國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
4
Correspondence. It also reviews the two previous studies on the stress-to-tone relation of English-to-Mandarin loanwords. Chapter 3 offers the findings of the present corpus.
Chapter 4 provides an analysis from the perspective of Optimality Theory. Chapter 5 concludes this thesis and points out several further issues.
1.2 Sources of the Corpus
This thesis establishes a corpus of 833 tone loans of monosyllabic, disyllabic, and trisyllabic English words. The data are mainly from Guo Yu Ri Bao Wai Lai Yu Ci
Dian (1981). This dictionary collects the loanwords that are transliterated only
according to sounds. In addition, it is the more frequently used loanwords in
Mandarin Chinese that are collected. The borrowed language of each loanword is also listed in the dictionary. Moreover, it also distinguishes the immediate borrowed language from the original language(s). For example, the first word of 沙皇/ßa.xwaN/
(H.MH), the title of the former emperor of Russia, is immediately borrowed from czar in English. Czar in English is borrowed from czar in Polish. Czar in Polish is
borrowed from Tsar’ in Russian. Tsar’ in Russian is borrowed from Kaisar in Gothic.
Kaisar in Gothic is borrowed from Caesar in Latin. The present corpus incorporates loanwords in accordance with the listed borrowed language, namely English. In other words, the present corpus collects Mandarin loanwords adapted only from English.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
5
CHAPTER 2
LITERATURE REVIEW
2.1 Related Optimality Theory
2.1.1 Optimality Theory
Optimality Theory is proposed by Prince and Smolensky (1993/2004). In the
framework of Optimality Theory, constraints are universal, and grammars are
rankings of a set of constraints. Given an input, Generator (GEN) generates an infinite number of candidates. Evaluator (EVAL) selects the candidate that best satisfies the set of constraints, and this candidate thus becomes the output.
The constraints are mainly classified into two types, namely, faithfulness
constraints and markedness constraints. Faithfulness constraints impose requirements on identity between the input and the output; markedness constraints emphasize the well-formedness of the output. These two types of constraints are in conflict with each other. Constraint ranking differs from language to language. Therefore, different rankings of constraints produce language-particular outputs.
The following are three basic types of faithfulness constraints.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
6
(1) Basic faithfulness constraints (McCarthy and Prince 1995)
(i) MAX-IO: Every input element must have its correspondent one in the output.
(ii) DEP-IO: Every output element must have its correspondent one in the input.
(iii) IDENT-IO: Every element in the input and in the output must be correspondent.
The following are some markedness constraints relevant to tone.
(2) Markedness constraints relevant to tonal issues (Yip 2002) (i) *H: No high tones.
(ii) *CONTOUR: No contour tones.
(iii) ALIGH-R (H, STEM): Align every H tone with the right edge of some stem.
Furthermore, Yip (2002) postulates a tonal markedness hierarchy to capture cross-linguistic tonal distributions, as in (3).
(3) Tonal markedness hierarchy *RISE >> *FALL >> *H >> *L
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
7
In other words, the rising tone is the most marked tone, whereas the low tone is the most unmarked.
2.1.2 Relational Correspondence
Influenced by Correspondence Theory (McCarthy and Prince 1995), Hsieh (2007) proposes Relational Correspondence. Relational Correspondence is different from
‘element-based Correspondence’ in that it centers on the contour preservation of input-output and output-output mappings across two or more segments. This theory includes four types of relations governed by a family of faithfulness constraints, as shown in (4).
(4) RELATIONAL CORRESPONDENCE (RELCORR)
Let t1 be a tone value contained within Rime R, and S1 is the temporal span associated with t1. Let t2 be another tone value contained in Rime R, and S2 is the temporal span associated with t2. Let t1’ be the correspondent of t1 in Rime R’, and let t2’ be the correspondent of t2 in Rime R’.
S1 proceeds S2.
Let t1 = x, t2 = y, t1’ = a, t2’ = b.
