3. A Corpus-Based Analysis
3.6 Summary
國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
51
When English stress is located in the first or the last syllable in an English word, the left-edge pitch of its corresponding tone is not lower than that of the nearest or neighboring organic syllable. When the stress is located in the middle syllable, the left-edge pitch of its corresponding tone is not lower than that of the two organic syllables beside it.
3.6 Summary
In this chapter, we have examined the tonal adaptation of English-to-Mandarin loanwords from four perspectives. The four perspectives are respectively the
distribution of the four Mandarin tones in relation to English stressed and unstressed syllables, the interaction between the voicing feature and tone, the tone loans with different frequencies of use, and the distinction between the tones corresponding to the English stressed and unstressed syllables. First, the observation of the tonal distribution reveals that the H-tone is most preferred by English stress because it shows the most faithful mapping from the pitch accent H of English stressed syllables.
Besides, the L-tone is the least preferred since it shows the least faithful mapping from the pitch accent H. Second, as for the voicing feature of the onsets in the English stressed syllable, the effect of voice enhancement is observed to interact with the preference for the H-tone. Moreover, voice enhancement occurs on the left edge of the
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
52
corresponding tones. Namely, an English stressed syllable with a voiced onset prefers a tone that is initially low-registered, i.e. the MH-tone, in the loanwords, while an English stressed syllable with a voiceless onset favors a tone that is initially
high-registered, i.e. the HL-tone. Third, the inspection of the tone loans with different frequencies of use suggests a tonal preference relation of the English stressed
syllables. That is, the tone loans with the H-tone in the corresponding stressed syllables are more preferred and used more often than those whose corresponding stressed syllables are adapted with the MH-tone or the HL-tone, which in turn are more preferred and used more often than those whose corresponding stressed
syllables are adapted with the L-tone. Finally, the perceptual distinction between the stressed and unstressed syllables in the borrowed English words is retained on the left edge of the corresponding tones in their loanwords. The left-edge pitch of the tone corresponding to the English stressed syllable is usually not lower than that of the tone corresponding to the English unstressed syllable in a tone loan.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
53
CHAPTER 4
AN OPTIMALITY THEORY ANALYSIS OF STRESS-TO-TONE INTERACTIONS IN MANDARRIN LOANWORDS
4.1 Introduction
From the perspective of Optimality Theory, a stress-to-tone relationship in loanwords is viewed as an input-to-output correspondence (Silverman 1992, Leben, 1996, Kenstowicz 2004, Wu C. 2006, Wu C. 2007, Wu H. 2006).
We have seen several patterns of the tonal adaptation in Mandarin loanwords adapted from English earlier. In this chapter, I account for the patterns observed in Chapter 3 by establishing a grammar of stress-to-tone Mandarin loanwords through the hierarchical constraints. The analysis in this chapter is divided into three parts.
The first focuses on the interaction between the Mandarin tones and the onsets of the English stressed syllables. The second focuses on the tonal preference relation of the English stressed syllables. The third centers on the tonal distinction between the corresponding stressed and unstressed syllables in Mandarin loanwords.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
54
4.2. Voice Enhancement
Stress and tone are respectively two important features of stress languages, such as English, and tone languages, such as Mandarin Chinese. The observations of the relationship between English stress and its corresponding tone have revealed that the pitch accent H in English stressed syllables is retained when they are adapted into Mandarin loanwords. Since the L-tone accounts for only a minority of occurrences of the corresponding stressed syllables in the present corpus, we refer to them as
exceptions, and will not discuss the cases of this kind further.
The investigations into the tonal adaptation of the corresponding stressed syllables are seperated into two conditions. The first condition is that the H-tone, the MH-tone, and the HL-tone are all available for the tonal adaptation. The second condition is that only the MH-tone and the HL-tone are available. For readability, the correspondents of English stresse are presented in bold face, and the epenthetic syllables in the loanwords are underlined.
4.2.1 Voiced Onsets
In this section, we analyze the tones that correspond to the English stressed syllables with voiced onsets. The observations in Chapter 3 suggest that the pitch peak
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
55
of English stressed syllables is usually preserved by the H-tone, the MH-tone, or the HL-tone. The H-tone is [+H] on left and right sides. The MH-tone and the HL-tone are [+H] on the right and left sides respectively. The L-tone is least likely to be
associated with the corresponding stressed syllables because neither of its two sides is [+H]. From the perspective of Optimality Theory, the realization of the pitch accent H of the English stressed syllables in loanwords is confined by the two anchoring
constraints in (1) and (2).
(1) ANCHOR-L(H, σÂ): The H of the source is anchored to the left edge of the corresponding syllable in the loan.
