Chapter 4 Optimality Theory Analysis
4.2 Adaptation of Coda Consonants
4.2 Adaptation of Coda Consonants
This section approaches the substitution of French coda consonants. The coda segments include [p, t, k, b, d, g, s, f, ʃ, z, v, ʒ, ts, m, n, ɲ, l, ʁ]. As shown in the previous chapter, the coda consonants tend to be preserved in loanword adaptation, except for the consonant [ʁ]. According to Lin (2007), Mandarin allows only [n] and [ŋ] to fill the coda position. However, the coda position could also have [ɹ] in a rhotacized rime. Hence, the markedness constraint CODA CONDITION is relevant to the adaptation of coda consonants:
(36) CODA CONDITION:
Assign one violation mark for every coda segment that is not an alveolar nasal [n],
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
77
velar nasal [ŋ], or a retroflex [ɹ].
The rest of this section provides an OT analysis of French coda adaptation. Coda segments occupying a greater portion of the data will be presented first, followed by the segments with fewer occurrences.
First, let us examine the lateral codas. The codas [l] and [ʁ] have similar adaptations. The segment [l] has a dominant substitute of [əɹ] (46.21%), a null adaptation (28.03%), and a replacement of [lɤ] (23.48%). Similarly, [ʁ] can be deleted (58.68%) or substituted by [əɹ] (34.38%). Such a pattern shows that the coda segment [l] and [ʁ] can be deleted. Accordingly, the constraint MAX CODA-[l]/[ʁ] is proposed.
It is positioned below the cut-off.
(37) MAX CODA-[l]/[ʁ]:
Assign one violation mark for every input coda segment [l] and [ʁ] that does not have a correspondence in the output.
The adaptation from [l] to [əɹ] and [lɤ] as well as the substitute from [ʁ] to [əɹ]
suggests that vowel insertion is a strategy to compensate for the absence of [l]/[ʁ]
codas in Mandarin. Therefore, violation of DEP ([-consonantal]) is plausible. The constraint should be placed below the cut-off line. Constraints DEP ([-consonantal]) and MAX CODA-[l]/[ʁ] have identical rankings, since the illicit coda segments [l] and [ʁ] can be deleted or adapted as a CV/VC syllable. The definition of DEP ([-consonantal]) is offered in (38), and tableaux (39)-(41) explicate the candidate selection.
‧
Assign one violation mark for every output vowel/glide segment that does not have a correspondence in the input.
(39) ‘Malmaison’ [mal.mɛ.zɔ̃] 馬爾梅松 [ma.əɹ.mej.soŋ] ≈ 馬勒梅松 [ma.lɤ.mej.soŋ] ≈ 莫梅森 [mwɔ.mej.sən] ‘the place Malmaison’
Input: [mal] CODA phonotactics and several features are crucial in candidate selection. Candidate (d) has a forbidden coda [l], making it a loser. From (e) to (g) different constraints are decisive. Violation of DEP [nasal]onset eliminates candidate (e). MAX [coronal] [-nasal] C
rules (f) out. IDENT [strident] hinders (g) from being the winner. Candidates (a)-(c) only incur violations below the cut-off line. They are chosen as possible variants in response to the illicit coda [l].
Unlike the coda segment [l] that has three variant adaptations, the substitution of [ʁ] allows only two possibilities. [ʁ] is either deleted or replaced by the [əɹ] syllable.
‧
However, another plausible candidate [pej.əɹ] (candidate (b)) is eliminated.
