Unlike other vowels in Korean, the vowel [Û] is unique in that it is subject to Û/zero alternation in suffixation, and to insertion in loanwords. Within rule-based frameworks, these two processes are characterised as Û-deletion rule and as Û-insertion, respectively (e.g. S.-C. Ahn 1985, 1998; Kim-Renaud 1982, among others). Consequently, Underspecificaton Theory (Archangeli 1984) treats the vowel [Û] as completely unspecified in lexical representations (e.g. H.-S. Sohn 1986). What these previous analyses have not dealt with, however, is the distribution of [Û] in mono-morphemic words, partly because morpheme-internal [Û] is static, in the sense that it is not subject to deletion or insertion. But, there is positive evidence that this vowel must be also treated differently from other vowels, because the occurrence of [Û] is highly constrained morpheme-internally, in suffixation and in loanwords, as we will see throughout the thesis
In this chapter, I present a general background to the vowel [Û] in Korean. In 1.1,
the segmental inventory of Korean is introduced. In 1.2, I briefly review how previous analyses have dealt with the processes regarding [Û] in suffixation. In 1.3 and 1.4, I examine the distribution of [Û] in mono-morphemic words and in suffixation where the occurrence of the vowel [Û] is not random: the context in which [Û] occurs is strictly regular. In 1.5, I consider the occurrence of the vowel [Û] in loanwords. Finally, in 1.6, I put forward the proposal that the Û-deletion and the Û-insertion rules can be unified if we assume that [Û] is construed as the phonetic interpretation of an empty nucleus. 1.1. The segmental inventory of Korean
I assume that Korean has 8 vowels and 19 consonants underlyingly. (1)
/i/, /e/, /E/, /Û/, /a/, /u/, /o/, //
(i) Lenis obstruent: /p/, /t/, /k/, /s/, /c/1, /h/
(ii) Aspirated obstruent: /pH/, /tH/, /kH/, /cH/
(iii) Tensed obstruent: /p’/, /t’/, /k’/, /s’/, /c’/
1 The symbols /c, c’, cH/ represent palato-alveolar affricates. According to the IPA, they are symbolised by /tS, tS’, tSH/ respectively. However, in this thesis, the symbols /c, c’, cH/ are used for notational convenience.
Regarding the phonetic quality of the eight vowels, /i/, /e/ and /E/ are front; /Û/ and /a/ are mid; /u/, /o/ and // are back. Note that the phonetic value of // is an unrounded back vowel which is transcribed as [¨] in the IPA. 2 Obstruents are classified according to three manner types, referred to as lenis, aspirated and tensed, as in (1). However, there is a gap for the coronal fricatives /s/ and /s’/: an aspirated /sH/ is not present. /s/ has special properties. Phonetically it is aspirated (Kagaya 1974), but phonologically it behaves like other lenis obstruents (Kim-Renaud 1974, Iverson 1983, among others. For a detailed discussion on this matter, see Chapter 3). Regarding the liquids, [r] and [l] are complementarily distributed: [r] occurs intervocalically and [l] elsewhere. A number of phonologists have proposed that the underlying segment of liquids in Korean is /l/. However, Y. Heo (1995) and S.J. Rhee & Heo (1998) argue that /r/ is the underlying segment. The debate concerning underlying segments will be discussed in Chapter 2. 1.2. Previous treatments of the vowel [Û] in suffixation Û/zero alternation in suffixation has been a highly controversial topic. The debate has focused on the issue as to whether or not the vowel [Û] is a part of the suffixes in question. Those who claim that [Û] belongs to a set of suffixes with Û/zero alternations, argue that it is deleted in appropriate contexts (e.g. S.-C. Ahn, 1985, 1991, Kim-Renaud, 1982, B.-G. Lee 1976, and H.-S. Sohn, 1986 among others). Those who propose that [Û] is unspecified in the underlying representation argue that it is inserted in particular contexts (e.g. H.-P. Choi 1937, C.-W. Kim 1968, S.-H. Kim 1992, and Y.-S. Kim 1985 among others). These contexts are shown below. (2)
/s’is/ [s’isÛmj] [s’isÛni] ‘to
2 In the relevant literature on Korean phonology, the back unrounded vowel is conventionally represented by the schwa symbol //, which I follow in this thesis. Bearing this in mind, the vowel // in Korean behaves differently from that of English, Dutch, French etc., in which this vowel occurs in metrically weak position. 3 The underlying representations in (2) are based on the ‘standard’ syllabification, i.e. stem-final consonants are syllabified as ‘coda’. 4 Note that the stem-final liquid is deleted. For an analysis of the phenomenon, see Chapter 4.
