Previous research on loanword adaptation generally agrees on the fact that when loanwords from a source language (henceforth, Ls) are introduced into a borrowing language (henceforth, Lb), they are usually altered to comply with Lb’s phonotatic constraints. However, remains an outstanding controversy regarding loanword adaptation as to the role of phonology and phonetics. From a phonology-only approach (LaCharite & Paradis 2005), the Lb-internal phonological grammar is attributable to any alternations of loanwords. Given this, any changes in loanword adaptation are purely phonological, subsequently making subphonemic information unnecessary. On the other hand, a phonetics-only approach (Silverman 1992, Kang 1996, Peperkamp & Dupoux 2003) assumes that loanword adaptation takes place at the level of speech perception, not at the UR → SR mapping level. In this framework, maximizing the perceptual similarity between Ls and Lb is the main factor to induce Lb outputs; misperception of Ls enhances loanword adaptation. However, recent studies (Smith 2006,2009; Yip 2006) try to show that loanword adaptation cannot be attributable only to the Lb-internal phonological grammar or perception-only module. Instead, all information Lb speakers have, drawn from Ls and Lb grammar, orthography, perception and other factors, might influence loanword adaptation such that Lb speakers produce specifically posited-Ls representations (henceforth, pLs) different from Lb inputs. Then, these forms have a corresponding relationship with the adapted loanword. Based on Smith’s (2006, 2009) integrated approach, which incorporates mostly the Lb-internal phonological grammar, perception, and orthography, this paper suggests that, in addition to these, both morphological knowledge of the Ls grammar and non-loan phonological rules might produce loanword doublets or variations.
This paper is organized as follows. In section 2, different approaches (i.e., phonological, phonetic, and integrated) into loanword adaptation are investigated. Section 3 introduces problematic data in English loanwords, which support the assumption that neither perception-only no phonology-only methods can sufficiently account for variations of English loans. Then, the paper introduces a new analysis which is sensitive to source-similarity effects and is characterized as having a corresponding relationship (i.e., SB (Ls-to-Lb) correspondence) between Lb outputs and the representations of Ls words perceived by Lb speakers influenced by perception, orthography, and Lb-internal grammar. A conclusion and further remarks are given in Section 4.
A great deal of recent theoretical attention in loanword adaptation has been paid to input forms to the loanword module, in terms of whether it should be handled by phonetic cues or Lb speakers’ phonological grammar. The phonological model in loanword adaptation (Yip 1993, LaCharite & Paradis 2005, Shinohara 2004, following Sapir 1925) supposes that Lb speakers interpret the Ls sounds based on their phonemic inventory; when the Ls sounds do not exactly align with or match the Lb sounds, they are reshaped into the closest alternative phonemes in Lb to better satisfy Lb phonotactics. In this model, speakers who first borrow given words (i.e., bilingual speakers) play a critical role in terms of selecting the underlying form of Ls words. This is diagramed in (1).
(1) Phonology-only approach (Kenstowicz 2006)
In this phonological model, perception does not play a role in the process of adaptation, but Lb speakers’ particular knowledge to the phonological structure of Ls does. However, the phonology-only approach has a drawback since there are many aspects of loanword adaptation that the Lb-internal grammar cannot explain. For example, in Japanese, unsyllabifiable consonants are resolved by means of deletion, not epenthesis (Smith 2009) as in /kak-r / → [ka.kպ] ‘write (nonpast)’ and /jom-sase/ → [jo.ma.se] ‘read (causative)’; loanwords undergo epenthesis rather than deletion in Japanese as in /kri:m/ → [kպ.ri:.mպ]. Similar to Japanese, Korean uses epenthesis repairs for loanwords, while native phonology prefers deletion or feature changing repairs (Kang 2003). The unresolved issue here is what factors affecting loanword adaption are different from native repair strategies. The alternative in this approach is to add more language-specific constraints to the Lb phonological grammar.
Another competing approach based on phonetics is mostly based on speech perception, proposed by several scholars (Steriade 2002, Kang 2003, Peperkamp & Dupoux 2003). In this framework, perception plays an important role in loanword adaptation. Especially, when Lb speakers misperceive Ls sounds, the specific repair strategy undergoes. This is seen in (2).
