The Sound System of French and Mandarin - 以優選理論分析國語中的法語借字

Chapter 1 Introduction

1.2 The Sound System of French and Mandarin

國

立政治大學

‧

N a tio na

l C h engchi U ni ve rs it y

According to Broselow (2006), the phonological constraints in loanword substitutions do not necessarily come from the borrowing language. The reason is that the alternations may be absent or lack obvious data in native phonology. Kenstowicz (2003) indicates that perceptual minimal change is the key factor of loanword substitutions. The present analysis aims to suggest that loanword adaptations do not form a separate grammar from the universal one. Instead, the loanword grammar is based on the constraints that are latent in the native language.

This thesis establishes a French loanword corpus. It contains 719 loanwords. The present analysis uses the framework of Optimality Theory (Prince and Smolensky 1993/2004). The rank ordering model of EVAL (henceforth ROE, Coetzee 2006) is used to account for the variant forms. This thesis focuses on the segmental adaptations of French consonants and nasal vowels, and aims to argue for a unitary constraint system.

The rest of the thesis is organized as follows. Chapter 2 provides a review of the literature, including (a) the theoretical framework provided by Optimality Theory (OT) and the rank ordering model of EVAL, (b) the relevant phonological bases in French and Mandarin, and (c) the previous studies of loanword phonology that relate to this thesis. Chapter 3 establishes a French loanword corpus. Segmental substitutes are coded in the corpus. Chapter 4 is an OT analysis of the segmental substitutes and gives an OT account of the variant adapted forms. Chapter 5 summarizes and concludes the thesis.

1.2 The Sound System of French and Mandarin

The sound inventories of French and Mandarin are listed in tables (2)-(4) and (7).

French is a language of many varieties (e.g. Belgian French, Canadian French, Swiss

‧

國

立政治大學

‧

N a tio na

l C h engchi U ni ve rs it y

French, etc.). I focus on the French spoken in France and choose Parisian French as the standard for transcription.

French has an oral-nasal contrast in the vowel system. Table (2) shows there are fifteen oral vowels in French. According to Fagyal et al. (2006), the vowels [ɪ, ʏ, ʊ]

are used by speakers of Canadian French and French spoken in Louisiana. The [ɑ]

vowel is present only in Belgian French, Swiss French, and French spoken in the rural areas of France. These vowels are marked in parentheses and they are not used for transcription.

(2) French oral vowels (adapted from Fagyal et al. 2006)¹

There are three nasal vowels [ɛ̃, ɔ̃, ɑ̃] in French. Fagyal et al. (2006) indicates that [œ̃] is only used in Quebec French and French spoken in the southern part of France. Therefore, it is not included in the present research.

Mandarin does not have oral-nasal contrast in the vowel system. Lin (2007) suggests that Mandarin has five vowel phonemes [i, y, u, a, ə]. Other vowels occur in specific contexts. Table (3) lists ten vowels that are proposed by Lin. The apical vowel [ɨ] is not included and will be discussed in Chapter 2.

1 The vowels in parentheses are not used in Parisian French, and are therefore excluded in the transcription.

‧

國

立政治大學

‧

N a tio na

l C h engchi U ni ve rs it y

(3) Mandarin vowels (based on Lin 2007)

The consonants of French are illustrated in (4). According to Valdman (1976), French voiceless stops [p, t, k] are unaspirated in syllable-initial position, and medial position with a stressed vowel. The voiceless stops stay unaspirated after [s] segment.

It is possible to have aspirated voiceless stops in syllable-final position. However, some scholars hold that all French voiceless stops are unaspirated (Fagyal et al. 2006).

Such phonetic variations are not the major focus in this study. Therefore, syllable-final stops are coded unaspirated in the corpus. The velar nasal [ŋ] occurs only in English loanwords such as ‘camping’ and ‘parking’. Hence it is not included in the present study. Fagyal et al. (2006) suggests that rhotics for French include the apical [r], uvular trill [R], uvular fricative [ʁ], velar approximant [ɰ], and voiceless uvular fricative [χ]. The detailed classifications show that immediate phonetic environment is pivotal in Fagyal’s analysis. Such phonetic variations are not present in Mandarin. Thus this thesis adopts Rose’s (1999) description of French consonant inventory, and only the uvular fricative [ʁ] is used to represent the rhotic segment in French. There are four affricates [ts, dz, tʃ, dʒ] in French. Only [ts] is observed in the corpus. The remaining affricates are marked in parentheses and excluded from the discussion.