(i) The ‘greater than’ relation: If x > y, then a > b (abbr. RELCORR(x > y))
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
8
(ii) The ‘less than’ relation: If x < y, then a < b (abbr. RELCORR(x < y)) (iii) The ‘equal to’ relation: If x = y, then a = b (abbr. RELCORR(x = y)) (iv) The ‘non-equal to’ relation: If x ≠ y, then a ≠ b (abbr. RELCORR(x ≠ y))
In this thesis, it is observed that the relative pitch height of the stressed and unstressed syllables of an English word is retained in the Mandarin tone loan. The concept of Relational Correspondence is applied to stress-to-tone mappings in 4.4.
2.1.3 Anchoring
Distinct from derivational phonology, which views a reduplicant as the result of the left-to-right or the right-to-left copying of the segments in its base, Optimality Theory posits that the segments of a reduplicant stand in correspondence with those in the base. Therefore, McCarthy and Prince (1995) propose anchoring constraints to reinterpret the concept of the direction of copying in derivational phonology.
Anchoring constraints refer to correspondence between the targeted position in the base and the segment in the same position of the reduplicant, as given in (5).
(5) LEFT-ANCHOR (Base, Reduplicant): The left edge of the reduplicant corresponds to the left edge of the base (McCarthy and Prince 1995).
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
9
According to Beckman (1999), anchoring constraints mark acoustically or perceptually prominent positions, namely, privileged positions. Moreover, it has been reported cross-linguistically that the left edge is the privileged position (Beckman 1999, Nelson 2003, Lunden 2004)
2.2 Previous Studies
It is found cross-linguistically that adaptation strategies in loanwords seek to maximize the perceptual and acoustic similarities between the sources and the adapted forms. At the same time, the surface forms have to conform to the phonological regulations of the borrowing language. In literature, there are two arguments for the tonal adaptation of Mandarin loanwords adapted from English, namely, Wu C. (2006) and Wu H. (2006). In this section, we review these two studies, and look into some problems arising therein.
2.2.1 Wu C.’s (2006) Analysis
According to Wu C. (2006), the corresponding stressed syllables show a tendency to be assigned high tones, that is, the high level tone and the high falling tone. On the contrary, they tend not to be assigned the low tone. High tones have
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
10
similar characteristics to stress in terms of pitch and length. From her perspective, high tones are the prominent units for tone languages, and stress is the prominent unit for stress languages. Compared with the rising tone and the low tone, both the high level tone and the high falling tone have a high pitch in the beginning, resulting in a more prominent effect in perception. As for the rising tone and the low tone, the former is perceptually more prominent because it ends with a higher pitch.
With regard to the unstressed syllables, Wu C. (2006) considers the epenthetic syllables and the unstressed syllables the same category. Their tonal assignments do not exhibit any preference in general. In addition, an epenthetic syllable with an /s/
onset tends to be associated with a high level tone or a high falling tone, while one with a /t/ or a /d/ onset prefers a high falling tone.
From the perspective of Optimality Theory, Wu C. (2006) proposes the constraints as follows.
(6) FAITH-PROMINENCE: The prominent unit of an input must correspond to the
prominent unit of its output.
(7) σT: Every syllable in the outputs must have a tone.
(8) * PROMINENCE/H：The prominent unit in an output does not associate with a high level tone.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
11
(9) * PROMINENCE/HL：The prominent unit in an output does not associate with a high falling tone.
(10) * PROMINENCE/MH：The prominent unit in anoutput does not associate with a rising tone.
(11) * PROMINENCE/L：The prominent unit in an output does not associate with a low tone.
The constraint ranking is shown in (12).
(12) FAITH-PROMINENCE, σT >> * PROMINENCE/L, * PROMINENCE/MH >>
* PROMINENCE/HL, * PROMINENCE/H
With these ranked constraints, the tone loan of guitar is evaluated in (13). The symbol ‘≅ ’ indicates English stress. Thus, ‘σ1σ≅2’ means that the second syllable is stressed.
