J East Asian Linguist (2012) 21:267–304 DOI 10.1007/s10831-012-9089-4
Adaptation of English complex words into Korean Mira Oh
Received: 16 March 2009 / Accepted: 5 January 2012 / Published online: 11 April 2012 © Springer Science+Business Media B.V. 2012
Abstract Most previous studies related to loanword adaptation have centered on segmental mappings between source and loanword sounds in morphologically simplex words (LaCharite´ and Paradis, Ling Inq 36:223–258, 2005; Kang, Phonology 20:1–56, 2003). However, few have considered the adaptation of complex words, specifically words made of multiple free morphemes. Examining the adaptation of complex English words into Korean, the present study makes four claims. First, it proposes that each component of a complex word is a unit for loanword adaptation in calculating sound mappings. Second, it suggests that each component word is a stem, whereas a loanword as a whole is categorized as a nominal word in Korean. Third, apparent single-unit adaptation is possible only when the first component allows variable final vowel epenthesis at the end of the first component word; this is analyzed in terms of split-base effects. Fourth, the allophonic realization of phonemes plays a crucial role in loanword adaptation. These claims are empirically supported by loanwords found via the National Academy of Korean Language (NAKL 1991) and Google searching (March–June 2011). Furthermore, this study provides an explicit formalization of the analysis of complex loanwords within Optimality Theory (Prince and Smolensky, Opimality Theory: Constraint interaction in generative grammar, 1993). The present study contributes to the literature of loanword phonology by shedding light on several issues. First, the study proposes a model for the adaptation of complex loanwords wherein both morphological structure and the phonetic information of the source language play important roles. There has been intense debate about the effect of input information on loanword adaptation, broadly split between a phonological view (LaCharite´ and Paradis, Ling Inq 36:223–258, 2005) and a perceptual view (Silverman, Phonology 9:289–328, 1992; Steriade, in: Hume and Johnson (eds.) The role of speech M. Oh (&) Department of English Language and Literature, Chonnam National University, 77 Yongbongro, Bukgu, Gwangju 500-757, South Korea e-mail:
[email protected]
123
268
M. Oh
perception in phonology, 2001; Peperkamp and Dupoux, Proceedings of the 15th International Congress of Phonetic Sciences, 2003). The adaptation of complex English words in Korean supports the perceptual approach, in that the allophone realized in a complex word determines the loanword sound. It also reveals that sound mappings in loanwords are determined by the morphological structure of source words. There has been a discussion about whether the morphosyntactic base is isomorphic to the phonological base (Steriade, Lexical conservatism and the notion base of affixation, 2000). Component-by-component analysis of the apparent single-unit adaptation provides support for the existence of a split-base effect as well as for the lexical conservatism proposed by Steriade (Lexical conservatism and the notion base of affixation, 2000). Finally, it makes a contribution to evaluation of the internal structure and morphological category of complex loanwords, which has been rarely considered in the literature on loanword phonology. Keywords Complex English words · Loanword adaptation · Morphological structure · Component-by-component analysis · Vowel epenthesis · Segmental mappings · The split-base effect
1 Introduction The goal of this paper is to investigate the adaptation of complex English words, consisting of multiple component words such as “make up” or “all in”, into Korean.1 In contrast to previous research on loanword adaptation, which has typically focused on segmental mappings between the source word and loanword sounds of morphologically simple words (LaCharite´ and Paradis 2005; Kang 2003), the present paper addresses a number of intriguing questions about the adaptation of complex words, including the question of what the unit of analysis for loanword adaptation that can determine sound mappings is. It also addresses whether complex loanwords behave in the same way as simplex loanwords in terms of vowel epenthesis and segmental mappings, the extent to which morphological and phonetic information in complex words from a source language influence the adaptation of complex words, and the morphological structure of complex loanwords. To address these questions, this study provides a quantitative analysis of the adaptation of English loanwords in Korean by employing a list of loanwords from the NAKL (National Academy of the Korean Language 1991) and from internet search results (Google; March–June 2011). The NAKL list contains approximately 5,000 English loanwords gathered from six daily newspapers and nine magazines published in Korea throughout 1990. The results of quantitative study indicate that simplex and complex loanwords behave differently in terms of segmental mappings and vowel epenthesis. At this point, it is necessary to clarify how the term “complex word” is defined in this paper. In morphology, a complex word is typically a single word consisting of 1
“Make up” has various meanings in English. It could be a noun, as in “to put on make up”; in American English, an adjective, as in “a make-up exam”; or a verb, as in “the couple will make up after their argument.” In this study, “make up” is considered to be a noun or an adjective. As discussed later, loanwords are adapted as a noun regardless of the part of speech in the source language.
123
Adaptation of English complex words
269
more than one morpheme, such as the English words “unhappy” or “stopping”. However, this paper is not about the borrowing of complex words of this nature. Instead, by “complex words” we specifically mean three types of words: compound words such as “rush hour”, verb-particle constructions or phrasal verbs such as “give up”, and nominalized forms such as “change up”. When investigating the borrowing of nominalized forms of phrasal verbs such as “make up”, “catch on”, and “cut off”, we propose that each component word in complex words in English is borrowed independently of the other but that the surface form of whole complex words in English still influences sound mappings in complex loanword adaptation. This argument is extended to compound-type words such as “rush hour” in Sect. 5. For ease of reference, we refer to the first and second components as Components 1 and 2. We consider only forms in which Component 1 ends with a consonant and Component 2 begins with a pronounced vowel or /l/. To begin with, Component 1-edge consonants and word-medial consonants are mapped differently. For example, “all in” is not adapted as *[ollin] but as [orin] after turning stem-final /l/ to [r] between vowels, whereas “olive” is adapted as [ollibɨ] in Korean. In other words, if “all in” were borrowed as a whole, the /l/ would be geminate. This suggests that a morphologically complex word is not borrowed as a single unit but that each component word is parsed and adapted separately and independently of the other component as follows: (1)
“all in”: /ol/ + /in/ → [orin]
Further, a systematic analysis of the corpus data reveals that the Component 1-final obstruents, which have different manners of articulation, trigger different types of vowel epenthesis as follows: (2)
Adaptation of English words a. Component 1 fricative and affricate-final: vowel epenthesis-always Simplex loanword Complex loanword “give” [kibɨ] “give up” [kibɨəp] h h “catch” [k ecˇ i] “catch on” [khecˇhion] b. Component 1 stop-final i. Vowel epenthesis-always Simplex loanword Complex loanword h h “start” [sɨt at ɨ] “start up” [sɨthathɨəp] ii. Vowel epenthesis-never Simplex loanword Complex loanword “back” [pek] “back up” [pegəp] iii. Variable vowel epenthesis Simplex loanword Complex loanword h “make” [meik ɨ] ~ [meik] “make up” [meikhɨəp] ~ [meikhəp]
123
270
M. Oh
As shown in (2a), vowel epenthesis always occurs when Component 1 ends with a fricative or an affricate. By contrast, a stop-final Component 1 exhibits all three types of vowel epenthesis, that is, vowel epenthesis-always, vowel epenthesis-never, and variable vowel epenthesis. The simplex form is reflected in the complex loanword in (2a) and (2bi). However, when vowel epenthesis-never applies, the Component 1-final stop undergoes voicing before a vowel-initial Component 2 as shown in (2bii). Furthermore, when vowel epenthesis variably applies in the simplex form, the Component 1-final stop is not subject to voicing, as given in (2biii). This study considers why vowel epenthesis always occurs for a fricative or affricate-final Component 1, and why in vowel epenthesis-never cases, there is a voicing alternation for the Component 1-final stop. It also evaluates why variation is allowed in complex loanwords if and only if it is allowed in simplex loanwords. Finally, the question of why there is no voicing alternation in variation forms, when it does occur in vowel epenthesis-never cases is also addressed. To address these questions, we first demonstrate that with limited exceptions, the final consonant of Component 1 patterns with word-final consonants but does not do so with word-medial consonants in respect to sound mappings and vowel epenthesis in complex loanword adaptation. In so doing, we claim that complex loanwords are adapted on a component-by-component basis and that, furthermore, the limited exceptions to the component-by-component analysis can be explained by the splitbase effect and lexical conservatism (Steriade 2000). To further support a component-by-component analysis, we provide evidence for the correlation between the epenthesis pattern in singletons (e.g. “back” and “make”) as well as for that in complex words (e.g., “back up” and “make up”). We also demonstrate that some degree of perceptual similarity is enforced at the level of the whole complex word. Furthermore, by investigating the asymmetry between the final consonant of Component 1 and that of the loanword as a whole with respect to phonological alternations, we propose that whereas each component word is categorized as a stem, the complex loanword as a whole is morphologically categorized as a nominal word. These results shed light on three issues surrounding loanword adaptation. The first of these issues is which factors influence sound mappings in loanword adaptation. Most previous studies of loanword adaptation have centered on investigating how different sound mappings are conditioned by the prosodic environment within a word (Kenstowicz and Suchato 2006; Kim and Curtis 2000; Kang 2003). By contrast, the present study examines how sound mappings are determined by the morphological structure of source words, proposing that the same consonant can be adapted differently depending on whether it is morpheme-internal, component-straddling, or word-final. As shown in the adaptation of “all in” as [orin] (*[ollin]) and “olive” as [ollibɨ], we argue that the sequence of phones has one rendition if it is considered complex and another if the same sequence of phones is considered simple. This suggests that the morphology of a borrowed word should be taken into consideration in examining loanword adaptation. The phonology and phonetics of the source language also play a crucial role in loan adaptation (LaCharite´ and Paradis 2005; Silverman 1992; Steriade 2001;
123
Adaptation of English complex words
271
Peperkamp and Dupoux 2003). We considered to what extent loanword adaptation is phonological or perceptual. Component-by-component analysis of complex loanword adaptation raises the question of whether the output of a complex word as a whole from the source language can influence complex word adaptation. The investigation of complex English loanwords in Korean, such as “cut off”, indicates that the surface form of the whole complex word from the source language influences sound mappings. Specifically, “cut off” may be loaned as *[khəthophɨ] in that alternate loanwords for “cut”, [khəthɨ] and [khət], can induce the apparent single-unit adaptation as *[khəthophɨ] in the same way in which “make up” is adapted as [meikhəp]. However, the intervocalic /t/ is flapped in “cut off” in English, and the voiced allophone is mapped into [d] in [khədophɨ] to the exclusion of *[khəthophɨ] in Korean. Such output-to-output correspondence between the whole complex word from the source language and the complex loanword supports the perceptual view, not the phonological view, of loanword adaptation, in that the allophones of the source language, e.g., voiced flaps in English, can determine loanword sounds in the adapting language. Second, we consider the base for complex loanword adaptation. If a componentby-component analysis of complex loanword adaptation holds, each component is expected to be loaned as an independent base in the lexicon and combined with other components. “Make up” would thus pose a challenge to componentby-component analysis in that it can be adapted as [meikhəp] or [meikhɨəp] but not as *[meigəp], which is the form that would be predicted by component-bycomponent analysis: /meik/ + /əp/ after obstruent voicing between sonorants. Following Steriade’s (2000) theory of a split-base effect in which the phonological base functions as a base for word formation as well as a morphosyntactic base, we argue that the reason why [meikhəp] for “make up” shows a blend of surface properties of the listed allomorphs [meikhɨ] and [meik] is that the listed allomorph [meikhɨ] functions as a phonological base for the expression of phonologically desirable properties. To be specific, when [meik] and [əp] are combined, the compound loanword gives rise to [meikhəp] as opposed to *[meigəp]. Such an adaptation serves a dual purpose. One is to be faithful to the listed allomorph, [meik], in that it does not have an epenthetic vowel after the stop. The other is to be faithful to the aspirated stop of the other allomorph, [meikhɨ]. The theory of lexical conservatism dictates that the aspirated consonant must be found in a pre-existing allomorph because the grammar disprefers the use of forms that are novel or that lack any lexical precedents. In this way, given that multiple alternate loanwords can be listed in the lexicon of the adapting language, speakers can consult not only the morphosyntactic base but also the distinct allomorph as the phonological base to improve the sound mapping of source word sounds. In the course of our discussion of this topic, a formal analysis couched within the OT framework is provided (McCarthy and Prince 1995). Thirdly, we can assess the morphological category of the base for complex loanword adaptation. Simplex loanword-final segments are always placed at the end of the word, and they are not informative about the morphological category of the loanword. However, in complex loanword adaptation, Component 1 is followed by Component 2. By demonstrating the asymmetry between the Component 1-final
123
272
M. Oh
consonant and the final consonant of the loanword as a whole with respect to the phonological alternation, we claim that whereas each component of the borrowed complex word, e.g., [X] and [Y], is morphologically categorized as a stem, the complex loanword as a whole is borrowed as a single nominal phonological word: [[X]stem [Y]stem]phonological word. This argument can answer the question of the level at which vowel epenthesis occurs. Given that differences in the pattern of vowel epenthesis between Component 1-final fricatives/affricates and Component 1-final stops result from different bases in the Korean lexicon, we can conclude that vowel epenthesis takes place at the stem level. The rest of this paper is organized as follows: Section 2 provides a brief review of Korean phonology and the theoretical background that is relevant to this study. Section 3 discusses the domain of loanword adaptation and provides support for component-by-component analysis of complex loanword adaptation, i.e., the theory that proposes that each component of a complex word is loaned independently of other components. Section 4 examines the asymmetrical patterns of final stop and non-stop obstruents of loanwords with respect to vowel epenthesis, which can explain the disparity between the Component 1-final stop and the Component 1-final fricative or affricate in terms of the possibility of apparent one-unit mappings. Furthermore, counterexamples to component-by-component analysis are analyzed in terms of the split-base effect and lexical conservatism (Steriade 2000). This section also investigates the morphological structure of complex loanwords and provides evidence that whereas each component word is categorized as a stem, the complex loanword as a whole is morphologically categorized as a nominal word. Section 5 presents our findings in light of various theoretical issues surrounding loanword adaptation and offers a conclusion.