(2) ANCHOR-R(H, σÂ): The H of the source is anchored to the right edge of the corresponding syllable in the loan.
With these two constraints, we can avoid the selection of the L-tone.
The examination begins with the first condition. It is found that not all of the three tones are selected. The H-tone and the MH-tone are chosen alternatively. From a statistical point of view, the H-tone accounts for 60.34% of the data while the
MH-tone for 39.66%. According to voice enhancement (Hsieh and Kenstowicz 2006,
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
56
Kim 2006), a voiced onset has a lowering effect on the tone of the following vowel.
Since the left-edge tone of the MH-tone is low-registered, as discussed in 3.3.1, selection of the MH-tone in loanwords becomes legalized by the constraint indicated in (3).
(3) *[+voiced]/[+H]-: Assign one violation mark to an initially high-registered tone if the onset of the stressed source syllable is [+voiced].
Under the framework of Optimality Theory, these two types of tonal adaptations, namely, the H-tone and the MH-tone, may be accounted for via re-ranking
*[+voiced]/[+H]- with ANCHOR-L(H, σÂ) and ANCHOR-R(H, σÂ), as is illustrated below.
First, we take a look at the loanwords in which the English stressed syllables are adapted as the H-tone. Take valve /vQÂlv/ for example. In the input, the onset is [+voiced]. It has two tone loans, namely, 伐 /fa/ (H) and 閥 /fa/ (MH). We discuss 伐 /fa/ (H) first, and will discuss 閥 /fa/ (MH) later. The analysis of the tonal adaptation of 伐 /fa/ (H) is given in (4), where the constraints ANCHOR-L(H, σÂ), ANCHOR-R(H, σÂ), and *[+voiced]/[+H]- are introduced to evaluate the four Mandarin tones. To obtain this output, ANCHOR-L(H, σÂ) and ANCHOR-R(H, σÂ) must outrank
*[+voiced]/[+H]-.
‧
In tableau (4), candidate (b) has serious violation of ANCHOR-L(H, σÂ) because the pitch accent H in the English stressed syllable is not anchored to the left edge of the corresponding tone in the loanword. As a result, it is ruled out. Candidate (d) has serious violation of ANCHOR-R(H, σÂ) because the pitch accent H is not anchored to the right edge of the corresponding tone. In consequence, it is removed. Candidate (c) is worse than candidate (b) and (d) because the pitch accent H is neither anchored to the left edge nor anchored to the right edge. Thus, it is ruled out as well. Candidate (c) violates *[+voiced]/[+H]- because it is an initially high-registered tone. It is correctly selected as the optimal candidate, as *[+voiced]/[+H]- ranks lower than the other two constraints, ANCHOR-L(H, σÂ) and ANCHOR-R(H, σÂ). Therefore, valve /vQÂlv/ is alternatively adapted as 伐/fa/(H).
Moreover, the selection of the H-tone in correspondence to the stressed syllables of the disyllabic and trisyllabic English words can be predicted by the same ranking,
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
58
such as 披頭四 /pHi.tHou.sˆ/ (H.MH.HL) adapted from Beatles /biÂÂtl6z/, 威化 (餅乾) /wei.xwa/ (H.HL) adapted from waffle /wAÂfl6/, and 巴波亞 /pa.pwo.ja/ (H.H.HL) adapted from Balboa /bAlbo´/. For more examples, please see (9i) and (11i) in Chapter 3.
Now we look at the loanwords where the MH-tone is selected for the
corresponding stressed syllables. Take the same input valve /vQÂlv/ for instance. The other output of it is 閥/fa/ (MH). In the input, the onset is [+voiced]. Its corresponding tone in the output is an MH-tone. To obtain the correct output, re-ranking of
*[+voiced]/[+H]- and the two anchoring constraints is necessary. That is,
*[+voiced]/[+H]- must dominate the two anchoring constraints, ANCHOR-L(H, σÂ) and ANCHOR-R(H, σÂ). See the following tableau.
(5) GYRB-M-007
/vQÂlv/ *[+voiced]/[+H]- ANCHOR-L(H, σÂ) ANCHOR-R(H, σÂ) a H *!
" b MH *
c. L * *!
d. HL *! *
In tableau (5), *[+voiced]/[+H]- is ranked higher than ANCHOR-L(H, σÂ) and
‧
ANCHOR-R(H, σÂ). Candidates (a) and (d) are initially high-registered tones, so they are ruled out, owing to the fatal violation to *[+voiced]/[+H]-. Candidates (b) and (c) are competitive candidates since they both obey *[+voiced]/[+H]- and violate
ANCHOR-L(H, σÂ). However, candidate (c) has one more violation than candidate (b), as it violates both ANCHOR-L(H, σÂ) and ANCHOR-R(H, σÂ). Therefore, it is ruled out.