‧

(4) French consonants (based on Fagyal et al. 2006)² bilabial labial-

palatal velar uvular

stop feature specification of French consonants is provided in (5) and (6). The former is a matrix of obstruents, and the latter is a matrix of sonorants. Only the segments that are present in the corpus are included in the matrixes.

2 The segments in parentheses are not included in the present study due to limited usage or allophonic nature.

‧

國

立政治大學

‧

N a tio na

l C h engchi U ni ve rs it y

(5) Feature specifications of French obstruents³

p t k f s ʃ ts --

b d g v z ʒ -- ʁ

[sonorant] - - - -

[continuant] - - - + + + + +

[nasal] - - - -

[lateral] - - - -

[spread glottis] - - - -

[strident] - - - - + + + -

[voice] -/+ -/+ -/+ -/+ -/+ -/+ - +

[labial] √ √

[coronal] √ √ √ √

[anterior] + + - +

[distributed] - - + -

[back] - - + - - + - +

[high] - - + - - - - -

3 The affricate segment [ts] carries [-continuant] and [+continuant] features on the same autosegmental tier. This is indicated by ‘+’, meaning that both features are present. There is a slash between ‘-’and ‘+’

in the [voice] specification of stops and fricatives. This shows that one of the segments is [-voiced] and the other is [+voiced].

‧

國

立政治大學

‧

N a tio na

l C h engchi U ni ve rs it y

(6) Feature specifications of French sonorants

m n ɲ l j w ɥ

[sonorant] + + + + + + +

[continuant] - - - + + + +

[nasal] + + + - - - -

[lateral] - - - + - - -

[labial] √ √ √

[coronal] √ √ √ √

[anterior] + - + -

[distributed] - + - +

[back] - - + - - + -

[high] - - + - + + +

Table (7) lists the consonant inventory in Mandarin Chinese. Many French consonants are absent in Mandarin consonant inventory. As a result, consonant substitution occurs in loanword adaptations. Lin (2007) uses [ɹ] to refer to the syllable-initial retroflex and the rhotacized vowel [əɹ]. However, a voicing contrast in the retroflexed fricatives [ʂ, ʐ] is observed in Taiwan Mandarin (Wan 2000). In order to capture this two-way contrast, the present research distinguishes the onset retroflex fricative [ʐ] from the coda segment [ɹ].

‧

國

立政治大學

‧

N a tio na

l C h engchi U ni ve rs it y

(7) Mandarin consonants (based on Lin 2007) bilabial labial-

dental

dental post- alveolar

alveolo- palatal

palatal velar

stop

+vcd

-vcd p p^h t t^h k k^h

fricative +vcd ʐ⁴

-vcd f s ʂ ɕ x

affricate -vcd

ts ts^h tʂ tʂ^h tɕ tɕ^h

nasal +vcd m n ŋ

approximant ^central w ɥ ɹ j ɥ w

lateral l

The feature specification of each Mandarin consonant is offered in (8) and (9).

The former is a matrix of obstruents, and the latter is a matrix of sonorants. Segment substitutes are better accounted for with such matrixes. Doing so also shows the feature similarity and disparity between Mandarin and French.

4 The table has a slight modification of the consonant chart in Lin (2007). While Lin uses the approximant [ɹ] for both onset and coda segments, we treat [ɹ] as the coda retroflex and [ʐ] as the onset retroflex.

‧

國

立政治大學

‧

N a tio na

l C h engchi U ni ve rs it y

(8) Feature specifications of Mandarin obstruents

p t k f s ʂ ɕ ts tʂ tɕ x

p^h t^h k^h -- -- ʐ -- ts^h tʂ^h tɕ^h --

[sonorant] - - - -

[continuant] - - - + + + + + + + +

[nasal] - - - -

[lateral] - - - -

[spread glottis]

-/+ -/+ -/+ - - - - -/+ -/+ -/+ -

[strident] - - - - + + + + + + -

[voice] - - - -/+ - - - - -

[labial] √ √

[coronal] √ √ √ √ √ √ √

[anterior] + + - - + - -

[distributed] - - - + - + +

[back] - - + - - + + - + + +

[high] - - + - - - + - - + +

‧

國

立政治大學

‧

N a tio na

l C h engchi U ni ve rs it y

(9) Feature specifications of Mandarin sonorants

m n ŋ l ɹ j w ɥ

[sonorant] + + + + + + + +

[continuant] - - - + + + + +

[nasal] + + + - - - - -

[lateral] - - - + - - - -

[labial] √ √ √

[coronal] √ √ √ √

[anterior] + + + -

[distributed] - - - +

[back] - - + - - - + -

[high] - - + - - + + +

‧

國

立政治大學

‧

N a tio na

l C h engchi U ni ve rs it y

Chapter 2 Literature Review

2.1 Optimality Theory

This section introduces the concepts of Optimality Theory (OT). Rank-ordering model of EVAL (ROE), a sub-theory of OT, is also discussed.