2 Background 2.1 Loanword adaptation Loanword research plays an important role in phonological theory, particularly because it reveals the role of phonetics and phonology between the source and the borrowing languages. There are two main approaches to loanword adaptation. The first approach can be termed the perceptual approach (Silverman 1992; Steriade 2001; Peperkamp and Dupoux 2003). According to the perceptual view, loanword adaptation is a process that maximizes the perceptual similarity between the loanword and the source word while satisfying the phonotactic constraints of the borrowing language, and the subphonemic realization of each language plays a crucial role in loanword adaptation. The second approach can be termed the phonological approach, and is best articulated by LaCharite´ and Paradis (2005). According to this view, because bilinguals, who have access to the phonology of both the source and borrowing languages, are typically the ones who perform loanword adaptation, the process of adaptation is inherently phonological, with little or no role for subphonemic variants in each language. According to the phonological view, a given phoneme of the source language is expected to be
123
Adaptation of English complex words
273
adapted uniformly regardless of its allophonic variation. Thus, the variable adaptation of a single phoneme would provide counterevidence to the phonological view. The case of English /s/ adaptation in Korean clearly demonstrates the sensitivity of segmental mapping to subphonemic information in the source language. English /s/, which is short in duration in the pre-consonantal position, is systematically borrowed as a lax /s/ in Korean, whereas English /s/ in all other contexts is loaned as the tense /s’/ in Korean because it is longer than the English /s/ in a cluster, e.g. “star” [sɨtha] vs “song” [s’oŋ] (Kim and Curtis 2000). The adaptation of an English word with a postvocalic word-final stop provides further evidence for the sensitivity of segmental mapping to subphonemic information in the source language. An English word with a postvocalic word-final stop variably triggers vowel epenthesis when borrowed into Korean (Kang 2003, p. 223). That is, for some words, /ɨ/ is consistently inserted after the word-final stop, e.g. “hip” [hiphɨ]. Another group of words show no vowel epenthesis, e.g. “pack” [phek]. For yet another group of words, forms with final vowel epenthesis and those without vowel epenthesis co-occur, e.g., “cut” [khəthɨ] ~ [khət]. Kang (2003) suggested that such variable vowel epenthesis maximizes perceptual similarity by being faithful to the optional release of word-final stops in English. On the other hand, the adaptation of the English /t/ in Korean supports the phonological approach because English /t/ is always mapped to [th] in Korean, although it is realized as a voiced flap in English and the Korean lax stop might be the best match to the sound; e.g., “water” [wəthə], *[wədə]. Likewise, segmental mappings in loanword adaptation can support both the perceptual similarity approach and the phonological approach. In this case, the intermediate position that takes into account a variety of factors to achieve the best match to the source word, including phonetics and phonology as well as orthography, can be pursued (Kenstowicz and Suchato 2006). Most previous studies on loanword adaptation have centered on segmental mappings between source word and loanword sounds of morphologically simplex words. In this regard, by focusing on the adaptation of consonants in complex words, this study sheds light on the issue of the factors that are relevant in sound mappings in loanword adaptation. It will be claimed that the morphology of the source words plays a crucial role in loanword adaptation and that the phonetic output of complex words as a whole in the source language still influences sound mappings in complex loanword adaptation. 2.2 Korean phonology In this section, we provide a brief review of the phonological system of Korean. An inventory of Korean consonants is shown in (3) below. (3) Inventory of Korean consonants (allophones in brackets) Labial Coronal Velar Glottal Lax stop p[b] t [d], cˇ [ǰ] k [g] Aspirated stop ph th, cˇh kh
123
274
M. Oh
Tense stop Fricative
p’
Nasal Liquid Glide
m
t’, cˇ’ s [∫] s’ [∫’] n l[r] y
k’ h ŋ w
Stops and affricates in Korean exhibit a ternary laryngeal distinction, i.e., lax, aspirated, and tense contrasts, whereas fricatives display a two-way laryngeal contrast, i.e., lax and tense contrasts. Lax stops are voiced between sonorants, and the liquid /l/ is realized as [r] in the onset. Korean syllables are maximally CVC (that is, they ignore the issue of long vowels and the status of on-glides). Sound sequences in a language are restricted by phonotactics. Onsets and codas display an unequal consonant distribution in Korean. All 19 consonantal phonemes (with the possible exception of the velar nasal) are pronounced in onsets, but only 7 can surface in codas: [p, t, k, m, n, ŋ, l]. To satisfy the coda constraint in Korean, neutralization occurs and consists of two processes: laryngeal neutralization in which consonants with marked laryngeal features (tense or aspirated) are neutralized to a lax stop, as shown in (4a), and place and manner neutralization in which /s, s’, h, cˇ, cˇh/ are neutralized into [t], as shown in (4b). (4)
Coda neutralization in Korean a. Neutralization in the laryngeal feature /suph/ [sup] “forest” /suph-i/ [suphi] h /puək / [puək] “kitchen” /puəkh/ [puəkhi] /pak’/ [pak] “outside” /pak’-i/ [pak’i] b. Neutralization in place and manner of articulation /nas/ [nat] “sickle” /nas’/ [nat] /nah/ [nat] “give a birth” /nacˇ/ [nat] /nacˇh/ [nat] “face” /nath/ [nat]
“forest” (Nom) “kitchen” (Nom) “outside” (Nom) “gave birth” “daytime” “each”
It is thus difficult to find a match in Korean for the rich set of coda consonants in English in the face of the Korean CVC syllable structure. The main interest of this paper lies in investigating the adaptation of Component 1-final consonants. The discussion of the borrowing of complex English words also requires a review of the morphological structure of Korean. The examples in (5) illustrate the asymmetrical behavior of native nominal and verbal paradigms. (5) a. Nouns i. /path/ [pat] ii. /kaps/ [kap] iii. /kaps/ [kap] b. Verbs i. /puth-ɨl/ ii. /əps-ta/ iii. /əps-i/
123
/path-ɨl/ [pathɨl] ~ [pasɨl] “field” (Acc) /kaps-to/ [kapt’o] “price-also” /kaps-i/ [kaps’i] ~ [kabi] “price” (Nom) [puthɨl] *[pusɨl] [əpt’a] [əps’i] *[əbi]
“attach” (Future) “not exist” (Decl) “not exist” (Adv)
Adaptation of English complex words
275
The coronal stop in the noun stem-final position undergoes assibilation before a vocalic suffix /ɨ l/, as in (5ai), whereas that in the verb stem-final position does not, as in (5bi). Korean does not allow any consonant cluster in the coda and one of the consonants in the stem final position deletes at the end of a word or before a consonant as shown in (5aii) and (5bii). However, consonant cluster simplification (CCS) applies optionally in the noun stem-final position, even before a vocalic suffix, as in (5aiii), but it does not apply in the verb stem-final position, as in (5biii). Likewise, the nominal and verbal stems behave differently with respect to phonological alternations. Such asymmetry between nouns and verbs has been analyzed in various ways. For example, Kenstowicz (1996) ascribed the asymmetry between nouns and verbs to the possibility of citing the stem in isolation. Specifically, he explained the variants in which CCS occurs even before a vocalic suffix in terms of Base-Identity because noun stems can stand alone and independently occurring forms can act as the Base. CCS cannot, however, be applied before a vocalic suffix in verbs because verb stems cannot occur in isolation and there is no Base for a comparison. On the other hand, Lee (2001) argued that the Base-Identity analysis does not explain why nouns have independent forms when verbs do not. He attributed the noun–verb asymmetry to the special status of nouns as prosodic words and contended that whereas Korean nouns are prosodic words, Korean verbs are not. Both of these analyses indirectly make reference to domain-specific faithfulness constraints for the noun category. According to Kenstowicz’s analysis, Base-Identity is applicable only when the noun stem is compared with the independently occurring form. Lee’s analysis of Korean nouns as prosodic words relies on the alignment constraint that requires the right edge of a prosodic word to coincide with the right edge of a syllable. The existence of such noun-faithfulness constraints suggests that the noun category is a salient domain (Beckman 1997; Smith 1997). Following Lee (2001), the present paper suggests that nouns and verbs constitute a different morphological category in Korean, specifically that nouns constitute prosodic words, and verbs constitute stems. Thus, it is clear that there are differences in the pattern between stem-final and word-final consonants.
3 The domain of complex loanword adaptation Previous studies in this area have focused less on the adaptation of complex words than on that of monomorphemic words. According to Ito’s (1990) study of the borrowing of English compounds into Japanese, each component of a compound word is adapted separately in loanword adaptation.2 However, this work addressed 2
Ito (1990) distinguished between word–word and stem–stem compounds in Japanese and argued that a distinction between two minimality requirements for stems and words also applies to the different patterning of single and compound loanword abbreviations in Japanese. For instance, monosyllabic members cannot occur in isolation, but they are allowed as compound members. For example, *jii (← jiinzu ‘jeans’) and *paN (← pantsu “pants”) cannot stand alone but are allowed only as compound members, as in [jii-paN] (← jiinzu pantsu “jeans pants”). Compound loanwords consist of two bounded stems that are permissible if they are composed of two morae. On the other hand, the word should at least be disyllabic. This study addresses those compound loanwords whose members are independent words.
123
276
M. Oh
only those compounds whose components are independent content words. On the other hand, the present study principally investigates the adaptation of complex English words into Korean when one component of a complex word is not an independent content word. In particular, it focuses on complex loanwords whose Component 1 ends with a consonant and whose Component 2 begins with a vowel or /l/. Through this, we investigate the unit of loanword adaptation over which adapters calculate sound mappings. To be specific, we questioned whether, for example, “up” is adapted by itself or as part of the preceding word in the adaptation of “make up”. In this section, by examining mainly the nominalized forms of phrasal verbs we argue that the domain of loanword adaptation is each component word of a complex word. Specifically, we examine those words that have a particle as the second component, such as “make up”. If a complex word in English is borrowed as a single unit into Korean, the complex word will behave like a single word. By contrast, if each component word of a complex word is adapted independently of the other component word, Component 1 and the complex word as a whole may behave differently. We argue for the component-by-component analysis, not for the singleton analysis, for complex loanword adaptation. To support this, we demonstrate the existence of Component 1- final consonant patterns with the adaptation of a word-final consonant, not a word-medial consonant, for both segmental mappings and vowel epenthesis. To begin with, the examples in (6) illustrate that only the Component 1-final consonant can trigger vowel epenthesis even before the Component 2-initial vowel, although such epenthesis can never occur word-internally. (6) a. Adaptation of English fricatives in Korean Simplex loanwords Complex loanwords Word-medial Word-final carnival [khanibal] drive [tɨraibɨ] drive in [tɨraibɨin] *[khanibɨal] *[tɨraibin] marshall [masˇal] wash [wəsˇi] wash up [wəsˇiəp] *[masˇial]3 *[wəsˇəp] b. Adaptation of English voiced stops in Korean Simplex loanwords Word-medial landing *[rendɨiŋ] [rendiŋ] jogging *[cˇogɨiŋ] ˇ [cogiŋ]
Complex loanwords Word-final hand [hendɨ]
handout [hendɨaut]4
log [rogɨ]
log out
[rogɨaut]
3
The vowel [ɨ] is a default epenthetic vowel in Korean. However, the vowels [i] and [u] are epenthesized after a palatal consonant and a labial consonant, respectively.
