The Adaptation of French Consonant Clusters in Vietnamese Phonology: An OT Account

1. Introduction

Loanword adaption in Vietnamese has been a topic of interest for different scholars during the last few decades. Most of the previous works on Vietnamese lexical borrowing focused on historical, etymological, sociolinguistic, and phonological descriptions of language borrowing from Chinese, French, and English in general. Among the popular works in Vietnam, K. Nguyen (2007) provides a systematic investigation of loanwords in Vietnamese. He presents some theoretical issues related to language contact, word borrowing and gives examples of Vietnamese loanwords. He proposes three main foreign language sources i.e., Chinese, French, and English. His study is a great contribution to the field of loanwords in Vietnamese, but the discussions are limited to the sociolinguistic situation and to the descriptive phonological aspects and illustrations of loanwords from these three main source languages in various fields. G. Nguyen (1998) in “Tu Vung Hoc Tieng Viet” (Vietnamese Lexicology) also touches upon lexical borrowing in the Vietnamese language. He gives the classification of loanwords in Vietnamese language and the assimilation in terms of phonology of these foreign origin words. Tran (2009) includes loanwords from different source languages with their etymology in his dictionary.

Section1 of this paper deals with Vietnamese loan word phonology with reference to French words and its previous literature. Section 2 is devoted to Vietnamese and French phonemic inventories and phonotactic patterns. Section 3 discusses the Phonological alternations, loan words and repair strategies with special reference to French loans in Vietnamese. An OT account of French loans in Vietnamese is explained in section 4 of this paper which is divided in to two subsections 4.1 and 4.2 with detailed representation of alternations in coda and onset clusters of the loan words in constraints based approach. Section 5 summarizes the main findings and implications of the paper.

2. Vietnamese Loan Word Phonology with Reference to French Words and its Previous Literature

The eighty years of French colonial rule resulted in the enrichment of the Vietnamese language by loanwords borrowed from French. Some scholars have studied language contact between Vietnamese and French. Barker (1969) provides the first paper on French loanwords in Vietnamese, taking account of the theoretical literature in the field of research. He examines the phonemic inventory of French, the donor language and that of the host language though his discussions are limited to a general description of phoneme adjustments. Vuong (1991) investigates words originating from French in Vietnamese. Huynh (2010) provides some statistics of French loanwords used among different Vietnamese age groups in recent years. A recent paper of Kang, Pham & Storme (2014) examines French loanwords in Vietnamese with the focus on adaptation of vowels into the native language. The paper provides two case studies of French vowel integration into Vietnamese. The study supports the role of knowledge of the source language phonology that adapters bring to adaptation and provides an empirical contribution to the field of loan phonology with a detailed description of French-Vietnamese vowel adaptation. Recently, the use of English words in Vietnamese means of public communication and newspapers has attracted attention among some researchers in Vietnam. To sum up, it is clear that there is no detailed, comprehensive study on the loanword phonology of Vietnamese done in the framework of Optimality Theory (Prince & Smolensky 1993). In other words, it is obvious that most of previous studies in Vietnamese loanwords are descriptive in nature and only list the changes that the sound segments or syllables undergo. The significant studies on loanwords in Vietnamese (Barker 1969, Huynh 2010, Kang, Pham & Storme 2014) have proposed rules of deletion/epenthesis or feature change to account for their adaptation into Vietnamese. For example, inserting the vowel /i/ to break up the onset cluster /sl/ in the word /slip/ > [silip] while deleting a consonant in a coda cluster as in /gɒlf/ which becomes [ɣon]. The problem with this approach is that it does not explain why the language prefers deletion in one case but epenthesis in another case to account for French syllable initial or final consonant clusters. Likewise, a consonant can be unparsed in some contexts but be parsed in other contexts. For example, when /film/ is adapted as [fim] the segment /l/ is deleted, however, /l/ is preserved and modified as /n/ in the word /kalk/ > [kan]. Such problems mentioned above can be reasonably solved in the OT framework in terms of constraints. The surface form that arises is the outcome of the constraint interaction and satisfaction. No specific rules are needed in the OT framework since “the candidate analyses, evaluated by the constraint hierarchy, are admitted by very general considerations of structural well-formedness” (McCarthy & Prince 1993: 5). No derivational steps are involved as the computation of the candidate set happens in parallel. Moreover, the OT framework mainly focuses on the surface or output form. Whatever be the underlying representation of the foreign loanwords (in French), they ultimately have to conform to the surface constraints of the recipient language (i.e., Vietnamese). The foreign forms will come out looking like the surface forms of the recipient language (Vietnamese). By inspecting the surface forms, the constraints and their ranking in a particular language can be worked out.

