CHAPTER 3 LIUJIA RAOPING HAKKA
3.4 Summary
國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
(37) The complete ranking of Liujia Raoping Hakka
*Rise
ID-t-L ID-LL ID-T-R
[Agree-t & *LL]ADJ(Yangping) [OCP- & *Fall]ADJ(Yangping) *T
[OCP- & *M]ADJ(Yangping)
[OCP- & *Fall]ADJ [OCP- & *M]ADJ
*M
*Fall *HH
[Agree-t & *LL]ADJ ID-T
Agree-t *LL OCP-
3.4 Summary
This chapter investigates the tone sandhi in Liujia Raoping Hakka. Tone sandhi in this language is positional. All the tones of the right syllables are preserved. In terms of the non-ru tone sandhi, the universal tonal markedness is the trigger of the tone sandhi. *Fall and *HH are ranked equally based on the markedness conflation (de Lacy 2004). However, it is not enough to account for the historically merged tones,
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
Yangping and Yangqu, which share an identical citation tone, but undergo different
tone sandhi patterns. Yangqu tone sandhi is totally positional whereas Yangping tone sandhi is partially positional and partially contextual. When Yangping is followed by Yin tones, the tone sandhi patterns are contextual. In order to account for the tone
sandhi of the merged tones under the OT framework, I argue for the indexed constraints (Pater 2007, 2010). On the other hand, the requirement of no citation tone exists in the ru tone sandhi. It is captured by *T following the idea of anti-faithfulness (Alderete 1999, 2001).
‧ 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
65
CHAPTER 4
NINGDU TIANTOU HAKKA
This chapter investigates the disyllabic tone sandhi in Ningdu Tiantou Hakka spoken in Jiangxi Province, China. The data are mainly from Huang (2010). Data from Li and Zhang (1992), Luo and Deng (1997), Liu (2001), Xie (2003), Chang (2003), and Huang (2006), are also considered. Most of the tone sandhi patterns are confirmed in the recoded data provided by Dr. Xiaoping Huang who is a native speaker of this language.
The rest of this chapter will be arranged as followed. Section 4.1 introduces the tone inventory and section 4.2 introduces tone sandhi patterns. Section 4.3 provides an OT analysis on the tone sandhi patterns. Finally, section 4.4 is the summary for this chapter.
4.1 Tone Inventory
There are seven citation tones in this language, namely Yinping HM, Yangping
‧
transcribed in terms of Yip’s (2001) system as well.In terms of Yinping, Huang (2010) transcribed it as 51. However, Huang used to specify it as 43 in 2006; moreover, other studies (Li and Zhang 1992, Luo and Deng 1997, Liu 2001, Xie 2003, Chang 2003) transcribed Yinping as 42 or 43. In Lin’s (2007) point of view, 51 is usually perceived as 53 in a speech. This tone starts in a high pitch range, and ends in a mid-pitch range. Yinping in this dialect displays a similar pattern; therefore, Yinping is transcribed as HM in this thesis. As for Yangping, Huang (2010) originally specified it as 324 which is different from his own study in 2006. Huang (2006) transcribed Yangping as 24. According to him and his recorded data, Yangping can be regarded as 24 or 13. The 3 in 324 can be considered a phonetic detail which is not significant in the phonological distinction. Furthermore, Duanmu (2007) mentioned that 24 is close to both 35 and 13. To determine its tone value, the four-contour system proposed by Yip (2001) is adopted. According to Yip (2001), since there is already a 35 tone in this language, the 24 should actually be transcribed as 13. Hence, Yangping is transcribed as LM. There is no distinction between Yinshang and Yangshang. It is originally transcribed as 214 (Huang 2010) and 213
(Huang 2006) respectively; thus, it is converted as MLM here. Also, it only surfaces
4 In addition to Huang (2006, 2010), other studies transcribe Yangqu as a high level tone. Nevertheless, according to the recorded data, Yangqu is more like a MH. It is closer to the transcription in Huang (2006, 2010).