2.1.1 Basic Tenets of OT

The theory of OT was proposed by Prince and Smolensky (1993/2004). Instead of using transformational rules, OT suggests a set of violable constraints to account for Universal Grammar (UG). All languages share the same set of constraints. The systematic differences between languages are explained by the language-specific rankings of universal constraints.

In OT, language forms are not generated by the continuous applications of transformation rules. Instead, a given input has an infinite number of possible outputs supplied by GENERATOR (GEN). These diverse outputs (the ‘candidates’) compete with each other to produce an optimal output. All possible outputs are evaluated by the language-specific constraint ranking that is contained in EVALUATOR (EVAL).Through the operation of EVAL, the winner candidate is obtained. The input-output mechanism in OT is not a serial process. All candidates go through GEN at the same time and they are simultaneously evaluated by EVAL.

The condition of becoming an optimal candidate is minimal violation. A candidate becomes an optimal output if it only violates the lower-ranked constraints or has minimum violations. Let’s say there are two constraints Con1 and Con2. The

‧

ranking relation between them is that Con1 ranks higher than Con2. Two candidates (a) and (b) compete with each other for becoming an optimal output. If candidate (a) violates Con1, it will be ruled out immediately. It does not matter whether candidate (a) violates Con2. However, if candidate (b) incurs violations of Con2 but satisfies Con1, it becomes the optimal candidate because of minimal violation.

The major categories of OT constraints are markedness constraints and faithfulness constraints. A markedness constraint requires the well-formedness of an output form. An output is well-formed if it conforms to the restrictions on phonological structures. For example, the markedness constraint *COMPLEX ONSET prohibits more than one segment in the onset position. A faithfulness constraint demands the similarity between the input and the output. Any input-output disparity will result in violations. For instance, MAX-segment forbids segment deletion in the output. A violation occurs if an input segment fails to find its correspondent in the output.

A graphic representation of OT is provided in (1):

(1) The operation of OT

‧

國

立政治大學

‧

N a tio na

l C h engchi U ni ve rs it y

As (1) indicates, OT generates the optimal output through a process of candidate selection. A given input has an infinite number of outputs produced by GENERATOR. These outputs are evaluated by EVALUATION, where a set of universal constraints are ranked. The evaluation is performed in a parallel manner. The optimal output is chosen by having minimal violation of the constraints.

In OT, the ranking of constraints and the evaluation of candidates are demonstrated in a ‘tableau’. A sample of OT tableau is given in (2).

(2) Sample OT tableau

Ranking: Constraint 1>>Constraint 2, Constraint 3>>Constraint 4

/input/ Constraint 1 Constraint 2 Constraint 3 Constraint 4

☞ a. candidate (a) *

b. candidate (b) *!

c. candidate (c) *! *

d. candidate (d) *!

e. candidate (e) *!

An OT tableau has one row for constraints and one column for candidates.

Constraints are listed in the first row. The leftmost constraint ranks the highest, followed by constraints that are ranked lower. The lowest-ranked constraint is placed in the rightmost column. If two constraints are mutually unranked, a dotted line is placed between the two constraint columns. This is shown in the ranking of Constraint 2 and Constraint 3. The candidates are displayed in the leftmost column. The first candidate is the optimal output, indicated by the symbol of a right-pointing hand ‘☞’.

Constraint violations are shown with asterisks. The exclamation marks show the fatal

‧

國

立政治大學

‧

N a tio na

l C h engchi U ni ve rs it y

violations incurred by the loser candidates. Shading accounts for the violations that do not affect the outcome. In tableau (2), candidates (b) and (c) violate Constraint 1, and they are both ruled out. The violation of Constraint 3 in (c) is no longer relevant. Its violation of Constraint 1 already decides the elimination. Candidates (d) and (e) violate Constraint 2 and Constraint 3, respectively. They are defeated in the competition. Candidate (a) incurs a violation of the lowest-ranked Constraint 4. It is selected as the optimal candidate. Its violation of Constraint 4 is shaded in gray because the null violations of the top-three constraints make it the winner.