In (2a), we observe that [Û] does not occur when the stem-final segment is either
a vowel or a liquid; otherwise, it does occur, as in (2b). For the proponents of Û-deletion, the underlying representations of the connective and the effective suffix are /Ûmj/ and /Ûni/, respectively. They propose rules in which the vowel [Û] is deleted between a vowel and a nasal, or a liquid. According to the proponents of Û-insertion, the underlying representations of these suffixes are /mj/ and /ni/, respectively. They utilise a rule in which the vowel [Û] is inserted between a consonant other than a liquid and a nasal.
Each approach has different empirical consequences. For vowel-final stems, as
in (2a), the deletion analysis treats the phenomenon as a vowel-hiatus effect; the suffix-initial [Û] is deleted when it follows another vowel. In liquid-final stems, the suffix-initial [Û] is deleted before a labial nasal. In this case, the implicit motivation for the deletion of [Û] would be to form a well-formed ‘coda’-onset sequence: [lm]. In other contexts, the suffix-initial [Û] is retained. On the other hand, the insertion approach assumes that the insertion of [Û] is attributed to euphonic
purposes, in the sense that the epenthetic [Û] between a consonant and a sonorant may contribute to ease of pronunciation, by avoiding two contiguous consonants (H.-P. Choi 1937: 167). The implication is that the epenthetic vowel functions as a boundary marker
between a stem and a suffix (S.-H. Kim 1992: 95).
I will now point out some empirical problems for both approaches. First, one
major problem with the deletion approach is that it is difficult to handle the asymmetrical distribution of nasal-nasal (henceforth NN) sequences morpheme-internally and in suffixation. Morpheme-internal NN sequences have no [Û] separating them. (3)
In (3), we observe that morpheme-internal NN sequences contain (partial) nasal
geminates, e.g. coronal-coronal, labial-labial, coronal-labial, labial-coronal, velar-labial and velar-coronal NN sequences, except velar-velar, labial-velar and coronal-velar.5 Unlike the process in suffixation (cf. (2b)), [Û] does not occur between two nasals morpheme-internally. But, if [Û]-deletion following a liquid creates a well-5 The distributional restriction is imposed on the velar nasal, i.e. this segment cannot occur in an onset position.
formed ‘coda’-onset cluster, then the deletion approach is unable to account for why the vowel [Û] is not deleted before a nasal in the connective and the effective suffixation, as in (3). Namely, morpheme-internal NN sequences are completely acceptable. The insertion approach also has a major empirical problem. There are exceptions in interrogative /ni/ and indicative /ne/ suffixation. Unlike the other nasal-initial suffixes in (2), they are exceptional in that Û-insertion does not occur, e.g. [anni] and [anne] ‘to hug’, [kamni] and [kamne] ‘to wind’. To treat them as exceptions to Û-insertion, however, would be counter-intuitive, because these are productive verbal inflectional processes. Another problem is that the Û-insertion analysis nonetheless requires the Û-deletion rule to account for the deletion of the stem-final [Û] in the stative forms (the stative suffix is /a/ or //6). Some examples of verbs ending in [Û] are /s’Û/ ‘to write’, /k’Û/ ‘to extinguish’, /t’Û/ ‘to float’; the stative form of each verb is [s’:], [k’:] and [t’:], respectively. Therefore, two conflicting processes are listed in the grammar within the Û-insertion approach. To summarise, both the insertion and the deletion approaches to Û/zero alternation lead to empirical problems. The deletion analysis cannot handle the asymmetrical behaviour of [Û] with respect to morpheme-internal NN sequences and those in suffixation. The Û-insertion analysis has both conceptual and empirical problems in that it still requires the Û-deletion rule to account for stative suffixation, where the stem-final [Û] is deleted. Furthermore, it treats highly productive interrogative and indicative suffixation as exceptional. 1.3. The distribution of [Û] in native mono-morphemic
words7 In this section, I consider the distribution of [Û] in native mono-morphemic words in order to show that its occurrence is not random. Consider morpheme-initial and -final [Û]. 1.3.1. Morpheme-final [Û] In Korean, mono-morphemic words may end in either a consonant or a vowel. Apart from verb stems which cannot occur in isolation,8 the occurrence of [Û] is quite rare in morpheme-final position.9 It indicates that [Û] shows a different behaviour 6 The shape of the stative suffix is determined by the stem-final vowel. If it is either /a/ or /o/, the suffix is /a/; otherwise, it is //: e.g. /mak/ ‘to block’, [maka]; /nok/ ‘to melt’, [noka]; /mk/ ‘to eat’, /mk/; /cuk/ ‘to die’, [cuk] and so on. 7 This section owes much to Y. Heo (1995). 8 It is interesting to note that mono-syllabic verb stems ending in the vowel [Û] begin with either a tensed or an aspirated consonant, e.g. [s’Û] ‘to write’, [k’Û] ‘to extinguish’, [tHÛ] ‘to sprout’ and so on. On the realisation of the vowel [Û] in these stems in suffixation, see Y. Heo (1995). 9 According to Y. Heo (1995: 25), there are only three examples, all of which are pronouns, e.g. [kÛ] ‘he’, [jnÛ] ‘other (person)’ and [nÛ] ‘which’.