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
as MLM when it is in the prosodically final position. It follows the contour formation of Bao (1999), who took Pingyao as the example, as in (1).
(1) Contour formation C ]
h l h
In (1), the concave tone is from the falling tone in the phrase final position.
Following Duanmu’s (2007) viewpoint, each TBU can carry just one tone. Therefore, the final tone “M” is realized by the extra mora in the phrase final position.
As in chapter three, the internal structures of the tones in this dialect are given in (2).
(2) The internal structure of Ningdu Tiantou Hakka tones
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
In (2), the level tones are not in the tone inventory. The high level tone HH only surfaces as a sandhi tone in this language. The mid-level tone and the low level tone are presented in order to clarify the tonal structures of possible candidates. Besides, the mid-level tone in this language is specified as a high registered low tone according to Yip (2001). Although the mid-level tone could be specified as either [Hr, l] or [Lr, h], [Hr, l] better explains the tone sandhi patterns in this language. In this dialect, the constraints, *Fall and *Concave, in the verb-object construction trigger HM and MLM to undergo tone sandhi. Then ID-t-L preserves the left tone melody of the input tone and form the optimal output, HH. If MM is specified as [Lr, h], the optimal output, HH, could not be successfully selected. Therefore, MM is considered [Hr, l] in this dialect.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
4.2 Tone Sandhi Patterns
Tone sandhi in this language is construction sensitive. Disyllabic words with different grammatical constructions undergo different tone sandhi processes. The grammatical constructions of disyllabic subject-predicate and verb-complement words do not undergo tone sandhi. In the verb-object constructions (henceforth construction ), tone sandhi takes place in the left syllable. On the other hand, in the modifier-head and conjunction constructions (henceforth construction ), tone sandhi may take place in either side or both sides of the disyllabic words. According to Huang (2010), disyllabic subject-predicate and verb-complement words do not undergo tone sandhi because these two constructions are structurally looser than other constructions.
However, Huang (2010) did not explain how their constructions are looser. Besides, if the internal construction influences tone sandhi patterns, what is the mechanism that decides different tone sandhi patterns in construction and construction . Additionally, if the so called looser constructions do not undergo tone sandhi, it faces difficulties in explaining Pingyao tone sandhi. Lin (2012) investigated the construction sensitive tone sandhi in Pingyao. In Pingyao, subject-predicate and verb-object constructions display the same tone sandhi pattern whereas modifier-head, conjunction and verb-complement construction display the other tone sandhi pattern.
It shows that the claim proposed by Huang (2010) is not necessarily true. In this case,
‧
language specific phenomenon. The tone sandhi patterns of construction and construction will be discussed respectively in the following section of this chapter.4.2.1 Tone sandhi patterns of construction
The disyllabic tone sandhi patterns of construction are illustrated as follows:
(3) Tone sandhi patterns of construction 2nd σ
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
tone sandhi occurs only when Yinping HM and Shang MLM are in the first syllable in construction . Then the requirement of no falling tones and no concave tones will be regarded as the motivation for tone sandhi in construction . It follows the universal tonal markedness tendency postulated by Yip (2001, 2002). In this tonal markedness tendency, contour tones are more marked. Moreover, the rising tone is more marked than the falling tone. If the falling tone is not allowed in the surface, the more marked rising tone should have been forbidden. In (3), the tone sandhi patterns obviously do not follow this tendency, i.e. the rising tones do surface. This tonal preservation tendency could be considered a specific tendency for this construction in this dialect.
It only occurs in construction . Generally, the tone sandhi patterns of construction are similar to the tone sandhi patterns in Liujia Raoping Hakka, except for the preservation of rising tones. Tone sandhi in both dialects can be regarded as positional tone sandhi. The motivation for tone sandhi is not due to adjacent tones but the position. The position decides the syllables which undergo tone sandhi.
In addition to the motivation of tone sandhi, the mechanism of the emergence of the high level tone is preserving the left tone melody of the input tone, as in (4).
