segment deletion will never be the optimal substitute. Thus, MAX is omitted to save the space in the following tableaux.
(22) Input: [ɾaita:] ‘lighter (noun)’
IDENT[dorsal] is inactive and is omitted here. Candidate (f), though is the most faithful output, is ruled out for its violation of the segmental markedness constraint *ɾ is fatal.
Candidate (b) is ruled out, as its violation of * zV[-hi] is fatal. Candidates (a), (c), and (d) violate IDENT[coronal], IDENT[voice], and IDENT[nasal], respectively, and are ruled out. Candidate (e) is selected as the optimal output at the expense of IDENT[MR].
‧
Next, consider the Japanese fricative [ʃ], which is substituted by [s]. To correctly predict [s] as the optimal output, we need MAX[Place].
(23) MAX[Place]: Assign one violation mark for every input that has no correspondent in the output in [Place].
(24) Input: [mẽ:ʃi] ‘calling card’
*ʃ IDENT[voice] IDENT[coronal] IDENT[MR] MAX[Place]
a. mẽpĩ *! [cont] (28)) are allophones before the vowel [i]. Taiwanese speakers who are affected by the pronunciation of Mandarin accent may pronounce [Ûi] instead of [si] (and [tÛi] instead of [ʦi], and [ʒi] instead of [zi]). Such alternation often occurs in the young generation.
Here we only consider the substitute [si] (aslo [ʦi] and [zi]).
IDENT[dorsal] and IDENT[nasal] are irrelevant and are omitted in the tableau (24).
Candidate (f) is ruled out as its violation of the segmental markedness constraint *ʃ is fatal. Candidate (a) violates IDENT[coronal] and IDENT[MR], and is eliminated.
Candidate (c) differs with the input in the manner of articulation, and is ruled out.
Candidate (d) is eliminated as it differs with the input in [voice]. Candidate (e) incurs one violation of MAX[Place], since [h] is voiceless. Candidate (b) satisfies all the constraints in the tableau, and is chosen as the optimal output.
Next, consider the fricative [ç], which is substituted by [h] in Taiwanese.
‧
IDENT[coronal] and IDENT[nasal] are irrelevant and are left out. The most faithful output candidate (e) is ruled out as its violation of *ç is fatal. Candidates (a), (b), and (c) incur at least one violation of the IDENT constraints, and are eliminated. Candidate (d) is chosen as the optimal output at the expense of violating MAX[Place], which is outranked by the other constraints.
Next, consider another fricative [ɸ], which is also substituted by [h].
(26) Input: [ɸɯɾe:mɯ̃] ‘frame’
IDENT[nasal] is inactive and is omitted in the tableau. Candidates (b), (c), and (d) violate IDENT[voice], IDENT[coronal], and IDENT[dorsal], respectively, and are ruled out.
The most faithful output candidate (f) contains an illegitimate segment [ɸ], which violates the segmental markedness constraint *ɸ, and is eliminated. Candidate (e) defeats candidate (a) by substituting a segment that is consistent with the input in
‧
[continuant] (satisfy IDENT[MR]). It is therefore chosen as the optimal output.
Let us now take a look at the Japanese affricates [ʧ], [ʤ], and [dz]. Consider first the voiceless affricate [ʧ], which is mostly substituted by [ʦ].
(27) Input: [ʧi:sai] ‘small’
*ʧ IDENT[voice] IDENT[coronal] IDENT[MR]
a. sisaj [del rel]!
☞ b. ʦisaj
c. lĩsaj *! [del rel]
d. kisaj *! [del rel]
e. ʧisaj *!
IDENT[dorsal] and IDENT[nasal] are inactive and are left out. The most faithful candidate (e) is eliminated by the segmental markedness constraint *ʧ. Candidates (c) and (d) are ruled out by IDENT[voice] and IDENT[coronal], respectively. Candidate (b) is selected over candidate (a) as the optimal output because the later fatally incurs one violation of IDENT[MR].
Finally, consider voiced affricates [ʤ] and [dz]. The voiced fricative [z] is the dominant substitute of both segments in Taiwanese. Actually, the two affricates show a similar adjustment process, for they are identical in every feature of that is
employed in the present study.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
75
IDENT[dorsal] is inactive and is omitted here. Candidate (e) is eliminated for it contains an illegitimate segment *ʤ. Candidates (a), (b), and (d) violate the constraints
IDENT[coronal], IDENT[nasal], and IDENT[voice], respectively, and are excluded.
Candidate (c) is selected as the optimal output, as it satisfies all the higher ranked constraints at the expense of violating IDENT[MR].
4.2.2. Substitution of nasals [m] and [n]
We presented the details of the substitution of Japanese consonants in the onset position that are illicit in the Taiwanese sound system in last section. However, there are some Japanese consonants, such as the labial [m] and the alveolar [n], that are licit in Taiwanese sound system, but still change to other consonants.