4
Inflectional suffixes follow the complex loanword as a whole, e.g., /hendɨaut/+ /i/ [hendɨausi] “handout” (Nom). The coronal stop /t/ in the noun stem-final position undergoes assibilation before a suffix.
123
Adaptation of English complex words
277
c. Adaptation of English affricates in Korean Simplex loanwords Word-medial pitcher [phicˇhə] *[phicˇhiə] manager [meniǰə] *[meniǰiə]
Complex loanwords Word-final catch [khecˇhi]
catch on
change [cˇheinǰi] change up
[khecˇhion] *[khecˇhon] [cˇheinǰiəp] *[cˇheinǰəp]
d. Adaptation of consonant clusters Simplex loanwords Word-medial Word-final quilting [khwilthiŋ] built [pilthɨ] *[khwilthɨiŋ]
Complex loanwords built in [pilthɨin] *[pilthin]
The data items in the simplex word-final position in (6) above illustrate how Component 1 would be borrowed in isolation. English voiced stops, fricatives, affricates, and consonant clusters before a vowel in word-medial position can never trigger vowel epenthesis in loanword adaptation, whereas the corresponding consonants in the final position of a simplex word and Component 1 can. For instance, “carnival” is always loaned as [khanibal] but not as *[khanibɨal], the simplex word “drive” is borrowed as [tɨraibɨ], and the complex word “drive in” is adapted as [tɨraibɨin] but not as *[tɨraibin]. Likewise, Component 1-final consonants do not pattern with word-internal consonants but with simplex wordfinal consonants with respect to vowel epenthesis. Such differences in the adaptation mode for English voiced stops, fricatives, affricates, and consonant clusters before a vowel, which depend on their positions, support the claim that each part of the complex word is loaned separately, not as a single unit. The second argument for component-by-component adaptation, as opposed to singleton adaptation, lies in different segmental mappings between Component 1-final consonants and morpheme-internal consonants. If the whole loanword is adapted as a unit, Component 1-final consonants are adapted in the same way as word-internal consonant in the prevocalic position. However, as shown in (7), differences in the ways in which English /l/ is adapted support the componentby-component adaptation.
123
278
M. Oh
(7) Adaptation of English /l/ in Korean Simplex English loanwords Word-internal: [ll] [ [r] ([ indicates preference.) tulip [thyullip] [ [thyurip] play [phɨllei] [ [phɨrei] calendar [khallendə] [[kharenda]
Complex English loanwords: Word-final: [r] *[ll] [l] all [ol] all in [orin] *[ollin] goal [k’ol] goal in [k’orin] *[k’ollin]5 hole [hol] hole in one [horinwən] *[hollinwən] all in one [orinwən] *[ollinwən]
The examples in (7) show that the final consonant /l/ of Component 1 is not adapted as [ll] but as [r] before the vowel-initial member. By contrast, [ll] is preferred to [r] as the sound corresponding to word-internal /l/. The disparity between word-medial adaptation as geminate [ll] and end-of-Component 1 adaptation as [r] supports component-by-component adaptation rather than singleton adaptation. Under a component-by-component analysis, Component 1 would be borrowed in isolation. Thus, “all” would be borrowed as [ol] first and then combined with [in] “in” to yield [orin], not *[ollin] when considering “all in”. In Korean, /l/ surfaces as [r] between vowels, e.g., /sal-i/ [sari] “skin” (Nom). The different ways in which English /r/ is adapted also favor componentby-component adaptation: (8) Adaptation of English /r/ in Korean Simplex English loanwords Complex English loanwords: Ø *[r] Word-internal: [r] Word-final (coda): Ø rotary [rothari] car [kha] catering [khethəriŋ] park [phakhɨ]
cheer up [cˇhiəəp] *[cˇhiərəp] overeat [obaithɨ] *[obarithɨ]
The word-internal /r/ in English is adapted as [r] before a vowel, whereas /r/ in the word-final or coda position is not mapped to [r]. Component 1-final consonant /r/ patterns with word-final /r/ in that it is not mapped to any sound before a vowelinitial member. Because of this, the different adaptations of /r/ depending on its position also support component-by-component adaptation as opposed to singleton adaptation. For example, in “cheer up”, “cheer” (Component 1) would be borrowed as [cˇhiə] first and then combined with [əp] “up” to yield [cˇhiəəp], not *[cˇhiərəp]. Further supporting evidence derives from the adaption of the sequence of /rl/. A word-internal /rl/ in English is always adapted as [ll] in Korean, e.g., “darling” [talliŋ] and “Beverly (Hills)” [pebəlli]. However, complex loanwords illustrate different behaviors with respect to [ll] gemination. For example, “overload” is loaned as [obərodɨ] but not as *[obəllodɨ]. English /l/ in the onset gets adapted as [r] 5
“Goal in” is not an English expression. It is a Korean invention of what could be an English expression. As a reviewer suggests, such an example demonstrates the abstract nature of the perceptual similarity, in that the actual expression “goal in” would be foreign to native English speakers.
123
Adaptation of English complex words
279
in Korean, e.g., “line” [rain]. Then “overload” is loaned as [obərodɨ] as a combination of [obə] and [rodɨ]. To summarize this section, the asymmetry between Component 1-final consonants and simplex word-internal consonants with respect to vowel epenthesis and segmental mappings supports the claim that each component word is borrowed separately.
4 Adaptation of complex words 4.1 Asymmetry between word-final stops and non-stops in loanword adaptation In the previous subsection, we demonstrated that Component 1-final fricatives and affricates can trigger vowel epenthesis before a vowel-initial component: (9)
give up [kibɨəp] *[kibəp]; catch on [khecˇhion] *[khecˇhon]
By contrast, when the first component ends with a stop, singleton adaptation is also possible before a vowel-initial component: (10)
make up [meikhɨəp] ~ [meikhəp], log in [rogɨin] ~ [rogin]
Given that each component of a complex word is adapted first and then combined with the adjacent component word in the adaptation of complex loanwords, such differences in patterns can be explained by the fact that different consonants may induce different rates of vowel epenthesis at the word-final position. When an English word ending with a fricative or an affricate is adapted in Korean, a vowel is uniformly inserted after the final obstruent. By contrast, English wordfinal voiceless stops are adapted in three different ways: as a lax stop, as an aspirated stop with an epenthetic vowel, or as an alternating loanword between a final lax stop and an aspirated stop with an epenthetic vowel (Kang 1996; Kang 2003). English word-final voiced stops also induce epenthesis variably. This is illustrated in (11). (11) a. Adaptation of English words ending in a fricative or an affricate (NAKL 1991) scarf [sɨkhaphɨ], boss [posɨ], switch [sɨwicˇhi], shirts [sˇəcˇhɨ]6 b. Adaptation of English words ending in a voiceless stop (NAKL 1991) i. Vowel epenthesis: peak [phikhɨ], week [wikhɨ] ii. Variable vowel epenthesis: jeep [cˇip] ~ [cˇiphɨ], cake [kheik] ~ [kheikhɨ] iii. No vowel epenthesis: kick [khik], back [pek], look [ruk] c. Adaptation of English words ending in a voiced stop (NAKL 1991) i. Vowel epenthesis: league [rigɨ], bed [pedɨ] ii. Variable vowel epenthesis: pyramid [phiramit] ~ [phiramidɨ], zigzag [cˇigɨcˇek] ~ [cˇigɨcˇegɨ] iii. No vowel epenthesis: Arab [arap], big [pik] 6
The quality of an epenthetic vowel is not determined by the place of the articulation of Korean consonants but by that of the English input (Oh 1992). Here /ɨ/ is inserted when the English input ends with an alveolar fricative or an affricate, e.g., “sports” [sɨphocˇhɨ], whereas /i/ is chosen when the English input ends with a palatoalveolar consonant, e.g., “coach” [khocˇhi].
123
280
M. Oh
Such differences in the ways in which a word-final stop can be adapted may be ascribed to the degree of the stop release and the duration of the vowel preceding the stop in the source language (Kang 2003; Oh and Kim 2006). This paper now examines the rate of vowel epenthesis after a postvocalic obstruent in loanwords. The data on loanwords with a word-final postvocalic obstruent were drawn from NAKL (1991). To begin with, examination of the NAKL list (1991) showed that English words ending with a nonstop obstruent are uniformly adapted with vowel epenthesis (/f/: N = 20, /v/: N = 26, /s/: N = 190, /z/: N = 36, /ɵ/: N = 2, /∫/: N = 18, /Ʒ/: N = 6, /ts/: N = 8, /ʤ/: N = 1, /t∫/: N = 18, /ʤ/: N = 22). By contrast, English words ending in a stop are adapted in three different ways: vowel epenthesis, variable vowel epenthesis, and no vowel epenthesis (11b, c). The rate of each realization is given in (12). (12) The rate and number of final-stop renditions based on Kang’s (2003) analysis of NAKL (1991) Loanword realization
Vowel epenthesis
Variable vowel epenthesis
No vowel epenthesis
Total
Word-final stop in English p
b
t
d
k
g
Avg.
18 %
44 %
59 %
99 %
26 %
75 %
50 %
(N = 13)
(N = 4)
(N = 91)
(N = 68)
(N = 31)
(N = 18)
(N = 225)
4%
0%
14 %
1%
1%
4%
6%
(N = 3)
(N = 0)
(N = 21)
(N = 1)
(N = 1)
(N = 1)
(N = 27)
78 %
56 %
27 %
0%
73 %
21 %
44 %
(N = 56)
(N = 5)
(N = 42)
(N = 0)
(N = 87)
(N = 5)
(N = 195)
100 %
100 %
100 %
100 %
100 %
100 %
100 %
(N = 72)
(N = 9)
(N = 154)
(N = 69)
(N = 119)
(N = 24)
(N = 447)
Likewise, the corpus data from the NAKL list (1991) showed the asymmetry between the adaptation of English word-final fricatives or affricates and that of English word-final stops with respect to vowel epenthesis. Given that each component of a complex word is adapted first and then combined with the adjacent component word in the adaptation of a complex word, different patterns of complex loanwords depending on the Component 1-final consonant can be expected. Specifically, if Component 1 ends with a fricative or an affricate, it triggers vowel epenthesis with no variation, and a single loanword is listed in the Korean lexicon. Then it is combined with a following component, yielding a complex loanword. This point is exemplified in (13).