2.1.2 Rank-Ordering Model of EVAL (ROE)

Coetzee (2006) indicates that language phenomena are more than categorical.

Therefore, generative phonology has difficulty explaining the variant linguistic forms.

Classical OT treats one of the possible outputs as the actual output and leaves the rest undiscussed. Instead of discarding the non-optimal candidates, Coetzee argues that they could be ranked in terms of well-formedness. The ‘more well-formed’

non-optimal candidates are more likely to be the best candidates than the ‘less well-formed’ counterparts.

There are three major tenets in Coetzee’s model. First, all candidates are evaluated by EVAL. Therefore, every constraint could evaluate every candidate.

Second, EVAL contains a ranking hierarchy to account for the well-formedness of the candidate set. Language users have knowledge of this ranking, and their linguistic performance is influenced by the knowledge of the ranking hierarchy. Third, apart from the best candidate, other candidates are within the language users’ access. The position in the ranking hierarchy decides whether a candidate is selected as the output.

A higher-ranked candidate is more likely to be the output. The more well-formed

‧

國

立政治大學

‧

N a tio na

l C h engchi U ni ve rs it y

candidate tends to be selected. It also has higher frequency than the less well-formed candidates.

To account for variations, Coetzee modifies the ranking of the constraint set in classical OT. A critical cut-off line divides the constraint set into two parts: constraints above the cut-off line and constraints below the cut-off line. The cut-off line is placed in a position that meets two requirements. None of the variant forms violates the constraints above the cut-off line. The loser candidates are eliminated due to the violations above the cut-off. Violation above the cut-off line leads to candidate elimination, while violation below the cut-off only determines the frequency of the candidates. An example is given in (3).

(3) A sample tableau with cut-off line

/i/ DEP [-high] DEP [+back] MAX(segment) *σ̃/{i, u}

☞₁ a. ĩ *

☞₂ b. ∅ *

c. ũ *! *

d. other Ṽ *!

In (3), the input /i/ has two variants [ĩ] and null adaptation. The constraints *σ̃/{i, u} and MAX (segment) are placed below the cut-off line so that [ĩ] and vowel deletion are harmonic. The ranking below the cut-off line is MAX (segment)>>*σ̃/{i, u}.

Candidate (a) violates the lower-ranked *σ̃/{i, u} so that it is the first optimal candidate. It has a higher frequency than the second optimal candidate. There are two loser candidates, [ũ] and other non-nigh vowels. The constraints DEP [-high] and DEP [+back] are positioned above the cut-off line so that the two candidates will not be the optimal outputs. In general, fatal constraints are placed above the cut-off line, while non-fatal constraints are violable and located below the cut-off line.

‧

國

立政治大學

‧

N a tio na

l C h engchi U ni ve rs it y

There are three possible situations when the critical cut-off line divides the constraint set. Each of the scenarios is presented schematically in (4).

(4) Tableaux of the critical cut-off line (based on Coetzee 2006) (4a) Variation

(4b) No variation I (4c) No variation II

In (4), there are three possible situations. (4a) shows that more than one candidate is selected as the output. Cand 1 and cand 2 only incur violations of the non-fatal constraints, and variations occur. The first candidate has a higher frequency than the second candidate. (4b) suggests that only the first candidate is harmonic. The reason is that it violates only a single constraint below the cut-off line. (4c) differs from (4b) in the violation below the cut-off. Both candidates in (4c) violate the constraints above the cut-off line. No violation below the cut-off line is observed.

Therefore, the more well-formed candidate of the two is chosen as the optimal output.

‧

國

立政治大學

‧

N a tio na

l C h engchi U ni ve rs it y

2.2 Bases of Mandarin and French Phonology

This section presents the phonological bases in Mandarin and French that are relevant to this research.

2.2.1 Mandarin

In Mandarin, the maximal syllable structure is CGVX (Lin 2007). The X segment in the coda can be a consonant or a vowel. The coda consonant can only be filled by the alveolar nasal [n], the velar nasal [ŋ], or [ɹ] in a rhotacized rime. The rime segments must be identical in the features of [back] and [round]. Table (5) provides some examples of segment combination in Mandarin rime.

