The Sound Systems
3.2.3 Place of articulation
國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
92
appear syllable-initially, whereas unaspirated stops appear as the second member of an onset cluster after [s]. Refer to (59).
(59) Allophones of English stops
Phoneme Aspirated Unaspirated
/p/ [.p t.], pot [.sp t.], spot /t/ [.t p.], top [.st p.], stop /k/ [.k l.], kill [.sk l.], skill
It is interesting to observe adaptation of English stops into TM, as both English and TM have aspirated and unaspirated stops, but only English has voiced stops. That is, the mappings have to be asymmetric. If, more uncontroversially, aspirated stops are to map to their identical aspirated counterparts, it seems mysterious how English unaspirated and voiced stops are interpreted by TM adapters. As we will see, perception again is largely concerned with the mapping of stops, where VOT plays a crucial role in their asymmetric mapping.
3.2.3 Place of articulation
Traditionally, there are a set of binary features that are capable of specifying the major places of articulation of consonants, i.e. [labial], [coronal], and [dorsal]. A comparison between English and TM in this respect can be made in what follows.
For labial consonants, except that English has the voiced bilabial stop [b] and the voiced labio-dental fricative [v], the two languages bear a great similarity under this category. English has the bilabials [p, b, m, w] and the labio-dentals [f, v]. TM has the bilabials [p, p , m, w, ] and the labio-dentals [f].
English voiced bilabial stop [b] is mapped to either the voiceless [p] or the aspirated counterpart [p ] in TM. In addition to its voiceless counterpart [f], as just
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
93
mentioned, English [v] is occasionally interpreted as the labialvelar glide [w] in TM.
The two languages show a larger discrepancy in coronal consonants, however.
On the one hand, English has two interdental consonants that are absent in TM, [', ;], which are also unusual in the world’s languages. English interdental consonants are predominantly interpreted as the closest [s] in TM. On the other hand, TM is rich in dental sounds, including [t, t , s, ts, ts , n, l]. The exact places of articulation of the stops [t, t ], fricative [s], nasal [n], and approximant [l] reveal great individual/dialectal differences, which can be optionally classified as alveolar sounds (as in English) if produced by placing the tip or blade of the tongue slightly behind the upper teeth before the alveolar ridge, though in general they are made more forward by TM speakers than their English counterparts. Unlike TM, English [t, d, n, s, z, , l]
are more uncontroversially categorized under [alveolar]. English obsturents with this place of articulation are mostly mapped to their dental counterparts in TM
English has five palatal consonants [%, #, t%, d#, j], but TM only has the glides [j, ] in this position. Between the places of alveolar and palatal, TM has two places that English does not have: post-alveolar and alveolo-palatal. The four post-alveolar consonants [ , t , t , ] are also called retroflex sounds as production of them involves curling of the tongue for the underside to touch or approach the post-alveolar region. The retroflex [ ] is traditionally classified under post-alveolar whereas it is currently identified as alveolar in English (Hayes 2009). Additionally, TM has alveolo-palatal consonants [$, t$, t$ ], which are viewed as allophones due to their restrictive context to appear (preceding [i, j] or [y, ]), though as mentioned, it remains controversial as to which set of consonants they are allophones of (see (56) for details). English palatal obstruents are alternatively mapped to TM post-alveolars or alveolo-palatals.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
94
Toward the back regions, English has four velar consonants, i.e. [k, (, , w], while TM has five, i.e. [k, k , x, , w]. English velar stops are mapped to their voiceless and aspirated counterparts in TM. Mapping of nasals, however, depends greatly on the preceding vowel. The ‘rhyme harmony’ phenomenon (Hsieh, Kenstowicz, and Mou 2009) will be discussed in detail in Section 5.10. For fricatives, TM has a velar fricative [x], while English doesn’t. Analogously, English has a glottal fricative that TM does not have, [h]. It thus makes great sense that English [h] is interpreted as the most similar [x] in TM. A noteworthy aspect is that, as noted above, TM [x] is gradually replaced by [h] due to the influence of Taiwanese Southern Min (Hsiao 2011), which more or less reinforces the rationality of this mapping tendency.
