Chapter 3 Corpus
3.4 Summary and Comparison with Miao’s (2005) Analysis
國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
48
analysis and examples are given in (35) and (36), respectively. Nasal vowels are adapted in a similar way. The variant substitutions of [ɑ̃] are provided for illustration.
(35) Statistics of nasal vowel substitutions
French segment Mandarin structure Number Total Percentage
ɛ̃ Vn or Vŋ 70 71 98.59%
Nasal deletion 1 1.41%
ɔ̃ Vn or Vŋ 100 100 100%
ɑ̃ Vn or Vŋ 154 159 96.86%
Nasal deletion 5 3.14%
(36) Examples of nasal vowel substitutions
a. Angers [ɑ̃.ʒe] [an.ʐu] 安茹 ‘place name’
b. Angers [ɑ̃.ʒe] [ɑŋ.tɕjɛ] 昂傑 ‘place name’
c. Manche [mɑ̃ʃ] [mwɔ.ɕy] 莫須 ‘place name’
3.4 Summary and Comparison with Miao’s (2005) Analysis
This chapter builds a French loanword corpus. It also provides the substitution patterns of onsets, codas, and nasal vowels. Based on the data, onset and coda segments tend to be preserved and adapted. The only exception is [Vʁ] structure, which has coda deletion as the major substitute. Nasal vowels are mostly adapted as an oral vowel followed by a nasal segment. The patterns of loanword adaptation are provided in (37)-(39).
‧
(37) Adaptation pattern of obstruent segments French
‧
(38) Adaptation pattern of sonorant segments French
deletion 8.13% adaptation 41.32%
(39) Adaptation pattern of nasal vowels
French nasal vowels Mandarin loans Percentage
ɛ̃ adaptation 98.59% adaptation patterns of each consonant type. The differences are presented in (40) and (41). The former displays the differences in major onset substitutes, and the latter shows the different patterns in major coda adaptations. The onset adaptations between French and other foreign languages have striking similarity. Features of [spread glottis], [voice], and place of articulation can be changed, if the manner of articulation is preserved. The only difference lies in the adaptation of the French liquid [ʁ] and the
‧
[lateral] is not a prominent feature in loanword adaptation.(40) Comparison of the major onset substitutes with Miao’s (2005) study28 French
stops aspiration voiceless stops
stops aspiration
affricates place affricates place
voiced stops
stops voicing voiced
stops
stops voicing affricates voicing, place affricates voicing, place voiceless
fricatives
fricatives place voiceless fricatives fricatives voicing, place fricatives voicing, place affricates voicing, place,
aspiration
Coda adaptations have more differences between French and other foreign
28 Shading suggests the different adaptation patterns in this study and Miao’s (2005) research.
29 Miao (2005) investigates three foreign languages: English, German, and Italian.
‧
languages. This is demonstrated in (41).
(41) Comparison of the major coda substitutes with Miao’s (2005) study French
stops aspiration voiceless stops
fricatives place voiceless fricatives
fricatives voicing, place voiced fricatives
fricatives voicing, place affricates voicing, place affricates voicing, place
nasals nasals place nasals nasals place and other languages. French coda voiceless stops can be deleted in Mandarin, whereas other foreign languages only allow the aspirated or unaspirated counterparts. Also, French voiceless fricatives can be replaced with voiceless affricates of different place
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
53
and [spread glottis] feature. Other foreign languages only allow fricatives of different places of articulation. For French [l] and [ʁ], both deletion and [ɹ] can be the substitutes. This suggests that the change of place, [lateral], [approximant], and [sonorant] do not eliminate the variant forms. In the case of [l] and [r] across the different foreign languages, the vowel [ɚ] is the major adaptation. This shows that the change of [consonant] feature in liquid segments is less important in English, German, and Italian.
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
54
Chapter 4
Optimality Theory Analysis
This chapter offers the analysis of Optimality Theory (Prince and Smolensky 1993/2004; McCarthy 2008) in order to explain the substitution patterns of French loanwords in Mandarin. The ROE model proposed by Coetzee (2006) is used for analyzing loanword variations. Constraint rankings account for output selection.