from other vowels in that the latter occur freely in this position. The first generalisation regarding morpheme-final position, then, is as follows. (4) The vowel [Û] may not occur finally: *(C)Û# 1.3.2. The vowel [Û] in the initial nuclear position
The vowel [Û] may occur morpheme-initially, as in (5a), or as the initial nucleus,
[tÛlc’uk] ‘blueberry’ [tHÛcip] ‘hitch’ [sÛrasoni]
[nÛkÛl] ‘sickening’ [nÛtHari] ‘agaric
(6) The vowel [Û] may occur as the first nucleus: #(C)Û. 1.3.3. Morpheme-internal [Û] Although [Û] does not occur morpheme-finally, it may nevertheless occur before a morpheme-final
consonant, as shown in (7). (7) [sasÛm] ‘deer’ [rÛn] ‘adult’ [c’acÛ @ ‘irritation’
[mEtÛp] ‘knot’ [katÛk] ‘full’ [tikÛt]
(7) illustrates that [Û] occurs before a single final consonant. Interestingly, note
that [Û] may not occur before a single consonant followed by another vowel, although other vowels may occur freely. Consider the following examples. (8) [akas’i] ‘young
In (8), notice that the underlined vowels may alternate with another vowel, but
not with the vowel [Û]: (due to lack of minimal pairs, consider only the underlined sequences) [akas’i] vs. [akis’i] ‘young lady’,10 [k’urmi] vs. [kwit’urami] ‘cricket’; [tarEk’i] vs. [krik’i] ‘to be restrained’; [sirEki] vs. [toriki] ‘Dutch treat’. In other words, forms like *[akÛs’i], *[k’urÛmi], *[anÛk’op] etc., are not possible words in Korean. The implication is that the vowel [Û] may not occur before a consonant followed by another vowel. This is not consistent with (5), because [Û] may occur in the initial nucleus when it is followed by a consonant and another vowel. However, note that there is a significant difference between the data: [Û] occurs in the initial nucleus in (5) and in the non-initial nucleus in (8). At this stage, we can make the preliminary generalisation that [Û] has a different distribution in the initial and the non-initial nuclear positions, i.e. the vowel [Û] in the initial nucleus position cannot be absent (see Chapter 3). However,
[Û] does occur before two consonants followed by a vowel as shown in
[mindÛlle] ‘dandylion’ [kotÛlkE] ‘horsebell’
It appears that the two consonants in (9) after [Û] can constitute ‘coda’-onset
clusters within ‘standard’ syllabification (S.-C. Ahn 1985, H.-S. Sohn 1986). Also, as in (8), if the underlined vowels are not taken into consideration, the two neighbouring consonants would have formed ‘coda’-onset clusters. That is, [aks’i], [k’ulmi], [ank’op], [alt’ap], [talk’i] and [silki] are possible words in Korean.11 This indicates that [Û] may not occur between two consonants in ‘coda’-onset clusters. It implies that the occurrence of [Û] is sensitive to the distribution of surrounding consonants. Before we investigate the occurrence of [Û] with respect to the distribution of surrounding consonants any further, the observations made thus far can be summarised as follows. (10) (a) [Û] may not occur morpheme-finally. (b)
[Û] in the first nuclear position may occur before a single consonant followed by another vowel, but not in other positions.
[Û] may occur before two consonants followed by
10 These two forms are synonyms. 11 Note that [l] rather than [r] occurs in ‘coda’ position. This is due to the allophonic relation between the two consonants, as discussed in 1.1.
Given the generalisations in (10), it is fair to say that [Û] has a rather limited distribution compared to other vowels in Korean. 1.3.4. The vowel [Û] and its surrounding consonants The generalisation in (10b) requires further refinement with respect to the occurrence of morpheme-internal [Û], because this vowel occurs before a consonant followed by another vowel, as shown in (11). (11) (a)
Between a nasal or a lenis obstruent, and a liquid
[cinÛrmi] ‘fin’ [htÛre] ‘trash’
(b) Between a lenis obstruent and a nasal
[nakÛne] ‘stranger’ [kocÛnk] ‘silent’
(d) Between a tensed or an aspirated obstruent, and a lenis obstruent
[kEk’Ûsi] ‘tidily’ [tlkHÛtk] ‘click’
In contrast to (8) and (9), if the underlined [Û] were absent, its surrounding
consonants would not form ‘coda’-onset clusters: e.g. *[cinrmi], *[htre] and *[pusrm]. These forms are not possible in Korean. Thus, the occurrence of [Û] is sensitive to the distribution of its surrounding consonants depending on whether or not they may serve as ‘coda’-onset clusters. The revised generalisation regarding (10b) is as follows. (12) [Û] may not occur between ‘coda’-onset clusters morpheme-internally. The condition in (12) regulates the absence of [Û] in (9) and its presence in (11).
