The Corpus and Patterns
4.5 Simplex codas
國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
135
only the [f ] sequence occurs in a high frequency in the corpus. The patterns are shown in (90).
(90) Adaptation patterns of the “fricative-[ ]” sequences L2 seq. L1 struct. Percentage Example (L2 → L1)
[f ] [.f1V1.l2] 94.12% (16/17) [.f o .do .] Frodo → [.fo.lwo.two.] 佛羅多 [.f1] 5.88% (1/17) [.f st.] Frost → [.fo.s!.t .] 佛斯特
[v ] [.w12] 100% (1/1) [.æ.v l.] Avril → [.ai.wei. .] 艾薇兒 [' ] [.s1V1. 2] 66.67% (2/3) [.( . .] Guthrie → [.k .s . wei.] 葛瑟瑞
[.s1] 33.33% (1/3) [.d a v. u.] Drive Thru → [.t .lai.su.] 得來速
[s ] [.s1V1.l2] 100% (1/1) [.s .læ .k .] Sri Lanka → [.s .li.lan.k a.] 斯里蘭卡
[% ] [. 1V1. 2] 100% (1/1) [. k.] Shrek → [. . wei.k .] 史瑞克
Overall, retention of both segments through vowel insertion for the fricative and lateralization for the retroflex approximant remains the primary strategy, accounting for 86.96% (20/23) across the board. A small 8.7% (2/23) of them undergo retention of the fricative but elision of [ ], as can be observed in the [f ], [v ] and [' ] sequences. Notably, in “[.æ.v l.] Avril → [.ai.wei. .] 艾薇兒”, the labial [w] in the output can be said to be the mapping of the labiodental [v], the retroflex [ ] through gliding, or both.
4.5 Simplex codas
Turning to syllable-final codas, we have seen that in Chapter 2, English permits a wide range of consonants to be the coda, including obstruents, whereas TM has a rather limited set of permissible consonants to appear in coda: only the alveolar nasal [n], velar nasal [ ], and retroflex approximant [ ] (when the nuclear vowel is a schwa) are allowed to be the coda consonant. In addition, English allows as many as four
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
136
consonants to be the coda (e.g. “[.t ksts.], texts”), while clusters are strictly prohibited in both syllable margins in TM. It follows that except for a simplex [n], [ ], and post-schwa [ ], all the other English syllable-final consonants have to undergo adjustments to fit the severe restriction of TM on the coda position. In this section, the attention is centered on the adaptation of simplex codas. Crucially, as elaborated on in 2.6.1, monosyllabic loanwords are intentionally left out throughout this section to avoid the prosodic influence of binary feet in TM.
4.5.1 Stops
The two languages involved have similar sets of stops, as has been seen in (74) and listed below again for convenience.
(91) Stops of English and TM (also (74))
bilabial alveolar/dental velar
[–vcd, –asp] p p t t k k
[–vcd, +asp] p p t t k k
[+vcd] b d
The only difference lies in the absence of voiced stops in TM. For stop onsets, we make the prediction that except for faithful mappings of English [p, t, k] to TM [p, t, k] and English [p ,t , k ] to TM [p , t , k ] respectively, Englishvoiced stops [b, d, (] should be recognized as [p, t, k], since they are only slightly distinct in terms of VOT. In observing stop codas, on the other hand, as voiceless consonants are unaspirated in coda by default, it is expected that unlike onsets, in instances of retention, English voiceless unaspirated stops will mainly be mapped to TM aspirated stops, yet English voiced stops will mostly be mapped to TM voiceless unaspirated stops. The prediction is made on the ground that though it may seem ideal for an
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
137
English voiceless unaspirated stop to be recognized as an identical TM sound, the cost, however, would be to allow an English voiced stop to map to a TM aspirated stop, rendering more discrepancies between this pair in voicing and aspiration. This is obviously a worse result if adaptation occurs to a large extent in perception.