Consider first the substitution of the labial nasal [m]. The nasal feature of the Japanese [m] is deleted when the segment is in the onset of a NVN syllable. In Taiwanese, two nasal sounds cannot co-exist within the same syllable (Chung 1996, Hung 2002). The onset labial nasal in such kind of syllable therefore changes to a non-nasal [b] in loanwords. The constraint in (29) is adopted to avoid such kind of syllable structure. This constraint has to be ranked higher than IDENT[nasal] to exclude the most faithful substitute [m]. The symbol N stands for nasal segments.
(29) *[NVN]σ: Assign one violation mark for every output that has nasals in both onset and coda position in the same syllable.
Regarding the nasalization of vowels, they are clearly nasalized before /ɴ/ in Japanese (Vance 2008), as in [ɾẽɴ] ‘ream’. As for the vowels that follows the nasal consonants like [mõ:ta:] ‘motor’, no literature is found in discussing the nasalization of such vowels. However, when there is a nasal consonant in the onset position, it is
‧
unlikely that the nucleus possesses an oral vowel. On the contrary, it may be likely that a combination of an onset non-nasal consonant and a nucleus nasalized vowel will happen. Therefore, it is assumed in the present study that the vowel that follows a nasal consonant will be nasalized in real pronunciation.
When the Japanese nasalized vowels are loaned to Taiwanese, their [nasal]
feature may either be deleted or be maintained. The nasalized vowel will lose the [nasal] feature when being followed by a nasal coda, for example, [ɾẽɴ] → [len]
‘ream’. In contrast, the [nasal] feature will be preserved when the vowel is in an open syllable, as in [mõ:ta:] → [mõta] ‘motor’. According to Hung (2002, 2005), there is a CV nasal harmony constraint in Taiwanese, as in (30). This constraint excludes syllables like [lẽn] and [mo.ta] in real pronunciation.
(30) C-V N.H.: Assign one violation mark for every output whose onset and nucleus do not agree in [nasal]. (Hung 2002)
With the new constraints proposed above, consider first the modification of the labial [m] in the onset position in an NVN syllable.
(31) Input: [hoɾɯ
mõɴ] ‘hormone’
‧
Substitution of the uvular [ɴ] will be discussed in next section, and is left out here.
*[NVN]σ is undominated, since an NVN syllable is illegitimate in Taiwanese.
Candidates (a), (b), and (c) show that outputs that differ with regard to input in the voice feature and the place feature, such as [dorsal] and [coronal], will never be the optimal output. Based on this, the only possible substitute of the voiced labial nasal [m] is the voiced labial stop [b]. Candidate (h) violates MAX by deleting segment [m], and is ruled out. MAX is violated if it is a segment rather than a feature that is deleted.
Therefore, candidates (a) to (g) do not incur violations of this constraint. As the most faithful output, candidate (e) is still ruled out since it violates the undominated constraint *[NVN]σ. Candidate (d) repairs the problematic NVN structure by deleting the nasal feature of the onset consonant. By doing so however, it violates C-V N.H. and IDENT[nasal]. Candidate (f) eliminates the nasal feature of the vowel, but violates both
*[NVN]σ and C-V N.H.. The optimal output is candidate (g), for it satisfies all the higher ranked constraints at the expense of violating IDENT[nasal]. This shows that IDENT[nasal] has to be outranked by all of the other constraints listed in the tableau.
Next, consider the labial [m] in the onset position in a CV syllable.
(32) Input: [mõ:ta:] ‘motor’
Constraints *[NVN]σ and IDENT[voice] are inactive and are omitted. Candidate (a) incurs one violation of C-V N.H. for it deletes the nasal feature of the onset consonant,
‧
which also violates the lower ranked constraint IDENT[nasal]. Candidate (b) satisfies the CV harmony constraint by deleting the nasal feature of the vowel, but violates IDENT[nasal] twice, and is also ruled out. The two candidates show that any deletion of the nasal features in an Nṽ syllable is fatal. Candidate (f) is eliminated due to the violation of MAX. The most faithful output candidate (c) defeats candidates (d) and (e) and is chosen as the optimal output, for it is identical with the input in the place
feature (that is, IDENT[dorsal] andIDENT[coronal]) but the other two do not.
Next, consider the alveolar [n] in the onset position in either an NVN syllable or a CV syllable, as in tableaux (33) and (34) respectively.
(33) Input: [nõŋki] ‘careless’
C-V N.H. MAX IDENT[dorsal] IDENT[coronal] IDENT[nasal]
a. lõ.ɡi.suʔ *! *
The alveolar [n] functions in a way similar to the labial [m] in evaluating candidates.
The difference is that for the alveolar [n], the output and the input shall be identical in the [coronal] feature rather than in the [labial] feature. In addition, the nasal [n]
‧
changes to the non-nasal [l] instead of [b] in an NVN syllable.