Typology of interdental fricatives with reference to loanword adaptation

Consequently, we will examine differential substitution in various languages and try to explain how our current analysis can account for the typology of interdental fricatives in loanwor

Typology of interdental fricatives with reference

to loanword adaptation

Juhee Lee(Kyung Hee University)

Lee, Juhee 2006 Typology of interdental fricatives with reference to loanword

adaptation Studies in Phonetics, Phonology and Morphology 12.1 127-148 In this

paper, I will discuss the adaptation of interdental fricatives in a loanword context Cross-linguistically, this sound is not common in many languages and, therefore, this fact leads us to substitute certain sounds either [t] or [s] Consequently, we will examine differential substitution in various languages and try to explain how our current analysis can account for the typology of interdental fricatives in loanword adaptations Moreover, we consider that the asymmetric pattern observed in Dutch is due to the prominence level of position and we analyze within the positional markedness ranking Finally, we consider a variety of factors in the most problematic

case, Korean, a language in which the adaptation pattern is rather unstable (Kyung

Hee University)

Keywords: interdental fricatives, loanword, substitution, variation, positional faithfulness,

positional markedness, optimality theory.

1 Introduction

In this paper, the sound on which we focus the bulk of our attention isvoiced-voiceless pairs for the dental fricative in English Cross-linguistically,the English interdental fricative is not a common sound If you place theblade of your tongue on either the inside or the edge of the upper teeth,allowing the appropriate narrow gap for friction, you will get the beginning

sound of thigh if you do not voice, and the sound at the beginning of the thy

(Roca and Johnson 1999)

Notice that the English spelling system does not distinguish betweenthese two sounds Thus, the phonetic alphabet symbols are [] (the Greek

letter “theta”), for the voiceless sound in thigh, and [] (the Old English letter “eth”), for the voiced sound in thy.

These dental fricatives are uncommon sounds in many languages, whichmeans simply that they do not exist in the consonant inventory Therefore,when we adopt an English loanword that contains interdental fricativesounds, we need to change into other sounds that which existed in the

 This research was supported by the Kyung Hee University Research Fund in 2005 20051021) An earlier version of this paper was initially presented for the winter conference (2005-12-03) from the Phonology-Morphology Circle of Korea at Seoul National University

(KHU-of Education and the winter conference (2006-02-06) from the Linguistics Society (KHU-of Korea at Kyung Hee University I wish to thank Young-Suk Kim, Sang-Cheol Ahn, and Mira Oh for the helpful comments and suggestions I am also grateful to the reviewers for the suggestions and comments Of course all errors are mine.

native phonological system In general, they are realized as different types,depending on language Let us first consider the following in Table 1.

Table 1 Stops, fricatives and affricates

Typologically, in general, // is turned into either [s] or [t] in loanwords.Cross-linguistically, speakers will adopt the unmarked s or t to conform totheir own native phonological system Numerous research studies havebeen conducted regarding phonological theory as it relates to the differentaspects of various languages (Menyuk 1968, Schmidt 1977, James 1986,Ioup and Weinberger 1987, Weinberger 1990, Hancin-Bhatt 1994, Paradisand LaCharité 1997, Eckman 1977, Lombardi 2000, Cho and Lee 2001,Lee and Cho 2002, Ahn 2003 and many others)

(1) Attested substitutions for English interdentals

[t] substitution: Thai, Russian, Hungarian, Serbo-Croatian, Tagalog,

Moroccan Arabic, Quebec French, and Xhosa.[s] substitution: Japanese, German, Egyptian Arabic, and European

French

As shown in (1), two different substitutions for the interdental fricativesare observed in various languages Japanese and German replace // with[s], while Thai, Russian, and Hungarian substitute [t] for the same segment

1 In child language, different children produce different substitutes, i.e either [f, v], [s, z], [t, d] The selection may depend on the stage of development.

However, English loanwords in Korean show different behaviorcharacteristics Two different substitution patterns are visible for theinterdental fricative // We shall also bring up this issue in section 3.4.2,comparing for the case of French dialectal variation Then we shall move todiscuss additional factors such as the Dutch loanword, which is sensitive tosyllable position.

In exploring these issues, I shall employ the concept of OptimalityTheory (McCarthy and Prince 1993) for the overall analysis of Englishinterdental substitution For that, let me briefly introduce the previousanalysis for the immediate section

2 Previous studies

As explained earlier, English interdental fricatives cause problems forlearners who speak different native languages (French, Russian, Japanese,Spanish, etc.)