Whether an output survives depends on the position of the cut-off line. Fatal violations above the cut-off lead to elimination, while violations below the cut-off result in variation.
As shown in the previous chapter, Mandarin speakers adopt different strategies when dealing with French loanwords. Thus, sections 4.1 and 4.2 respectively discuss onset adaptations and coda substitutions. The adaptations of nasal vowels, issues regarding minimal word and mora deletion, and the discussion of non-phonological factors are presented in 4.3, 4.4, and 4.5, respectively.
4.1 Adaptation of Onset Consonants
Mandarin speakers allow variations when they adapt French onset segments. The present corpus shows that onset consonants are mostly preserved. Therefore, the constraint MAX ONSET is proposed to avoid onset deletion.
(1) MAX ONSET: Assign one violation mark for every input onset segment that does not have a correspondence in the output.
Violations of MAX ONSET induce candidate elimination. The constraint is
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
55
positioned above the critical cut-off line and stays undominated. The following subsections, 4.1.1 and 4.1.2, respectively present the analysis of illicit onset segments and the substitutions of shared segments between French and Mandarin.
4.1.1 Illicit Onset Consonants
This section investigates the substitution of illicit onset segments. Consonants [b, d, g, ∫, v, z, ʒ, ʁ, ɲ, ŋ] are included in the discussion. Segments occupying the major proportion of the data will be discussed first, followed by the segments of fewer occurrences.
Several constraints are placed above the cut-off line because the violations will lead to ill-formedness. The constraints are listed below.
(2) Consonant markedness constraint:
Assign one violation mark for every illicit onset segment in Mandarin, including [b], [d], [g], [v], [z], [ʃ], [ʒ], [ʁ], [ŋ] and [ɲ].
(3) Mandarin phonotactics:
Assign one violation mark for every illicit segment combination in Mandarin, such as
*[fi],*[ki], and *[ky], etc.
(4) MAX [continuant]:
Assign one violation mark for every input [continuant] feature that has no correspondent in the output.
(5) MAX [back]:
Assign one violation mark for every input [back] feature that has no correspondent in the output.
(6) MAX [coronal] [-nasal] C:
Assign one violation mark for every input [coronal] feature of a non-nasal consonant
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
56
that has no correspondent in the output.
(7) DEP [nasal] onset:
Assign one violation mark for every output onset [nasal] feature that has no correspondent in the input.
(8) [*[-distributed] & IDENT ([anterior])] fricativesegment:
Assign one violation mark for every output fricative segment that has a different [anterior] feature from the input and thus becomes [-distributed].
The consonant markedness constraint and Mandarin phonotactics constraint function to rule out illicit segments and illegitimate segment combinations. MAX [continuant] prohibits the deletion of the [continuant] feature. Therefore, a fricative segment could not become a stop segment in the output. McCarthy (2008) suggests that the input lexicon should not be restricted to language-specific regulations.
Grammar alone suffices to account for the generalizations. Whether the features are specified in the underlying structures are not important under the framework of Optimality Theory. In the case of [ɹ]-to-[l] adaptation, the constraint MAX [continuant]
is used. MAX [continuant] does not have the problem of feature specification from the perspective of OT. Therefore, it is used to account for loanword adaptation.
MAX [back] disallows the deletion of the [back] feature. Hence, the uvular fricative [ʁ] cannot have an alveolar liquid [ɹ] in the output. MAX [coronal] [-nasal] C
ensures that a non-nasal consonant preserves the [coronal] feature. For example, the alveolar liquid [l] cannot become the velar fricative [x] in the output. As for DEP [nasal] onset, it prevents the insertion of [nasal] feature in the output. Therefore, an input [ba] cannot have an output of [ma].
Łubowicz (2005) points out that a restricted local conjunction needs to fulfill two requirements. First, the locus of violation needs to be identical between the conjuncts.