In summary, the following table shows the combinatorial possibilities of two
consonants with respect to the intervening vowel [Û] (Y. Heo 1995: 32). (13) The occurrence of morpheme-internal [Û] between two consonants (C1 and C2)
Two examples do not conform to (13). The vowel [Û] invariably occurs after a
lenis fricative /s/, even though a tensed obstruent follows [Û]. According to ‘standard’ syllabification, these two types of consonants would have formed ‘coda’-onset clusters without the intervening [Û]. Two examples of words of this type are shown in (14). (14) [usÛk’wang] ‘funny’
The presence of [Û] between a lenis fricative /s/ and a tensed obstruent in native
Korean is limited to only two words, which may be treated as exceptions. However, as we will see later in section 1.5, the vowel [Û] in the same context occurs frequently in loanword adaptation. In particular, complex ‘coda’ clusters involving /s/ from English such as wasp
are phonetically realised as [wasÛpHÛ], [tHesÛtHÛ] and [tisÛkHÛ], respectively. Note that [Û] intervenes between a lenis fricative and an aspirated stop, which shows approximately the same distribution as the data in (14). Thus, the presence of [Û] in this context requires attention in native Korean. Therefore, we must seek a proper account of the occurrence of [Û] related to the special nature of fricatives.
Thus far, we have described the presence and the absence of the vowel [Û] in
mono-morphemic words. In the next section, I will consider verbal suffixation where Û/zero alternations occur. 1.4. The vowel [Û] in verbal suffixation As mentioned earlier, verbal stems in Korean cannot occur in isolation. They require at least one suffix. In general, we can distinguish two types of verbal suffixes depending on whether or not Û/zero alternations take place. In this section, the discussion focuses on the set of suffixes involved in Û/zero alternations. The remaining set of suffixes is dealt with in Chapter 4. Consider the following verbal suffixes with Û/zero alternation, as shown in (15), e.g. the intentive suffix alternates [r] with [Ûr], the connective [mj] with [Ûmj], and the honorific [si] with [Ûsi]. (15) (a)
/kapH/ [kapHÛr] [kapHÛmj] [kapHÛsi]
/pEtH/ [pEtHÛr] [pEtHÛmj] [pEtHÛsi]
/c’ocH/ [c’ocHÛr] [c’ocHÛmj] [c’ocHÛsi] ‘to
[tak’Ûr] [tak’Ûmj] [tak’Ûsi] ‘to
/ps/ [psÛr] [psÛmj] [psÛsi]
[c’icÛr] [c’icÛmj] [c’icÛsi] ‘to
/kr/ [kll] [klmj] [ksi] ‘to
In (15), we observe that [Û] occurs if the stems end in either an obstruent or a
nasal. It does not occur if the stems end in either a vowel or a liquid. As seen above in (13), the occurrence of [Û] in suffixation contexts is roughly similar to in mono-morphemic contexts. However, there are cases where Û-realisation is inconsistent with (13). As discussed in section 1.2, the nasal-final stems behave in a rather different way. Morpheme-internally, we do not encounter [Û] between two nasals. The relevant data are repeated from (3) in (16). However, in suffixation contexts, [Û] occurs between two nasals, as in (15b). (16) [simmani]
So far, we have investigated the occurrence of [Û] in both mono-morphemic and
suffixation contexts. What we have observed is that [Û] does not behave arbitrarily. Its distribution is controlled by the presence or the absence of following vowels and the quality of the surrounding consonants. One exception would seem to involve the morpheme-initial position, where the occurrence of [Û] does not seem to be affected by the presence of a following vowel. In addition, the distribution of [Û] in mono-morphemic words is identical to that in verbal suffixation but slightly different, with respect to nasal-final stems. In the next section, I discuss the occurrence of [Û] in loanword adaptation from English. 1.5. The vowel [Û] in loanwords When a foreign word is adopted in another language, its phonetic shape usually undergoes phonological changes in conformity with the segmental and the syllabic system of the host language. The segmental system deals with the transformation of segments of the source language which are lacking in the segmental inventory of the host language. The syllabic system deals with the (re-)adjustment of segments or the repair strategy which may violate language-specific syllabic conditions (Paradis 1988, Paradis & Lacharité 1997 and Yip 1993 among others). Typical instances of 12 In liquid-final stems, alternative phonetic forms are possible in child speech: [krÛr] in the intentive; [krÛmj] in the connective; [krÛsi] in the honorific.
syllabic readjustments involve deletion of consonants and insertion of vowels. In this section, I briefly investigate these two aspects of loanword adaptation, focusing on the occurrence of [Û].