(5) Examples of Mandarin rime (based on Lin 2007)

Features Examples

Licit rime [-back] [-round]/ [-back] [-round] [an]

[+back] [-round]/ [+back] [-round] [ɑŋ]

Illicit rime [+back][+round]/[-back] [-round] *[un]

[-back][+round]/[+back] [-round] *[yŋ]

In (5), rime segments sharing the same [back] and [round] features are licit forms, such as [an] and [ɑŋ]. Disparity of the two features will lead to illicit rime combinations, such as [un] and [yŋ].

Mandarin places restrictions on the syllable containing an [f] segment. Wan and Jaeger (2003) list several possible co-occurrences of onset [f] and rimes. Limited segment combinations are observed. The onset [f] can only be followed by [u], [ej], [ən], [oŋ], [ow], [a], [an], and [ɑŋ]. The combination of *[fi], *[fɨ], *[fy], *[fɛ], *[fɔ], and *[fɤ] are disallowed in Mandarin. According to Lin (2007), Mandarin labial

‧

國

立政治大學

‧

N a tio na

l C h engchi U ni ve rs it y

consonants cannot precede a mid-vowel in a CV syllable. That is why *[fɛ], *[fɔ], and

*[fɤ] are illicit. However, [f] can co-occur with [ɔ] if the labio-velar glide [w] appears between [f] and [ɔ], making a legitimate syllable [fwɔ].

Mandarin has restrictions on the syllable structure of coronal fricatives and affricates. Alveolo-palatals [tɕ, tɕʰ, ɕ] can only be followed by a high front vowel or glide. For example, syllable [tɕi] and [tɕy] are possible in Mandarin, while *[tɕo] and

*[tɕe] are not. By contrast, dental affricates/fricative [ts, ts^h, s], post-alveolar affricates/fricative [tʂ, tʂʰ, ʂ, ʐ], and velars [k, k^h, x] cannot appear before a high front vowel or glide. For instance, [ku] and [kən] are licit syllables in Mandarin, yet *[ki]

and *[ky] are not. Such restrictions show the palatalization constraint.

Lin (2007) proposes that only five vowels [i, y, u, a, ə] are phonemic in Mandarin. Other vowels are allophonic because they only occur in specific contexts.

In Lin’s analysis, the three high vowels [i, y, u] are phonemes. The two high vowels [ı]

and [ɩ] have limited distribution. The apical vowel [ı] can only be preceded by the dental affricates [ts]/ [ts^h] and the dental fricative [s]. The other apical vowel [ɩ]

follows only after post-alveolar consonants [tʂ , tʂʰ, ʂ] and [ʐ]. Cheng (1973) proposes a high unrounded central vowel [ɨ] to cover the two apical vowels. He treats [ı] and [ɩ]

as two variants of the high vowel [ɨ]. As phonetic details of [ı] and [ɩ] are not the major focus in this study, we follow Cheng’s classification and use [ɨ] to refer both apical vowels. According to Lin (2007), the mid vowel [ə] has four allophones [ə, e, o, ɤ]. The low vowel [a] has three allophones [a, ɑ, ɛ]. Based on these classifications, the present research uses eleven vowels in the transcription: [i, y, u, ɨ, a, ɑ, ɛ, ə, e, o, ɤ].

2.2.2 French

In French, the onset position and the coda position maximally allow three

‧

國

立政治大學

‧

N a tio na

l C h engchi U ni ve rs it y

segments, such as ‘strident’ [stʁidɑ̃] (shrill) and ‘aster’ [astʁ] (star). The nucleus can be filled by a diphthong or a glide-vowel sequence (Fagyal et al. 2006). Figure (6) presents a simple illustration of the syllable structure in French.

(6) Simple and branching syllable structures in French (based on Fagyal et al. 2006) Syllable

Onset Rime

Nucleus Coda (p) (ʁ) ɛ (i) (t) (ʁ)

paix ‘peace’ C V [pɛ]

est ‘is’ V [ɛ]

ère ‘era’ V C [ɛʁ]

père ‘father’ C V C [pɛʁ]

prêtre ‘priest’ C C V C C [pʁɛtʁ]

être ‘to be’ V C C [ɛtʁ]

The syllabification of French follows the Maximal Onset Principle. For example, given a group of segments [ɛkstʁɛ], the syllabification should be [ɛk.stʁɛ] instead of [ɛks.tʁɛ]. This is because the Maximal Onset Principle tends to maximize the number of onset segments. As mentioned earlier, French has at most three segments in the onset position. Therefore, the onset segment is [stʁ] rather than [tʁ].