This paper aims to shed light on the loanword adaptation phenomenon, with special reference to French loans in Vietnamese, which will be a contribution to an understanding of the internalized phonology of the Vietnamese language. The research endeavors to provide a basis of description of the language for any further research. The scope of the study is limited to the source language, French and the Vietnamese we examine is standard Vietnamese, i.e., the Hanoi dialect. The following research questions will guide our study: Firstly, how are foreign consonant phonemes adapted into Vietnamese? Secondly, how do Vietnamese speakers nativize foreign syllable structures that are incompatible with their native phonology?

For the purpose of the present study, the main source of data is from print materials such as the paper by Barker (1969), the list of French loanwords in the thesis of Huynh (2010), the book on loanwords of K. Nguyen (2007), dictionary of loanwords in Modern Vietnamese by Tran (2007), French-Vietnamese dictionary and Vietnamese-French dictionary. The selected list of words was checked in pronunciation by some Vietnamese who are speaking Hanoi dialect. They are mostly lexical items introduced into Vietnamese after the French came to Vietnam. Lexique 3.80 (New et al. 2001), which is available at <http://www.lexique.org> is used as a source of French phonological transcription. The phonological transcription of the Vietnamese French loans is provided by Kirby with the aid of vPhon, a Python script that converts Vietnamese scripts to IPA transcriptions (Kirby 2008) and from the SEAlang Library Vietnamese Lexicography, which is available at <http://sealang.net>.

3. Vietnamese and French Phonemic Inventories and Phonotactic Patterns

In this section we are going to focus on the Vietnamese and French phonemic inventories and phonotactic rules governing onset and coda formations now.

4. Phonological Alternations, Loan Words, and Repair Strategies with Special Reference to French Loans in Vietnamese

In this section, we will concentrate on the phonological repair strategies the Vietnamese phonology adapts while borrowing the French consonant clusters at the syllabic level. In other words, we are going to highlight the phonotactic adjustments that the recipient language (Vietnamese) undergoes to conform to its syllable structure rules. For instance, the CC clusters of French may undergo either vowel epenthesis or deletion when entering Vietnamese. These two repair strategies are discussed below:

5. An OT Account of French Loans in Vietnamese Phonology

In this section, we are going to look at the coda and onset clusters of French loans in Vietnamese within the OT model of constraint rankings although our analysis is confined to the repair strategies that Vietnamese loan phonology exhibits at the syllabic level. The prime focus of our interpretation in our study is the nativization of foreign consonant structures which are not allowed in the native language. So far this issue has not been perceived from the perspective of OT which can explain universal typology across languages in the world on an explicit canvas with suitable rational justification. By using the constraint based approach within OT we will show here how a uniform set of constraints can account for phonological adaptation at the syllabic and segmental level. Loan word adaptation and constraint based approach also bear close affinity with strength asymmetries evident in the patterning of segmental speech sounds not only in a specific language but also in languages across the world. Dutta (2012) claims that phonological strength relations can play a crucial role in the patterning of phonological features not only in the areas pertaining to language acquisition, pitch accent patterns and tonal phenomenon but also to loan word adaptation. The constraint ranking approach in OT can better represent this phenomenon in a comprehensive and clear fashion.

Here we have noticed that the French loanwords strictly conform to the Vietnamese syllable structure patterns which imply that the markedness constraints of Vietnamese are higher ranked and they act as a sieve through which the French loanwords pass.

