CHAPTER 4 NINGDU TIANTOU HAKKA
4.3.1 OT analysis of Tone sandhi in construction
4.3.2.2 The first syllable tone sandhi
國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
(32) Temporal ranking 3
ID-T-R(Lr) OCP-Hr *Concave
ID-T-L(Lr) *Hr ID-MH
*IsRise [ID-T-L(Hr) & *Fall]SEG *Rise ID-T-L(Hr) *Fall ID-T-R(Hr)
4.3.2.2 The first-syllable tone sandhi
As for the first-syllable tone sandhi, it occurs when MLM is followed by a low register tone. The trigger can simply be regarded as *Concave. However, the ranking in (32) encounters some problems in accounting for the mechanism of the first-syllable tone sandhi.
‧
(33) Shang MLM-Yinqu ML in the temporal ranking 3 MLM-ML → HM-ML
[fo tsien] ‘rocket’
MLM-ML [Lr, hlh]-[Lr, hl]
OCP-Hr *Concave ID-T-L(Lr) ID-MH *Hr *IsRise [ID-T-L(Hr) & *Fall]SEG *Rise *Fall
a.MH-ML
In dealing with the first-syllable tone sandhi, ID-T-R(Hr) and ID-T-L(Hr) are not influential. In this case, they are not considered in this section. In (33), only four candidates are discussed because other candidates are ruled out by OCP-Hr, ID-T-R(Lr) and *Concave. Candidate (a) is ruled out by *Hr because it contains the high register tone, MH. Candidate (b) should be the optimal output, but it is wrongly ruled out by *Hr. In this case, candidates (c) and (d) are incorrectly predicted as the optimal output forms. The case of candidate (c) is similar to that in (25). However, the solution proposed in (26) does not work here because the input forms are different.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
Notice that both candidates (c) and (d) consist of two tones which have same register.
To rule out candidates (c) and (d), OCP-Register is a possible solution.
(34) OCP-Register (OCP-Reg):
Assign one violation mark for adjacent tones which have the same register.
Nevertheless, OCP-Reg has to ranked lower than *Rise so that ML-ML, the optimal output of HM-MH, will not be eliminated. Due to the low ranking, OCP-Reg is not enough to rule out candidates (c) and (d). A conjoined constraint is posited in (35).
(35) [OCP-Reg & *Rise]ADJ:
Assign one violation mark for adjacent tones which violate both OCP-Reg and
*Rise (Lin 2011).
The constraint in (35) is from Lin (2011). This constraint is originally used to explain the tone sandhi in Dongshi Hakka. Here it is posited to rule out candidate (d) in (33). It has to be ranked higher than both ID-MH and *Hr. It is ranked higher than
*Hr in order to successfully chose candidate (d) as the optimal output and rule out
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
candidate (c) in (29). On the other hand, it ranked higher than ID-MH to ensure that ML-ML, the optimal output of HM-MH, is not eliminated. The tableau in (36) shows
how [OCP-Reg & *Rise]ADJ works.
(36) [OCP-Reg & *Rise]ADJ rules out LM-ML MLM-ML → HM-ML
[fo tsien] ‘rocket’
MLM-ML [Lr, hlh]-[Lr, hl]
OCP-Hr *Concave ID-T-L(Lr) [OCP-Reg & *Rise]ADJ ID-MH *Hr *IsRise [ID-T-L(Hr) & *Fall]SEG *Rise OCP-Reg *Fall
☞b.HM-ML
[Hr, hl]-[Lr, hl] * * **
?c.ML-ML
[Lr, hl]-[Lr, hl] * * * **
d.LM-ML
[Lr, lh]-[Lr, hl] * *! * * *
In (36), candidate (d) is ruled out by [OCP-Reg & *Rise]ADJ because both LM and ML are low register tones and there is a rising tone in the tonal pair. However, the optimal output still loses to candidate (c). Notice that the differences between the tones in candidates (b) and (c) are the register of HM and ML and the pitch in the intersyllabic position. Since the register of the output falling tones is more marked,
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
the pitch in the intersyllabic position is considered. The pitch in the intersyllabic position of candidate (c) is rising. In order to rule out candidate (c), *IsRise has to dominate *Hr somehow. Whenever the rising pitch in the intersyllabic position is penalized, there is a citation tone in the tonal pair. In other words, *IsRise also applies to the context where there is a citation tone in the tonal pair. The conjoined constraint is posited in (37).
(37) [*IsRise & *T]ADJ:
Assign one violation mark for adjacent tones which violate both *IsRise and *T.
*T in (37) follows the concept of anti-faithfulness (Alderete 1999, 2001) and forbids citation tones in the output. If it is not conjoined with *IsRise, it is meaningless to posit this constraint because its ranking is low. It has to be dominated by both *Rise and [ID-T-L(Hr) & *Fall]SEG so that MH-ML, the optimal output of MH-HM, will not lose to the candidate, ML-ML. The low ranked *T cannot solve the problem in (33) and (36) alone; hence, [*IsRise & *T]ADJ is necessary. [*IsRise &
*T]ADJ does not only target this tonal pair specifically, but also rules out the candidate, ML-MLM, and makes possible the emergence of HM-MLM, the optimal output of MLM-MLM.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
In terms of the ranking of [*IsRise & *T]ADJ, it has to dominate [OCP-Reg &
*Rise]ADJ; otherwise, ML-LM, the optimal output of ML-HM, will lose to the candidate, ML-ML. The residual problem in (36) is solved and demonstrated in (38).