The situation here is similar to the [ʁ] adaptation in the onset position. As mentioned previously, [ʁ] is likely to possess two inputs. One is the French input [ʁ], and the other is the English input [ɹ]. Tableau (40) carries the input of [ʁ], allowing only the null adaptation (candidate (a)) and the replacement of [xɤ] (candidate (e)). demonstrates that both null adaptation (candidate (a)) and [əɹ] substitution (candidate
‧
eliminated from the candidate set.(41) ‘Quimper’[kɛ̃.pɛɹ] 坎培[kʰan.pʰej] ≈ 坎佩爾[kʰan.pʰej.əɹ] ‘the place Quimper’ (English input)
Input: [pɛɹ] CODA
In the case of [n] adaptation, there are three possible substitutes: [Vn], [Vŋ], and [na]. When the input contains at least two syllables, the substitutes tend to be [Vn]
and [Vŋ]. If the input is monosyllabic, the adaptation is usually [na], with vowel epenthesis rather than segment deletion. The adaptation of multisyllabic input is provided in (42).The candidate selection of monosyllabic input is presented in (49).
As tableau (42) shows, only the first two candidates are the winners. Candidates (c)-(g) are eliminated. Candidate (c) deletes more than one mora from the input. It violates a new constraint MAXμ2. Candidate (d) does not preserve the nasal feature in the coda position. It incurs a violation of MAX [nasal] coda and it is eliminated.
Syllable [tin] in (e) is ruled out. The reason is that the onset and the coda cannot both
‧
be [+anterior]. This is captured in the constraint *SHARE [+anterior] onset& coda. Both (f) and (g) violate CODA CONDITION. They are excluded from being the winner.
As the input [ɡi.jɔ.tin] is trisyllabic, Mandarin speakers tend to adapt the nasal coda rather than inserting a vowel after the nasal segment. More about this will be discussed in 4.4. Candidates (a) and (b) conform to Mandarin phonotactics. They are chosen as the optimal candidates. Tableau is provided in (42). Relevant constraints are defined in (43)-(45). even though the input is [iŋ], speakers of Taiwan Mandarin could pronounce it as [in], [iŋ], [jiəŋ], or [jiN] (nasal approximant). However, Lin also notes that the syllable [tin] is one of the accidental gaps in Mandarin syllable types.
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
82
(44) MAX [nasal] coda:
Assign one violation mark for every input [nasal] feature in the coda segment that does not have a correspondence in the output.
(45) *SHARE [+anterior] onset& coda: Let x ∈ onset, and y ∈ coda.
Assign one violation mark for every pair of x and y that is linked to the same token of [+anterior].
The adaptation of monosyllabic input is different. Tableau (49) indicates that the input [sεn] has a disyllabic adaptation. Besides the violation of MINIMAL WORD, candidates (b)-(d) also violate other constraints. The faithful output [sεn] is illicit, since in Mandarin the rime [ɛn] can only appear after the glide [j]. Hence, the constraint *εN is active to rule out candidate (b). Lin (2007) points out that Mandarin rime segments must have features identical in backness. The rime in [sεŋ] differs in [back] feature. Therefore, candidate (c) is ruled out by IDENT [back]-rime. Candidate (d) has violations of MINIMAL WORD, CODA CONDITION, and MAX [nasal] coda. It is considered a loser candidate. As for (e) and (f), they differ from the input in [nasal]
and [labial] features, respectively. Hence, they are ruled out of the competition.
Eventually, only (a) is the optimal candidate. The constraints *εN, IDENT [back]-rime, and MINIMAL WORD are defined in (46)-(48). The tableau is given in (49).
(46) *εN:
Assign one violation mark for every [εn] syllable that is not preceded by the glide [j].
(47) IDENT [back]-rime:
Assign one violation mark for every rime that has segments of different [back]
features.
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
83
(48) MINIMAL WORD:
Assign one violation mark for every word that does not have at least two syllables.
(49) ‘La Seine’ [la.sεn] 塞納(河) [saj. na] ([xɤ]) ‘the river Seine’
Input: [sεn] MINI WRD
CODA COND
ID [back]- rime
*εN MAX [nasal] coda
ID [lab]
DEP ([-cons])
☞ a. saj.na *
b. sεn *(!) *(!)
c. sεŋ *(!) *(!)
d. sεt *(!) *(!) *(!)
e. saj.ta *! *
f. saj.ma *! *
Similar to the adaptation of onset segments, the coda segment [ɲ] is replaced by [n]. Tableau (50) provides the illustration. The new constraint MAX [high] is defined in (51).