Despite a violation of ANCHOR-L(H, σÂ), candidate (b) is correctly selected as the best candidate. As a result, valve /vQÂlv/ is alternatively adapted as 閥/fa/ (MH).
In addition, this ranking also predicts the selection of the MH-tone in
correspondence to the stressed syllables in disyllabic and trisyllabic English words, such as 迪士尼/ti.߈.ni/ (MH.HL.MH) adapted from Disney /dIÂznI/,and 維他命 /wei.ta.miN/ (MH.H.HL) adapted from vitamin /vaÂIt´mIn6/. For more relevant examples, please see (9ii) and (11ii).
From tableaux (4) and (5), we can see that the ranking relation between
*[+voiced]/[+H]- and the two anchoring constraints is crucial in selection of the two alternative tonal tones.
Furthermore, each of the syllables in the English words is adapted with one of the four Mandarin Chinese tones. Theoretically speaking, this phenomenon occurs under the requirement of assigning a tone to each of the corresponding syllables in the loanwords. According to Yip (1989), a tone-bearing unit in Mandarin Chinese is a
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
60
syllable. Therefore, we propose an undominated constraint as (6).
(6) *TONELESS: Assign one violation mark to syllables without a tone.
I use the symbol ‘ ’ to refer to a toneless syllable. Consider tableau (7), where ANCHOR-L(H, σÂ) and ANCHOR-R(H, σÂ) outrank *[+voiced]/[+H]-, and tableau (8), where *[+voiced]/[+H]- dominates ANCHOR-L(H, σÂ) and ANCHOR-R(H, σÂ).
(7) GYRB-M-007
/vQÂlv/ *TONELESS ANCHOR-L (H, σÂ)
ANCHOR-R
(H, σÂ) *[+voiced]/[+H]-
" a. H *
b. MH *!
c. L *! *
d. HL *! *
e. *!
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
61
(8) GYRB-M-007
/vQÂlv/ *TONELESS *[+voiced]/[+H]- ANCHOR-L (H, σÂ)
ANCHOR-R (H, σÂ)
a. H *!
" b. MH *
c. L * *!
d. HL *! *
e. *!
Candidate (e) in tableaux (7) and (8) is ill-formed because it violates *TONELESS, which is ranked high. As a result, it is removed. The competition among the four candidates from (a) to (d) in tableaux (7) and (8) can be seen in tableaux (4) and (5) respectively.
Alternative tonal outputs, namely, the H-tone and the MH-tone, can be obtained via constraint re-ranking, which is permitted by the cophonology model, as given in (9).
‧
(9) Grammar lattice of tonal adaptation in stress-to-tone Mandarin loanwords (preliminary)
Master Ranking
*TONELESS >>
Cophonology A Cophonology B
TONELESS >> TONELESS >> *[+voiced]/[+H]- >>
On the one hand, cophonology A selects the H-tone. On the other hand, the MH-tone is chosen under the evaluation of cophonology B. Cophonology A predicts the tonal adaptation that presents the most faithful stress-to-tone correspondence.
Cophonology B selects the tonal adaptation which results from the interaction between Mandarin tones and English onsets.
As Table 4.1 shows below, 35 out of the 58 English stressed syllables with voiced onsets are adapted with the H-tone, which can be chosen as the optimal output under cophonology A. 23 of them are associated with the MH-tone, which can be chosen under cophonology B. In sum, all of the 58 syllables can be predicted by cophonologies A and B, with a 100.00% accuracy rate.
ANCHOR-L(H, σÂ), ANCHOR-R(H, σÂ), *[+voiced]/[+H]-
‧
Table 4.1: Tones in Correspondence to English Stressed Syllables with [+voiced]
Onsets (The First Condition) H, MH, HL
Percentage 60.34% 39.66% 0%
Now, let us turn to the analysis of the second condition. The examination suggests that the MH-tone is more likely to be assigned to the corrrsponding stressed syllables than the HL-tone when the H-tone is not available. We adopt three criteria to define the non-available tone. First of all, orthography in Mandarin Chinese plays a role in tonal adaptation. That is, a systematic gap or an accidental gap may block the occurrence of the H-tone syllables, as governed by (10).
(10) *GAP: Assign one violation mark to a syllable that is a systematic gap or an accidental gap.