In the previous literature, Menyuk (1968), following many other observers,reports that /, / are the sounds mastered last and substituted most frequently

by English native speakers Menyuk further reports that it is the distinctivefeatures  strident (differentiating /, / from /s, z/) that are the features lastmastered by the native speakers of English

Eckman (1977), on the other hand, proposes the principle of related-to-markedness in the Markedness Differential Hypothesis (MDH).According to Eckman, the areas of difficulty that a language learner willencounter can be predicted on the basis of a systematic comparison of thegrammars of the native language, the target language and the markednessrelations stated in universal grammars, such as in:

difficulty-(2)

(a) Those areas of the target language, which differ from the nativelanguage and are more marked than the native language, will bedifficult

(b) The relative degree of difficulty of the areas of the target language,which are more marked than the native language, will correspond

to the relative degree of markedness

(c) Those areas of the target language, which are different from thenative language, but are not more marked than the native language,will not be difficult

(Eckman 1977: 61)Schmidt (1977) also studies fricatives, in this case the interdentals // and//, for the loanword adaptations in Arabic contexts He argues that some ofhis Egyptian Arabic subjects have been exposed to Classical Arabic, whichitself possesses these fricatives in its repertoire That is, an historical rule

merged these three segments with the stop series /t, d, d/ as presented in(3):

(3) Classical Arabic Egyptian colloquial Arabic

Words that were then borrowed (or-reborrowed) from the classical into thecolloquial lexicon instead underwent a new sound change, merging theinterdentals /, / with the sibilants /s, z/ and creating a new emphaticsibilant /z/ as the reflex of // as shown in (4):

(4) Classical Arabic Egyptian colloquial Arabic

‘to mention’ akar zakar

In contemporary Egyptian Arabic, there are numerous lexical triplets withinterdental fricative, sibilant and stop variants, e.g., /a:li/~/sa:lis/~/ta:lit/,

‘third’ For this reason, Schmidt considers the possibility that his subjects’problems may derive in part from the same kinds of developmental sourcesthat affect the L1 acquisition of these sounds

Schmidt also points out that // and // are well documented as ‘thesounds mastered last and substituted most frequently by English nativespeakers This may explain why L2 learners from virtually all L1backgrounds have problems with these two sounds In the study of Arabicspeakers of English, Schmidt shows how his subjects’ pronunciation of //and // undergoes variations, with transfer of /s/ and /z/ occurring incolloquial but not in formal speech

Moreover, Gatbonton (1978) discusses the fact that in the first phase ofacquisition, a learner first uses one incorrect phonological form in allcontexts, and then later introduces another form, which is used in freevariation with the first, while in the second phase, each form is gradually

restricted to its own context It is defined as a phonological core approach, for example a common sound [m] is phonological core vs German [x] relatively rare and thus it is called non-core or peripheral Therefore, //

and // are non-core and they are subjected to change as core sounds.From the aspect of acquisition, Brown (2000) studies the interrelationbetween speech perception and phonological acquisition from infant to adult.Furthermore, Kenstowicz and Kisseberth (1979) also cite evidence that boththe French and Serbo-Croatian languages lack the non-strident interdentalfricatives  and , and when native speakers of these languages produceEnglish words with these problematic segments in them, /s/ and /z/ aresubstituted by the French speakers, while /t/ and /d/ are substituted by theSerbo-Croatian speakers Consequently, they conclude that the phonological

structures of these two languages evidently must treat dental fricatives andstops differently.

Weinberger (1990) refers to this type of error phenomenon as differentialsubstitution as presented below:

Table 2 Differential substitution (data cited from Weinberger 1990)

According to the universal marking conventions adopted from Chomskyand Halle (1968), the most unmarked continuant is /s/, and it therefore willuniformly replace the English interdental fricative // This will prove inthe Japanese case but fail in the Russian as well as the Mandarin and Dutchcases Hence, in the next immediate section, I shall analyze all the differentkinds of typological patterns, as presented in Table 2, within OptimalityTheory for the theoretical tool

3 Analyses for differential substitution

As discussed in the previous section, much work has been done in theliterature However, in Rule-Based Theory, it is difficult to explain suchsubstitution in a principled manner Since L1 has no interdental sound, achild cannot acquire a rule changing it either to [t] or [s] Therefore, this is

an obvious case of what the L2 acquisition calls transfer Of course, L1must be crucial in the sense that speakers of different languages usedifferent substitutions In this section, I shall suggest that OptimalityTheory allows a solution for our discussion

3.1 [t] substitution

As explained earlier, Thai, Russian, Hungarian, Serbo-Croatian, Tagalog,Moroccan Arabic, Quebec French, and Xhosa are the language type for [t]substitution when they are faced with interdental fricative sounds Therealization of [t] is the case of showing that in a more unmarked segment itplays a crucial role for the adaptation.