‧
2.2.2 Wu H.’s (2006) Analysis
Wu H. (2006) investigates the monosyllabic and disyllabic Mandarin loanwords adapted from English. The major observations are as follows. First, the falling tone tends to be chosen for the monosyllabic Mandarin loanwords. Second, the high level tone and the rising tone are most preferred by the stressed syllables of disyllabice English words. The high falling tone, however, is the least preferred. Third, the low tone is usually associated with the epenthetic syllables.
Wu H. (2006: 238) argues that the strategies of tonal adaptation in Mandarin loanwords are influenced by the onsets and the biased native lexicon. The stressed syllable of an English word is the site where the H pitch is located. Therefore, in terms of tonal adaptation of disyllabic English words, the choice of the high level tone for the corresponding stressed syllables is straightforward. In contrast, the choice of
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
13
the rising tone needs some explanation. Wu H. (2006) observes that stressed syllables whose onsets are sonorants or voiced stops tend to be associated with the rising tone.
Therefore, Wu H. (2006) suggests that the choice of the rising tone results from the influence of depressor consonants, namely, sonorants and voiced obstruents. In general, the depressor consonants have a lowering effect on a following tone. This phenomenon is called ‘tonal depressing effect’, and is commonly found in languages.
2.2.3 The Differences in Previous Studies
In terms of the tones that correspond to the English stressed syllables, Wu C.
(2006) finds that the high level tone and the high falling tone are preferred. However, Wu H. (2006) observed that the high level tone and the rising tone are preferred, and that the high falling tone is the most disprefferred. Regarding the unstressed syllables, Wu C. (2006) includes epenthetic syllables, while Wu H. (2006) does not discuss them. This thesis reexamines the relationship between stress and tone in
English-to-Mandarin loanwords.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
15
CHAPTER 3
A CORPUS-BASED ANALYSIS
3.1 Basics of the Corpus
The present corpus contains a collection of Mandarin loanwords, with a total of 833 tone loans from the monosyllabic, disyllabic, and trisyllabic English words. The data are mainly collected from Guo Yu Ri Bao Wai Lai Yu Ci Dian (1981). This paper focuses on three kinds of information, namely, the tones that correspond to the
English stressed and unstressed syllables, the English segments, and the different tone loans of each English word.
This corpus codes the tones that correspond to the English stressed and unstressed syllables for the investigation of the relationship between English stress and Mandarin tones. Moreover, it codes the segments of English words for the
examination of the interaction between segmental features and tones. Finally, it codes the different tone loans of each English word for the comparison of the more and less frequently used tone loans.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
16
The present corpus utilizes the format in Microsoft Office ExcelTM 2003, as given below.
(1) The present corpus
Microsoft Office ExcelTM 2003 is familiar to the public. The advantages of it are as follows. First, this program is convenient for the establishment of a small-sized database. It also serves as a convenient tool to update information and to sort tones according to the corresponding stressed or unstressed syllables. Second, due to its portability, the present corpus can be operated as long as a computer is available.
Third, the corpus can be shared easily for academic research.
In this corpus, the monosyllabic, disyllabic, and trisyllabic English words are
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
17
respectively coded in separate categories. Each of the English words is serially numbered, as in (2), (3), and (4).
(2) Serial #: GYRB-M-006 (3) Serial #: GYRB-D-008 (4) Serial #: GYRB-T-010
The abbreviation ‘GYRB’ stands for Guo Yu Ri Bao Wai Lai Yu Ci Dian. The following letters, ‘M’, ‘D’, and ‘T’, stand for the monosyllabic, disyllabic, and trisyllabic English words, respectively. The number that follows ‘M’ in (1) refers to the sixth English word in the monosyllabic category. The number following ‘D’ in (2) refers to the eighth English word in the disyllabic category. The number following ‘T’
in (3) refers to the tenth English word in the trisyllabic category.