*GAP is an undominated constraint. It is ranked as high as *TONELESS. Now
‧
consider the input lime /laÂIm/ and its output 萊姆 (MH.L). There is no Chinese character associated with the syllable /laI/ and the H-tone. Thus, the H-tone /laI/ is a gap. Both cophonologies A and B are able to choose the MH-tone when the H-tone syllable is not available. For example, cophonology A is shown in tableau (11).
(11) GYRB-M-054
In tableau (11), candidate (a) is a gap, violating *GAP, which is ranked high. As a result, it is ruled out. Consequently, candidate (b) is the optimal output. lime /laÂIm/ is thus adapted as 萊姆 (MH.L), in which the the second syllable is an epenthetic syllable.
Second, the word frequency also plays a role in determining the selection of Chinese characters in the loanwords. In addition, most of the loanwords are adapted from English nouns. The selection of Chinese characters is thus influenced by the
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
65
parts of speech of Chinese characters as well. The observations in Chapter 3 show that a Chinese character with a low word frequency is avoided. In addition, a verb, an interjection, an onomatopoeic word, or a noun referring to kinship is also prevented from adaptation in loanwords. Therefore, we propose the following constraint.
(12) VALIDWORD (VLWD):
Assign one violation mark to a syllable that is not a valid word.
To be more specific, a word with a low word frequency, a verb, an interjection, an onomatopoeic word, or a noun referring to kinship is not a valid word for
loanwords. Like *TONELESS and *GAP, VALIDWORD is also an undominated constraint. Consider the input dyne /daÂIn/ and its output 達因 (MH.H) in tableau (13).
‧
Regarding candidate (a), only the word 搭/ta/ (H) can be found under the prohibition against Chinese characters with the low word frequency. However, 搭/ta/ (H) is a
verb, which is not a valid word in the loanwords. Consequently, candidate (a)is ruled out due to the violation of *VLWD. As for candidate (b), 達/ta/(MH) is found.
Compared to 搭/ta/ H), 達/ta/(MH) is more than a verb. In addition to a verb, meaning ‘to reach’, it can also be an adjective, meaning ‘high ranking’, or an adverb, meaning ‘all over’. As a result, candidate (b) is the optimal output.
Third, Chinese characters with [+negative] meanings only are rarely found in loanwords. In other words, the majority of the Chinese loanwords adapt Chinese words with positive or neutral meanings. Thus, we propose the *[+negative]
constraint in (14).
‧
(14) *[+negative]: Assign one violation mark to a syllable with [+negative] meanings only.
In the tableau above, candidate (a) violates *[+negative] because only Chinese characters with [+negative] meanings are found, such as 淤 /y /(H), meaning ‘to become silted up’, 瘀 /y/ (H), meaning ‘bruise’, and 迂 /y /(H), meaning
‘obstinately observing old rules or ideas’.Thus, cacdidate (a) is removed. Candidate (b) is the optimal candidate.
As Table 4.2 shows, 28 out of the 36 English stressed syllables with voiced onsets are adapted with the MH-tone. Only eight of them are associated with the
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
68
HL-tone. The accuracy rate is 77.78%.
Table 4.2: Tones in Correspondence to English Stressed Syllables with [+voiced]
Onsets (The Second Condition) MH, HL
MH HL
Monosyllabic EW 5 3
Disyllabic EW 21 3
Trisyllabic EW 2 2
Total 28 8
Percentage 77.78% 22.22%
4.2.2 Voiceless Onsets
In this section, we analyze the tones that correspond to the English stressed syllables with voiceless onsets. The observations in Chapter 3 indicate that the H-tone is the most preferred by the English stressed syllables with voiceless onsets in the first condition, and that the HL-tone is the more preferred than the MH-tone in the second consdition.In general, the MH-tone is avoided neither in the first or the second condition. The H-tone and the HL-tone are initially high-registered, whereas the MH-tone is initially low-registered. Therefore, I propose the *[-voiced]/[-H]-
‧
(16) *[-voiced]/[-H]-: Assign one violation mark to an initially low-registered tone if the onset of the stressed source syllable is [-voiced].
This *[+voiced]/[+H]- constraint is ranked as high as ANCHOR-L(H, σÂ) and ANCHOR-R(H, σÂ). Consider this example, the input bikini /bIkiÂnI/ and its output 比基 尼 /pi.t˛i.ni/ (L.H.MH). As for the input, the onset of the stressed syllable is
[-voiced]. Its corresponding tone in the output is the H-tone. See the following tableau.
In tableau (17), the irrelevant constraints are omitted. Candidate (e) violates the
‧
undominated constraint *TONELESS because it has no tone. Due to this fatal violation, it is ruled out. Candidates (b), (c), and (d) all incur at least one violation of the lower ranked constraints, but candidate (a) does not. Consequently, candidate (a) is the optimal candidate. For more examples, please see (13) in Chapter 3.