In a similar vein, what is fundamental to [t] substitution is the samething that we see in child phonology That is, the unmarked manner ispreferred Let us consider following through by examining Russian data inorder to understand this type of adaptation

As explained earlier, the adaptation of [t] is the case of the emergence of

the unmarked manner, therefore, we need to employ some context-free

markedness constraints That is, *Cont and *Stop, which must be ranked

lower than the faithfulness constraint, *Ident(Cont), by the emergence of

the unmarked ranking as defined in (6) and (7).

(6) *Cont: Fricative is not allowed

*Stop: Stop is not allowed

*:  is not allowed

(7) Ident(Cont): Corresponding segments in input and output have identical

values for the feature [continuant]

With these constraints, the grammar of [t] substitution is as in (8)

(8) * >> *Cont >> *Stop >> Ident(cont)

In the tableau, *Cont is ranked higher than *Stop since stops are lessmarked than fricatives This is supported by sound system generalizations

as well as child phonology in general In L1, children go through a stagewhere fricatives are absent and are replaced by the stops (Locke 1973,

Vihman 1996) Moreover, Maddieson (1984) maintains that all languageshave stops, but not all languages have fricatives.

Thus, the ranking of *Cont >> *Stop is universal The faithfulnessconstraint Ident(cont) is ranked lower in this type of substitution since the

unmarked output is determined by the emergence of the unmarked ranking.

That is, * >> *Cont >> *Stop and, therefore, Ident(cont) does not play animportant role for the adaptation of [t].2

However, Lombardi (2000) proposes a different ranking for the case athand.3 Let us consider the following tableau in (9)

(9) * >> Ident(Manner) >> *Cont >> *Stop

As seen above, in Lombardi (2000), Ident(Manner) is ranked higher than

*Cont and *Stop This constraint Ident(Manner) implements stop, cont, andstrident as manner features Thus, candidate b has violated Ident(manner),since it changes [cont] to [stop], and candidate c also has violated it by the factthat candidate c changes the value of the [strident]

Nonetheless, as shown in (8), markedness constraints *, *Cont, and *Stop

are ranked above faithfulness constraint Ident(Manner) by the emergence of

therefore, claim that the ranking shown in Lombardi (2000) is not compatiblewith her proposed unmarked substitution

2 Wenk (1979) performed an experiment that tested the relationship of speech style and overall language proficiency affecting the choice of substitution Wenk suggests the following developmental stages.

Stage 1: f (Child phonology: often substitute f for )

//*TH*Cont*StopIdent(Manner)  a *!b t**  c s*! (Ahn uses *TH instead of * .)

4 Ahn (2003) argues that this is contradictory with her analysis because the unmarked output must be decided by the natural ranking of markedness constraints rather than intervening faithfulness constraint.

3.2 [s] substitutionJapanese, German, Egyptian Arabic, and European French are all the typefor [s] substitution when they adapt this English // The adaptation of [s]

is the case in which the faithfulness constraint is ranked higher thanmarkedness constraints In order to understand the result of our analysis,let us first consider following data in (10)

(10) Japanese loanwords

bath [basu] mother [maza:]

thank [saku] the [za]

thirty [sa:ti] smooth [sumu:zu]

death [desu]

In Japanese, the English interdentals are realized as [s]/[z] As already seen insection 3.1, the sound substitutions shown in (10) are also accountable by theundominated segmental structure constraints in the form of * and theconstraint * must be satisfied at the cost of Ident[cont] Let us consider thefollowing tableau in (11)

(11) * >> Ident(cont) >> *Cont >> *Stop

However, Lombardi (2000) proposes a reranking of these markednessconstraints Let is consider the following in (12)

the Manner feature does not provide an adequate answer Furthermore, areranking of markedness constraints, from *Cont >> *Stop to *Stop >> *Cont,

is not desirable in this case

Ahn (2003), on the other hand, proposes a more elaborate analysis but Ideem that the constraint *Strident may not be needed Let us consider thefollowing in (13)

“fallacy of perfection,” since the optimal output violates other constraints such

as *Strident

However, I contend that the motivation of *Strident is unclear since theuse of *Strident incurs the violation Occam’s Razor Without using moreconstraints, constraint ranking, * >> Ident(cont) >> *Cont >> *Stop,correctly singles out candidate c That is, the simple is better in grammar.Moreover, there is no explanation why *Strident dominates *Cont I,therefore, argue the necessity of this constraint *Strident is redundant