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
57
Second, there is a causal relation between the conjunction of the faithfulness constraint and the markedness constraint. The violation of the faithfulness constraint results in the violation of the markedness constraint. The conjoined constraint in (8) has a domain of fricative segments. If an output fricative has a different [anterior]
feature from the input, which leads to the [-distributed] feature of the fricative segment, it will be ruled out by the conjoined constraint. Since the violation of
*[-distributed] C and IDENT [anterior] will only bring variations, they are placed under the cut-off line. The definition of each constraint will be provided below.
Violations of some constraints will only result in variations. These constraints are placed below the cut-off line:
(9) IDENT [lateral]:
Assign one violation mark for every input-output disparity of the [lateral] feature.
(10) IDENT [anterior]:
Assign one violation mark for every input-output disparity of the [anterior] feature.
(11)*[sonorant]:
Assign one violation mark for every sonorant segment.
(12) *[-distributed] C:
Assign one violation mark for every [-distributed] consonant.
The feature [lateral] plays a minor role in the adaptation process. This is shown in the change of laterality when the English input [ɹ] is substituted by [l]. Similar adaptation pattern is found in Miao (2005). In Miao’s study, English [ɹ] is usually replaced by Mandarin [l] in the onset position. Such adaptation allows the input-output difference of laterality. Consequently, the constraint IDENT [lateral] is positioned below the cut-off line. Violations of IDENT [lateral] are not fatal.
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
58
IDENT [anterior] is placed below the cut-off line. Violation of it leads only to variations. For example, the fricative onset [z] could become [ɕ] if the following vowel were [i] or [y]. A difference of the [anterior] feature does not result in elimination. As for *[sonorant], it ensures that some licit sonorant segments such as [l]
and [m] will not be ruled out in candidate selection. For instance, the alveolar liquid [ɹ]
could become a lateral liquid [l]. Such adaptation violates *[sonorant], but the violation is not fatal. *[-distributed] C prohibits any [-distributed] consonant in the output. Violation of the constraint will not lead to elimination. Given a foreign input [ʒ], the optimal output could be [ts], which carries a [-distributed] feature. Therefore,
*[-distributed] C is placed below the cut-off line to allow variations.
Let us first examine the adaptation of the alveo-dental lateral [l] and the uvular fricative [ʁ]. The segment [l] tends to be faithfully retained. The dominant substitutes of [ʁ] are the dental lateral [l] (59.69%) and the velar fricative [x] (24.69%). This shows that the preservation of [continuant] is important. As a result, the constraint MAX [continuant] should be placed above the cut-off line.
In terms of feature differences, [x] is more similar to [ʁ] than the segment [l].
However, the adaptation of [l] is more frequently observed than [x]. Why is that? As French is not a dominant foreign language in Taiwan, it is argued that French loanwords have two inputs, English and French. The [ʁ]-to-[l] adaptation presents a similar strategy when Mandarin speakers substitute the English alveolar liquid [ɹ]
with [l]. According to Miao (2005), Mandarin speakers tend to replace the English liquid [ɹ] with [l] (94.34%). Such similar pattern shows that the [ʁ]-to-[l] adaptation is likely to result from an English input. On the contrary, the substitution of [x] from [ʁ]
suggests a French input. In terms of feature identity, segments [x] and [ʁ] share more features than the consonants [l] and [ʁ]. Therefore, two sets of ranking are needed to account for the adaptations of the [ʁ] segment.
‧
provided in (13) and (14), respectively.(13) ‘Le Marais’ [(lə).ma.ɹɛs] 瑪雷(區) [ma.lej.(tɕʰy)]
‘the Marais district’ (English input)30 31 Input:
Candidates (b)-(f) have respective violations above the cut-off line, and they are ruled out. Candidates (c)-(f) incur violations of the undominated constraints. Candidate (c) violates both DEP [nasal] onset and MAX [continuant]. The former constraint ranks higher than the latter. Therefore, (c) is defeated immediately by the DEP constraint.