To explain the native syllable inventory of Vietnamese, we can use the blanket constraint OK-, introduced by Yip (1993). She introduced this constraint as a package constraint to deal with four different constraints at the level of the syllable in Cantonese. The first one is a constraint that deals with onsets. It is a combination of a constraint that blocks onset deletion (the Onset Principle formulated by Itô 1989) and another that avoids consonant clusters in the onset position i.e., * [CC. The second constraint of the package that she mentions is NUCLEUS, which means nuclei must always be vocalic. Word-finally, however, in Cantonese, nuclei may be any [+son]. The third one is a constraint which restricts the set of possible codas to /p, t, k, m, n, ŋ, w, j/. The fourth constraint is MINSYLL, which means the minimal syllable in Cantonese is bimoraic i.e., [µµ]_MINσ. Likewise, for Vietnamese, we propose the OK- constraint that is also a package of constraints related to do with syllable structure. (i) Firstly, Vietnamese does not allow either a complex onset or a complex coda. (ii) Secondly, the nucleus is always vocalic, as Vietnamese does not allow syllabic consonants. (iii) Thirdly, Vietnamese restricts the set of possible codas (like Cantonese) to /p, t, k, m, n, ŋ, w, j/ only. In addition to this, Vietnamese also has restrictions on what can occur in the onset position. Regarding onset segments, Vietnamese only allows loanwords with sounds which are present in the native language (except /p/, which is accepted in some French loanwords). This means that the French segments like /g/, /ʁ/, /ʃ/, /ʒ/, /j/, and /ɥ/ are not allowed in loanwords in onset position. However, the constraint MINSYLL in the package constraint of OK- in Yip (1993) is not observed in Vietnamese as it can have mono-moraic words like [ba] ‘three’, [ve] ‘to return’, or [mɛ] ‘mother’, i.e., CV syllables.

Postulating the ‘OK-’ constraint implies that Vietnamese has to introduce some repair mechanisms to accommodate French words that violate any of these constraints. Onset and coda clusters are not allowed by the phonotactic rules of Vietnamese phonology. Particularly, syllabic consonants can’t find a place in Vietnamese and certain disallowed segments in Vietnamese are either deleted or altered.

The next important feature that we have observed here in this data set is that a particular segment is subject to deletion in either an onset cluster or a coda cluster of the French loanwords and thereby it conforms to the phonotactic patterns governing syllabification of the native Vietnamese. The tables displaying the changes are listed below for the convenience of the readers:

5.1. Alternations in Coda Clusters and Constraint Based Approach in OT

In the first list of obstruent + obstruent coda clusters, we find that the segment that is allowed according to Vietnamese phonology remains and the other is deleted (the third constraint of the OK- constraint). When one of them is licit and the other is illicit (e.g., calque), we observe retention of both segments. The illicit segment /l/ is modified to /n/ and the /k/ is retained. The complex coda is avoided by epenthesizing a vowel at the end of the word.

When both the segments are illicit in the coda position and when both are obstruents, it is the first that deletes and the second consonant undergoes phonological changes. What interests us is the list of sonorant + obstruent clusters, where both are disallowed. In this case, it is the sonorant that survives (the /l/ is changed to /n/) when it precedes the obstruent. In cases, of the reverse order, where there is a violation of the Sonority Sequencing Principle (SSP), it is the obstruent that survives.

We introduce two new constraints to account for these facts; namely, PARSE-SALIENT/ MIMIC-SALIENT and ALIGN-R.

In Vietnamese, we believe the threshold of saliency, is based on the Sonority Sequencing Principle developed by Clements (1990), Dost (2004), and others. The sonority scale shows that the obstruents are the least sonorous elements, nasals are more sonorous than obstruents, liquids are more sonorous than nasals but less than glides and vowels are the most sonorous elements.

From the data, it is clear that obstruent sounds can freely delete in Vietnamese while all other sounds (namely, sonorants) are above the salient threshold and have to be parsed. Jacobs & Gussenhoven (2000) objected to the use of salience as a feature in accounting for the application or non-application of a rule as the notion of salience is language specific and not a universal feature. Hence, to avoid using the notion of salience (which is language specific), we could instead rename the constraint PARSE SALIENCE as a more general constraint MAX-SON. The sonorant survives because of the constraint, namely MAX-SON.

In the coda position, we note that deletion is a more favorable strategy than epenthesis (epenthesis only happens to the word “calque”). When we have two illicit segments in the coda position, it is the first member of the cluster that is deleted.

This shows us that we need the constraint ALIGN-R, which is a high ranked one in Vietnamese.

In order to keep the feature in the output similar to the input, we also require the constraint IDENT-F shown in (5).