(2) Perception-only approach (Kenstowicz 2006)
This perception-only approach assumes that all adaptation processes occur during perception, and how the Ls acoustic signals are perceived by Lb speakers determines the outputs of loanwords. In this model, native speakers tend to distort the perception of nonnative forms when the nonnative forms are illicit in Lb; they cannot distinguish such forms from similar legal forms (Depoux et al. 1999; Smith 2006, 2009). For example, Japanese speakers perceive both illicit VCCV sequences and well-formed VCպCV as VCպCV since they perceive illicit VCCV as VCպCV with an illusory vowel (Depoux et al. 1999, Smith 2009). Therefore, as for Japanese Lb speakers, the input of English cream is /kպ.ri:.mպ/ with an epenthesized vowel; therefore, in fact no epenthesis in the UR → SR mapping is required in the perception-only model. However, this perception-only approach cannot account for a different phonological status between an epenthetic vowel in loanwords and a non-epenthetic vowel in native words. For example, Japanese speakers distinguish an inserted vowel from an Ls-based (i.e., non-epenthetic) vowel in the process of accent assignment in loanwords from French (Shinohara 2000, Smith 2006).
Due to the drawbacks of both the phonology-only and perception-only approaches toward loanword adaptation, this paper adopts Smith’s (2006, 2009) integrated approach modulated by phonology, perception and even other factors such as orthography, the Lb internal grammar, or the Ls grammar of Lb speakers. Interestingly, this paper suggests that Lb-speaker knowledge of the Ls-morphological grammar and of native phonological rules is another important factor in loanword adaptation(Silverman 1992). This approach considers not only cases in which Lb-illicit structures that are avoided in non-loans in order to conform to Lb phonotactics but also cases in which loanwords surface unfaithfully from Ls to do so. Incorporating the Lb speaker’s internal grammar, perception, and other factors in loanword adaptation, this approach facilitates the correspondence relation between Lb outputs and Ls outputs as perceived by Lb speakers (i.e., the pLs representation), indicated by the symbol | |. This is formalized as the SB (Ls-to-Lb ) correspondence, illustrated in (3) (Smith 2009).
Again, the pLs representation is the borrower’s posited representation of the source-language form and part of the Lb speaker’s phonological system, controlled by perceptual information, orthographic information, explicit knowledge of the Ls grammar and so on (Smith 2009). Now consider productive English loanword doublets in Japanese loanword adaptation repaired by either epenthesis or deletion, as discussed in Smith (2009).
The doublets in loanword adaptation in (4) are attributed to the existence of the pLs forms gleaned from orthography as well as from auditory perception. Again, the optimal form is chosen through the evaluation of faithful constraints on the SB correspondence relation between the pLs representation and potential Lb candidates. Consider the evaluation of the Ls input /griserin/ → the pLs |griserin| → the output [gպ.ri.se.ron].
Due to the fact that the anti-deletion constraint Max-SB is highly ranked, candidate (5b), which deletes |g| in the pLs representation, cannot be the winner. Additionally, the candidate in (5c) is ruled out because of the violation of a markedness constraint banning a complex onset. What is important in this approach is that the Lb-internal phonology is also relevant for loanword adaptation as when Lb phonotactic restrictions drive adaptation; the high ranked markedness constraints, such as CodaCon and *σ[CC, play a crucial role to get the optimal output.
As for another permissible output for glycerine [ri.se.ron], Smith (2009) posits the different pLs form |riserin| rather than |griserin|. If so, why do the Japanese speakers establish two different pLs forms? Smith’s argument departs from the fact that, when Ls orthography is available, the pLs representation is |griserin| to comply with the loan’s spelling, while the pLs representation becomes |riserin| when orthography is not provided. In other words, some Japanese Lb speakers may not perceive certain non-salient consonants in auditory borrowing (i.e., when orthographical information is not given), and they may establish the pLs form |riserin| for glycerine. Based on this assumption, consider the tableau in (6).
Different from the tableau in (5), candidate (6b) is selected as an optimal form since Max-SB is not incurred in this scenario. Rather, Dep-SB determines the output; candidate (6a) is ruled out due to the violation of Dep-SB.
Now consider the evaluation of the native word /jom-sase/ → [jomase] ‘read (causative)’. This is seen in (7).
Since non-loan words do not have pLs representations, they vacuously satisfy SB correspondence constraints, resulting in deletion, not epentheisis.
As seen in (7), candidate (7c) wins; (7a) and (7b) cannot be the winning candidates since (7a) incurs the violation of highly-ranked CodaCon and (7b), in which a vowel is inserted, violates Dep-IO. As seen in (5)- (7), accepting the orthographic influence on the pLs form in Japanese can account for epenthesis in the face of perceptual deletion. This strongly suggests that factors other than phonology and perception also should be considered in loanword adaptation.