Apart from Sino-Korean words, most loanwords in Korean come from English.
In terms of ‘standard’ syllabification, English has a rather more complex syllable structure than Korean. English has branching onsets and codas, while Korean does not (S.-C. Ahn 1991). Thus, when English words with complex consonant clusters are adopted in Korean, it is reasonable to assume that various repair strategies are required to readjust segmental sequences into the syllable structure of Korean. One prominent strategy is to insert [Û] to break up consonant sequences in order to preserve maximal segmental identity in comparison to the source language. Thus, we expect multiple Û-insertion in various positions in English loanwords. However, as we will see below, like in native Korean, the occurrence of [Û] is highly regulated in that it is not inserted in every consonant cluster. We do not see any insertion in certain types of consonant clusters. In this section, I describe the contexts in which [Û] does and does not occur in English loanword adaptation, and compare them with those of native Korean.
Before we investigate the occurrence of [Û] in loanwords, I sketch how English consonants are adopted into the consonant system of the borrowing language. First, consider the following consonants in non-final, pre-vowel position. English voiceless stops are realised as aspirated stops, and voiced ones are converted into lenis stops in Korean. (17) English
With respect to other obstruents, inter-dental fricatives /T/ and /D/ are realised as
[s] and [t], respectively. The coronal fricatives /s/ and /z/ are realised as [s] and [c]. The labial fricatives /f/ and /v/ are realised as [pH] and [p]. The palato-alveolar fricative /S/ and /Z/ are realised as [sj] and [c]. The affricates, /tS/ and /dZ/ are as [cH] and [c]. Thus, [c] in Korean corresponds to /z/, /Z/ and /dZ/ of English. The relevant examples are shown in (18). (18) thinner
is a brand of cigarette in Korea.
The three nasals of English are realised in Korean without modification, as
The English liquids /l/ and /r/ are adopted in rather complex ways in Korean. First, initial /l/ and /r/ in English are realised as [l] in Korean. Thus, there is no segmental distinction between these two consonants in initial position in Korean due to the allophonic relation between [l] and [r], as mentioned in section 1.1. In internal position, however, they are realised as [ll] and [r], respectively. Consider the following examples. (20) lemon
Regarding the initial position, as in lemon
, there is a distributional
constraint on liquids in native Korean: liquids cannot occur in initial position.14 If this constraint were active in loanword adaptation, then serious confusion would arise between liquid-initial and vowel-initial words; both types of words would begin with a vowel. The constraint on liquids is suspended in loanword adaptation, probably to minimise lexical confusion.
With respect to the internal position, internal /l/ in English is represented as [ll]
in Korean, while /r/ is realised as [r]. It seems that the English orthography, irrespective of the number of l
’s being spelled, does not affect the realisation of internal /l/ as [ll] in English loanword adaptation, as in cola
15 One important observation is that a vowel always follows [ll]. This observation is supported by the word film
where [Û] occurs between /l/ and /m/, i.e. [pHillÛm]. Unlike /l/ and /r/ in initial position, the segmental identity of English internal /l/ and/r/ is more or less preserved.
To summarise, the following schema illustrates how English consonants are
converted into Korean in initial and internal position.
14 In fact, the pronunciation of my surname is [li], but the actual pronunciation is [i] due to the fact that surnames in Korean occur initially rather than finally as in western names, i.e. [l] is dropped in initial position. Interestingly, note that North Korean does not have this constraint. Therefore, the same surname is realised as [li] without l-dropping. 15 As we see later in Chapter 3, the sequence [ll] is possible morpheme-internally in native Korean: [klle] ‘dustclothes’, [plle] ‘insect’, [tallE] ‘to soothe’, [nalljp] ‘quick’ etc. However, note that a single [l] occurs only in final position.
In the next section, I deal with final consonants as well as complex consonant
clusters where we observe multiple Û-insertion. 1.5.2. The distribution of [Û] in loanwords In terms of ‘standard’ syllabification, Korean does not allow complex onsets and codas. Accordingly, when a language which has complex consonant clusters, like a word is borrowed from English, [Û] is inserted in various positions in order to conform to Korean syllable structure. Thus, the prime sites of the occurrence of [Û] are branching onsets and codas where [Û] appears between two consonants in the former case, and after two consonants in the latter. Some relevant examples of each case are shown in (22) and (24), respectively.