Let us go beyond the segmental changes in feature and discuss the broader retention/deletion distributions of the excess coda consonant. As mentioned in Chapter 1, the “Preservation Principle” (Paradis and LaCharité 1997, LaCharité and Paradis 2005) in loanword adaptation maintains that the segmental information in the source be maximally preserved in the output form, unless the cost of preservation is too extreme. The principle is claimed to be universal in the sense that it is the adapter’s intention to faithfully copy the detailed pronunciation of a foreign word when this word is newly coined in his/her native language. It hence tempts us into making the prediction that in our situation, an excess consonant coda in English should undergo vowel insertion to form an additional syllable in TM. The adaptation patterns in question are given in (92).
(92) Adaptation patterns of stop codas a. [p]
TM Percentage Example (L2 → L1)
[.p V.] 50% (8/16) [.(æ.l p.] Gallup → [.kai.lwo.p u.] 蓋洛普 Ø 25% (4/16) [.h p.h p.] hip hop → [.$i. xa.] 嘻哈
[.pjV.] 18.75% (3/16) [.t% p.l n.] Chaplin → [.t wo.pje.lin.] 卓別林 [.fV.] 6.25% (1/16) [. æ.l p.] Ralap → [.lei. .fu.] 雷爾夫
b. [t]
TM Percentage Example (L2 → L1)
Ø 56.82% (25/44) [.d#æ.k t.] Jacket → [.t$ja.k .] 夾克
[.t V.] 40.91% (18/44) [.hæm.l t.] Hamlet → [.xa.mu.lei.t .] 哈姆雷特 [.tsV.] 2.77% (1/44) [.do .n t.] donut → [.two.na.ts .] 多拿茲
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
138
c. [k]
TM Percentage Example (L2 → L1)
[.k V.] 68.33% (41/60) [.mo .ze . k.] mosaic → [.ma.sai.k .] 馬賽克 Ø 25% (15/60) [.l .d# k.] logic → [.lwo.t$i.] 邏輯 [.kV.] 3.33% (2/60) [.sto . k.] Stoic → [.s!.two.k .] 斯多葛 [.xV.] 1.67% (1/60) [.mju.n k.] Munich → [.mu.ni.xei.] 慕尼黑 [. V.] 1.67% (1/60) [.zu. k.] Zurich → [.su.li. .] 蘇黎世
d. [b]
TM Percentage Example (L2 → L1)
[.pV.] 100% (4/4) [.( b.s n.] Gibson → [.t$i.pu.s n.] 吉布森
e. [d]
TM Percentage Example (L2 → L1)
Ø 53.13% (17/32) [.æn.d d.] Android→ [.an.t wo.] 安卓 [.tV.] 37.5% (12/32) [. d.w n.] Edwin → [.ai.t . n.] 愛德恩 [.t V.] 9.38% (3/32) [.k ju.p d.] Cupid → [.t$ jou.pi.t .] 邱比特
f. [(]
TM Percentage Example (L2 → L1)
[.kV.] 62.5% (5/8) [.d .l s.] Douglas → [.t u.k .la.s!.] 道格拉斯 Ø 37.5% (3/8) [.d#i.t .b .] jitterbug → [.t$i.lu.pa.] 吉魯巴
With respect to the pattern of retention/deletion, the result barely corresponds to our expectation in that overall, retention (the mapping of “CV”) makes up only 60.98% (100/164) of all, with the remaining 39.02% being deletion. Individually, 44.74% (34/76) of English alveolar stops [t] and [d] undergo retention, which is the lowest proportion among all stops, whereas 80% (16/20), the highest among all, of L2 bilabials [p] and [b] are retained through vowel insertion.
Turning to the segmental changes, we see that much to our expectation, in cases where the stop is retained, a significant 93.06% (67/72) of all English unaspirated
‧
[.t wo.pje.lin.] 卓別林”. Moreover, as expected again, among cases where the stop coda is retained, English voiced stops map to voiceless unaspirated stops 87.5%(21/24) of the time, as in “[.d .l s.] Douglas → [.tau.k .la.s!.] 道格拉斯”. There seems no remarkable difference in featural changes concerning the distinction in voicing and place of articulation.