3.3 Vowels
The vowel constitutes the nucleus of a syllable, and is defined with the properties of height, backness, and roundness. As this dissertation does not intend to probe into TM adjustment of English vowels, we simply give the fundamental background knowledge in this respect. Generally, English has a richer inventory of vowels than TM. They are presented in (60) and (61), respectively. English has thirteen vowels, and TM has nine, with all allophones temporarily included.
(60) English vowels (based on Hayes 2009)
front central back
unrounded unrounded rounded
high tense i u
lax
mid tense e o
lax /
low æ a
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
95
(61) TM vowels (based on Chao 1968, Cheng 1973, Lin 2007b)
front central back
unrounded rounded unrounded rounded
high i y u
mid e o
low a
Note that in English, the low central vowel [a] should not be considered to be phonemic as it occurs in diphthongs only. In TM, comparatively, recent scholars agree that TM has five phonemes /i, y, u, , a/ (Duanmu 2000, 2007; Lin 2007b). This statement is grounded on the following observations, though slight disagreement on allophonic classification still exists. For mid vowels, on the one hand, [e] appears in the diphthong [ei] in open syllables, but not in syllable-final position, unless preceded by palatal glides [j, ]. The variant [ ] appears in open syllables, but cannot follow a palatal glide. Moreover, [o] occurs in the diphthong [ou] or is preceded by labials in open syllables. Finally, schwa [ ] appears exclusively before nasals [n, ]. For low vowels, on the other hand, [a] exists in open syllables, before [n] in closed syllables, and in the diphthong before [i]. The lowest [ ] appears in the diphthong ending in [u]
and before [ ] in closed syllables.
What is also worth mentioning is that TM high vowels [i, y, u] are not contrastive with their glide counterparts [j, , w]. To be more specific, when a high vowel precedes a mid or low vowel, which is more sonorous, the latter is parsed as the nucleus, and the high vowel becomes its glide counterpart and is assigned to the onset position.
There is a high vowel that is not listed in the inventory but does exist in spoken TM, i.e. the apical vowel [!]. It appears syllable-finally after dental ([ts, ts , s]) and post-alveolar ([t , t , , ]) fricatives, affricates, and approximant, all of which are
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
96
apical consonants. Scholars (Chao 1968; Duanmu 2000, 2007; Lin 2007b) have treated this vowel as the voiced extension of the preceding consonant into the nuclear position, rendering the latter a syllabic consonant. This dissertation follows these scholars and considers [!] to be a syllabic approximant, not a real vowel.
In addition to monophthongs, English has five diphthongs, and TM has four, as shown below.
(62) English diphthongs
[ ] [ ]
mid e o
mid
low a a
(63) TM diphthongs
[i] [u]
mid ei ou
low ai u
A handful of differences between English and TM diphthongs are worth noting.
First, English diphthongs end in lax vowels [ ] or [ ], while TM diphthongs end in tense vowels [i] or [u], though acoustically those in the latter do not precisely reach the height of [i] or [u] (Lin 2007b). Moreover, the low front [a], being a phoneme in TM, occurs more freely both in the diphthong [ai] and as a monophthong, whereas it is restricted to being part of the diphthongs [a ] and [a ]. Next, TM diphthong [ u]
differs from English [a ] in that the former starts with a low back vowel but the latter begins with a low front vowel. Finally, the diphthong inventories of the two languages are quite similar, except that English has [ ], which is absent in TM, involving back-to-front movement of the tongue.
Adaptation of vowels in TM loanwords from English reveals greater instability
‧
compared with that of consonants. This may be closely correlated to the fact that each TM character (a syllable) acts as a minimal unit in meaning (a morpheme), and thus choice of characters may depend on the semantic function it carries. See (64) for example.