1In addition to orthographical information, Yip (2002) argues that visual information on lip rounding and jaw height may affect the pLs representation. Moreover, it is reported that the degree of speaker, bilingualism or syntax can influence adaptation (Silverman 1992).
Previous studies in English loanword adaption in Korean phonology argue that the repair strategy to resolve illicit-consonant sequences is epenthesis, as in picnic [pikhƚ-nik]. However, little research has been done regarding loanwords that take a different repair strategy, such as feature changing (i.e., nasalization or lateralization). This is shown in nickname [niŋneim] or hotline [hallain]/[hannain]. This paper is to analyze Lb data where either epenthesis, nasalization, or/and lateralization apply and analyze them in a uniform manner. Following Smith’s (2006, 2009) SB correspondence, this paper proposes that perception and the Ls grammar to which Lb speakers are exposed affect loanword adaptation. More importantly this paper aims to show that, in addition to the influence of phonetics and Lb speakers’ Ls knowledge onto the pLs representations, the phonological constraints active in Lb (i.e., delateralization, nasalization, lateralization) also influence the pLs representations.
First, before considering data containing obstruent-plus-sonorant sequences in loanwords, this paper introduces Lb data containing such sequences.
(8) Obstruent-sonorant sequence (Tak 1997, Davis & Shin 1999)
As seen in (8), all obstruent-nasal and noncoronal-liquid sequences undergo nasalization. Interestingly, in Korean there are no attested nouns ending in a coronal obstruent /t/ except for the name of the Korean alpha-bets t and l. Therefore, there exists an example that can support what inevitably happens when a coronal obstruent and a liquid occur contiguously in that order (i.e., /tikƚt+liƚl/→ [tikƚlliƚ]); lateralization applies to this sequence.
Additionally, in Korean grammar, the underlying coronal nasal-liquid sequence (/nl/) and the reverse order (/ln/) undergo lateralization as in /nanlo/ → [nallo] ‘heater’ and /milnap/ → [millap] ‘wax.’ On the other hand, the non-coronal nasal-liquid sequences (i.e., /ml/, /ŋl/) undergo nasalization rather than lateralization as in /kamli/ → [kamni] ‘supervision’. The data are drawn from Tak (1997) and Davis & Shin (1999).
To account for syllable contact situations within the framework of the optimality-analytic approach, Davis & Shin (1999) propose two important constraints. One is a language specific constraint Similarity (i.e.,*[+sonorant, +coronal], [+sonorant, +coronal]), which disallows sequences of coronal sonorant consonants, and the other is SyllableContact (SyllCon), which bans rising sonority over a syllable boundary. These two constraints along with other Max and Identity constraints produce the right output in Korean non-loans. This is seen in (10) and (11).
In non-loan phonology, SyllCon, Ident-Onset[sonorant] and Similarity are undominated; any candidates in (10) and (11) which violate these constraints (i.e., (10a), (10d) and (11a)) are ruled out. Interestingly, (11b) and (11c) do not violate Similarity since the sequences n.n/l.l share their structures. As for the tableau in (10), candidate (10c) is realized as an optimal form since candidate (10b) violates constraint Ident-Place. On the other hand, candidate (11c) becomes the winner in spite of the violation of the Max-[nas] since Max-[nas] is ranked lower than Max-[lat].
Now, consider the loanword data containing obstruent-sonorant sequences; those in the left column undergo vowel epenthesis, while those in the right column feature changing (either nasalization or lateralization).
In spite of the widely accepted proposal that epenthesis is a repair strategy in English loanword adaptation in Korean phonology, there are many loans where Korean non-loan processes (i.e., nasalization or/and lateralization) are applied.
Before moving into the SB correspondence model, consider the evaluation of picnic and nickname in loanword adaption with the Lb constraint ranking.
As seen in (13), the non-loan constraint ranking produces the right output [niŋneim] for the loan nickname. However, as for tableau (14), undesirable candidate (14c), influenced by nasal assimilation, could be chosen as a winner; in fact candidate (14d) in which a vowel is epenthesized is realized as the real output.