4.5.2 Fricatives
As has been discussed in 4.3.2, there exist more discrepancies between English and TM in fricative consonants than in stop consonants. The inventories of the fricatives in the two languages are repeated below.
(93) Fricatives of English and TM (also (76)) labio-
Except for this, all English consonants can serve as the coda. On the one hand, as TM allows no coda other than [n], [ ], or [ ] when the nuclear vowel is a schwa, an English consonant except the three in this position should either be retained via vowel epenthesis or simply be ignored. On the other hand, since the only fricatives that the two languages have in common are the labialdental [f] and alveolar [s], consonants that are not [f] or [s] will inevitably undergo featural changes in addition to vowel
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
140
insertion if the adapter is to retain the fricative. Given the high noise of friction in the utterance of fricative sounds, it makes great sense to expect that a TM adapter’s primary strategy for an English fricative coda is to retain it by epenthesizing a vowel to form an independent syllable. The adaptation patterns of fricative codas are shown below.
(94) Adaptation patterns of fricative codas a. [f]
TM Percentage Example (L2 → L1)
[.fV.] 75% (3/4) [.d#o .s f.] Joseph → [.t$ jau.s .fu.] 喬瑟夫 Ø 25% (1/4) [.sm .no f.] Smirnoff → [.s!.mei.lwo.] 思美洛
b. [']: unattested
c. [s]
TM Percentage Example (L2 → L1)
[.sV.] 74.55% (82/110) [.bi.d#is.] Bee Gees → [.pi.t$i.s .] 比吉斯 [. V.] 15.45% (17/110) [.h s.k .] husky → [.xa. .t$ i.] 哈士奇 Ø 8.18% (9/110) [.p æ. s.] Paris → [.pa.li.] 巴黎
[.t V.] 1.82% (2/110) [.m n.t o s.] Mentos → [.man.t wo.t u.] 曼陀珠23
d. [%]
TM Percentage Example (L2 → L1)
[$V.] 37.5% (3/8) [.k æ .m n.] Cashman → [.k ai. y.man.] 凱許曼 [t$ V.] 25% (2/8) [. .p a .] Apache → [.a.p a.t i.] 阿帕契
Ø 25% (2/8) [. (.l .] English → [.i .t$i.li.] 英吉利
[ V.] 12.5% (1/8) [.k æ .m .] cashmere → [.k . .mi. .] 喀什米爾
23 This adaptation is, to a certain degree, semantically involved as 珠 ([.t u.]) refers to pearls or beads in TM, which is easily associated with the classic round shape of this brand’s candy. .
‧
Once again, the adaptation patterns correspond to our prediction to a large degree in that, on the one hand, of the 161 fricative codas, 143 are retained through inserting a vowel to form an additional syllable, accounting for 88.82%. On the other hand, in the instances where the fricative stays, at least the manner feature of [+delayed release]
is copied in the processes of transformation from a fricative to another fricative or from a fricative to an affricate, with minor exceptions found in syllables with [v]-codas. Places of articulation are copied rather loosely, which can be understood by the diverse pattern found in (94c), even though [s] is a consonant shared by both languages.
‧
Comparison between the affricates of the two languages is given below again.
(95) Affricates of English and TM (also (79))
alveolar alveolo-palatal palatal retroflex TM, respectively, with obligatory vowel insertion in compliance with the native phonology, by virtue of the absence of the two sounds in TM. Both pairs are in the neighborhood of [t ]/[d ]. The alveolar pair [ts ]/[ts] is less likely for L2 [t ]/[d ] to map to, since they are farther away from [t ]/[d ] compared to the other two affricate pairs in TM. The patterns are shown in (96).
(96) Adaptation patterns of affricate codas
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
143
[.t V.] 20% (1/5) [.( .n d .] Greenwich → [.( .lin.wei.t .] 格林威治
Affricate codas in the corpus are rarely found. In general, once again consistent to our prediction, a majority of the affricate codas undergo vowel insertion to be retained, as found in six out of nine (66.67%) items. Regarding featural alternations in cases of retention, English [t ]/[d ] are shifted either a bit forward to [t ]/[t ] or a bit backward to [t ] in TM, both of which are neighboring sounds to English [t ]/[d ] in terms of places of articulation. They are never adapted into [ts ]/[ts], corresponding to our prediction.