3.3 Positional prominence

As previously illustrated in Table 2, section 2, there are some languageswhere the adaptation of interdental fricatives is sensitive to the syllableposition Let us recapitulate these cases

5 Mandarin does not allow the nasal [m], any liquids, or any obstruents in coda position Possible Syllable codas (Broselow et al 1998)

a Mandarin: glide, nasal (n, )

b English: glide, liquid, nasal (m, n, ) voiceless obstruent, voiced obstruent

In the literature, James (1984) discusses the matching levels for Dutch.His hypothesis for this discrepancy is due to the strong and weak position.Let us consider the following in (15):

(15) Matching levels for Dutch (James 1986)

a Phonological: gross distributional relatedness to a number of Ctypes in the native language

b Phonetic place – dental/alveolar e.g.) [t, d, s, z]

c Articulatory: tip advancing, raising, etc e.g.) [t, d]

James argues that the structural basis of transfer in syllable initial position

is articulatory, hence, the dental segment will replace the interdentalfricative In word-final position, however, Dutch speakers analyze theproblematic segment at the phonetic level, specifically for manner, andreplace the interdental fricative with [s] This is because word-finalposition is considered the “weak” position, and fricatives are presumed to

be weaker than stops

Interestingly, the appearance of these stops and fricatives is also found

in varieties of African-American Vernacular English (hence, AAVE), theinterdental fricatives are realized as either alveolar stops or labiodentalfricatives, depending upon the location of the fricative in the word Thus,the data shown in (15) is not loanword specific, but rather universal Let usconsider the data in (16)

(16) Positional variation in AAVE

thin [n]  [tn] bath [b]  [bf]

this [s]  [ds] brother [brr]  brv]

In (16), at the beginning of the word, the interdental fricative // is realized

as [t], whereas word-medially or word-finally, the interdental fricative //

is realized as [f] For the same pattern, at the beginning of the word, // isrealized as [d] while word-medially or word-finally, same sound // isrealized as [v]

Note that these rules are quite consistent In AAVE, it is ungrammatical

to pronounce // as [t] unless it is at the beginning of the word (bath

*[bt]), and it is ungrammatical to pronounce // as [f] if it is at thebeginning of a word (thin *[fn]) The same holds true of voicedinterdental fricatives Although the social marking of English interdentalfricatives // and // is well-established in sociolinguistic studies(Labov1966), their precise role as ethnic marking is still in dispute

Nonetheless, there would be no question that perceptually [f, v] are frequentlyconfused with [], [] even for native speakers of English (Hancin-Bhatt 1994).This fact also leads us to understand error patterns in child phonology

Let us now turn to the major question of Dutch and Mandarin How do weaccount for their substitution in OT? Lombardi (2000) argues that the

preference for coda fricatives over coda stops emerges in acquisition, and thatcan be seen as the result of the emergence of additional markedness effects onstops and fricatives That is, with the context-free markedness constraints

*Cont and *Stop, other constraints relevant to the coda are needed

(17) Mora[stop] >> *Mora[cont] >> *Mora[son]

As seen in (17), with the formulation of markedness hierarchy by Moren(1999), Lombardi uses *Mora[stop] and *Mora[cont] since this ranking of

*Mora[stop] >> *Mora[cont] does not affect onsets Therefore, her account

of positional variation is shown as below

Although the tableau above seems to single out the desired candidate, there

is a fundamental problem in using a series of *Mora[F] constraints in (17)

As mentioned in her paper, Lombardi is also aware of the potentialproblem of using the *Mora[F] constraint since there are other languageswhich have no evidence for mora in the coda Thus, this ranking shouldonly be visible in languages where codas are moraic.6

Thus, I argue that the problem of positional prominence in Dutch cannot

be solved by positional faithfulness7 (Beckman 1998) For example, theconstraint ranking of Ident-Coda[cont] >> Markedness >> Ident[cont] is notworking since the [cont] feature value of the coda is not maintained at all.Thus, we need to employ other constraints, which have no conflict for thecoda status in other languages I, therefore, contend that positional markedness(see Zoll 1998, Ahn 2003, Smith 2002, 2004) constraints will be effective,

6 Ahn (2003) also argues the problem of *mora[F] constraint.

7 According to Beckman (1998), there is positional privilege Privileged positions are initial syllables, stressed syllables, syllable onsets, roots and long vowels.