(22) (a) Branching onset: Cr cluster
In (22a), we observe that all underlined Cr clusters, i.e. /pr/, /tr/, /kr/, /fr/, /br/, /dr/, /gr/ are resolved by the vowel [Û], as in (22a). With respect to the Cl clusters in (22b), the vowel [Û] occurs between an obstruent and /l/; /l/ is also realised as [ll] in
Korean, as discussed in (20). The occurrence of [Û] in this position is intended to maximally preserve the segmental identity of the source language. Note that the context in which [Û] occurs in (22a) coincides with the native Korean words in (11a) in section 1.3.3, where [Û] occurs between an obstruent and a liquid: [htÛre] ‘trash’; [sinapÛro] ‘gradually’; [sikÛrci] ‘to vanish’. The examples in (22) also illustrate that [Û] appears in the first syllable as for native Korean in (5) in section 1.3.1, e.g. [kÛrEto] ‘even if’; [s’Ûreki] ‘rubbish’; [sÛrasoni] ‘lynx’. This strongly indicates that even in loanwords the behaviour of [Û] is not random at all. In addition, consider loanwords in which [Û] occurs between two aspirated stops. The source of these examples is usually related to natural science terminology or to German. As noted in (13) above, these cases are absent among native words. (23) [aks’eropHÛtHol]
Consider complex ‘coda’ clusters in final position.
In (24), we observe that [Û] does not appear between two consonants, but it does occur finally. Recall that in native Korean this vowel does not occur in final position. The occurrence of [Û] in final position seems obligatory when consonant clusters precede it, i.e. *[pHlÛp] or *[pHlÛpHÛ]. As discussed in section 1.3.3, [Û] does not occur between ‘coda’-onset clusters. As we see in (25) below, the presence of final [Û] in loanwords is also affected by the quality of the preceding consonant, i.e. it does not occur after certain types of segments.
In order to see how the quality of consonants determines the presence of [Û] in
final position, consider the following English loanwords that end in a single consonant. (25) (a) Nasal-final
[Û] does not appear after nasal- or liquid-final words, as in (25a) and
(25b). The r-final English words are of interest. Note that the segment /r/ is not realised on the surface, as in the word car
in Korean. The same phenomenon is observed in internal position where /r/ occupies ‘coda’ position followed by an onset, e.g. carpet
[kHapHet] or [kHapHetHÛ], target
[tHaket] or [tHaketHÛ], and so on. This implies that Korean adopts a non-rhotic (or British) accent in English loanword adaptation.16 In other words, the treatment of coda-/r/ of English loanwords is the same as that of the British accent: coda-/r/ is deleted before a consonant or in final position. This is surprising in the sense that the influence of the American accent prevails in Korea. For instance, English education is based upon American English and Korean adopts the American accent with respect to vowel-quality, e.g. dot
) [tat(kHm)] and hip hop
[hip hap]. Non-rhoticity in English loanwords would be due to the nature of /r/ in Korean, i.e. /r/ does not occur in ‘coda’ position. I will discuss the matter regarding to the segment /r/ in more detail in Chapter 3 and 5.
Regarding the stop-final words in (25c), at first glance, the generalisation seem
to be that [Û] occurs irrespective of whether final stops are voiceless or voiced in the source language. In fact, the presence of [Û] in final position may be a significant indicator to distinguish native from loan words in Korean. It is reasonable to assume that Koreans perceive some degree of foreignness with [Û] in final position, since there are almost no words ending in this vowel. With respect to the voiceless stop-final words, however, we observe that there are alternations between an aspirated and a lenis stop: aspirated stops occur with the following [Û], lenis ones without it. What these alternations suggest is that the occurrence of [Û] is sensitive to the quality of a preceding consonant. Parallel cases arise in English words ending in complex codas where /s/ is involved.
Unlike (24), where [Û] does not occur in ‘coda’-onset clusters in English, the examples in (26) show that it is not the case. The vowel [Û] mediates between /s/ and a following aspirated stop, despite the fact that these consonant sequences are a well-formed ‘coda’-onset ones in English. Once again, this strongly suggests that the quality of surrounding consonants affects the occurrence of [Û] in loanwords as well as native words, as discussed in section 1.3.3, e.g. [usÛk’wa @ ‘funny’ and [mesÛk’p] ‘to feel sick’, as in (14). The relevant generalisation is that the vowel [Û] 16 For a general description of the difference between the British and the American accent, see Wells (1982). Another instance of non-rhoticity in English loanword adaptation is found in Cantonese which has been influenced by British English historically and socially (Silverman 1992, Yip 1993). However, S.J. Rhee (2000) argues that non-rhoticity of both languages in English loanword adaptation is mainly due to phonotactic constraints on /r/ rather than social and historical factors.
must occur after the fricative /s/ followed by aspirated or tensed obstruents in both native and loan words.