‧
(50) ‘Gascogne’ [ɡas.kɔɲ] 加斯科涅 [tɕja.sɨ.kʰɤ.njɛ] ‘the place Gascogne’
Input: [kɔɲ] *[ɲ] CODA violating IDENT [back]-rime is fatal. Candidate (f) and (g) are ruled out because they have different [labial] and [nasal] features from the input, respectively. Hence, only (a) and (b) are possible outputs.
Next, let us consider the adaptation of voiceless fricatives. The segment [s] has a dominant substitute of [sɨ] (80.33%). Other adaptations such as [ʂ], [ɕ], and deletion are the minority forms. Miao (2005) points out that among the obstruent coda segments, fricative codas are most distinctive and are not subject to deletion. The new
‧
Assign one violation mark for every coda [s] segment that is not adapted as [sɨ].
Since coda [s] has only one adaptation form, the constraint CODA [s]
faithfulness stays undominated. Tableau (52) displays the candidate selection process.
Only the substitute [sɨ] is a licit adaptation. Candidates (b)-(e) violate CODA [s]
faithfulness, and hence the elimination results. Even though candidate (f) is a faithful adaptation, it is ruled out because of the violation of MINIMAL WORD and CODA CONDITION.
The adaptation of [f] coda shows that vowel insertion is the sole strategy.
Tableau (55) indicates that candidate (a) is the only optimal output. Candidates (b)-(e) Input:
‧
violate the new constraint IDENT [f]. All of them are eliminated. The faithful output (f) has an illicit coda segment; it also violates MINIMAL WORD. As a result, candidate (f) is defeated by candidate (a). The new constraint is defined in (54). Ranking is provided in (55).
(54) IDENT [f]:
Assign one violation mark for every input segment [f] that is not [f] in the output.
(55) ‘Craffe’[kʁaf] 克哈夫[kʰɤ.xa.fu] ≈ 卡夫[kʰa.fu] ≈ 卡拉弗[kʰa.la.fu] ‘the
Tableau (56) presents the constraints and ranking. It shows that candidates (a)-(d) and (f) are the optimal outputs. Even though candidate (d) has a low frequency, it is still chosen as the optimal candidate. Candidates (e), (g), and (h) violate the conjoined constraint, and they are eliminated. MAX [back] disallows candidate (i) from being
‧
consonant markedness constraint.(56) ‘Avranches’ [a.vʁɑ̃ʃ] 阿弗宏許[a.fu.xoŋ.ɕy] ≈ 阿夫朗什[a.fu.lɑŋ.ʂɤ]
Compared with the adaptation of onset [z] segment, the substitutes of [z] codas are relatively few in number. While the onset [z] segment allows four adaptation forms [s], [ɕ], [ts], and [tɕʰ], the coda [z] has only two possibilities [s] and [ts]. This phenomenon may result from the palatalization process. The substitutes [ɕ] and [tɕʰ]
are mostly adapted from [zi], yielding the palatalized syllables [ɕi] and [tɕʰi]. In the case of [z] coda adaptation, palatalization is unnecessary. Therefore, the coda segment [z] is mostly replaced by [s] or [ts]. Tableau (57) illustrates the candidate selection
‧
In (57), candidates other than (a) and (b) are eliminated. For (c), it has an illicit coda segment, and it is defeated. Candidates (d)-(f) violate the conjoined constraint, making them losers. Finally, syllables [sɨ] and [tsɨ] survive due to null violations above the cut-off line.
Although the onset [v] has a dominant substitute of [w], the adaptation of [f] is prevailing in the case of coda [v]. When coda [v] is preceded by [i], the adaptation tends to be [wej] instead of [fu]. The present corpus has limited occurrences of [iv]
syllables. Other syllable types appear more often. Such distribution explains why the substitute of [fu] is dominant in the coda adaptation.