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
(4) Construction tone sandhi mechanism HM-HM → HH-HM
[kʰwai tʃʰa] ‘drive the car’
In (4), the left tone melody of the first HM is . It is preserved to form a level tone. The left tone melody preservation follows the tonal target (Yip 2001) as it is argued in the last chapter. The example of the initial taget is in (5).
(5) The initial target of Yinping ʃɔŋ HM ‘fragrant’
In (5), there is a short high plateau in the initial of the tone. It follows the
‧
due to the initial tonal target in this dialect.The competition among the high level tone and other level tones will be demonstrated in the OT analyses in 4.3.
4.2.2 Tone sandhi patterns of construction
Tone sandhi patterns in construction are more complicated than it is in construction . Both left and right syllables may undergo tone sandhi. The disyllabic
tone sandhi patterns of construction are illustrated as follows:
(6) Tone sandhi patterns of construction
2nd σ
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
In (6), both the first and the second syllables may undergo tone sandhi. All the low register tones, LM, ML and L, have no change, except for MLM. It displays a tendency that low tones in this dialect tend to stay invariant. In terms of the second-syllable tone sandhi, it takes place when high register tones are in the second syllable. This kind of tone sandhi results from the avoidance of two adjacent high register tones or the requirement of no tone rising in the syllable boundary. The mechanism is given in (7) and (8).
(7) The avoidance of two adjacent high register tones MH-HM → MH-ML
[tʰien tin] ‘light’
In (7), it is an example of the second-syllable tone sandhi. The trigger of the tone sandhi is the avoidance of two adjacent high register tones. The high register tone in the second syllable is then dissimilated into a low register tone. The other type of the second-syllable tone sandhi is in (8).
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
(8) No tone rising during the syllable boundary ML-HM → ML-LM
[tʃʰi tʃʰa] ‘car’
H
M M M M L pitch rise L L
In (8), the right tonal feature of the first syllable is L while the left tonal feature of the second syllable is H. The pitch across these two syllables is an abrupt rising. In order to prevent the rising pitch in the intersyllabic position, the second syllable is changed to LM. This phenomenon only takes place when there is one tone staying unchanged in the tonal pair during the tone sandhi process, as in (8)5. However, the tonal pair in which both tones undergo tone sandhi does not display the pattern in (8).
As for ML-ML, the optimal output of HM-MH, it does not display the pattern in (8) because ML-ML here does not contain any citation tones. This analysis is based on Lin (2006) and Zhang (1999). Lin (2006) mentioned that, in terms of articulation, it is difficult for speakers to produce intersyllabic tone melodies with wide distance apart.
Besides, Zhang (1999) pointed out that the duration of the intersyllabic position is shorter than a syllable. Hence, it is harder for the rising pitch to be realized in the
5 The unchanged tonal pair, ML-MH, will be discussed in 4.3.2.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
intersyllabic position. On the other hand, the tone sandhi of ML-HM is not regarded as the requirement of the agreement of intersyllabic tone melodies because it will face difficulties in the OT framework. If so, the optimal output, ML-LM, will be eliminated by the candidate, ML-ML.
As for the first-syllable tone sandhi, it occurs when MLM is followed by low register tones. Since MLM needs an extra TBU to realize the last tonal feature M, it only occurs in the second syllable. The mechanism is shown in (1). Therefore, the tone sandhi may be triggered by the requirement of no concave tone in the non-phrase final position.
Last, the both-syllable tone sandhi takes place in two contexts. One is when HM is followed by a high register tone. The other is when MLM is followed by a high register tone. The triggers for the tone sandhi patterns are the avoidance of two adjacent high register tones, the requirement of no high tones and the avoidance of MLM in the non-phrase final position. First of all, the avoidance of two adjacent high register tones could trigger the tone sandhi when HM is followed by a high register tone. Second, the requirement of no high tones prevents the sandhi tones from becoming high register tones. Therefore, both syllables undergo tone sandhi. Since most of the sandhi tones in construction are low-registered, this requirement exists.