(64) Influence of semantic factors
English input TM output Mapping
a. [.p æn.d .] panda → [.p .ta.] 胖達 [æ] → [ ] b. [.p æn.do. .] Pandora → [.p an.two.la.] 潘朵拉 [æ] → [a]
In (64), while the vowel of the first syllable in both cases is [æ], it is mapped to the low back [ ] in (64a) but to the low front [a] in (64b). This one-to-many asymmetry is not purely coincident but attributable to the word meanings the chosen characters carry. In (64a), the adapter deliberately uses the character 胖 ([.p .], ‘fat’) to create the conceptual connection between the loanword and the chubby figure of pandas16. In (64b), however, the adapter chooses the character 潘 ([.p an.]), a renowned Chinese family name, as the first character of this female name, since unlike English, family name comes first in Chinese17.
Another observation in TM loanwords from English that more or less influences the process of loanword production is the existence of certain preferable characters in foreign word transliteration, which, contrary to the semantic factors just mentioned, leads to a many-to-one asymmetry in vowel mappings. Take (65) for example.
loanwords from English, where still a number of patterned variations are discovered and analyzed.
This dissertation will not go into her details and lay the focus on consonants.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
98
(65) Influence of preferable characters
English input TM output Mapping
a. [.ni.s n.] Neeson → [.ni.$yn.] 尼遜 [i] → [i]
b. [.n l.s n.] Nelson → [.ni. r.s n.] 尼爾森 [ ] → [i]
c. [.na .æ.( .r .] Niagara → [.ni.t$ja.la.] 尼加拉 [a ] → [i]
In (65), the three vowels [i], [ ], and [a ] in the sources consistently map to the high front [i]. We assume that they are not perceptually interpreted as the same vowel, as the three vowels show a large discrepancy in perception. What crucially leads to this asymmetry is that conventionally, TM native speakers tend to use the character 尼 ([.ni.], ‘nun’), among many others, in transliteration of foreign names.
Still another factor that may be influential to the vowel adaptation is that a foreign input is subject to the legality of syllable combination of the borrowing language. For any given syllable that is not possible in the native language, even if all the segments are separately permissible, either the consonant or the vowel undergoes adjustment in word-loaning processes. Refer to (66) for illustration.
(66) Influence of syllable legality
English input TM output Mapping
a. [.k .t .] Kitty → [.k ai.ti.] 凱蒂 [ ] → [ai]
b. [.k d.m n.] Kidman → [.t i.man.] 基嫚 [k ] → [t$]
The best mapping to the English sequence [.k .] in TM should be [.k i.], as TM has no lax vowels. However, as has been discussed earlier in (56), velar stops can never procede a high vowel, and it follows that the input [.k .] has to undergo segmental change either through a greater alternation of the vowel or through the consonant. This is what happens in (66). In (66a), the adapter adopts the first strategy and the nuclear vowel is changed to the lower diphthong [ai], whereas in (66b), the
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
99
second strategy is adopted and the onset consonant is adapted into the affricate [t$].
Finally, it is undeniable that TM loanword adaptation is influenced by English orthography to a certain degree. The adapter may still take spelling into account if he or she sees the written form of the target word. Let us take (67) for example.
(67) Influence of orthography
English input TM output Mapping
a. [.h .v d.] Harvard → [.xa.fo.] 哈佛 [ ] → [a]
b. [.h p.k ns.] Hopkins→ [.xwo.p u.t$in.s!.] 霍普金斯 [ ] → [o]
Both pronounced as [ ], the first vowel in Harvard is adapted to [a], while that in Hopkins is adapted to [o] in TM, which is assumed to result from the adapter’s judgment on the spelling of the source words18.
Despite the great irregularity of vowel adaptation, a number of general tendencies can still be observed in the following subsections.