The linearty of English syllables ar ecoded. ‘1st’, ‘2nd’, and ‘3rd’ are
respectively labeled to the first, second, and third English syllables. In addition, each of the English syllables is labeled as 1 , 2, or 3, meaning ‘primary stress’, ‘secondary stress’, or ‘unstressed’, respectively. The examples are given below.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
18
(5) Coding of English stress
Serial # EW 1st 2nd
GYRB-D-028 bingo 1 3
GYRB-D-157 cartoon 3 1
The abbreviation ‘EW’ stands for ‘English word’. As for bingo, English stress is located in the first syllable, so it is coded as 1 in the column referring to the first syllable. Its second syllable is unstressed, so it is coded as 3 in the column indicating the second syllable.
With regard to the coding of the four Mandarin tones, this study follows the convention in Wu C. (2006), in which the four tones in Mandarin Chinese are respectively represented as H, MH, L, and HL, as shown in (6). ‘LW’ stands for
‘Mandarin loanword’. The examples in (5) are further illustrated in (6).
(6) Coding of tones in Mandarin loanwords
Serial # EW 1st 2nd LW 1st 2nd
GYRB-D-028 bingo 1 3 賓果 H L
GYRB-D-157 cartoon 3 1 卡通 L H
Concerning the coding of the English segments, the IPA symbols are adopted. In addition, the segments are coded on the basis of syllable structure, namely, onset,
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
19
nucleus, and coda. The following is an example of a monosyllabic English word.
(7) Coding of English segments
EW Serial #
Spelling Onset Nucleus Coda
GYRB-M-039 ton tH √ n
The present corpus is different from the corpora established in the previous studies in that it collects as many tone loans of each English word as possible. This assumption is that the different tone loans of each English word may reflect an acceptable range of tone loans, as they show the alternations in tonal choices.
Therefore, in some cases, there may be a one-to-one relationship between an English word and its corresponding tone loan, whereas in other cases, there may be a
one-to-many relationship between an English word and its tone loans. In cases of the one-to-many relationship, each of the loan forms is labeled numerally, as exemplified in (8).
(8) Coding of different tone loans of ‘martini’
Serial # EW LW1 LW2
GYRB-T-023 martini 馬丁尼 L.H.MH 馬蒂尼 M.HL.MH
‧
In the example above, the segmental information is omitted. The borrowed English word, martini, has two tone loans, respectively labeled as LW1 and LW2.
Statistically speaking, there are 120 monosyllabic English words in the present corpus.
73 of them show a one-to-one relationship between the English word and the tone loan, while 47 have at least more than one Mandarin tone loan. In total, there are 188 tone loans adapted from the monosyllabic English words. In addition, there are 298 disyllabic English words. 193 of them show a one-to-one relationship between the English word and the tone loan, while 105 of them show a one-to-many relationship.
Totally, there are 450 tone loans adapted from the diyllabic English words. Moreover, there are 102 trisyllabic English words. 53 of them have only one tone loan, while 49 have more than one tone loan. There are totally 195 tone loans adapted from the trisyllabic English words.
In this study, the distinction between a one-to-one and a one-to-many
relationship is based on tonal differences rather than orthographic ones. Take 迪斯可 /ti.sˆ.kHF/ (MH.H.L) and 狄斯可 /ti.sˆ.kHF/ (MH.H.L) for example. They appear as two different entries in the dictionary. However, both are adapted from the same English word disco /dIÂsko/, and they have same segmental sequences and same tonal pattern. The only difference between them lies in orthography. That is, the only one difference between them is the Chinese character that corresponds to the English
‧
(MH.H.L) are considered one loan form in the present corpus. The reason for this distinction is that the main concern of this study is the tonal adaptation in the loanwords. The orthographic alternations are not a target for discussion, so they are not taken into account.
3.2 Distribution of the Four Tones
3.2.1 Loanwords from Monosyllabic English Words
Table 3.1 summarizes the statistical result of the tonal adaptation of monosyllabic English words.