123
Adaptation of English complex words
281
(13) Complex English loanwords whose Component 1 ends with a fricative or an affricate (NAKL 1991) a. “focus on”: [phokhəs’ɨ] + [on] → [phokhəs’ɨon] *[phokhəs’on] b. “watch out”: [wəcˇhi] + [aut] → [wəcˇhiaut] *[wəcˇhaut] The first components, “focus” and “watch”, have the singleton loanword bases [phokhəs’ɨ] and [wəcˇhi], and they are combined with the components [on] and [aut], respectively. On the other hand, if Component 1 ends with a stop, the loanword can be listed in the Korean lexicon in two different ways. One way is as a singleton base either with or without an epenthetic vowel, e.g. “bed” [pedɨ] and “back” [pek]. The other way is for the loanword to be listed with two allomorphs as a result of variable vowel epenthesis, e.g. “cake” [kheik] ~ [kheikhɨ]. Then, the component-by-component adaptation predicts three types of complex loanwords depending on the availability of vowel epenthesis. This prediction is exemplified below in (14). (14) Adaptation of complex loanwords whose Component 1 ends in a stop a. Vowel epenthesis-never “back up” [pegəp] *[pekhəp] cf. “back” [pek] h “pop up” [p abəp] *[phaphəp] cf. “pop” [phap] b. Vowel epenthesis-always “break up” [pɨreikhɨəp] *[pɨreikhəp] cf. “break” [pɨreikhɨ] “speed up” [sɨphidɨəp] *[sɨphidəp] cf. “speed” [sɨphidɨ] c. Variable vowel epenthesis “make up” [meikhɨəp] ~ [meikhəp], *[meigəp] cf. “make” [meikhɨ] ~ [meik] “log in” [rogɨin] ~ [rogin] cf. “log” [rogɨ] ~ [rok] “Back” is listed as [pek] in the lexicon, whereas “back up” is adapted as [pegəp] (\-[pek] + [əp]) through the intersonorant voicing of obstruents in (14a). “Break” is adapted only as [pɨreikhɨ], and “break up” is loaned as [pɨreikhɨəp] in (14b). On the other hand, “make” has dual allomorphs listed in the lexicon, [meikhɨ] and [meik], and “make up” is adapted either as [meikhɨəp] or [meikhəp], as shown in (15).7 (15) a. “make up” [meikhɨ] + [əp] → [meikhɨəp] b. “make up” [meik] + [əp] → [meikhəp] Note that the complex loanword in (15b) is loaned as [meikhəp], not as *[meigəp], which would be incorrectly predicted when using component-by-component analysis. If [meik] were combined with [əp], the final consonant [k] would undergo obstruent voicing between sonorants. On the surface, this example appears counter to the component-by-component analysis that we propose as a model for the 7
The NAKL list (1991) does not include the loanword for “make.” However, a Google search indicates that both “메이크” \meikhɨ[ and “메익” or “메잌” \meik or meikh[ are possible, although the allomorph [meik] “make” occurs far less frequently than the other allomorph [meikhɨ].
123
282
M. Oh
adaptation of complex words. However, we show in the next subsection that this example can, in fact, be accounted for by the component-by-component analysis within the two-base approach. At this point, we can examine two data sets to determine whether the asymmetrical adaptation of complex words depending on the Component 1-final consonant is supported. One data set is drawn from NAKL (1991). The NAKL list does not contain many complex loanwords; rather, it consists mainly of simplex loanwords from English. Thus, to complement the NAKL list we also collected more complex loanwords by conducting Google searches. The NAKL list is very faithful to normative conventions. Thus, to conduct a quantitative study of variable loanword alternates for each complex word in the NAKL list, the rates of loanword adaptation with or without an epenthetic vowel were calculated through Google searches. Appendix A shows the full list of complex loanwords whose Component 1 ends in an obstruent and whose Component 2 begins with a vowel, alongside Google search results.8 Spellings for Google searches were chosen as follows: for “catch on,” we searched for “캐치온” (\khe.cˇhi.on[) with an epenthetic vowel in Component 1 and “캐촌” (\khe.cˇhon[) without an epenthetic vowel in Component 1.9 For a voiced stop case such as “dug out” “더그아웃” (\tə.kɨ.a. ut[) with an epenthetic vowel in Component 1was searched. However, there are potentially two different choices for the loanword without an epenthetic vowel: “더 가웃” (\tə.ka.ut[) and “덕아웃” (\tək.a.ut[). These two choices were both searched because they are pronounced in the same way, i.e. [təgaut]. On the other hand, when Component 1 ends with a voiceless stop, we searched more possibilities. Specifically, for “make up,” we searched for “메이크업” \me.i.khɨ.əp[ with an epenthetic vowel at the end of Component 1 and for “메이컵” \me.i.khəp[ and “메익업” \me.ik.əp[ without an epenthetic vowel at the end of Component 1. However, we did not include “메잌업” \me.ikh.əp[ since it is not evident whether this would surface as [meigəp] after coda neutralization or [meikhəp] without coda neutralization. Figure 1 summarizes the overall rate of alternate complex loanwords averaged across the NAKL data and Google search results by Component 1 ending in a stop or non-stop obstruent. The rate of the availability of alternate loanwords between Component 1-final C (without an epenthetic vowel) and CV (with an epenthetic vowel) is substantially higher when Component 1 ends with a post-vocalic stop than when it ends with a non-stop obstruent (χ2 = 34.85; df = 1, p \ 0.0001). In other words, Component 8
Unlike singleton loanwords, homophones between native words and loanwords in the case of complex loanwords were rarely identified by Google searching. However, when homophones for a complex loanword were found, the search term was entered with the complex word in English. “Rent a car” was excluded from the NAKL data. An article was always treated as part of the component word. Thus, “rent a car” was borrowed as “렌터카” \renthəkha[. The rendition of articles warrants further analysis.
9 \[ represents the orthographic representation. There are alternate spellings for English loanwords, particularly for English vowels. For example, “catch on” can be spelled as either “케치온” \khechion[ or “캐치온” \khεchion[. However, the Appendix shows only the two representative forms relevant to the purpose of the discussion: one form with vowel epenthesis and the other without vowel epenthesis at the end of the first component.
123
Adaptation of English complex words
283
Fig. 1 The rate of alternate loanwords in complex loanwords
1-final fricatives and affricates are significantly more likely to trigger vowel epenthesis than the corresponding final stop. The substantial difference in the rate of alternate loanwords depending on the Component 1-final consonant from the source language supports the componentby-component analysis of complex loanword adaptation. That is, the fact that a fricative terminating Component 1 is always loaned with vowel epenthesis follows naturally from component-by-component analysis, in which each component is parsed morphologically, loaned independently of the other, and then combined with the adjacent component. If this is true, then component-by-component analysis of complex loanword adaptation can also account for the reason why the rate of alternate loanwords allowing both Component 1-final C and CV is significantly higher when Component 1 ends with a stop than when it ends with a fricative or an affricate. Before proceeding, more empirical support for the central prediction made by component-by-component analysis can be provided, i.e., the correlation between the epenthesis pattern in singletons (e.g.,“back” and “make”) and that in the complex words (e.g.,. “back up” and “make up”). For this purpose, we used a two-step process. First, we selected from the Appendix those data items whose Component 1 ends in either a voiced or a voiceless stop and divided them into three groups as demonstrated in (14); “vowel epenthesis-never”, “variable vowel epenthesis”, and “vowel epenthesis-always”. A word was categorized as variable if each of the two variants occurred at least 1% of the time. Next, we took the Component 1 word from each item and used Google searching to determine how often it was adapted with final C (no epenthetic vowel) and CV (epenthesis). Twenty-one simplex words came from “vowel epenthesis-never” complex items, 26 simplex words from “variable vowel epenthesis” items and 17 simplex words from “vowel epenthesis-always” items. According to component-by-component analysis, the singleton’s behavior should correlate with the classification of the complex words it belongs to. For example, “back” occurs in “back up” which never has epenthesis. So, “back” as a singleton should also show little epenthesis. For words that allow the apparent single-unit adaptation, i.e., “make up”-type words, we expect the singleton (“make”) to have both epenthesis and no-epenthesis.
123
284
M. Oh
Fig. 2 The rate of alternate loanwords in simplex loanwords used as Component 1
Figure 2 demonstrates how the first component is borrowed in isolation according to the three categories of complex word, showing that the prediction is borne out. The percentages shown are averages over the set of simplex words (types) for each category: for example, the 90% figure is an average over the 17 simplex words found in the “epenthesis-always” category. Indeed, the complex word class significantly affected the rate of vowel epenthesis of Component 1 in isolation (χ2 = 32.56, df = 2, p \ 0.0001). Both pairwise comparisons using Fisher’s exact test are also significant: Component 1 words in the “vowel epenthesis-never” class exhibit significantly less vowel epenthesis than the “variable vowel epenthesis” class (p \ 0.0001,) and the rate for the “variable vowel epenthesis” class is significantly lower than for the “vowel epenthesis-always” class (p = 0.0039). In other words, the rate of simplex epenthesis follows the order of “vowel epenthesis-always” [ “variable vowel epenthesis” [ “vowel epenthesisnever”. That is to say, the epenthesis pattern in singletons (e.g. “back” and “make”) is closely correlated with that in complex words (e.g. “back up” and “make up”). Table (16) compares component-by-component analysis with the singleton analysis based on the discussion thus far. (16)
Singleton analysis
Component-bycomponent analysis
Component 1-final fricative
Incorrect prediction; *[phokh s’on] “focus on”
Correct prediction;
“Make up”-type words
No variation is incorrectly predicted.
Variation is predicted, but *[meig p] is incorrectly predicted.
“All in”-type words
Incorrect prediction; *[ollin]
Correct prediction; [orin]
Analysis Loan pattern
[phokh s’ on] “focus on”
The table shows that component-by-component analysis can clearly explain various patterns of complex loanwords but cannot explain “make up”-type words. Then, the question becomes how such “make up”-type words can be accounted for under the component-by-component analysis. The next section addresses this
123
Adaptation of English complex words
285
question and demonstrates that this apparent contradiction can be explained in terms of the split-base effect. 4.2 The split-base effect in complex loanword adaptation It was found that epenthesis always occurs at the end of Component 1 if it ends in a fricative or an affricate but only variably if it ends with a stop. It was also revealed that the epenthesis patterns in singletons are closely related to those in the complex words. These two findings fit with component-by-component analysis. However, those complex loanwords whose Component 1 ends with a voiceless stop, as shown in (17), cannot be fully explained by the component-by-component analysis. (17) Complex English loanwords whose Component 1 ends with a voiceless stop a. make up [meikhəp] ~ [meikhɨəp] *[meigəp] b. check in [cˇhekhɨin] ~ [cˇhekhin] *[cˇhegin] c. wake up [weikhəp] ~ [weikhɨəp] *[weigəp] “Make up” is adapted as either [meikhəp] or [meikhɨəp] but not as *[meigəp] (NAKL 1991). Given that “make” is adapted as either /meikhɨ/ or /meik/ and combined with the following member /əp/, we expect only [meikhɨəp] or *[meigəp], which in fact does not occur. In other words, using component-by-component analysis can clearly predict one of the alternate loanwords, [meikhɨəp], i.e., /meikhɨ / + /əp/ → [meikhɨəp], but the occurrence of [meikhəp] and the non-occurrence of *[meigəp] are puzzling. There can be two possible explanations for this result. First, the [meikhəp] form of “make up” may be derived by an /ɨ/-deletion process in Korean: /meikhɨ/ + /əp/ → [meikhəp]. In Korean, the stem-final /ɨ/ is always deleted before a vowel initial suffix, e.g., /s’ɨ-ə/ → [s’ə] “write” (imperative form). However, /ɨ/-deletion never applies to nouns, as shown in (18). (18) Lee (2001, p. 378) a. kɨ +eke [kɨege] *[kege] “to him” b. pəsɨ+ esə [pəs’ɨesə] *[pəs’esə] “in the bus” The /ɨ/-deletion does not apply between the loanword /pəs’ɨ/ and the vowel-initial suffix /esə/, and thus, [meikhəp] “make up” cannot be derived from /meikhɨ/ + /əp/ through /ɨ/-deletion because loanwords are always categorized as nouns regardless of whether English source words are adjectival or verbal in character in English, as shown in (19) (Oh 1995). (19) Data from Oh (1995) and the NAKL (1991) a. Native Korean words nole-hata [norehada] “sing” “song-do”
123
286
b. Simplex loanwords smart-hata [sɨmathɨhada] real-hata [riəlhada] drive-hata [tɨraibɨhada] mark-hata [makhɨhada] c. Complex loanwords make up-hata [meikhəphada]
M. Oh
“be smart” “be real” “drive” “mark” “make up”
As shown in (19a), /hata/ “to do” is attached to a noun, and thus the whole word is turned into a verb in Korean. The examples in (19b) illustrate that adjectives and verbs such as “smart”, “real”, “drive”, and “mark” in English are uniformly treated as nouns because /hata/ is directly attached to them. Because of this, the epenthetic vowel /ɨ/ at the end of the English loanword /meikhɨ/, which is treated as a noun, cannot be subject to /ɨ/-deletion. Notice that the complex loanword is also combined with /hata/ as shown in (19c). Furthermore, the /ɨ/-deletion analysis cannot explain why /ɨ/ is never deleted in the complex loanwords for which Component 1 ends with a fricative or an affricate. A second possible explanation can be found in the relative frequency of complex loanwords (Hay 2001).10 If the complex word “make up” occurs more frequently than each component word, then [meikhəp] can be categorized as a single unit instead of the two individual units of “make” and “up”. This explanation has two disadvantages. First, it cannot account for why such single-unit adaptation occurs only when Component 1 ends with a stop. As discussed in Sect. 3, the English /l/ is not adapted as [ll] at the end of Component 1, although word-internal /l/ is more likely to be loaned as [ll]: [orin], *[ollin] “all in” vs. [ollibɨ] “olive”. Further, an English word-final fricative or affricate is always rescued by an epenthetic vowel, and this type of apparent singleton adaptation rarely occurs when Component 1 ends with a fricative or an affricate.11 The other shortcoming of the frequency-based account derives from making a frequency comparison between the complex word and its individual units. For example, the Google hit count for [meikhəp] was 3,600,007, whereas it was 10,600,000 for [əp].12 However, one may point out that the more pertinent frequency comparison would be between “make up” and “make” because the focus of the discussion is on how Component 1 behaves with respect to vowel epenthesis. The Google search results indicate that the loanword for “make” 10
I am grateful to Bruce Hayes and Adam Albright for suggesting this possibility.