Why does picnic undergo epenthesis, while nickname undergoes nasalization? To account for the different behavior of these loanwords, this paper facilitates the SB correspondence model, another instantiation of correspondence theory (McCarthy & Prince 1995). Correspondence relations have been proposed for various S1-S2 pairs such as IO correspondence, BR (base-reduplicant) correspondence (McCarthy & Prince 1995), OO (output-output) correspondence (Benua 1997) and so on. The SB correspondence refers to the correspondence relations between Ls output forms perceived by Lb speakers (pLs forms) and Lb outputs. The pLs representations are attributed to auditory perception, an explicit Ls grammar, and even an Lb native grammar. One faithfulness constraint on the SB relation is Max-SB, which bans deletion in Lb outputs from the corresponding pLs representation. On the other hand, Dep-SB penalizes insertion in Lb outputs from the corresponding pLs representation. These two SB constraints outrank Max-IO and Dep-IO since Max-SB and Dep- SB are vacuously satisfied for nonloans.
The primary argument of this paper is that the different realizations or variations of the loans are generally attributed to the different pLs representations. For example, as for the loan picnic, Lb speakers perceive /k/ in Ls as being released. In other words, as discussed byParker (1977) and Kang (2003), Korean Lb speakers perceive English released stops as the corresponding stops followed by [ƚ]. For that reason, this paper posits that |pikhƚnik| is the pLs representation for the loan picnic. What is interesting is that epenthesis applies to only momomorphemic loans such as icnic, Kaplan, and Atlanta. If so, why do the loans in the left column in (12), such as walkman, booklet, potluck, behave differently from those in the right column in (12), such as nickname, bookmark, good night? This paper conjectures that loans such as walkman, booklet, potluck are morphonologically distinct from nickname, bookmark, good night. Kawahara and Kao (2010) suggest that loanword morphology might be processed differently from Ls morphology by investigating a new Japanese suffix [- zu]. This is a loanword from the English plural –s, which is now used to form group names as in Seibu Lions. This suffix assigns an accent on root-initial syllable out of accordance with the universal tendency to assign an accent on syllables next to the affixes (Revithiandou 2008, Kawahaha & Kao 2010). According to Kawahara and Kao (2010), the origin of initial accentuating behavior is drawn from the assumption that Japanese Lb speakers perceived Seibu Lions as a compound like [[wordX]- [wordY-zu]]; instead of a universal affix accent assigning rule, a compound accent rule, which assigns an accent to the initial syllable of the second member of the compound, was applied and resulted in [seibu-ráion-zu],not *[seibu-raión-zu]. This clearly shows that loanword morphology could be separate from the morphology of the Lb since the suffix [-zu] behaves differently from other native suffixes and since the loan Lions is treated as a monomorphemic word and displays initial accentuating behavior. Subsequently, this paper assumes that Korean Lb speakers perceive walkman, booklet, and potluck as a monomorphemic words since they have no linguistic pressure to break them into two separate morphemes; the loanwords do not have comparable loan pairs such as walkman vs. *sleepman, booklet vs. *magazinelet, and potluck vs. *panluck. In contrast, when Korean Lb speakers encounter the loans in the right column in (11), such as nickname, bookmark, and pop music, they analyze them as compounds: word[word[nick]##word[name]], word[word[book]##word[mark]], and word[word[pop]##word[music]]. Given this assumption, it is proposed that the loans perceived as monomorphemic words undergo epenthesis, while the loans analyzed as compounds are influenced by feature changing (i.e., nasalization or/and lateralization). Therefore, the morpheme internal obstruent of the obstruent-plus-sonorant sequences is systematically released, resulting in vowel epenthesis. However, when loanwords are considered to be made up of more than one distinct morpheme or word, Korean speakers gear up to retrieve morphological information from the explicit Ls grammar, which may be different from the real Ls grammar. Unlike previous research, this paper assumes that pLs representations can be overridden by the Lb speaker’s morphological knowledge on the Ls grammar. As for the loan nickname, Korean Lb speakers analyze it as a compound (i.e., word[word[nick]##word[name]], even though nick is not used as an isolation, because the Lb speakers acknowledge that a nickname is a type of a name comparable to a real name. Therefore, Korean Lb speakers perceive /k/ in nickname as being unreleased; the pLs representationis posited as |nik.ne.im|.
Following Smith (2006, 2009), the relevant constraints are proposed in (15).
The evaluation of picnic with the pLs |pi.khƚ.nik| and of nickname with the pLs |nik##ne.im| is given in (17) and (18), respectively.