In (58), the syllable [fu] (candidate (a)) is the winner candidate. Even though [wu]
(candidate (b)) has very low frequency, it is still one of the optimal candidates in the tableau. The illicit coda segment in (c) precludes it from being the winner. A violation
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
89
of IDENT [labial] disallows candidate (d) to be optimal. Candidates (e)-(g) violate MAX [continuant].They are defeated by candidates (a) and (b).
(58) ‘Nièvre’ [njɛvʁ] 涅夫勒 [njɛ.fu.lɤ] ‘the place Nièvre’
Input: [njɛvʁ] *[v] MAX[cont] ID[lab] ID[voi] DEP([-cons])
☞ a. fu.lɤ * *
☞? b. wu.lɤ *
c. ɛvʁ *!
d. xu.lɤ *! * *
e. pu.lɤ *! * *
f. mu.lɤ *! *
g. ku.lɤ *(!) *(!) * *
The segment [ʒ] has four variants [tɕ, ʐ, tʂ, ʂ] in onset adaptation. However, it only allows [ʐ] and [tɕ] in coda adaptation. We can still observe the preference of [+continuant], [-anterior] and [+coronal] features in the adaptation of [ʒ]. Details of candidate selection are provided in tableau (59).
In (59), [ʐ], [tɕ], [tʂ], and [ɕ] (candidates (a)-(d)) are the optimal outputs. The first two candidates are the dominant outputs in the present corpus. The remaining candidates are considered possible outputs in the current ranking set. Candidates (e) and (f) violate the conjoined constraint. Therefore, they are defeated. For (g), it has different [strident] and [back] features from the input, making it a loser. The fatal violation of *[ʒ] eliminates (h). Only the first four candidates are possible outputs among the candidate competition.
‧
Next, let us consider the adaptations of coda stop segments. The segments [t] and [k] share similar adapted forms. They are either replaced by aspirated counterparts or deleted, depending on the number of syllables. There are few occurrences of [p]
adaptation. Generalizations are unlikely. Therefore, the substitution patterns of [t] and [k] are discussed. The former segment serves as an example.
Tableau (60) gives an example of coda [t] adaptation in a monosyllable.
Candidate (a) does not violate any constraints above the cut-off. It is the optimal candidate. The illicit coda segment in (b) precludes it from the winner position.
Besides, it also violates MINIMAL WORD. Candidate (c) does not conform to the MINIMAL WORD constraint, making it a loser. As for (d), it has a different [strident]
feature from the input. Therefore it is ruled out. Candidate (e) has a fatal violation of MAX [coronal] [-nasal] C. It is defeated, too.
‧
preference for voiceless unaspirated counterparts. Since the adaptation pattern is similar for the three coda segments, [d] is taken as an example.(61) ‘Gide’[ʒid] 紀德 [tɕi.tɤ] ‘the writer Gide’
As indicated in (61), the monosyllabic input [ʒid] is adapted as a disyllabic word [tɕi.tɤ]. The remaining candidates have violations above the cut-off. Consequently,
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
92
elimination occurs. Candidate (b) has a fatal violation of MINIMAL WORD and CODA CONDITION. Candidate (c) is monosyllabic, violating the MINIMAL WORD constraint. DEP [nasal]onset rules out candidate (d). As for (e), it has an extra [continuant] feature. This violates the new constraint DEP [continuant]. Finally, (f) is out because it deletes the [coronal] feature, leading to the violation of MAX [coronal]
[-nasal] C. Candidate (a) is the optimal output. The new constraint is defined in (62).
(62) DEP [continuant]:
Assign one violation mark for every output [continuant] feature that does not have a correspondent in the input.
Lastly, let us discuss the adaptation of the affricate [ts]. There are few instances of [ts]. Only four occurrences are observed. Generalizations of coda adaptation are therefore unlikely to be derived. Accordingly, no substitution pattern could be offered.
This limitation may be resolved when more occurrences of [ts] are available.