11
The loanwords for “push up” and “close-up” can be counterexamples to the generalization that Component 1-final fricatives always trigger vowel epenthesis since they have variants without vowel epenthesis, e.g., “푸샵” \phusyap[ and “클로졉” \khɨlloǰəp[. The final vowel of the variant [phu∫ap] for “push up” suggests that the loanword is introduced through Japanese since “up” would be always loaned as [əp] if the source word came from English; cf. “푸시업” \phusiəp[. But “클로접” \khɨlloǰəp[ still remains as a counterexample to the generalization. 12
[əp] may include Korean and Sino-Korean homophones, e.g., /əp/ “work” in Korean. Thus, we searched for the English loanword for “up” by only listing both “업” \əp[ in Korean and the English word “up”. For a fair comparison, we also searched the English loanword for “make up” by listing both “메이컵” \meikhəp[ in Korean and the English word “make up”. However, the Google hit count for “make up” was taken (and given in the Appendix) only when Korean spellings were given without the English word.
123
Adaptation of English complex words
287
alone is used far less frequently than that for “make up”. From this perspective, this example may favor the frequency-based account. However, the example of “wake up” clearly argues against the frequency-based account. “Wake up” has two alternate loanwords, [weikhəp] and [weikhɨəp]. The Google hit count for [weikhəp] was 452, 204 whereas it was 696,000 for [weikhɨ].13 Thus, the frequency account cannot explain the apparent singleton adaptation of [weikhəp] “wake up” because here the occurrence of the component word outnumbered that of the complex word. The question thus remains of how the occurrence of [meikhəp] and the nonoccurrence of *[meigəp] will be explained under the component-by-component adaptation. At this point, we can re-examine the cases of complex English loanwords whose Component 1 ends in a stop. For ease of reference, the examples in (14) are repeated in (20). (20) (=(14) Adaptation of complex loanwords whose Component 1 ends in a stop a. Vowel epenthesis-never “back up” [pegəp] *[pekhəp] cf. “back” [pek] h “pop up” [p abəp] *[phaphəp] cf. “pop” [phap] b. Vowel epenthesis-always “break up” [pɨreikhɨəp] *[pɨreikhəp] cf. “break” [pɨreikhɨ] “speed up” [sɨphidɨəp] *[sɨphidəp] cf. “speed” [sɨphidɨ] c. Variable vowel epenthesis “make up” [meikhɨəp] ~ [meikhəp] *[meigəp] cf. “make” [meikhɨ] ~ [meik] “log in” [rogɨin] ~ [rogin] cf. “log” [rogɨ] ~ [rok] The first generalization we can draw from the examples in (20) is that the aspirated stop is avoided as a loanword sound if the listed allomorph does not contain an aspirated stop: [bek] + [əp] → [pegəp] *[pekhəp] “back up”. In other words, *[pekhəp] is not possible because there is no surface allomorph in which the aspiration is evident for “back”. The second generalization is that the complex loanword [meikhəp] “make up” demonstrates a blend of surface properties of the listed allomorphs [meikhɨ] and [meik]. The first generalization can be easily explained by component-by-component analysis. This section focuses mainly on the second generalization. We contend that [meikhəp]-type cases provide empirical support for Steriade’s (2000) theory of a split-base effect and the idea of lexical conservatism. Steriade (2000) did not conflate the concepts of phonological, morphosyntactic, and semantic bases. Instead, the work separated bases into phonological and morphosyntactic bases. That is, it claimed that the morphosyntactic base is not isomorphic to the phonological base of the word and further contended that the speaker obtains access not only to the morphosyntactic base but also to the phonological base for word formation. This is referred to as the split-base effect, examples of which are given in (22). Further, the work argued for lexical conservatism, which evaluates candidates 13 We searched for the English loanword for “wake up” and “wake” by listing both “웨이컵” \weikhəp[ and “웨이크” \weikhɨ[ in Korean and the English word “wake up”.
123
288
M. Oh
by searching the lexicon for some non-specific and listed allomorphs of a morpheme possessing any phonological property, including stress patterns and segmental properties. Lexical conservatism reflects a set of grammatical preferences that disfavor the use of forms that are phonologically novel or lack lexical precedents (Steriade 2000). In this account, to form a complex adjective such as remédiable, the speaker must consult not only the verb rémedy, which serves as the morphosyntactic base, by satisfying -able’s subcategorization needs but also the adjective remédial, which serves as the phonological base because it lends its stress pattern to the -able form. There is a question as to why the split-base effect is needed. By adopting the stress pattern of remédial, the -able form is improved stress-wise; the lapsed string is diminished compared to *rémediable. Likewise, the split-base effect always arises when the morphosyntactic base of affixation lacks a phonological property that is desirable in the derivative. In such cases, the derivative may adopt the phonology of a distinct listed allomorph. However, it is important that only the listed allomorphs are accessible to the speaker for the purpose of word formation. Steriade (2000) also conducted a survey to compare –able forms based on similar verbs such as parody and remedy. She found that ‘remédiable’ is preferred over other variants, while ‘párodiable’ is much more preferred over *paródiable. She accounts for the different preference for these two –able forms in terms of the splitbase effect and lexical conservatism: Stress is able to shift from rémedy to remédible because the word remédial is already known to speakers, whereas the shift from párody to *paródiable has no such lexical support. The relevant constraint is formulated in (21). (21)
(Steriade 2000)
Lex [±stress]: For any stressed syllable σ in the target form T(μ), there is a correspondent σ’ in some listed allomorph, L(μ), such that σ’ is stressed and the nuclear vowels of σ and σ’ are identical. With the constraint ranking of Lex [±stress] over *Lapse σσσ, which disallows three consecutive unstressed syllables, the following tableaux illustrate how remédiable is possible but *paródiable is not. (22) (Steriade 2000) a. Listed allomorphs: rémedy, remédy- in remédial Lex [±stress] *Lapse σσσ ☞remédiable ok: remédial rémediable ok: rémedy *! b. Listed allomorphs: párody but not *paródial Lex [±stress] *Lapse σσσ ☞párodiable ok: párody * paródiable *!
123
Adaptation of English complex words
289
The tableaux in (22) illustrate that the stressed syllable of the verbal base can be changed only when there is a listed allomorph with a shifted stress pattern. It is possible to further examine how the split-base effect and lexical conservatism can explain [meikhəp]-type words. According to Steriade (2000), a lexical precedent for the phonological base is searched for in the lexical paradigm. However, in the adaptation of complex words, alternate allomorphs listed for a single source word are consulted. The complex loanword [meikhəp] “make up” exhibits a blend of surface properties of listed allomorphs [meik] and [meikhɨ]. On the one hand, it is faithful to [meik] in that it does not have an epenthetic vowel after the final stop. On the other hand, the aspirated stop [kh] is taken from the other listed allomorph [meikhɨ]. When such multiple allomorphs are lacking, single-unit adaptation is not possible.14 Thus, in the absence of *[pekhɨ] for “back”, *[pekhəp] “back up” is not possible. In other words, singleton adaptation hinges on the availability of variable allomorphs. The correlation between the epenthesis pattern in singletons and the epenthesis pattern in the complex words was discussed in the previous section. Before presenting an analysis of complex loanword adaptation couched within the OT framework, we provide a list of our relevant constraints in (23), wherein KK and EK stand for Korean to Korean and English to Korean, respectively. (23) a. KK- Lex C[aspiration]: Any aspirated C in the output should have a correspondent C’ in some listed allomorph of μ in Korean that is identical to C in aspiration. b. No IntSon VL: No voiceless lax obstruents are allowed between sononants. c. EK-Ident [voice]: The voicing specification of the output must be faithful to that in the English input. d. Max-V: Every vowel in the input has a correspondent in the output. e. Dep-V: Every vowel in the output has a correspondent in the input. We can examine how the constraints in (23) interact to yield the optimal outputs for the three optimal complex loanwords discussed above.
14 The analysis proposed in this paper predicts that for complex loanwords whose component 1 ends in a voiceless stop, if epenthesis is attested, a variant without epenthesis but with a lenis stop should not be possible. However, as a reviewer points out, “스트라잌아웃” \sɨ.thɨ.ra.ik.a.ut[ is attested in addition to “스트라이크아웃” \sɨ.thɨ.ra.i.khɨ.a.ut[. We notice that the Component 1 is adapted as a 4 syllable word, “스트라잌” \sɨthɨraik[. It may be later syllables in long words are positionally weak and speakers may allow neutralizing a contrast in late syllables in long words.
123
290
M. Oh
(24) a. English input: [meɪkʌp]; Listed allomorphs for the Korean base: [meik] ~ [meikhɨ], [əp] KK-Lex EK-Ident Max-V Morphosyntactic base: No IntSon C[aspiration] [voice] [meikhɨ], [əp] VL ☞ i. [meikhɨəp] ii. [meikhəp] *! iii. [meigəp] *! iv. [meikəp] *! b. English input: [meɪkʌp]; Listed allomorphs for the Korean base: [meik] ~ [meikhɨ], [əp] Morphosyntactic base: No KK-Lex EK-Ident Max-V Dep-V [meik], [əp] IntSon C[aspiration] [voice] VL i. [meikhɨəp] *! h ☞ ii. [meik əp] iii. [meigəp] *! iv. [meikəp] *! c. English input: [bækʌp]; Listed [pek], [əp] Morphosyntactic base: No [pek], [əp] IntSon VL i. [pekhɨ əp] ii. [pekhəp] ☞ iii. [pegəp] iv. [pekəp] *!
allomorphs for the Korean base: KK-Lex EK-Ident Max-V Dep-V C[aspiration] [voice] *! *!
* *
In (24a), Candidate (24aiv) fatally violates the un-dominated No IntSon VL constraint, which prohibits any voiceless obstruents between sonorants. Candidate (24aiii) is not optimal in that it does not satisfy highly ranked EK-Ident [voice]; that is, the voiced stop [g] is not faithful to the corresponding sound [k] in the English input. Candidates (24ai) and (24aii) satisfy KK-Lex C[aspiration] in that the listed allomorph [meikhɨ] contains the aspirated stop. However, Candidate (24aii) is suboptimal in that it violates Max-V; that is, the final vowel of the morphosyntactic base is lost. Thus, Candidate (24ai) is the optimal output. In (24b), candidates (24bi) and (24bii) satisfy KK-Lex C[aspiration] as the aspirated stop has a correspondent in the listed allomorph [meikhɨ]. Unlike in (24a), however, Candidate (24bii) surpasses Candidate (24bi) because Candidate (24bi) violates Dep-V by inserting a vowel that is not in the morphosyntactic base.