Candidates (d), (e), and (f) violate MAX ONSET, the conjoined constraint, and *ɹ ONSET, respectively. They are defeated. As for (b), it is eliminated by MAX [coronal]
[-nasal] C. Candidate (a) only has violations below the cut-off line, and it is chosen as the
30 Only the syllable containing the onset segment [ɹ] is relevant to the discussion; therefore, vowel and coda segments are not analyzed in this tableau. All tableaux in this thesis follow this convention.
31 Hsiao (2011) mentions that in Mandarin the retroflex obstruent [ʐ] occupies the onset position, while the retroflex liquid [r] fills the coda position. Here I adopt Hsiao’s classification of the onset retroflex segment .The constraint *ɹ onset is used to rule out the [ɹ] segment in the onset position.
‧
candidate becomes the velar fricative [x]. In (14), Candidates (b)-(e) have respective violations above the cut-off line. They are ruled out. The violation of the conjoined constraint makes (b) a loser. Besides the conjoined constraint, candidate (c) also violates DEP [nasal] onset. The two violations eliminate (c). Candidate (d) has an onsetless syllable, making it an unfavored candidate. Candidate (e) has an illicit onset segment [ʁ], and it is eliminated by the consonant markedness constraint. Finally, the null violations allow (a) to become the optimal candidate.(14) ‘Le Marais’ [(lə).ma.ʁɛ] 瑪黑(區) [ma.xej.(tɕʰy)] ‘the Marais district’ (French input)
Tableaux (13) and (14) have similar set of constraints. The differences lie in the markedness constraint *ɹ onset in (13), *[ʁ] in (14) and some MAX constraints concerning [coronal] and [back] features. The optimal candidate goes to [l] if the loanword has an English input [ɹ]. Otherwise, the optimal candidate is [x] from a French input [ʁ]. Despite the difference in loanword origin, the same constraint set
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
61
can be utilized to obtain the optimal candidate.
Next, let us consider the substitutes of French voiced stops [b], [d], and [g]. Each of the segments has two substitutes, the unaspirated stop [p, t, k] and the aspirated stop [ph, th, kh]. The data shows an overwhelming preference for the unaspirated voiceless stops (98.36% for [b]; 95.41% for [d]; 64.58% for [g])32. Here we observe that both the feature [voice] and the feature [spread glottis] alter. This suggests that the two features are relatively unimportant in loanword adaptation. Therefore, IDENT [aspirated] and IDENT [voice] should be placed below the cut-off line. Since the adaptations of [b] and [d] are similar, [b] is taken as an example for illustration.
(15) IDENT [aspirated]:
Assign one violation mark for every input-output disparity of the [aspirated] feature.
(16) IDENT [voice]:
Assign one violation mark for every input-output disparity of the [voice] feature.
Tableau (17) shows that violating the consonant markedness constraint, MAX-ONSET or DEP [nasal] onset is fatal. Therefore, candidates (c)-(e) are ruled out.
The constraints IDENT [aspirated], IDENT [voice], and *[sonorant] do not decide the optimal candidates. Violations of these constraints will only lead to variations.
Candidates (a) and (b) are the optimal candidates. The former has higher frequency than the latter.
32 The adaptation from [g] to [tɕ] accounts for 31.25%.
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
62
(17) ‘ballet’ [ba.lε] 芭蕾 [pa.lej]
Input: [ba] *[b] MAX-ONS DEP[nas]onset ID[asp] ID [voi] *[son]
☞ a. pa *
☞?33 b. pʰa * *
c. i *!
d. ba *!
e. ma *! *
The situation becomes more complicated in the adaptation of [g]. Similar to the adaptations of [b] and [d], the segment [g] has a dominant substitute of a voiceless unaspirated counterpart. However, [g] could also be replaced by the affricate [tɕ].
Tableau (18) provides an example of [ga] adaptation. Although [ka] is predicted to be an output, it has null occurrence in the corpus. Such situation may result from the limited character type of [ka] in Mandarin.
