Perceptual weakness of codas

國

立 政 治 大 學

‧

N a tio na

l C h engchi U ni ve rs it y

167

for minimally two syllables in a word.

5.2.3 Perceptual weakness of codas

As opposed to syllable onsets, syllable codas, being the syllable-final element, are generally viewed as a perceptually weak position within the syllable. From the perspective of articulation, as consonants in coda position lack release burst in many languages, reliable acoustic information of phonological contrasts is thus reduced to a large degree. Contrary to onsets, again, syllable codas allow a less marked inventory, and pervasively, they show less faithfulness to the underlying structure, as codas are more likely to undergo neutralization, usually to a default segment, or assimilation to the following onset consonant. What is noteworthy in this category, moreover, is that the prosodic preference for disyllabic words in TM will cast a significant effect on adaptation strategies, especially those of codas, for monosyllabic loanwords (more details of TM disyllabicity will be discussed in Section 5.13). To avoid such an influence and lay the focus on positional effects, in the survey of the retention/deletion of English codas we intentionally exclude monosyllabic data and consider polysyllabic source words only. The retention/deletion rates of coda consonants, obtained from our loanword data, seem to well reflect the perceptual weakness of syllable codas.

Unlike the uniformly high retention frequencies of all onset consonants, the retention/deletion patterns of English coda consonants demonstrate a more divergent distribution. First, for English stop codas, retention only accounts for 60.98%

(100/164), and up to 39.02% (64/164) of them undergo elision. As for fricative codas, the retention rate is close to that of second/third onsets, as up to 88.82% (143/161) of them are retained, yet still lower than the nearly perfect retention rate of simplex onsets. For affricates, the retention rate lowers again to 66.67% (6/9), though tokens

‧ 國

立 政 治 大 學

‧

N a tio na

l C h engchi U ni ve rs it y

168

of this type are rather few. Moreover, being the well-formed codas in TM, nasals are retained in 98.67% (444/450) of the cases through faithful retention, featural change, or vowel insertion. Of the two English liquid codas, the retention rate of the lateral [l]

is the low 55.26% (63/114), and even worse, that of the retroflex [ ] reaches the lowest 11.06% (22/199). In summary, English simplex codas are retained in 70.92%

(778/1097) of the cases, which is the lowest retention rate among the three positions within a syllable.

Perception is not completely universal, however, as it relies greatly on the L1 sound inventories and syllabification patterns. As shown in Figure 1 reveals, perceptual processing involves both acoustic perception and L1 phonology; that is, perception is phonological, too. This can be illustrated by comparing the loanword adaptations of two languages with different syllable structures but from the same source language. For example, in Cantonese, unaspirated stops [p, t, k] can appear syllable-finally, and like English, they are unreleased in this position. In the literature where a narrow phonetic transcription is adopted, the unreleased closure of the stop coda is symbolized by an upper right corner added to the stop, as in [.s/p2.] (濕, ‘wet’), [.s/t2.] (膝, ‘knee’), and [.s/k2.] (塞, ‘to plug up’) (Bauer and Benedict 1997).

Consider (113), drawn from Bauer and Benedict (1997), in which we list all Cantonese adaptations of English words with simplex stop codas appearing in their book.

(113) Cantonese adaptations of English loanwords with simplex stop codas (Bauer and Benedict 1997)

a. Faithful retention

Stop Item (L2 → L1)

[p/b] [.k æp.] cap → [.k *p.] 帽

[.k æp.t n.] captain → [.k 7p.t @n.] 頓

[.t%ip.] cheap → [.t%i7p.]

‧

‧ 國

立 政 治 大 學

‧

N a tio na

l C h engchi U ni ve rs it y

170

[.w t.] watt → [.w 7k.]

[k/(] [.k o .dæk.] Kodak → [.k 7.ta7t.] 柯達

Being permissible coda consonants in both languages, the unreleased stop codas in English are mostly mapped to the identical stops in their Cantonese adapted forms, as shown in (113a), with a few minor exceptions, as shown in (113b), where the manner feature [–continuant] remains but the place changes. On the one hand, deletion of the stop coda is rarely found, since it is well perceptible to Cantonese speakers. On the other hand, an epenthetic vowel for the retention of a stop coda is rendered redundant in this scenario, as stop codas are structurally well-formed in Cantonese. This strengthens the stance that L1 phonology joins the loanword adaptation process early in the level of perceptual processing.

However, in Korean, where an unreleased stop is also allowed, a vowel is variably inserted after the postvocalic stop. The variability is shown in (114).

(114) Korean adaptations of English loanwords with simplex stop codas (Kang 2003)

No insertion Insertion

a. [.k w k.] quick → [.k wik.] [.wik.] week → [.wi.k .]

b. [.k k.] kick → [.k ik.] [.m (.] MIG → [.mi.k .]

c. [.k k.] kick → [.k ik.] [.h t.] hit → [.hi.t i.]

[.t p.] tip → [.t ip.]

Unlike Cantonese, where an L1-phonologically permissible L2 stop is more systematically retained by remaining in the coda position, the examples in (114) reveals more variability. It either remains in coda or serves as the onset of an epenthetic vowel. In this respect, Kang (2003) postulates that the seemingly variable patterns of vowel insertion is in fact largely bound by a few phonological factors. He holds that vowel insertion tends to apply when the nuclear vowel is long or tense,

‧ 國

立 政 治 大 學

‧

N a tio na

l C h engchi U ni ve rs it y

171

rather than short or lax (114a), the stop is voiced in English rather than voiceless (114b), and the stop is coronal instead of non-coronal (114c). Despite the variation, in any situation, the coda stop is well perceived in the perceptual processing of the Korean adapter. Surprisingly enough, in his experiment among the 447 English loanwords with a postvocalic stop listed in the appendix, none of them undergoes the process of coda-deletion (225 consistently adapted with final vowel insertion, 195 consistently adapted without final vowel insertion, 27 showing variation of both).

Again, like Cantonese, this reinforces our assumption that L1 phonology plays a crucial role in perception. The comparison in (115) summarizes the adaptation strategies of stop codas in the three languages mentioned.

(115) Stop coda adaptation of TM, Cantonese, and Korean

TM Cantonese Korean

faithful retention * √ √

unfaithful retention * √ √

vowel insertion √ * √

deletion √ * *

In general, TM allows no stops as the coda, which follows that faithful retention is unlikely to occur. Variation lies in whether the stop is retained via vowel insertion (e.g. [.slo .væk.] Slovak → [.s!.lwo.fa.k .] 斯洛伐克) or simply ignored through stop deletion (e.g. [.slæ.v k.] Slavic → [.s!.la.fu.] 斯拉夫). In Cantonese and Korean, where stops are licit codas, faithful retention is found in both languages (e.g.

[.p æn.k e k.] pancake → [.pa7n.k ek.] 班戟 in Cantonese and [.k k.] kick → [.k ik.] in Korean), while only in the latter vowel insertion is also discovered (e.g.

[.wik.] week → [.wi.k .]). Deletion is not (or at most rarely) found in the two languages owing to the well-formedness of stop codas in their phonotactics.

‧ 國

立 政 治 大 學

‧

N a tio na

l C h engchi U ni ve rs it y

172