123
Adaptation of English complex words
291
(24c) illustrates a case in which only a single base, rather than multiple bases, is listed in the Korean lexicon for Component 1 of a complex word. In (24c) Candidates (24ci) and (24cii) are suboptimal because they violate KK-Lex C [aspiration], in that the voiceless aspirated stop [kh] has no correspondent in the Korean base. Candidate (24ciii) is selected as the optimal output although it violates EK-Ident [voice] because it contains the voiced stop [g] as a correspondent to the voiceless stop [k] in English. This demonstrates that KK-Lex C[aspiration] outranks EK-Ident [voice]. Thus, it can be said that the first piece of evidence for the validity of the lexical conservatism analysis is that the existence of a [meikhəp]-type variant can be predicted by the existence of both [meikhɨ] and [meik]. A second piece of evidence comes from English words ending in a fricative or an affricate. The question is how the constraint ranking established thus far can account for cases such as “catch on”, whose Component 1 is listed as a singleton base with an epenthetic vowel in the Korean lexicon. The tableau in (25) illustrates such an example. (25) English input: [kæʧƆn]; Listed allomorphs for the Korean base: [khecˇhi], [on] Morphosyntactic base: No IntSon KK-Lex EKMax-V [khecˇhi], [on] VL C[aspiration] Ident[voice] ☞ a. [khecˇhi on] b. [khecˇhon] *! h c. [k eǰion] *! The complex word “catch on” is adapted as [khecˇhion] because Candidate (25a) satisfies all the relevant constraints. That is, the reason [t∫]-final words behave differently is that a *[khet] (← /khecˇh/) form never exists as a base for “catch”. Component-by-component analysis can clearly explain the following examples in (26): (26) a. make up: /meikhɨ/+/əp/ → [meikhɨəp] b. cut off: /khəthɨ/+/ophu/→ [khəthɨophu] However, as discussed earlier, [meikhəp] is also possible as an alternate loanword by virtue of the split-base effect and lexical conservatism. Then we can predict the same type of loanword for “cut off”, because both [khət] and [khəthɨ] are listed as allomorphs in the Korean lexicon (NAKL 1991). However, the examples in (27) do not support this prediction. (27) a. cut off: /khət/+/ophɨ/ → [khədophɨ] *[khəthophɨ] *[khəsophɨ] cf. make up: /meik/+/əp/ → [meikhəp] b. cut -ɨl (Acc): /khət/-/ɨl/ → [khəsɨl] *[khədɨl] The examples in (27) are informative in three respects. First, as shown in (27a), the Component 1-final /t/ does not undergo assibilation turning /t/ to [s], indicating
123
292
M. Oh
that Component 2 is not morphologically categorized as a suffix because only nominal suffixes can trigger assibilation, as shown in (28a). (28) a. /soth-ɨl/ [sothɨl] ~ [sosɨl] “pot” (Acc) h b. /sot /+/an/ [sodan] *[sothan] *[sosan] “inside of a pot” The example in (28b), /soth/ + /an/, is a compound in Korean, and each component of a compound constitutes a separate prosodic word that can account for the coda neutralization turning any laryngeal consonants into a lax stop. Second, the example in (27b) shows that “cut” is adapted as a noun by itself in that the loanword-final /t/ undergoes assibilation before the accusative suffix. Third, in contrast to [meikhəp] for “make up”, “cut off” is never adapted as *[khəthophu]. This raises the question of why the aspirated stop mapping of the voiceless stop of the source form is possible for “make up” but not for “cut off” and what induces differences between “make up” and “cut off” with respect to sound mapping. The discussion thus far has not clarified whether EK-Ident [voice] compares loanwords with the whole word or just with the first subword in English. The nonoccurring complex loanword *[khəthophu] suggests that the complex word of a source language as a whole needs to be taken into consideration. American English coronal stops are phonetically realized as voiced flaps in certain environments, e.g., “writer” [wraɨDər]. Thus, the adaptation of “cut off” as [khədophu], not as *[khəthophu], follows naturally from the established constraint ranking. The tableau in (29) illustrates how the interaction between the constraints derives [khədophu], not *[khəthophu], as the optimal output. (29) English input: [kʌDƆf]; Listed allomorphs for the Korean base: [khət] and [khəthɨ], [ophu] Morphosyntactic base: No IntSon KK-Lex EK-Ident Max-V [khət], [ophu] VL C[aspiration] [voice] English input: [kʌDƆf] a. [khəthophu]
*!
☞b. [khədophu] c. [khətophu]
*!
*
Candidate (29a) fails to satisfy EK-Ident [voice], but Candidate (29b) satisfies EK-Ident [voice], which requires the voiced stop in English to be mapped into the voiced sound of the output regardless of whether the sound surfaces in listed allomorphs. Note that the voiced stop is not phonemic in Korean. We have thus far demonstrated that the complex loanword [meikhəp] is formed based on two distinct allomorphs [meikhɨ] and [meik]. The independently occurring allomorph [meikhɨ] permits the satisfaction of KK-Lex C[aspiration] when the morphosyntactic base [meik] is combined with [əp], yielding [meikhəp]. When such an allomorph is lacking, singleton adaptation is not possible. Thus, in the absence of *[pekhɨ] for “back”, *[pekhəp] “back up”, is not possible. The allomorph of [meik]
123
Adaptation of English complex words
293
also plays a role in deciding [meikhəp]. The independent existence of [meik] is what allows [meikhəp] in the first place. The presence of [meik] is indispensible in complex loanword adaptation, even though not all of its phonological properties are carried over to the complex loanword. Likewise, to form a complex loanword such as [meikhəp], not only [meik] but also [meikhɨ] is needed. In other words, the speaker consults not only [meikhɨ], which serves as a phonological base for providing the voiceless aspirated stop for the voiceless stop from English source word sounds, but also must consult [meik], which serves as a morphosyntactic base for providing the segmental composition of the loanword. Then the adaptation of “cut off” as [khədophɨ] in Korean suggests that the output-to-output correspondence between the complex source word as a whole and the corresponding loanword plays a crucial role. In the customary sense of the term base, the base of loanword adaptation is either the underlying representation or output of the source form. However, for complex loanword adaptation, two types of output-to-output constraints are involved: One constraint is an English output-to-Korean output constraint such as EK-Ident [voice], and the other is a Korean output-to-Korean output constraint such as KK-Lex C[aspiration]. Likewise, complex loanword adaptation is subject to the phonotactics of the adapting language, output-to-output constraints between source words and loanwords, and those between loanwords. 4.3 The morphological structure of complex loanwords Few studies have examined the questions surrounding the internal structure and morphological category of complex loanwords. For example, the problem of whether the morphological category of each component word is the same as that of the complex loanword as a whole has not been addressed. In this section, we maintain that complex words in English are loaned as single nominal words but that each component word making up the complex word is treated as a stem, not as a separate prosodic word. To support this argument, we examine the assibilation turning /t/ to [s] in prevocalic position in nominal paradigms. Both Korean nouns and loanwords undergo assibilation as illustrated in (30). (30) a. Native nouns /path/[pat] b. Loanwords “ticket” [thikhet]
/path- ɨl/[pathɨl] ~ [pasɨl]
“field” (Acc)
‘ticket’- ɨl [thikhesɨl], “ticket” (Acc) *[thikhethɨl] “David” ‘David’- ɨl [teibisɨl] “David” (Acc)15 [teibit]~ [teibidɨ] or [teibidɨrɨl] *[teibidɨl]
Because of the morphophonemic restriction against underlying /t/-final nouns in Korean, alternation between [t] in an isolated form and [s] in prevocalic position is 15 In Korean, the accusative case marker is /ɨl/ after a noun ending in a consonant but is /lɨl/ after a noun ending in a vowel.
123
294
M. Oh
attested. This raises the question of at what level the /t/-[s] alternation takes place. Following Sohn (2001) and Davis and Kang (2006), we contend that the assignment of /s/ to [t]-final loanwords such as “cut” is done at the word level, not at the stem level. To begin with, Sohn (2001) attributed the assignment of /s/ to [t] final loanwords to the phonotactic constraint of the loanword stratum and contended that the native Korean stratum allows for four coronal consonants in the word-final position, i.e.. [s, cˇ, th, ch], but that the loanword stratum allows for only the fricative [s]. Similarly, Davis and Kang (2006) argued that the assignment of /s/ to loanwords such as “cut” reflects a word-level output-to-output anti-correspondence constraint. They maintained that the loanword-final [t] is lexicalized as [s] in Korean because loanwords are completely integrated into the Korean morphophonological system and explained the similarity between native Korean nouns and loanwords with respect to assibilation in terms of a word-level output-to-output anti-correspondence constraint that forces word-final [t] to alternate in inflected forms. Further, to realize the consistent alternation between an unaffixed noun ending in [t] with an inflected form having [s], they proposed an anti-faith [-continuant] constraint, stating that if two words are in an output–output relationship, then the corresponding sound of a word-final [-continuant] segment must not be [-continuant]. They maintained that the borrowing of the English word-final [t] as lexical [s] in Korean indicates the complete integration of loanwords into Korean morphophonology. Assuming that the /t/-[s] alternation occurs at the word level, not at the stem level, we now turn to the question of whether the morphological category of each component word is the same as that of the complex loanword as a whole. Specifically, we ask whether when adapting “cut off”, the base for “cut” is /khət/ or /khəs/ in addition to /khəthɨ/. If /khət/ is the base for “cut”, each component would be morphologically categorized as a stem, whereas if /khəs/ is the base, each component would be a word. Examples are provided in (31). (31) a. cut off:: [khət]stem [ophu]stem → [khədophu]phonological word *[khəsophu] b. cut–ɨl (Acc): [[khət]stem]phonological word - ɨl] → [khəsɨl]phonological word *[khədɨl] cf. cutting [khəthiŋ] *[khəsiŋ] c. log out–ɨl (Acc): [rogɨ]stem [aut]stem → [rogɨ aut]phonological word - ɨl] → [rogɨausɨl] *[rogɨaudɨl] As discussed above, both Sohn (2001) and Davis and Kang (2006) assigned /s/ to loanword-final [t] at the word level, not at the stem level. Thus, a loanword-final /t/ is adapted as [s] before a vowel-initial Korean suffix because “cut” is loaned as a nominal word by itself, i.e., /khət - ɨl/, [khəsɨl] “cut- ɨl (Acc)”, as shown in (31b). On the other hand, the output in (31a), [khədophu], suggests that the [t]-final loanword exists at the stem level. Further, the example in (31c) demonstrates that the complex loanword as a whole for “log out” behaves as a nominal word since the complex loanword-final /t/ also undergoes assibilation. However, the output in (31a), [khədophu], cannot provide full support for the position that the base for “cut” is /khət/, not /khəs/, because the other possible base, /khəs/, would also give rise to [khədophu] after its neutralization at the word level and
123
Adaptation of English complex words
295
combination with the following component, i.e., /khəs/ → [khət] + [ophu] → [khədophu]. In other words, the complex loanword for “cut off” cannot provide decisive evidence for determining the base for Component 1 for “cut off” between /khət/ or /khəs/. Instead, the complex loanword for “post it” supports the claim that each component is categorized as a stem. “Post it” is loaned as either [phosɨthɨit] or [phosɨthit].16 If the base for “post” is /phosɨs/ as well as /phosɨthɨ/, we can expect *[phosɨdit], as per [khədophu] “cut off”; i.e., /phosɨs/ → [phosɨt] + [it] → *[phosɨdit]. The nonoccurrence of *[phosɨdit] provides strong evidence for the claim that [t]final loanwords exist. This fact also implies that each component is loaned at the stem level. The output of [phosɨthit] “post it” provides another example of [meikhəp]-type words. The tableau in (32) illustrates how the interaction between constraints can produce the optimal output. (32) English input: [poʊstıt]; Listed allomorphs for the Korean base: [phosɨt] ~ [phosɨthɨ], [it] EK-Ident Max-V Morphosyntactic base: No IntSon KK-Lex VL C[aspiration] [voice] [phosɨt], [it] a. [phosɨdit]
*!
☞ b. [phosɨthit] c. [phosɨtit]
*!