The SB correspondence constraints, which compare pLs forms and outputs, especially Max-SB and Dep-SB, induce the right optimal forms, as seen in (16) and (17). Since these two SB constraints are undominated, any candidates that incur a violation of Max-SB and Dep-SB cannot be the winner. Therefore, candidates (16c) and (17b) are ruled out; of course, candidates that disobey SyllCon cannot be the winner as in (16a) and (17a). As seen in (16) and (17), Korean Lb speakers have a consistent tendency to release an obstruent of the obstruent-sonorant sequences within a morpheme or a simple word, while over the morpheme or word boundary, the release of an obstruent followed by a sonorant is banned.
Then why does cutline undergo epenthesis while hotline undergoes a feature changing process? Interestingly, hotline is realized as either [hallain] or [hannain]. As discussed before, following Kawahaha & Kao (2010), we conjecture that Korean Lb speakers analyze loans differently from the way that Ls speakers do. If Korean Lb speakers could not find any comparable loan that they semantically connect, they analyze them as a monomorphemic word; in that case, vowel epenthesis applies. In contrast, if they happen to match loans with another semantically connected loan, Korean Lb speakers analyze them as a complex word; in such a case, nasalization or lateralization applies. For example, cutline is not analyzed as a compound or bimorphemic to most speakers since it does not have a loan pair that supports it is a compound; its pLs representation is posited as |khʌthƚlain|. On the other hand, the loan hotline can be analyzable as two elements hot and line. Based on semantically related loans such as cold war or hot link, Korean speakers differentiate hotline from cutline in terms of their morphological representation in such a way that hotline is analyzed as a bimorphemic word, but not a compound, since this loan does not have an exact comparable loan pair like coldline. Then, Korean speakers assign a morpheme boundary over hot and line; it is analyzed as word[hot#line]; the pLs representation for hotline can be proposed as |hot.la.in|.3
Different pLs representations as in (18) and (19), |khʌthƚlain| and |hat#la.in|, respectively, yield different optimal loans. Candidate (18b) undergoes vowel epenthesis and is realized as the optimal form since it is the faithful candidate to the pLs representation, while (19b) with vowel epenthesis cannot be the winner since its pLs representation contains an unreleased obstruent |t|. Therefore, even though (19c) does not obey Ident-SB[son] because of the application of lateralization, it is selected as the optimal form.
Then, why do some of the loans have variations between nasalization and lateralization as shown in hotline? To account for the doublets or variations in (20), this paper assumes that explicit Ls knowledge and the native phonological rule of the Korean Lb speakers take a crucial role. Korean has a language-specific constraint *w[l which mitigates the realization of the word-initial [l] , resulting in [n]. The data is given in (20) from Tak (1997).
Word-initial Delaterlazation, which enforces phonotactics in Korean native language, also affects some of loanwords, yielding doublets or variations. This is seen in (21).
Assuming that delaterlization may occur in loans, this paper suggests that if Korean Lb speakers analyze loans into compounds consisting of two explicit words (i.e., word[word[X]##word[Y]] whose second element begins with /l/, the loans may undergo delateralization, resulting in [n]. Noteworthy is that the loan hotline is pronounced as either [hallain] or [hannain] even though lateralization is the more frequent repair strategy. As mentioned before, Korean Lb speakers begin to analyze hotline into hot and line based on another loan cold war with a similar semantic correlation; its pLs is represented as |hat.la.in|. However, some Korean Lb speakers begin to analyze it as a compound and yield the alternative pLs representation is |hat.na.in| where delaterlization applies. In other words, if hotline is perceived as a compound, the word-initial /l/ of the second word line might undergo delateralization as it is reflected in the pLs representation. And if it is perceived as a bimorphemic word, it does not undergo delateralization, which influences a word-initial /l/, resulting in [hallain] as an optimal form. To account for this, two constraints Max- SB[lat] and Max-SB[nas] are introduced with the ranking Max-SB[lat] >> Max-SB[nas]. The final constraint is given in (22).
The different pLs representations for the outputs [hannain] and [hallain] are proposed as |hat.na.in| and |hat.la.in|, respectively. The difference is drawn from how Korean Lb speakers analyze them based on their Ls grammar. In (23), candidate (23c) is selected as a winner since (23a) and (23b) incur a violation of SyllCon and Dep-SB, respectively. Additionally, (23d) disobeys Max-SB[nasal] since a nasal in a second syllable is realized as a lateral in the output form. In contrast, as for the tableau in (24), candidate (24d) becomes a winner since it obeys all SBcorrespondence constraints including SyllCon.