According to the calculation in Tseng (2013), the syllable type [ka] does not appear in high frequency. Out of the 393 syllable types, [ka] is ranked number 356. It has only 18 attestations. Besides, very limited usage is observed in the Sinica Corpus of Accumulated Word Frequency (現代漢語語料庫詞頻統計). The syllable [ka]55 only appears eight times in the corpus; [ka]35 is used nine times. There is no observation of the syllable [ka]214 nor[ka]51. Such a phenomenon indicates that the syllable type [ka] is rarely observed in Modern Mandarin. The [ka] syllable usually goes through palatalization and becomes [tɕja]. Cheng (1973) points out that historically the word ‘card’ has two readings: [ka] 214 and [tɕhja] 214. This suggests that the palatalization process is present historically and synchronically. In the present
33 The symbol ‘☞?’ suggests an output that is possible within the constraint set, yet the output either accounts for a low percentage of the present corpus or does not appear in the data.
‧
consequence of a non-phonological factor. More of this will be discussed in 4.5.Tableau (18) illustrates the process of candidate selection. Candidate (b) is ruled out by the constraint *[g]. MAX ONSET eliminates the onsetless syllable in (e).
Candidate (f) violates both *[ŋ] and DEP [nasal] onset. It is a loser in the tableau. Here we regard the affricate [tɕ] as a contour segment having ‘edge effects’ (Hall, 2007).
Namely, [tɕ] carries both the [-continuant] feature and the [+continuant] feature on the same autosegmental tier. In this regard, (a) does not have a different [continuant]
feature with the input. The constraint IDENT [coronal] is placed below the cut-off so that candidate (a) is immune from elimination. Finally, (a), (c), and (d) are possible outputs. Despite the null occurrence of candidate (c), it is chosen because of non-phonological factors. The constraint IDENT [coronal] is defined after the tableau.
(18) ‘La Garonne’ [la.ɡa.ʁɔn] 加隆(河) [tɕja.loŋ.(xe)] ‘the river Garonne’
Input: [ga] *[g] *[ŋ] DEP[nas] Modern Mandarin. Therefore, candidate (c) is marked with a question mark, showing that the syllable [ka] is not common in Mandarin.
‧
Assign one violation mark for every input-output disparity of the [coronal] feature.
Next, let us examine the adaptation of the fricative [∫]. There is an overwhelming preference of [ɕ] (70.15%). [∫] could also be replaced by the retroflex [ʂ] despite its relatively low frequency (22.39%). This adaptation pattern demonstrates the importance of the [continuant] feature, showing that MAX [continuant] is active.
(20) ‘Chinon’ [ʃi.nɔ̃] 曦農 [ɕi.noŋ] ‘the place Chinon’
Two new constraints IDENT [strident] and IDENT [distributed] occur in tableau (20). The former constraint is placed above the cut-off line because violation of it will bring elimination. The latter constraint is located below the cut-off line; violation of it
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
65
will only result in variations. They are defined before the description of the candidate selection process.
(21) IDENT [strident]:
Assign one violation mark for every input-output disparity of the [strident] feature.
(22) IDENT [distributed]:
Assign one violation mark for every input-output disparity of the [distributed] feature.
As shown in (20), the undominated constraint *[ʃ] makes candidate (k) a loser.
The conjoined constraint rules (e), (g), and (h) out. Candidate (j) has an onsetless syllable, and it violates MAX ONSET. The violations of MAX [back] and IDENT [strident] eliminate candidate (i). Candidates (a)-(d) and (f) are the harmonic candidates.
Candidates [ɕi] and [ʂɨ] are the actual outputs observed in the present corpus.
Even though candidates [ʐɨ], [tɕhi], and [tʂh ɨ] also possess the feature of [-anterior], [+coronal], and [+continuant], they are not the actual outputs. Nevertheless, they are predicted to be the possible outputs in the current constraint ranking. According to McCarthy (2008), “we aren’t concerned with the phonology of the actual words of the language so much as the possible words.” Namely, the grammar of OT places more emphasis on possible outputs instead of actual outputs. Candidates (c), (d), and (f) are predicted to be the real outputs if more data are collected.