French spelling has a set of systematic correspondences between the phonemes and a group of letters. Some relevant phonological rules are listed here. According to Valdman (1976), single letters a, i, u, p, b, d, f, y, m, n, l represent the phonemes [a, i, y, p, b, d, f, j, m, n, l], respectively. Letter combinations ch and gn stand for the phoneme [ʃ] and [ɲ]. The letter c is pronounced as [s] when it occurs before the letters i and e; the pronunciation of c becomes [k] when it is followed by the letters a, o, and u. For instance, the word ‘cinéma’ (‘movie’) is realized as [si.ne.ma], while ‘cou’

‧

國

立政治大學

‧

N a tio na

l C h engchi U ni ve rs it y

(‘neck’) is pronounced as [ku]. Similar patterns can be observed in the spelling of g, which has two pronunciations [ʒ] and [g]. The word ‘gifle’ (‘cheek’) has a pronunciation of [ʒifl], while ‘gamin’ (‘youngster’) is realized as [ga.mẽ]. The glide [j]

has five spellings, i, il, ile, ill(e),and y. Examples include ‘iode’ (‘iodine’) [jɔd], ‘seuil’

(‘threshold’) [sœj], ‘sillon’ (‘furrow’) [sijõ], ‘fille’ (‘daughter’) [fij], and ‘voyons’

(‘see-first person plural’) [vwa.jɔ̃].

2.3 Previous Studies on Loanwords in Mandarin

This section discusses some studies on Mandarin loanwords, mostly the segment substitutions of English loanwords in Mandarin.

Miao (2005) investigates the Mandarin loanwords borrowed from English, German, and Italian. The focus is on consonant substitutions. Miao suggests that the foreign forms need to fulfill two tasks when they enter Mandarin phonology. First, the loans need to find the closest phoneme in the borrowing language. Second, the loans need to adjust the phonological structure to become licit forms in the native phonology. Miao adopts Steriade’s (2002) Perceptibility-map (P-map) hypothesis. She suggests that the perceptual salience of the segment features decides if the segment is preserved. Manner features are perceptibly more salient, so they tend to be retained.

Place features and [voice]/ [spread glottis] features tend to be sacrificed because they are less distinct. Miao proposes the ranking hierarchy in terms of feature identity between the input and the output. IDENT ([manner]) ranks higher than IDENT ([place]), which dominates IDENT ([voice]) and IDENT ([aspirated]). Obstruent adaptations tend to preserve manner of articulations. Most obstruent loans are faithfully retained, and deviations in voicing/aspiration are allowed. Nasals tend to be faithfully preserved. Change of place feature is sometimes observed in the coda

‧

國

立政治大學

‧

N a tio na

l C h engchi U ni ve rs it y

position. As for sonorants and glides, they tend to be faithfully adapted. Deviations of features such as [lateral] and [sonorant] are found. In sum, Miao suggests that perceptual similarity is the key factor of phoneme substitutions.

Wu (2008) discusses the phonological adaptations of English loanwords in Mandarin. The focus is on consonant substitutions. The analysis shows that the adaptation patterns are influenced by the positions of consonant segments.

Consonants in word-initial and word-medial positions tend to be retained or adapted.

On the contrary, consonants in word-final position are mostly deleted. In terms of perceptual saliency, onset consonants are more prominent than the coda counterparts.

Therefore, onset segments tend to be preserved. Some salient segments like [s] are bound to be preserved in both onset and coda positions. Similar to Miao’s (2005) analysis, Wu stresses the perceptual factors in loanword adaptation.

2.4 Research on French Loanwords

This section reviews some studies of French loanwords in different languages, including German, Japanese, and Fon.

Itô and Mester (2001) propose a type of stratum-specific faithfulness constraints to account for French loanwords in German. They argue that markedness constraints are fixed in the ranking hierarchy. By putting faithfulness constraints in different places of the ranking hierarchy, sublexicons (i.e. variations) appear. These sublexicons differ in the degree of nativization. For example, both [ʒ] and nasal vowels are absent in German inventory. Long lax vowels and coda [ʀ] segment are also non-native in German. When the French word ‘jongleur’ [ʒɔ̃glœ:ʀ] (juggler) is

在文檔中以優選理論分析國語中的法語借字 - 政大學術集成 (頁 14-0)