In addition, we note that two more faithfulness constraints, which are active in Vietnamese loanword phonology, are DEP-IO and MAX-IO.

The set of constraints analyzed earlier are presented once again below for the convenience of the readers from (8) to (11):

Now we are going to show the OT constraint rankings exhibited by the coda clusters in Vietnamese. The OT tableaus can better illustrate the ordering of these constraints which can display the patterns of obstruent + obstruent coda cluster (one licit and another illicit) in Vietnamese loan phonology.

The first process is one in which C₁ is a licit obstruent and C₂ is an illicit obstruent. The licit C₁ remains as coda of the original syllable and the illicit C₂ undergoes deletion. For this case, we consider the word /elips/, which becomes [elip]. To account for this phonological change, we require the above constraints. The ordering of the constraints is summarized in (12).

The tableau in Table 16 illustrates the explanation of the adaptation of the word /elips/, which changes to /elip/.

In /elips/ > /elip/, the ranking of DEP-IO >> ALIGN-R >> MAX-IO leads to the choice of candidate (b), in which only the segment /s/ in the coda cluster is deleted. Candidate (c) violates the highest constraint OK-, in particular, the coda licenses only /p, t, k, m, n, ŋ, w, j/ and candidate (a) also fails due to its violation of the highest constraint OK-, in particular, *Complex-Coda.

To simplify the obstruent + obstruent cluster with both illicit segments, we found that C₁ is deleted and C₂ is modified. This take place, for instance, in the adaptation of /kɔʁd/ > /kɔt/, where the first member of the cluster /ʁd/ is deleted and the second one is modified. In this case, we also require the same constraints such as OK-, MAX-SON, DEP-IO, IDENT-F, ALIGN-R, and MAX-IO. As shown in Table 17 below, the highest constraint OK- rules out most of the candidates, which are candidates (a), (b), (c), and (f). Particularly, candidate (a) is ruled out due to the fatal violation of OK- (i.e., *Complex Coda). Both Candidates (b) and (c) also violate the OK- (i.e., Coda License). Thus, candidate (d), on violating only one feature (voice) becomes the optimal one meanwhile candidate (e), violating two features (voice and place) is banned. This shows us that IDENT-F is a gradient constraint.

To explain the adaptation of sonorant + obstruent coda clusters, we consider the foreign word, /valv/, which becomes /van/ in Vietnamese. There is no vowel insertion betweenor after /lv/. The obstruent is deleted and the sonorant is changed to /n/, which has got the closer correspondence with the input segment. So, we will be using the constraints like OK-, MAX-SON, DEP-IO, IDENT-F, ALIGN-R, and MAX-IO to account for the phonological alterations. The adaptation of /valv/ > /van/ will be the explained by OT tableau in Table 18. As shown in Table 18 below, only candidates (d) and (e) obey the highest ranked constraint OK-σ. However, Candidate (e) is not allowed as it violates a high ranked constraint IDENT-F twice (it has two different features i.e., place and manner). Therefore, candidate (d) wins and becomes the optimal candidate as it violates only one feature (place).

To consider the words with illicit obstruent and sonorant segments in coda position, we take the example of the word /kabl/ which changes to /kap/. As mentioned earlier, the input consonant cluster violates the Sonority Sequencing Principle. Let us see if the constraints that we have posited so far can account for the output /kap/. As shown in Table 19, if we have only the constraints mentioned above, the optimal candidate could hypothetically be a form like /kan/ instead of the attested form /kap/.

In recent years, a great deal of attention has been devoted to the implementation of Mora Theory in an Optimality-theoretic framework (e.g., Sherer 1994, Zec 1995, Sprouse 1996, Broselow et al. 1997). When we look at the data given above, we note that the syllabic consonant /l/ which is the nucleus of the second syllable (and which would thus be assigned a mora) is not permitted to become the coda (or margin of a syllable in Vietnamese, in which case, it would lose its mora as coda consonants do not add to the weight of a syllable in Vietnamese.) There is a tension in the language between retaining the mora (as it is in the input) and obeying the constraints of Vietnamese.

It is resolved by deleting the segment itself. Hence, to account for this fact, we need to take recourse to another constraint i.e., IDENT-IO (µ) as shown in (13) below.

Let us see how we can now account for the facts. Let us assume that the liquid /l/, being a syllabic consonant in French, comes with a mora in the input. The explanation for /kabl/ > /kap/ is presented in Table 20 below.