Third, the avoidance of MLM in the non-phrase final position triggers the tone sandhi
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
of MLM in the first syllable. The high register tones following MLM change because of the requirement of no high tones.
4.3 OT analysis
According to 4.2, constructions and construction display very different tone sandhi patterns. In the framework of OT, because of different morphological structures, different constraint rankings may be proposed. It follows the idea of cophonology (Orgun 1996, 1998, 1999, Anttila 1997, Anttila and Cho 1998, Inkelas and Zoll 2005, 2007, among others) that different rankings do exist. Besides, the indexed approach (McCarthy and Prince 1995, Itô and Mester 1999, 2001, Alderete 2001, Pater 2000, 2007, 2010, among others) also works in dealing with different constructions with different tone sandhi patterns. However, cophonology fails to account for situations where the indexed approach could successfully solve. In this case, the indexed approach is adopted.
4.3.1 OT analysis of Tone sandhi in construction
To account for the tone sandhi in construction , several constraints will be posited in (9)-(15).
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
(9) IDENT-IO-T-R(Hr) (abbr. ID-T-R(Hr)):
Assign one violation mark for every Hr tone standing at the right edge which is different from its corresponding tone in the output.
(10) IDENT-IO-T-R(Lr) (abbr. ID-T-R(Lr)):
Assign one violation mark for every Lr tone standing at the right edge which is different from its corresponding tone in the output.
(11) IDENT-IO-T-L(Hr) (abbr. ID-T-L(Hr)):
Assign one violation mark for every Hr tone standing at the left edge which is different from its corresponding tone in the output.
(12) IDENT-IO-T-L(Lr) (abbr. ID-T-L(Lr)):
Assign one violation mark for every Lr tone standing at the left edge which is different from its corresponding tone in the output.
(13) IDENT-IO-t-L (abbr. ID-t-L):
Assign one violation mark for an output tone which the tone melody at the left edge of the tone melody level is different its corresponding input (Lin 2011).
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
(14) *Fall
Assign one violation mark for every output falling tone.
(15) *Concave
Assign one violation mark for every output concave tone.
Constraints in (9)-(12) are based on Lin’s (2012) work. Lin (2012) investigated the tone sandhi in Pingyao. In Pingyao, the tone sandhi may take place in both the first and the second syllable. Lin (2012) then posited the faithfulness constraints which target the register. However, in Pingyao tone sandhi, Lin (2012) defined the preserved syllable as “head” and the “head” follows the tonal preference tendency proposed by de Lacy (1999, 2002), i.e. the high tone is preferred in the head position.
However, as it is mentioned in chapter three, the “head” is easily misled to the stress issue. Besides, the preservation of register in is a different case in Ningdu Tiantou Hakka. In this dialect, low register tones are preserved in both construction and construction . In this case, the tonal preference tendency proposed by de Lacy (1999, 2002) fails to account for this. Accordingly, the tonal preservation is regarded as a typological difference.
In OT, the syllable preservation is due to the interaction among faithfulness
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
constraints and markedness constraints. Hence, in this dialect, the tone sandhi takes place in both syllable because of the interaction among faithfulness constraints, (9)-(13), and markedness constraints.
In this construction, the second syllable never changes; therefore, ID-T-R(Hr) and ID-T-R(Lr) are undominated. All the candidates whose tone in the second syllable undergoes tone sandhi must be ruled out by ID-T-R(Hr) and ID-T-R(Lr). In this case, the violation of ID-T-R(Hr) and ID-T-R(Lr) will no longer be discussed in the rest of this chapter. ID-t-L captures the fact that the optimal output forms preserve the left tone melody of the input form. *Fall and *Concave prevent the output falling tones and concave tones. *Concave needs to dominate ID-T-L(Lr) so that MLM may not surface in the first syllable. *Fall has to be dominated by ID-T-L(Lr). In this case,
*Fall will only influence high register tones. Due to that HM in the first syllable undergoes tone sandhi, *Fall dominates ID-T-L(Hr). In this case, *Fall only rule out high register falling tones instead of low register falling tones. The constraint ranking is given in (16).