4.5.4 Nasals
In 4.3.4, it has been shown that English and TM have the same inventory of nasals, given again in (97) below. The only difference, though, resides in the absence of [m]-codas in TM.
(97) Nasals of English and TM (also (81)) bilabial alveolar velar
m m n n
Being a well-formed coda in TM, an English [n] and [ ] in coda should presumably be perceived with little effort and thereby retained as the coda in the TM output. A rational prediction thus follows: English [n] is mapped to TM [n], English [ ] is mapped to TM [ ], and English [m] is randomly mapped to TM [n], [ ], or an independent syllable with [m] being the onset preceding an inserted vowel. The result is given below.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
144
(98) Adaptation patterns of nasal codas a. [m]
TM Percentage Example (L2 → L1)
[n] 64.47% (49/76) [.b k.h m.] Beckham → [.pei.k .xan.] 貝克漢 [ ] 23.68% (18/76) [.d .n m.] Denim → [.tan.ni).] 丹寧
[.mV.]11.84% (9/76) [. m.l t.] Omelet → [.ou.mu.li.] 歐姆力
b. [n]
TM Percentage Example (L2 → L1)
[n] 72.13% (251/348) [.fo .m .l n.] formalin → [.fu.ma.lin.] 福馬林 [ ] 26.72% (93/348) [.h . o . n.] Heroin → [.xai.lwo.i).] 海洛英 Ø 1.15% (4/348) [.mo .f n.] morphine → [.ma.fei.] 嗎啡
c. [ ]
TM Percentage Example (L2 → L1)
[ ] 70.83% (17/24) [.p .d ).] pudding → [.pu.ti).] 布丁 [n] 29.17% (7/24) [.s ).k .] sinker → [. n.k a.] 伸卡
As expected, 99.11% (444/448) of English nasal codas are retained as the coda or an additional syllable with an epenthetic vowel in TM. In terms of segmental mappings, likewise, the distribution fairly meets our prediction in that English [n] is chiefly mapped to TM [n] (72.13%, 251/348) and English [ ] mainly to TM [ ] (70.83%, 17/24), as shown in (98bc). Yet, the adaptation pattern of English [m] does not reveal a normal distribution among L1 [n] and [ ], and [.mV.], not even close: the percentages are 64.47% (49/76) for [n], 23.68% (18/76) for [ ], and 11.84% (9/76) for [.mV.].
4.5.5 Liquids
We have seen that English and TM share the same set of liquid inventory, as listed again in (99).
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
145
(99) Liquids of English and TM (also (83)) alveolar alveolar/post-alveolar
l l
As has been discussed earlier in Chapter 3, in addition to nasals, the retroflex [ ] is the only consonant that can serve as a coda in TM, whereas the context is exclusively limited to the syllable [. .], where no onset consonant is allowed.
Because of this, it should be expectable that except for the [ ] in the [. .] syllable, a large fraction of liquid codas need to be interpreted as an independent syllable by TM adapters through vowel insertion, either ahead of (to form an [. .] syllable) or behind the liquid, if the Preservation Principle (Paradis and LaCharité 1997, LaCharité and Paradis 2005) holds true. The patterns are shown below.
(100) Adaptation patterns of liquid codas a. [( ) ]
TM Percentage Example (L2 → L1)
[ ] 100% (4/4) [.me . .] Mayer →[.mei. .] 梅爾
b. [l]
TM Percentage Example (L2 → L1)
[. .] 53.51% (61/114) [.o .n l.] O’Neal →[.ou.ni. .] 歐尼爾 Ø 44.74% (51/114) [. l.p æs.] all pass →[.ou.p a.] 歐趴 [.lV.] 1.75% (2/114) [.k ol.(e t.] Colgate →[.k u.lu.t$je.] 高露潔
c. [ ]
TM Percentage Example (L2 → L1)
Ø 90.72% (176/194) [.lu.z .] loser →[.lu. .] 魯蛇 [. .] 8.76% (17/194) [.k æ%.m .]