If we consider the tableau in Table 20, candidate (a) violates OK- (as /b/ is not permitted in the coda of a syllable and as syllabic consonants are not allowed). Candidates (b) and (c) violate the same constraint as the segments /b/ and /l/ are not allowed as coda in the native language. In (e), the segment /b/ has been deleted and the syllabic consonant /l/ becomes the coda of the previous syllable, changing to /n/ but losing its mora. Though /n/ is allowed in the coda position, in this case, there is violation of the constraint IDENT IO (µ) as a segment that had a mora in the input loses it in the output. Candidate (f) also violates the same constraint, as /l/ has now become the onset, thereby losing its mora again. Hence, candidate (d) is the optimal candidate. In this case, we have deletion of the segment /l/ itself, which would be a violation of MAX-SON as well as MAX-MORA (part of the MAX family) and not an IDENT-MORA violation.

Regarding the sonorant and obstruent cluster, we find only one case in our data that is, /kalk/ that becomes /kankɛ/. We attribute this to the constraint ALIGN-R, which disallows epenthesis at the right edge. In this case, the sonorant /l/ changes to /n/ and the obstruent is preserved. The constraints posited so far will not give us the right result as seen below:

From the tableau shown in Table 21, we can see that candidate (a) which has a complex coda, violating the highest constraint OK-, is banned in the native language. Both candidates (b) and (d) consist of the coda /l/, which also violates OK- (i.e., Coda licensing). Candidate (c), which deletes the sonorant /l/, violates another high ranked constraint MAX-SON. Candidate (e) violates IDENT-F as /l/ changes to /n/, DEP-IO because of epenthesis of the vowel and ALIGN-R. Hence, candidate (f) will be the optimal candidate. However, this is not the right result!

How is this word different from /valv/, which also has a sonorant followed by an obstruent? The only difference between a word like /kalk/, which becomes /kanke/ in Vietnamese and a word like /valv/ that becomes /van/ is the fact that in the former case, the obstruent /k/ is a licensed segment in Vietnamese as /k/ is a possible coda in Vietnamese (unlike the obstruent /v/ in /valv/ which is not permitted in coda position). In the word /valv/, we noted that the obstruent is lost and the sonorant is retained. In the case of /kalk/, however, both the sonorant and the licensed obstruent are retained. This is accomplished in the language by means of epenthesis of a vowel at the end. We noted that the ranking of constraints given for /valv/ would not give us the right results for /kalk/. Hence, we need to introduce a new constraint MAX-LICENSED SEGMENT.

The segment /k/ is a licit segment in the coda position. It thus constitutes a licensed segment. If the constraint MAX-LICENSED SEGMENT is higher ranked than IDENT-F, we obtain the right results. Table 22 shows below the derivation of of /kalk/ > /kankɛ/.

Having looked at the repair strategies used to accommodate coda clusters in Vietnamese, we now examine onset clusters in the following sub section.

6. Conclusion

The adaptation of foreign phonemes supports the role of the native language. The patterns of the process of phonologization are not random but quite systematic. In French loans adapted in Vietnamese, foreign segments which are not present in the native language, are completely banned. Phonotactic adaptations of illicit syllable structures showed that deletion is the favored option to avoid clusters in the coda position. Onset clusters, which are disallowed in native Vietnamese, are repaired either by the strategy of epenthesis or deletion. Hence we have postulated the constraint hierarchy for the Vietnamese loan phonology in this schemata: OK-σ >> IDENT-IO (µ) >> MAX-SON >> MAX-Licensed segment >> IDENT-F >> ALIGN-L/ ALIGN-R >> DEP-IO >> MAX-IO. This study has some practical implications for Vietnamese phonology and loanword phonology in general. The findings of the study reveal that the constraints needed and their rankings play an important role in loanword adaptation. As loan word adaptation is in a state of flux we find certain constraints that are floating, which leads to optionality in pronunciation. This study has more critical perspective on Vietnamese loan word phonology with more detailed and scientific phonological explanations in OT framework, in contrast to previous research on loan adaptation in Vietnamese (Barker 1969, Huynh 2010, K. Nguyen 2007) which gave descriptions in relation to general phonology and etymology.