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
(16) The constraint ranking of Construction ID-T-R(Hr) ID-T-R(Lr) ID-t-L *Concave
ID-T-L(Lr)
*Fall ID-T-L(Hr)
In (16), it demonstrates that ID-T-R(Hr) and ID-T-R(Lr) needs to dominate markedness constraints because tones in the second syllable never change. Moreover, ID-t-L does not have crucial ranking with markedness constraints. It only dominates ID-T to confirm that the optimal output is selected and the left tone melody of the input form is preserved. The tableau in (17), (18) and (19) provide examples in the ranking in (16).
‧
(17) Yinqu ML in the ranking ML-HM → ML-HM
[kua tsʰiaŋ] ‘sweep the tomb’
ML-HM is a tonal pair which does not undergo tone sandhi. In (17), level tones are discussed because they are possible outputs in construction . Moreover, the tableau in (17) illustrates that ID-T-L(Lr) needs to dominate *Fall so that candidate (c), the optimal output, will not be eliminated by candidate (f). ID-t-L rules out candidates (b), (d), (g) and (h) because the left tone -melody of these candidates is different from the left tone melody of the input form, which is . Candidate (e) violates *Concave
‧
due to MLM in the first syllable. Candidates (a) and (f) are eliminated by ID-T-L(Lr) because the tone in the first syllable is not identical to the input ML. Tableau (18) will diaplay the tone sandhi process of Yinping HM and the ranking between *Fall and
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
register low tone, [Hr, l]. If the mid-level tone is defined as [Lr, H], it will not violate ID-t-L and candidate (f) and (g) will tie. If so, the optimal output could not be predicted successfully. On the other hand, tableau (18) demonstrates that *Fall has to dominate ID-T-L(Hr); otherwise, candidate (a), HM-HM, will be wrongly predicted as the optimal output. Candidates (b), (d), (g) and (h) are ruled out by ID-t-L because the left tone -melody in the output is different from the left tone melody of the input form, which is . Candidate (e) contains MLM, so it violates *Concave. Candidate (a) and (c) contain two falling tones; therefore, they violate *Fall twice. Even candidate (f) violates *Fall once, it is still chosen as the optimal output. Furthermore, in (19), the tone sandhi pattern of Shang MLM and the ranking between *Concave and ID-T-L(Lr) will be illustrated.
‧
(19) Shang MLM in the ranking MLM-HM → HH-HM [sie sam] ‘wash clothes’
In (19), it demonstrates that *Concave should dominate ID-T-L(Lr). If there is no crucial ranking, candidate (e) and candidate (f) will tie. Besides, if *Concave is
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
violate *Fall twice. Since candidate (f) only violates *Fall only once, it is selected as the optimal output.
According to (14) and (15), it is obvious that the triggers for the positional tone sandhi in construction are *Fall and *Concave. The preservation of the first syllable is captured by ID-T-L(Lr) and ID-T-L(Hr).
4.3.2 OT analysis of Tone sandhi in construction
In construction , tone sandhi could take place in both first and second syllable.
In this case, constraints preserving the input tones have to target both syllable as in construction in 4.3.1. In this dialect, there is a tendency for low register tones to stay unchanged. In this case, the register must be considered. Four constraints posited in (9), (10), (11), and (12) are still needed in construction . They are ID-T-R(Hr), ID-T-R(Lr), ID-T-L(Hr) and ID-T-L(Lr) respectively.
4.3.2.1 The second-syllable tone sandhi
In this construction, the tonal markedness tendency (Yip 2001, 2002) still works.
The concave tone is the most marked. Rising tones are more marked than falling tones.
The constraints for concave tones and falling tones have already been posited in (14)
The constraints for concave tones and falling tones have already been posited in (14)