Consider (25d) where the English examples end in fricatives or affricates. Like
the stop-final words in (25c), final voiceless and voiced labial fricatives in English are realised as [pHÛ] and [pÛ] respectively. Regarding coronal fricative-final words, we observe that [Û] occurs finally after a fricative and an affricate, as in [psÛ] and [cEcÛ]. Recall from (21) that /z/ in English is realised as a lenis affricate [c] in Korean. Thus, the example jazz
provides another context in which [Û] occurs after an affricate.
Thus far, we have seen that [Û] occurs between consonant clusters, and, in final
position in English loanword adaptation. However, we note that there are other cases where the vowel [i] occurs in final position instead of [Û], as shown in (27). (27) sash
All these examples end in a palatal fricative or a palatal affricate in the source language. These consonants are converted into [si] for /S/ as in sash
, [ci] for /Z/ and /dZ/ as in beige
, respectively, and [cHi] for /tS/ as in coach
. It is plausible to assume that the realisation of [i] in this position is influenced by the preceding palatal consonant. In particular, the distinction between /Z/ and /dZ/ in English is merged into [ci] in Korean. Recall that the segment [c] is also used to represent English /z/, as in (25d). In this case, the vowel [Û] rather than [i] occurs in final position. Hence, the segmental distinction between /Z/ and /dZ/, and /z/ in English is made on the basis of which vowel occurs in final position.
Until now, I have considered English loanwords ending in a single consonant.
We observe that the occurrence of [Û] depends on the quality of preceding consonants, i.e. if they are a nasal or [l], then it does not occur; if they are fricatives or affricates, it does occur, while the vowel [i] occurs after English palatal consonants. Comparing (25) with (24), an important difference is noted in that the occurrence of [Û] is obligatory when it follows complex coda sequences in (24), but its occurrence depends on the quality of preceding consonants in (25). Thus, in loanwords, we do not have forms such as *[plpH], *[peltH], *[plkH], among others.
To sum up, the contexts in which [Û] occurs among loanwords are as follows.
(i) The vowel [Û] does not occur after nasals and liquids.
(ii) The vowel [i] occurs after palatal affricates and fricatives.
(iii) The vowel [Û] occurs after non-palatal fricatives.
(iv) The vowel [Û] occurs after stops; but it alternates with zero after words ending in voiceless stops; if zero alternants occur, the preceding stops must be lenis.
(i) The vowel [Û] occurs finally when complex clusters precede it.
(ii) The vowel [Û] does not occur between these two consonants if the first member is a sonorant (e.g. nasal or liquid); it does occur between the two if the first member is a fricative
(d) Words containing two contiguous aspirated segments
[Û] intervenes between these two segments.
1.6 An alternative view of the vowel [Û] in Korean Up to now, I have investigated the occurrence of [Û] in native Korean and loanwords. I have tried to show that the distribution of the vowel [Û] is not arbitrary in native Korean, i.e. it is highly regular in mono-morphemic as well as in suffixation contexts. The quality of neighbouring consonant(s) affects the presence or absence of [Û]. Similarly, the occurrence of [Û] in loanwords is also systematically constrained, i.e. its distribution depends on the quality of adjacent consonants. This section reviews an Û-insertion approach to loanwords. Then, I put forward an alternative view on the vowel [Û] by rejecting the previous approaches to it.
As I pointed out in section 1.2, both Û-deletion and Û-insertion analyses face
empirical problems. The deletion analysis is not able to handle the asymmetrical distribution of NN sequences in mono-morphemic and in suffixation contexts. The result of Û-insertion is to treat productive suffixation in the interrogative and the indicative as exceptional. With respect to loanwords, the main motivation of Û-epenthesis is to conform to Korean syllable structure which bars branching onsets and final complex codas. For instance, S.-C. Ahn (1998: 211-212), proposes the following four rules to account for Û-epenthesis. (29) Vowel epentheses in loanwords
1 DQ HPSW\ VORW IRU D YRZHO V\OODEOH & FRQVRQDQW |
17 The N with an empty slot refers to an unspecified nucleus that will get phonetic content. According to Ahn’s proposal, the phonetic content is provided by redundancy rules of vowel specification. I do not discuss this matter in this chapter; see Chapter 5 for a detailed discussion on this issue.
(d) Epenthesis in coda (III): default case
Putting aside rule (29b) for the time being, let us consider the empirical effects of these three rules. Rule (29a) is designed to break up a consonant cluster by Û-insertion since Korean allows only a single consonant in onsets. The rules in (29c) and (29d) treat complex coda sequences in final position. The former deals with lC and NC (N: nasal, C: obstruent) clusters such as lp
etc. The latter concerns sC clusters such as sp
. Interestingly, the insertion sites of these three rules are exactly the same as those for the vowel [Û] in native mono-morphemic words as discussed in 1.3. Some relevant native examples are repeated in comparison with loanword ones. (The sequences under comparison are underlined.) (30) (a) Rule (29a)
(30) indicates that the contexts in which [Û] occurs in the formulations of Û-insertion rules of (29) are exactly the same as the distribution of [Û] in native words. This illustrates that a rule-based approach to loanwords runs into the duplication problem
(Kenstowicz & Kisseberth 1977: 136), i.e. rules are added to the grammar which duplicates the morpheme structure constraints.