Candidate (32c) fatally violates the No IntSon VL constraint. Candidate (32a) violates EK-Ident [voice] in that the Component 1-final /t/ is not voiced in the English input. Candidate (32b) is the optimal candidate because it satisfies all of these constraints. Note that the example of “post it” is consistent with that of the complex loanword for “cut off” in that the Component 1-final coronal stop of “post it” is not flapped in English but the Component 1-final coronal stop /t/ in “cut off” is flapped and is mapped into [d] in Korean. Likewise, the necessity of referencing the whole complex word from a source word is also well supported. In contrast to “post it”, most cluster-final loanwords always trigger vowel epenthesis. For example, Component 1 of “built in” ends with a stop, but it is always loaned as [pilthɨ] because the final stop is released in English. Likewise, cluster-final stops of the first component almost always trigger vowel epenthesis. The Appendix provides a complete list of these examples. To summarize, we have shown that each component word is characterized as a stem but that the loanword as a whole is treated as a nominal word.
16 “Post”, as a loanword, was used far more often (Google hits: 26,200,000) than “post it”. This provides an argument against the frequency-based account of the apparent single-unit analysis of complex loanwords.
123
296
M. Oh
5 Discussion and conclusion This paper proposed a model for the adaptation of complex loanwords. Most previous studies of loanword adaptation have analyzed segmental mappings between the source word and loanword sounds of monomorphemic words. However, loanwords cannot be fully understood by relying only on such a segmental correspondence within a simplex word. The results of this study indicate that a consonant terminating Component 1 can be treated differently from both word-final consonants, e.g., “cut off” [khədophu] versus “cut” –i (Nom) [khəsi] and word-medial consonants, e.g., “party” [phathi]. Further, a stop terminating Component 1 behaves differently than the corresponding fricative or affricate in terms of vowel epenthesis, e.g., “make up” [meikhəp] vs. “focus on” [phokhəs’ɨon]. Based on analysis of adaptation of complex English words into Korean, we argued for a component-by-component analysis for complex loan adaptation. The finding that each component word is the domain for the loanword adaptation of complex words sheds light on the issue of vowel epenthesis. We asked at what stage vowel epenthesis occurs and whether it occurs with each component word of the complex word or with the complex word as a whole. Each component of the source form is parsed morphologically and loaned in the adapting language. Specifically, each component of the complex word “catch on” is loaned as [khecˇhi] for “catch” and [on] for “on.” Similarly, each component of “back up” is loaned as [pek] and [əp]. As shown in Sect. 4, the first component ending in a stop shows a significantly lower rate of vowel epenthesis than that ending in a fricative or an affricate. This asymmetry suggests that vowel epenthesis takes place within the limit of each inner component of the complex word. Further, given that each component is loaned as a stem, we can easily conclude that vowel epenthesis takes place at the stem level, not at the word level. The examination of the adaptation of nominalized forms of English phrasal verbs such as “make up” and “give up” in Korean (Sect. 3) indicates that component words of complex words are loaned separately, not as a single unit. We questioned whether the component-by-component analysis can be extended to the adaptation of other types of complex words. To begin with, the examples in (33) illustrate that acronyms pattern with nominalized forms of English phrasal verbs in loanword adaptation. (33) Acronym
Loanword
UFO (Unidentified Flying Object) [yuephɨo] *[yuepho] PLO (Palestine Liberation Organization) [phiero] *[phiello] LA (Los Angeles) [erei]~[ellei] (?)17 “UFO” is adapted as [yuepho] but not as *[yuepho] in Korean, supporting component-by-component analysis, wherein each English letter acts as a component part. 17 The acronym/abbreviation of “Los Angeles”, “LA” can be borrowed marginally with geminate [ll]. As a reviewer suggested, some people may treat it as a single unit if they do not recognize it as an acronym.
123
Adaptation of English complex words
297
The adaptation of compound words in English also provides support for component-by-component analysis. English /tl, dl, fl/ sequences are loaned differently depending on the position of /l/. When /l/ is located morphemeinternally after a consonant in English, as in “medley”, it is adapted as the geminate [ll], e.g., [medɨlli] “medley”. By contrast, the Component 2-initial /l/ of a compound word is more likely to be adapted as [r] when preceded by a consonant, e.g., [kaidɨrain] “guide line.” These different patterns are supported by a survey made using the Google search engine in Hangul, the Korean alphabetic script. The hit counts for each loanword are given in (34). (34) a.
Adaptation of /tl, dl/ Loanword sound Word
Loanword
Simplex word: [ll] [ [r]
Hangul (r) 1 %
medley
Percentage Total hits 289,780
Hangul (ll) 99 % Hitler
Hangul (r) 0 %
3,390,000
Hangul (ll) 100 % Compound word: guide line Hangul (r) 96 % [ll] \ [r] Hangul (ll) 4 % dead line Hangul (r) 99.9 %
1,338,000 93,904
Hangul (ll) 0.1 % ad lib
Hangul (r) 94 %
153,190
Hangul (ll) 6 % cutline
Hangul (r) 99.9 %
191,008
Hangul (ll) 0.1 % b. Adaptation of/fl/ Loanword sound Word
Loanword
Simplex word: [ll] [ [r]
Hangul (r) 9 %
Butterfly
Percentage Total hits 309,000
Hangul (ll) 91 % Compound word: safe line [ll] \ [r]
Hangul (r) 100 %
1,090,000
Hangul (ll) 0 % Likewise, the sequence of phones has one rendition if it belongs to a morphologically simplex word and another if the same sequence of phones constitutes a morphologically complex word. This suggests that English compound words are treated differently from simplex words in loanword adaptation. Likewise,
123
298
M. Oh
component-by-component analysis can be extended to account for the adaptation of English compounds in Korean. As with acronyms such as “LA”, the geminate adaptation of /l/ is also marginally possible in the adaptation of compound words, e.g., “ad lib” [edɨrip] ~ [edɨllip]. However, the existence of the geminate adaptation of /l/ in acronyms and compound words cannot repute the component-by-component analysis of loanwords because this adaptation is clearly asymmetrical for simplex and compound words with respect to the degree of /l/ gemination. Morpheme-internal /l/ in English is much more likely to be loaned as [ll], whereas /l/ at the beginning of Component 2 is more likely to be borrowed as [r], as shown in (34). This suggests that the component-bycomponent analysis also holds for complex loanwords in which Component 2 begins with a consonant. As discussed in Sect. 4, the adaptation of “cut off” as [khədophɨ] in Korean argues for output-to-output correspondence between the complex source word as a whole and the corresponding loanword. The rendition of an English flap sound of a complex word as a voiced stop in Korean has some bearing on the perceptual versus the phonological debate about loanword adaptation, and on the factors relevant to loanword adaptation. First, the discussion in this section provides support for the perceptual approach to loanword adaptation. Both flaps in English and voiced stops in Korean are allophones. However, they play a crucial role in determining the phonetic match. In other words, a given phoneme of the source language is not uniformly adapted as some corresponding phoneme in the borrowing language, as expected by the phonological approach. Rather, as predicted by the perceptual approach, a source language phoneme is borrowed differently according to its allophonic realization. Second, the adaptation of the allophonic flap in English as [d] in Korean is limited to morphologically complex words. Otherwise, we would incorrectly predict *[wədə] for “water” as opposed to [wəthə]. Notice that morphology is not part of the perceptual apparatus since a morpheme boundary is not something that is part of the acoustic signal. This argues for an analysis of loan adaptation that makes reference to both phonetic information and the morphological structure of the source word. To conclude, our study of how complex English words are borrowed in Korean has five key findings. First, each component word is loaned independently of the other component word, and they are then combined to yield the loanword as a whole. Second, parsing English loanwords into a [[X]stem [Y]stem]phonological word structure is crucial for explaining the asymmetry between the behavior of final stops of Component 1 and that of corresponding fricatives or affricates. Third, quantitative analysis of the NAKL list (1991) and of Google search results provides support for the observation that single-unit adaptation is possible only for stops, not for fricatives or affricates. Furthermore, it establishes a correlation between the rate of vowel epenthesis in an isolated component word and that in a complex loanword. Fourth, segmental mappings of [meikhəp]-type words support the split-base hypothesis and lexical conservatism (Steriade 2000); that is, the split-base hypothesis and lexical conservatism can explain why [meikhəp] “make up” exhibits a blend of the morphological base and the phonological base. Finally, the examination of complex loanwords provides support for the perceptual view of loanword adaptation. The adaptation of English complex loanwords in Korean
123
Adaptation of English complex words
299
demonstrates that the surface form of the complex word as a whole from the source language determines sound mappings. Acknowledgments This work was supported by the National Research Foundation of Korea Grant, funded by the Korean Government (327-2007-2-A00316). I would like to thank Kie Zuraw, Robert Daland, and Bruce Hayes for their discussions and comments on earlier versions of this paper. I would also like to thank the editors and three reviewers for their helpful comments and criticisms, which greatly improved the quality of the analysis and sharpened the clarity of the exposition. I am grateful to Changmun Yang for her valuable assistance in the corpus and Google data collection for this project. I am also grateful to Minsu Kim and Syejeong Kim for their help with the statistical analysis.
Appendix Appendix (*The Korean search term was entered with the English complex word. Otherwise, only the Korean spelling was used.) A: Google searches for the loanwords from NAKL (1991) a.
First component ending in a stop: The phonetic output for the final stop of the first component is given in brackets for loanword alternatives without vowel epenthesis.
English input
Vowel epenthesis
Percentage
No V epenthesis
Percentage
Total hits
back attack
배크 어택
0
백어택 [g]
100
191,000
back up
배크 업
0
백업 [g]
100
6,290,069
빅이벤트 [g]
100
6,550,000
0
12,803,840
92
7,167,000
big event
비그 이벤트
check out
체크 아웃
dugout
더그아웃
fade out
페이드 아웃
flute ensemble
플루트 앙상블
grade up
그레이드 업
graphic equalizer
그래피크 이퀼라이저
hit and away 히트 앤 어웨이 hit and run
히트 앤 런
0 100 8 100 99 100 0
첵아웃/체카웃 덕아웃/더가웃 [g] 페이다웃
0
299,000
플룻앙상블 [d]
1
102,917,000
그레이덥
0
2,050,000
100
293,000
그래픽이퀼라이저 [g]
100
힛앤어웨이
0
391,099
98
힛앤런 [d]
2
1,497,700
hot issue
하트 이슈
0
핫이슈 [d]
100
22,556,300
kick off
키크 오프
0
킥오프 [g]
100
13,800,000
make up
메이크 업
85
메이컵 [kh]
pick up*
래크 업
pipe organ
파이프 오르간
100
pipe organist
파이프 오르가니스트
100
0
15
23,300,007
100
6,468,230
파입오르간
0
353,000
파입오르가니스트
0
32,700
픽업 [g]
pop art
파프 아트
0
팝아트 [b]
100
18,000,000
pop artist
파프 아티스트
0
팝아티스트 [b]
100
15,600,000
123
300
M. Oh
English input Vowel epenthesis
Percentage
red apple
레드 애플
rock opera
로크 오페라
0
rocket engine 로케트 엔진
36
set up
세트 업
No V epenthesis
Percentage
레대플
100
1
Total hits
0
3,970,000
록오페라 [g]
100
1,290,000
로켓엔진 [d]
64
4,840,000
셋업 [d]
99
1,047,374,200 3,740,000
shoot off
슈트 오프
0
슛오프 [d]
100
shut out*
셔트 아웃
55
셧아웃 [d]
45
9,080
strike out
스트라이크 아웃
98
스트라익아웃 [g]
2
4,226,800
tag out
태그 아웃
택아웃
0
5,350,000
b.
100
First component ending in a fricative or an affricate
English input
Vowel epenthesis
Percentage
No V epenthesis
Percentage
Total hits
change up
체인지업
100
체인접
0
3,320,000
Christmas eve
크리스마스 이브
100
크리스마시브
0
76,100,000
클로접
9
7,876,000
코신
0
77,200 28,400,000
close up
클로즈업
course in*
코스인
91 100
drive in
드라이브 인
100
드라이빈
0
force out
포스아웃
100
포사웃
0
2,670,000
100
가소븐레인지
0
3,990,000
100
오파트
0
35,900
100
옴니버샐범
0
521,000
gas oven range 가스오븐레인지 off art
오프아트
omnibus album 옴니버스앨범 pitch out
피치아웃
100
피차웃
0
805,000
plus alpha
플러스 알파
100
플러살파
0
658,000,000
serve and dash
서브앤대시
100
서밴대시
0
183,000
c.