As Table 4.3 shows, 68 of the 92 English stressed syllables with voiceless onsets are adapted with the H-tone. Only nine of them are associated with the MH-tone, and 15 of them are adapted with the HL-tone. The accuracy rate is 73.91%.
Table 4.3: Tones in Correspondence to English Stressed Syllables with [-voiced]
Onsets (The First Condition) H, MH, HL
Now we turn to the examplesof the second condition, in which the H-tone is not available. Consider the input Hepburn (cut) /hEÂpbŒ’n/ and its output 赫本(頭 ) /xF.p´n/ (HL.L). In the input, the onset of the stressed syllable is [-voiced]. Its
‧
corresponding tone in the output is the HL-tone, as illustrated in tableau (18).
(18) GYRB-D-187 can be found under the prohibition against Chinese characters with the low word frequency. However, a verb is not a valid word in the loanwords. In consequence, candidate (a)is ruled out on account of the violation of *VLWD. Candidates (b) and (c) show more violations of the lower ranked constraints than candidate (d). Therefore, candidate (d) is the best output.
As Table 4.4 shows, eight of the 11 English stressed syllables with voiceless onsets are adapted with the HL-tone. Three of them are associated with the MH-tone.
The accuracy rate is 72.73%.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
72
Table 4.4: Tones in Correspondence to English Stressed Syllables with [-voiced]
Onsets (The Second Condition) MH, HL
MH HL
Monosyllabic EW 0 3
Disyllabic EW 2 5
Trisyllabic EW 1 0
Total 3 8
Percentage 27.27% 72.73%
4.2.3 Zero Onset
The examinations in Chapter 3 suggest that the H-tone is the most preferred tone for the onsetless English stressed syllables. That is, without any influence from English onsets, an H-tone is slected for a corresponding stressed syllable since it shows the most faithful mapping from the pitch accent H of the English stressed syllables. The analysis begins with the first condition. Consider the input Aesop(‘s
fables) /iÂsAp/ and its output 伊索 (寓言) /i.swo/ (H.L). As for the input, the stressed syllable is onsetless. Its corresponding tone in the output is the H-tone, as shown in tableau (19), where irrelevant constraints are omitted.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
73
(19) GYRB-D-281
/iÂsAp/ *TONELESS ANCHOR-L(H, σÂ) ANCHOR-R(H, σÂ)
" a. H
b. MH *!
c. L *! *
d. HL *!
e. *!
Tableau (19) reveals that the H-tone is the optimal candidate when the
corresponding stressed syllables are onsetless. For more examples, please see (15) in Chapter 3. As shown in Table 4.5, 11 of the 13 onsetless English stressed syllables are adapted with the H-tone. Only two of them are assigned the HL-tone. In other words, the corpus shows an 84.62% accuracy rate.
‧
Table 4.5: Tones in Correspondence to Onsetless English Stressed Syllables
(The First Condition)
Percentage 84.62% 0% 15.38%
The cases under the second condition are very few. That is, only one case is found, as shown in Table 4.6. Thus, it is not discussed here.
Table 4.6: Tones in Correspondence to Onsetless English Stressed Syllables
(The Second Condition)
‧
Thus, we can conclude the constraint rankings as in (20).
(20) Grammar lattice of tonal adaptation in stress-to-tone Mandarin loanwords (revised)
Master Ranking
*TONELESS, *GAP, VLWD, *[+Negative] >>
Cophonology A Cophonology B
TONELESS, *GAP, VLWD,
4.3 Frequencies of Use and the Tonal Preference Relation
The examinations of the present corpus in Chapter 3 indicate a tonal preference relation of the English stressed syllables. It is suggested that the H-tone is more preferred than the MH-tone and the HL-tone by English stressed syllables and that the MH-tone and the HL-tone are favored equally. Moreover, the L-tone is the least preferred tone. For example, the input hamlet /hQÂmlIt/ is adapted as 哈姆雷特
ANCHOR-L(H, σÂ), ANCHOR-R(H, σÂ), *[-voiced]/[-H]-, *[+voiced]/[+H]-
‧
/xa.mu.lei.tHF/ (H.L.MH.HL) or 漢姆雷特 /xan.mu.lai.tHF/ (HL.L.MH.HL). These two tone loans only differ in the tone that correspond to the English stressed syllable.
/xa.mu.lei.tHF/ (H.L.MH.HL) or 漢姆雷特 /xan.mu.lai.tHF/ (HL.L.MH.HL). These two tone loans only differ in the tone that correspond to the English stressed syllable.