Now consider another type of loanword containing sonorant sequences. Some loanwords containing these sequences have more than one pronunciation, especially when the second member of the sequences is a liquid, as seen in (25).
As discussed earlier, in non-loan phonology, the nl and ln sequences undergo lateralization, while noncoronal-plus-l/r sequences are subject to nasalization. If so, why do download and inline undergo nasalization instead of lateralization, and why do only and lonely undergo either nasalization or lateralization? To account for the doublets or variations in (25), this paper assumes that Korean Lb speakers perceive download as two separate words, undergoing delateralization with the pLs representation |da.un.no.dƚ|. As noted earlier, the requirement on this is that the loans have semantically connected loanwords such that Korean Lb speakers feel that they need to analyze them into further separate words. For example, the loan download is analyzed into down and load and load undergoes delateralizaiton; there exists another related loan upload, which undergoes nasalization [əmnodƚ], not [əphƚnodƚ].
Consider the evaluation of loans download and compare it with all night where the normal native phonological rule, lateralization, applies as in [ollaithƚ].
As seen in (26) and (27), Korean Lb speakers’ compound judgment on the loans download and all night takes a crucial role to produce pLs representations, resulting in delateralization and lateralization, respectively. As indicated in tableau (11), /nl/ and /ln/ sequences violate the undominated constraint Similarity in Korean; candidate (26a) and (27a) cannot be the winner. Since these are loan compounds, the pLs for download is |da.un.no.dƚ|, not |da.un.lo.di-|. Therefore, obeying Max-SB[nas], candidate (26c) wins. As for the tableau in (27), the candidate in (27d), which obeys Dep-SB and Max-SB[lat], is selected as a winner in spite of a vio-lation of Max-SB[nas] and Max-[nasal].
Interestingly, both loans only and lonely undergo lateralization and sometimes delateralization. As for loans only and lonely, this paper proposes that Korean Lb speakers might analyze both loans as bimorphemic words, producing [on#ly] and [lone#ly]. This is different from Ls grammar in which only is a simple monomorphemic word, not a bimorphemic word like word[X#ly]. I assume that Korean Lb speakers acknowledge that –ly, an English suffix deriving adverbs out of adjectives, is very productive in English and analyze only as a bimorphemic word. Therefore, only is adapted as [olli]. As for some Korean Lb speakers, they further analyze only and lonley as a compounds (i.e., word[word[on]##word[ly]] and word[word[lone]##word[ly]]; they are applied by delateralization, yielding another permissible output [onni] and [lonni]. Therefore, loans only and lonely can undergo either lateralization or delateralization, depending on the Korean Ls speakers’ morphological judgment on these loans.
2Kang (2003) and Davis & Kang (2006) argue that an Lb morphophonemic restriction plays a crucial role in loanword adaptation. According to them, in Korean, there is no underlyingly /t/-final nouns. Historically, they have evolved into /s/-final nouns as in */pat/~/pat-e/~/pat-ƚl/, /pat/~/pas-e/~/pas-ƚl/ ‘field’ (Kang 2003). Therefore, two different adaptation patterns show a different conjugation as shown below. (1) Adaptation of cut a. no vowel epenthesis khʌt khʌs-ƚl khʌs-e b. vowel epenthesis khʌtƚ khʌthƚ-lƚl khʌthƚ-e Even though there is no /t/-final noun, there is a case to support the view that the coronal obstruent-plus-liquid sequence undergoes lateralization. Therefore, this paper accepts the view that lateralization is a repair strategy for this illicit sequence in Korean loanword phonology. 3In spite of the fact that the loan outlet does not have any related loans that are semantically connected, this paper proposes that its pLs form should be |aut#let|, and therefore its loan is pronounced as [aullet]. We conjecture that the reason why it is treated as a bimorphemic loan, is based on the existence of the older loan outlet [authƚlethƚ], meaning ‘a receptacle for the plug of an electrical device’. To avoid a confusion of the older loan outlet, it is analyzed as a bimorphemic word and undergoes lateralization. Interestingly, Korean Lb speakers never perceive it as a compound since there is no semantically related loan with outlet. It cannot be the target of delateralization; *[aunnet] is never realized as a permissible output. 4The violations of Ident-SB[son] and Ident-[son] do not affect the evaluation of the loanwords; they are omitted in the tableaux. Additionally, as one of the reviewers pointed out, candidates (23c), (23d), (24c) and (24d) violate Ident-[son], which is not a fatal violation; it is not indicated in the tableaux.