First component ending in a cluster
English input
Vowel epenthesis
Percentage
No V epenthesis
Percentage
Total hits
best eleven
베스트일레븐
100
베스틸레븐
0
155,000
built-in-system 빌트인시스템
100
빌틴시스템
0
138,002
bunt and run
번트 앤 런
100
번탠런
0
826
crank in
크랭크인
100
크랭킨
0
236,000
text editor
텍스트에디터
100
텍스테디터
0
356,000
Windup
와인드업
100
와인덥
0
306,000
123
Adaptation of English complex words
301
B. Google search list a.
First component ending in a stop
English input Vowel epenthesis
Percentage
No V epenthesis
Percentage
Total hits
add in*
에드인
100
에딘
0
add up*
에드업
100
에덥
0
4,090
attack on
어태크 온
0
어택온 [g]
100
4,860,000
black eye
블래크 아이
0
블랙아이 [g]
100
357,000
black out
블래크 아웃
0
블랙아웃 [g]
100
52,741,900
break away*
브레이크 어웨이
66
브레이커웨이 [kh]
34
24,860
break in*
브레이크인
98
브레이킨 [kh]
2
2,948,400
25,700,000
break out
브레이크 아웃
100
브레익아웃
0
6,150,000
break up
브레이크 업
100
브레익업
0
6,870,000
check in
체크인
99
체킨 [kh]
1
7,907,200
check off
체크 오프
100
첵오프
0
24,700,000
cut in*
커트인
50
컷인 [d]
50
407,000
cut off
커트 오프
84
컷오프 [d]
16
1,823,000
cut up
커트 업
100
컷업 [d]
0
986,700
dark horse
다크 호스
100
닥호스
0
3,610,000
dig out
디그아웃
33
drop out
드로프 아웃
eat out*
이트 아웃
96
fade in
페이드인
100
get it on
게트잇온
0
겟잇온 [d]
100
200,000
get out
게트 아웃
0
겟아웃 [d]
100
4,240,000
get over
게트 오버
0
겟오버 [d]
100
3,150,000
get up
게트 업
0
겟업 [d]
100
67,905
head up
헤드업
100
0
4,201,520
heat up*
히트 업*
100
0
9,286
job interview
자브인터뷰
0
잡인터뷰 [b]
100
2,870,000
0
딕아웃 [g] 드롭아웃 [b]
67
1,320,000
100
1,450,000
잇아웃 [d]
4
32,980
페이딘
0
779,000
헤덥 힛업
job opening
자브오프닝
0
잡오프닝[b]
100
884,000
keep out
키프 아웃
0
킵아웃 [b]
100
229,000
knock out*
노크 아웃
0
녹아웃 [g]
100
377,830
0
락앤락 [g]
100
1,670,000
4
로긴 [g]
96
8,982,000,000
2
129,220,000
lock and lock 라크 앤 라크 log in
로그인
log out
로그 아웃
98
look out
루크 아웃
0
룩아웃 [g]
100
10,800,000
look up
루크 업
0
룩업 [g]
100
417,000,006
make it
메이크잇
34
1,602,202
mark up
마크 업
100
막업
0
4,310,127
meet up
미트 업
100
밋업
0
2,563,080
network
네트워크
88
넷워크 [d]
12
69,760,195
plug in
플러그인
36
플러긴 [g]
64
1,133,000,000
66
록아웃/로가웃 [g]
메이킷 [kh]
123
302
M. Oh
English input Vowel epenthesis
Percentage
No V epenthesis
Percentage
팝업 [b]
100
Total hits
pop up
파프 업
right up*
라이트 업
100
set off
세트 오프
90
셋오프 [d]
set out
세트아웃
100
shake it
쉐이크잇
87
shake up
쉐이크 업
100
shut up*
셔트 업
0
셧업 [d]
100
41,408
slip on
슬리프 온
0
슬립온 [b]
100
153,000
speed up
스래드업
100
스피덥
0
7,200,000
start up
스타트 업
100
스탓업
0
3,771,920
step into
스테프 인투
0
스텝인투 [b]
100
477,000
step up
스테프 업
0
스텝업 [b]
100
7,670,000
straight up*
스트레이트 업
0
163,047
take off*
테이크 오프
100
테익오프
0
21,300,399
take out
테이크 아웃
100
테익아웃
0
1,690,000
take over
테이크 오버
100
테익오버
0
24,384,500 1,940,000
0
100
라잇업
7,610,041
0
274,000
10
17,610,000
셋아웃 [d]
0
7,632,200
쉐이킷 [kh]
13
500,703
0
401,000
쉐익업
스트레잇업
trade off
트레이드오프
100
트레이도프
0
wake up*
웨이크 업
51
웨이컵 [kh]
49
927,204
work out
워크 아웃
100
웍아웃*
0
12,904,040
wrap up
래프 업
0
랩업 [b]
100
1,180,001
b.
First component ending in a fricative or an affricate
English input
Vowel epenthesis
Percentage
No V epenthesis
Percentage
Total hits
brush up*
브러시업
100
브러셥
0
1,540
catch on
캐치온
100
캐촌
0
3,940,000
catch up*
캐치업
100
캐첩
0
2,320,289
cross over*
크로스오버
100
크로소버
0
68,103,800
dress affair
드레스어페어
100
드레써페어
0
394,000
dress up
드레스업
100
드레썹
0
4,150,000
each other
이치아더
English Alphabet 잉글리시알파벳
100
이차더
0
445,000
100
잉글리샬파벳
0
302,000
face off
페이스오프
100
페이소프
0
1,980,000
face up*
페이스업
100
페이썹
0
8,190 6,780,000
focus on
포커스온
100
포커손
0
give away
기브어웨이
100
기버웨이
0
23,400
give in*
기브인
100
기빈
0
83,900
give up*
기브업
100
기법
0
13,900
live in
리브인
100
리빈
0
3,580,000
live out*
리브아웃
100
리바웃
0
19,500
123
Adaptation of English complex words
English input Vowel epenthesis
303
Percentage
No V epenthesis
Percentage
Total hits
Los Angeles
로스앤젤레스
100
로샌젤레스
0
5,160,000
love affair
러브어페어
100
러버페어
0
1,110,000
match up
매치업
100
매첩
0
740,000
move on*
무브온
99
무본
1
642,110
move out
무브아웃
100
무바웃
0
961,000
news anchor
뉴스앵커
100
뉴생커
0
14,900,000
pass away
패스어웨이
100
패서웨이
0
677,000
pass on
패스온
100
패손
0
5,700,000
pass out*
패스아웃
100
패사웃
0
6,440
pass over
패스오버
100
패소버
0
3,800,000
pass up
패스업
100
패섭
0
2,650,000
patch up
패치업
100
패첩
0
3,100,000
pitch up*
래치업
100
피첩
0
11,201
price index
프라이스인덱스
100
프라이신덱스
0
884,000
push on
푸쉬온
100
푸숀
0
1,390,000
push up
푸시업
93
푸샵
7
984,300
raise up
레이즈업
100
레이접
0
509,000
rush hour
러시아워
100
러샤워
0
356,000
thanks a lot
땡스어랏
100
땡써랏
0
9,000
touch and go 터치앤고
100
터챈고
0
34,900
touch out
터치아웃
100
터챠웃
0
2,870,000
touch up
터치업
100
터첩
0
3,090,000
wash up*
워시업
100
워셥
0
645,003
watch out
와치아웃
100
와차웃*
0
915,007
watch over
와치오버
100
와초버
0
2,580
what’s up*
와스업
와썹
1
637,540
c.
99
First component ending in a cluster
English input
Vowel epenthesis
Grand opening 그랜드오프닝 hand out
핸드아웃
jump up*
점프업
Percentage 100 99 100
No V epenthesis
Percentage
Total hits
그랜도프닝
0
231,000
핸다웃
1
471,420
점펍
0
3,190 2,104,600
post it
포스트잇
포스팃
1
round about
라 운드어바웃
100
라 운더바웃
0
208,200
round up
라 운드업
100
라운덥
0
439,000
Stand up
스텐드업
99
스텐덥
1
1,931,900
99
123
304
M. Oh
References Beckman, Jill N. 1997. Positional Faithfulness. PhD dissertation, University of Massachusetts, Amherst. Davis, Stuart, and Hyunsook Kang. 2006. English loanwords and the word final [t] problem in Korean. Language Research 42: 253–274. Hay, Jennifer. 2001. Lexical frequency in morphology: Is everything relative? Linguistics 39(6): 1041– 1070. Ito, Junko. 1990. Prosodic minimality in Japanese. In Parasession on the Syllable in Phonetics and Phonology, 213–239. Chicago: Chicago Linguistics Society, University of Chicago. Kang, Hyunsook. 1996. English loanwords in Korean. Studies in Phonetics, Phonology and Morphology 2: 21–48. Kang, Yoonjung. 2003. Perceptual similarity in loanword adaptation: English postvocalic word-final stops in Korean. Phonology 20: 1–56. Kenstowicz, Michael. 1996. Base identity and uniform exponence: alternatives to cyclicity. In Current trends in phonology: Models and methods, ed. J. Durand, and B. Laks, 363–394. Salford: University of Salford. Kenstowicz, Michael, and Atiwong Suchato. 2006. Issues in loanword adaptation: A case study from Thai. Lingua 116: 921–949. Kim, Soohee, and Emily Curtis. 2000. Phonetic duration of English /s/ and its borrowing in Korean. Japanese and Korean Linguistics 10: 406–419. Kwulipkwukeyenkwuwen [The National Academy of the Korean Language]. 1991. Oylaye sayong siltay cosa: 1990 nyendo [Survey of the state of loanword usage: 1990]. Seoul: NAKL. Lacharite´, D., and C. Paradis. 2005. Category preservation and proximity versus phonetic approximation in loanword adaptation. Linguistic Inquiry 36: 223–258. Lee, Yongsung. 2001. The noun-verb asymmetry in Korean Phonology. Studies in Phonetics, Phonology and Morphology 375–398. McCarthy, J., and A. Prince. 1995. Faithfulness and reduplicative identity. In Papers in optimality theory (University of Massachusetts Occasional Papers 18), ed. J. Beckman, L.W. Dickey, and S. Urbanczyck, 249–384. Amherst, Mass.: Graduate Linguistics Student Association. Oh, Mira. 1992. Palatal consonants, labial consonants, and vowel epenthesis in Korean. Korean Journal of Linguistics. 17: 141–161. Oh, Mira. 1995. A prosodic analysis of non-derived environment blocking. Journal of East Asian Linguistics 4(4): 261–279. Oh, Mira. 1996. Linguistic input to loanword phonology. Studies in Phonetics, Phonology and Morphology 2: 117–126. Oh, M., and H. Kim. 2006. A phonetic duration-based analysis for vowel epenthesis: English postvocalic word-final stops. Studies in Phonetics, Phonology and Morphology 12: 307–324. Peperkamp, S., and Dupoux, E. 2003. Reinterpreting loanword adaptations: the role of perception. Proceedings of the 15th International Congress of Phonetic Sciences, pp. 367–370. Prince, Alan, and Paul Smolensky. 1993. Opimality Theory: Constraint interaction in generative grammar. Ms, Rutgers University and University of Colorado. Silverman, D. 1992. Multiple scansions in Loanword Phonology: Evidence from Cantonese. Phonology 9: 289–328. Smith, Jennifer L. 1997. Noun faithfulness: on the privileged behavior of nouns in phonology. Ms, ROA242. Sohn, Hyang-sook. 2001. Optimization of word-final coronals in Korean loanword adaptation. Ms, Kyungpook National University. Steriade, Donca. 2000. Lexical conservatism and the notion base of affixation, Ms. University of California, Los Angeles. Steriade, Donca. 2001. Directional asymmetries in place assimilation: a perceptual account. In The Role of Speech Perception in Phonology, ed. E. Hume and K. Johnson, 219–250. New York: Academic Press.
123