Table 3.1: Tonal Adaptation of the Monosyllabic EWs
Tone Number Percentage
H 61 32.45%
MH 47 25.00%
L 33 17.55%
HL 47 25.00%
Total 188 ≈ 100.00%
As shown in Table 3.1, the H-tone accounts for the highest percentage of the data, namely, 32.45%. The L-tone takes up the lowest percentage, namely, 17.55%. The MH-tone and the HL-tone account equally for the same percentage of the data, that is,
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
22
25.00%. In other words, the H-tone is the most favored tone in correspondence to English stress, the L-tone is the least favored tone, and the MH-tone and the HL-tone are in between.
3.2.2 Loanwords from Disyllabic English Words
In this section, we take a look at the distribution of tones that respectively correspond to the stressed and unstressed syllables of disyllabic English words. First, see the statistical result of the English stressed syllables, as shown in Table 3.2.
Table 3.2: Tonal Adaptation of the Stressed Syllables in Disyllabic EWs
Tone Number Percentage
H 150 33.71%
MH 116 26.07%
L 71 15.96%
HL 108 24.27%
Total 445 ≈ 100.00%
The result shown in Table 3.2 is similar to the one in Table 3.1. Here, the H-tone is also the most preferred tone, accounting for 33.71%. The L-tone is the least
preferred choice, taking up 15.96%. Both the MH-tone and the HL-tone are in between. They account for a similar percentage, namely 26.07% and 24.27%
‧
Second, Table 3.3 provides the statistical result of the English unstressed syllables.
Table 3.3: Tonal Adaptation of the Unstressed Syllables in Disyllabic EWs
Tone Number Percentage
H 123 27.64%
MH 131 29.44%
L 68 15.28%
HL 123 27.64%
Total 445 ≈ 100.00%
As shown in Table 3.3, the L-tone accounts for the lowest percentage of the data, namely, 15.28%, while the MH-tone accounts for the highest percentage, namely, 29.44%. In fact, the pitch of the unstressed syllable is lower than that of the stressed one in an English disyllabic word. Thus, the statistical result above brings about a difficulty in explanation if we assume that the lower pitch in an English word corresponds to a lower tone in the Mandarin loanword. We will explore more on the distribution of the four Mandarin tones corresponding to the English unstressed syllables by examining the Mandarin loanwords from trisyllabic English words in the following section.
‧
3.2.3 Loanwords from Trisyllabic English Words
In this section, we examine the distribution of the tones that respectively
correspond to the stressed and unstressed syllables of trisyllabic English words. First, we look at the statistical result of the tonal adaptation of the English stressed syllables, as shown in Table 3.4.
Table 3.4: Tonal Adaptation of the Stressed Syllables in Trisyllabic EWs
Tone Number Percentage
Table 3.4 shows that the percentage for the H-tone is similar to that of the HL-tone. However, the percentage for the H-tone is still higher than that of the HL-tone. The L-tone accounts for the minority of the data. The percentages of the MH-tone and HL-tone are between the highest and the lowest percentages. In other words, the result shown here is consistent with the results shown in Table 3.1 and Table 3.2.
Concerning the tones corresponding to the English unstressed syllables, the
‧
statistical result is given in Table 3.5.
Table 3.5: Tonal Adaptation of the Unstressed Syllables in Trisyllabic EWs
Tone Number Percentage
H 102 27.15%
MH 93 24.73%
L 63 16.67%
HL 118 31.45%
Total 376 ≈ 100.00%
Table 3.5 shows that the L-tone accounts for the lowest percentage of the data, which is opposite to our expectation. Just as the dilemma presented in Table 3.3 in 3.2.2, the statistical result here also leads to a difficulty in explanation here if we assume that a lower pitch in an English word corresponds to a lower tone in the Mandarin loanword.
We will offer in 3.5 a solution to Table 3.3 and Table 3.5 by examining the data in the aspect of relative relation of the tones corresponding to the English stressed and unstressed syllables.
In general, the examination of the distribution of the four Mandarin tones
In general, the examination of the distribution of the four Mandarin tones