Finally consider rule (29b). This rule is designed to account for the occurrence
of [Û] in final position, when loanwords in the source language end in a voiced stop,
as in cube
[petÛ] and tag
[tHEkÛ].18 In fact, these examples constitute a set of words which is totally unique to loanwords, precisely because almost no native words end in the vowel [Û]. This rule, however, does not take into account alternations between aspirated and lenis stops, as illustrated in (25c), e.g. soup
[sup] or [supHÛ], cut
[kHt] or [ktHÛ] and cake
[kHeik] or [kHeikHÛ]. What these examples show is that there is an intrinsic relation between the laryngeal nature of the stops and the occurrence of [Û]: the presence of aspirated stops requires a following [Û], but the presence of lenis stops does not. Within Ahn’s rule-based formulation, several additional rules are required in order to capture such alternations between aspirated and lenis stops and between [Û] and zero. For instance, since English voiceless stops are converted into aspirated stops in Korean, a process in which aspirated stops become lenis in final position is needed, e.g. [sup], or another process, in which the vowel [Û] is inserted when it follows aspirated stops, is required, e.g. [supHÛ]. Such rules, however, do not seem to establish a non-arbitrary relation between the phonological processes in question and the contexts in which they take place. Concretely, these rules do not reveal the reason why lenis stops without [Û] occur in final position, and why aspirated stops require a following [Û] when they occur in final position.
Thus far, I have examined the Û-insertion approach to loanwords. Apart from
final [Û] in loanwords, the vowel [Û] in internal position shows that there is no phonotactic difference between native and loanwords. This implies that we can treat the internal [Û] in loanwords in the same grammar as native words. The previous Û-deletion and the Û-insertion analysis, however, appear not to capture the same distribution of [Û] in a unified way, precisely because these approaches assume that the native morpheme-internal [Û] is lexically specified.
However, there is an alternative way to achieve a unified treatment of the vowel
[Û]. This is to recognise an empty nucleus
in the lexical representation with the assumption that the phonetic interpretation of an empty nucleus is the vowel [Û] when certain conditions are satisfied, and that it is not audible otherwise. For instance, consider the words [pipÛrapHon] ‘vibraphone’ and [sinapÛro] ‘gradually’. Under this approach, the lexical representations are as follows. (31) (a) [pipÛrapHon] vibraphone
18 According to Ahn’s transcription, the phonetic forms are cube
[pedÛ] and tag
[tHEgÛ] where intervocalic voicing takes place. As we will see in Chapter 3, I argue that intervocalic voicing is not present in the phonology of Korean. Thus, throughout the thesis, all intervocalic obstruents are represented by lenis ones. In addition, close inspection of rule (29b) reveals that this rule comprises two processes: Û-epenthesis and intervocalic voicing. That is, after Û-epenthesis, intervocalic voicing occurs. 19 In particular, note that the final /n/ in (31a) is not syllabified. Full syllabification of given segmental strings will be discussed in Chapter 2 and onwards.
(31) gives partial lexical representations in which N2 and N3, in (31a) and (31b), respectively, are empty. The task is, then, to develop an algorithm of how to interpret these empty nuclei. If an adequate algorithm is available, the obvious advantage of this approach is that the grammar becomes simpler, because no Û-insertion and Û-deletion rules of the type in (29) are necessary in Korean phonology.
Regarding the algorithm for the interpretation of empty nuclei, I adopt
Government Phonology (GP), which provides a principles-and-parameters-based account, i.e. the Empty Category Principle. In the remaining chapters, I will investigate how this principle together with language-specific conditions determines the phonetic interpretation of empty nuclei in Korean.
A David A. Damari, O.D. stereopsis, which was reported as “30 sec- Jeannette Liu, O.D. onds of arc/normal.” For the examination Karen Bell Smith, O.D. plaints, but her father was concerned be-cause of decreased working distance at the Abstract Attention Deficit/Hyperactivity Disorder now characterized as a mental disorder are (ADHD) is one of the most studied, and most c
CURRICULUM VITAE SHARIK ALI Department of Biotechnology, Kurukshetra University, Kurukshetra, Haryana Personal Profile Name Sharik Ali Father Name Mr. Shakir Ali Date of Birth 29 May 1985 Sex Male Nationality Indian Permanent Add. 588, Rawatyana, Lalitpur-284403, Uttar Pradesh, India Educational Qualification 1. Doctorate, Ph.D. (Biotechnology) Pursuing from Kuruksh