Cashmere
→[.k a. !.mi. .]
喀什米爾
[.ja.] 0.52% (1/194) [. .k wa .] Aquair →[.a.k wei.ja.] 阿葵亞
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
146
First, in (100a), for the four instances of [. .] syllables, all of them (100%) are faithfully mapped to [. .] in TM with no exception. Second, somewhat against our presumption, the pattern of retention/deletion of [l]-codas in (100b) exhibits an approximately fifty-fifty distribution, though retention is still found in slightly more items (a combined 55.26% from [. .] plus [.lV.]). In (100c), where the [ ]’s are preceded by a vowel that is not a schwa, there is a significant tendency toward deletion—90.72% (176/194) of [ ]-codas are deleted, with the remaining 9.28%
(18/194) rescued by inserting a vowel or featural change, which fully runs counter to our prediction.
Taking a closer look at the corpus data, however, we may find the patterns of English liquid codas are largely correlated with the backness quality of the preceding vowel (an earlier discussion that deals with variable adaptation of English [ ]-codas in this respect appears in Lü (2013)). To see the precise patterns, we subcategorize the patterns of liquids in terms of the backness value of the preceding vowel. The result is given in (101) below, where, likewise, the target items under investigation are limited to polysyllabic ones to keep the patterns from the influence of the native prosody.
(101) Adaptation patterns of liquid codas by the backness of the preceding vowel a. (V[–back]) [l]
TM Percentage Example (L2 → L1)
[. .] 72.73% (40/55) [.t% l.si.] Chelsea → [.t$ e. .$i] 雀兒喜 Ø 27.27% (15/55) [.( l.b t.] Gilbert → [.t$i.pwo.t .] 吉伯特
b. (V[+back]) [l]
TM Percentage Example (L2 → L1)
Ø 61.02% (36/59) [.l .v .p ul.] Liverpool → [.li.u.p u.] 利物浦 [. .] 35.59% (21/59) [.t l.sto .]
Tolstoy
→ [.two. .s!.t ai.]
托爾斯泰
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
147
[.lV.] 3.39% (2/59) [.d m.b l.do .]
Dumbledore
→ [.t .pu.li.two.]
鄧不利多
c. (V[–back]) [ ]
TM Percentage Example (L2 → L1)
[. .] 66.67% (10/15) [.dæm.p .] Dampier → [.tan.p i. .] 丹皮爾 Ø 26.67% (4/15) [.( d.j .] Goodyear → [.ku.t .i.] 固特異 [.ja.] 6.67% (1/15) [. .k wa .] Aquair → [.a.k wei.ja.] 阿葵亞
d. (V[+back]) [ ]
TM Percentage Example (L2 → L1)
Ø 96.09% (172/179) [.k .t un.] Cartoon → [.k a.t o .] 卡通 [. .] 3.91% (7/179) [.ho .mon.] Hormone → [.x . .mo .] 賀爾蒙
The influence of the preceding vowel on the retention/deletion of the liquid coda is obvious here. In (101a), when the lateral [l] follows a [–back] vowel in the source word, 72.73% (40/55) of them are retained as [. .] in TM, while the remaining 27.27% (15/55) are simply ignored. When it is preceded by a [+back] vowel, on the contrary, it is deleted in majority, as found in 61.02% (36/59) of all, while in the rest 38.98% (23/59) it is preserved as [. .] or [.lV.].
Such a mirror image is also observed in [ ]-codas. In (31c), where the retroflex [ ] is preceded by a [–back] vowel, 73.34% (11/15, [. .] and [.ja.] combined) of them are retained, and deletion of it is found in 26.67% (4/15) of the cases. When it follows a [+back] vowel, the contrast is even more dramatic: deletion makes up 96.09%
(172/179) of all, with only 3.91% (7/179) of them retained.