The adaptations of voiced fricatives present more variations. The segment [v]
could be replaced by [w] or [f]. The segment [z] has four substitutes: [s], [ɕ], [ts], and [tɕʰ]. The fricative [ʒ] could be adapted as [tɕ], [ʐ], [tʂ], or [ʂ]. Let us first begin with the adaptation of [v]. The segment [v] is illegitimate in Mandarin, so the consonant markedness constraint is used to rule it out. Tableau (23) illustrates the candidate
‧
Assign one violation mark for every input-output disparity of the [labial] feature.
As indicated in (23), candidate (b) has one fatal violation of the consonant markedness constraint; it is ruled out immediately. MAX [continuant] eliminates (d).
DEP [nasal]onset rules (e) out. Both (c) and (f) are eliminated because of the violation of IDENT [labial]. *[v] eliminates candidate (g). As for (h), it has a syllable *[fi] that is illicit in Mandarin. Therefore, it is defeated in the competition. Finally, only the
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
67
first candidate is left to be the optimal candidate.
The voiced fricative [v] can also have two outputs in the adaptation. Tableau (25) provides the illustration.
(25) ‘Vendée’ [vɑ̃.de] 旺代[wɑŋ.taj] ≈ 馮戴[foŋ.taj] ‘the place Vendée’
Input: [vɑ̃] *[v] DEP[nas]onset MAX [cont] ID[lab] ID [voi]
☞ a. wɑŋ
☞ b. foŋ
c. voŋ *!
d. xoŋ *! *
e. poŋ *! *
f. moŋ *! *
g. loŋ *!
Both candidates (a) and (b) are possible outputs in the adaptation of [vɑ̃].
Candidate (c) incurs a violation of *[v], and it is ruled out. MAX [continuant] and DEP [nasal] onset eliminates (e) and (f), respectively. Candidates (d) and (g) both incur violations of IDENT [labial]. They are defeated in the competition. Tableau (25) shows that [w] and [f] are possible outputs in the adaptation of [v]. Both [w] and [f]
preserve the features of [labial] and [continuant].
The segment [z] can be substituted by [ɕ], [tɕʰ], [s], and [ts]. When [z] is followed by a high front vowel/glide, it is usually adapted as [ɕi] or [tɕʰi]. Otherwise, [z] is replaced with [s] or [ts]. Tableau (26) presents the candidate selection of [zi].
‧
(26) ‘Moïsi’ [mwa.zi] 莫(瓦)希 [mwɔ.(wa).ɕi] ‘the politician Moïsi’
Input: [zi] [*[-dis] &
The constraint BODYSHARE [high] CV[-back] is used to rule out the unpalatalized candidates (c) and (d). It is defined in (27).
(27) BODYSHARE [high] CV[-back]: Let x ∈ onset, y ∈ nucleus.
Assign one violation mark for every pair of x and y that do not share the feature [high]
when y is [-back].
The conjoined constraint makes (e), (g), and (h) losers. Candidate (f) violates the markedness constraint and it is ruled out. Candidate (i) is defeated because its [strident] feature does not match that of the input. Finally, (a) and (b) become the winner candidates. The reason is that they go through palatalization and conform to Mandarin phonotactics.
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
69
On the other hand, [z] is mostly substituted by [s] or [ts] when it appears before a non-high vowel/glide. The adaptation prefers an anterior and distributed substitute.
Therefore, the conjoined constraint eliminates (c), (d), and (f). Candidate (e) is defeated because of the marked onset segment [z]. Finally, only candidates (a) and (b) are harmonic candidates.
(28) ‘Besançon’ [bə.zɑ̃.sɔ̃] 柏桑松 [pwɔ.sɑŋ.soŋ] ‘the place Besançon’
Input: [zɑ̃] [*[-dis] & ID[(ant)]]fr seg *[z] ID [voi] ID [distr] *[-distr] C ID[ant]
☞ a. sɑŋ * *
☞ b. tsɑŋ * *
c. ʂɑŋ *! * * *
d. tʂɑŋ *! * * *
e. zɑŋ *! *
f. ʐɑŋ *! * *
In the corpus, four substitutes are chosen in the adaptation of [ʒ]. When [ʒ] is followed by the high front vowel [i], it tends to be substituted by the affricate [tɕ].
Otherwise, the segments [ʐ], [tʂ], and [ʂ] are the adaptations. However, the ranking predicts that other than the [tɕi] adaptation, four other substitutes are possible outputs.
Previously, the adaptation of [ʃi] shows more than one optimal output (see tableau 20).
It is suggested that [ʒ] and [ʃ] share similar substitutes when they precede the nucleus [i]. Tableau (29) exemplifies the process of candidate selection.
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
70
(29) ‘Gide’ [ʒid] 紀德 [tɕi.tɤ] ‘the writer Gide’
Input: [ʒid] [*[-dis] & ID[(ant)]]fr seg *[ʒ] MAX[cont] ID[voi] ID [distr]
☞ a. tɕi *
☞? b. ɕi *
☞? c. ʐɨ
☞? d. tʂɨ *
☞? e. ʂɨ *
f. ti *! * * *
g. ʒi *!
As indicated in (29), only candidates (f) and (g) are defeated. Their respective violations of the conjoined constraint and *[ʒ] are fatal. In the end, the first five candidates are possible outputs. Even though candidates (b)-(e) are not dominant in the present corpus, they are chosen as the optimal candidates within the constraint ranking.
When [ʒ] does not precede the high front vowel [i], it can be replaced by [ʐ], [tʂ], or [ʂ]. Tableau (30) illustrates the competition of different candidates.
In (30), the conjoined constraint eliminates (e) and (f). *[ʒ] prohibits the faithful candidate (h) from being the optimal output. As for (g), it has a different [strident] and [back] feature from the input. It is defeated. Therefore, candidates (a), (b), and (c) are the harmonic candidates. Despite its low frequency, candidate (d) remains one of the harmonic candidates in the ranking tableau.
‧
Finally, let us take a look at the adaptation of nasal segments. Since [m] and [n]
are present in Mandarin, they are preserved in loanword adaptation. The segment [ɲ]
is foreign, and it is mostly replaced by the alveolar nasal [n].
Tableau (31) illustrates the constraint ranking. The three undominated constraints
*[ŋ],*[ɲ], and MAX ONSET eliminate (c) (d), and (e), respectively. As for (b), it differs from the input in the feature [labial], and thus it is defeated. Finally, only (a) is the optimal candidate. Violation of IDENT [high] is not fatal.
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
72
(31) ‘Avignon’ [a.vi.ɲɔ̃] 亞維儂 [ja.wej.noŋ] ‘the place Avignon’
Input: [ɲɔ̃] *[ɲ] MAX ONS *[ŋ] ID [lab] ID [hgh]
☞ a. noŋ *
b. moŋ *! *
c. ŋoŋ *!
d. ɲoŋ *!
e. oŋ *!
4.1.2 Adaptation of Shared Segments
This section discusses the adaptations of shared segments between French and Mandarin. The identical onset segments include [p, t, k, s, f, l, m, n]. Nevertheless, only the first four segments allow deviant adaptations.
Let us begin with the substitution of French voiceless stops [p, t, k]. Each of the voiceless onsets has two substitutes, the faithful unaspirated stop [p, t, k], and the aspirated counterpart [ph, th, kh]. Since [p] and [t] have similar adaptations, the substitutes of [p] are provided as an example.
(32) ‘Picardy’ [pi.kaʁ.di] 庇卡底 [pi.kʰa.ti] ≈ 皮卡底 [pʰi.kʰa.ti] ‘the place Picardy’
Input: [pi] MAX ONS ID [asp]
☞ a. pi
☞ b. pʰi *
c. i *!
Tableau (32) shows that both (a) and (b) are optimal because none of them
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
73
violates the constraints above the cut-off line. Since (b) only violates IDENT [aspirated], it is still considered one of the optimal candidates. The difference between (a) and (b) lies in frequency: (a) is used more often than (b). Candidate (c) is ruled out because it violates MAX ONSET.
Besides the substitutes of [k] and [kh], the input segment [k] has another variant
Besides the substitutes of [k] and [kh], the input segment [k] has another variant