(38) [*IsRise & *T]ADJ rules out ML-ML MLM-ML → HM-ML
[fo tsien] ‘rocket’
MLM-ML [Lr, hlh]-[Lr, hl]
OCP-Hr *Concave ID-T-L(Lr) [*IsRise & *T]ADJ [OCP-Reg & *Rise]ADJ ID-MH *Hr *IsRise [ID-T-L(Hr) & *Fall]SEG *Rise *T OCP-Reg *Fall
☞b.HM-ML
[Hr, hl]-[Lr, hl] * * * **
c.ML-ML
[Lr, hl]-[Lr, hl] * *! * * * **
In (38), candidate (c) violates [*IsRise & *T]ADJ because the pitch in the intersyllabic position is rising and ML in the second syllable is a citation tone. In this case, candidate (b) is successfully predicted. The ranking for construction can be revised as follows:
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
(39) Temporal ranking 4
ID-T-R(Lr) *Concave
ID-T-L(Lr)
OCP-Hr [*IsRise & *T]ADJ
[OCP-Reg & *Rise]ADJ
*Hr ID-MH
*IsRise [ID-T-L(Hr) & *Fall]SEG *Rise
ID-T-L(Hr) *Fall *T OCP-Reg ID-T-R(Hr)
4.3.2.3 Both-syllable tone sandhi
In terms of the both-syllable tone sandhi, the triggers are the avoidance of two adjacent high register tones, the avoidance of MLM in the non-phrase final position and the requirement of no high tones respectively. In OT, these factors can be captured by OCP-Hr, *Concave and *Hr. Since these three constraints are already discussed, the ranking in (39) could also account for. The tableau in (40) will take HM-HM as an example and illustrate how the tone sandhi of both syllables works.
‧
(40) Yinping HM-Yinping HM in the temporal ranking 4 HM-HM → ML-ML
[fuŋ tʃʰa] ‘windmill’
HM-HM [Hr, hl]-[Hr, hl]
OCP-Hr *Concave ID-T-L(Lr) [*IsRise & *T]ADJ [OCP-Reg & *Rise]ADJ ID-MH *Hr *IsRise [ID-T-L(Hr) & *Fall]SEG *Rise *T OCP-Reg *Fall
a.HM-HM
In (40), candidate (a) consists of two high register tones, so it violates OCP-Hr.
Candidate (b) changes the input tone in the first syllable into a concave tone; thus, it is ruled out by *Concave. The second syllable of candidate (c) is preserved from the input. Also, there is a rising pitch across the syllable boundary. This is prevented by [*IsRise & *T]ADJ. [OCP-Reg & *Rise]ADJ ruled out candidate (d) because the two tones are in the same register and its second syllable is a rising tone. Candidate (e)
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
contains a high register tone, so it is ruled out by *Hr. In this case, candidate (f) is chosen as the optimal output. On the other hand, the tone sandhi of both syllables can occur in another situation, i.e. where MLM followed by a high register tone. It displays the same violation patterns as in (40).
4.3.2.4 Wrong predictions of [*IsRise & *T]ADJ
The ranking so far has successfully accounted for tones which undergo tone sandhi. However, the ranking is too powerful to influence tones which do not undergo tone sandhi. [*IsRise & *T]ADJ would incorrectly rule out the unchanged tonal pairs, LM-HM and ML-MH.
In terms of LM-HM, the LM is preserved by ID-T-L(Lr). As for HM, owing to the rising pitch in the intersyllabic position, it is incorrectly changed. LM-HM, LM-MH, LM-ML and LM-LM are all possible candidates.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
(41) Yangping LM-Yinping HM in the temporal ranking 4 lien LM tau HM ‘sickle’ → LM-HM
LM-HM [Lr, lh]-[Hr, hl]
OCP-Hr *Concave ID-T-L(Lr) [*IsRise & *T]ADJ [OCP-Reg & *Rise]ADJ ID-MH *Hr *IsRise *Rise *T
☞a.LM-HM
[Lr, lh]-[Hr, hl] *? * * * **
?b.LM-MH
[Lr, lh]-[Hr, lh] * ** *
?c.LM-ML
[Lr, lh]-[Lr, hl] * * *
?d.LM-LM
[Lr, lh]-[Lr, lh] * ** *
Notice that the constraints lower than [*IsRise & *T]ADJ is not influential in this condition except for *Rise and *T. Therefore, in addition to *Rise and *T, they are not discussed in this section. In (41), the tonal pair LM-HM should not undergo tone sandhi; however, the optimal output LM-HM is ruled out by [*IsRise & *T]ADJ. To rule out candidates (b) and (d), OCP-Rise is a solution.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
(42) OCP-Rise:
Assign one violation mark for adjacent rising tones.
Nevertheless, OCP-Rise has to be ranked lower than *Rise; otherwise, the unchanged tonal pair, MH-LM, will be incorrectly ruled out. If OCP-Rise is ranked this low, it cannot solve the problem in (41). Since there is a high register tone in the input, OCP-Rise is restricted to the context where the input high register tones is retained. To define such a situation, IDENT-High register should be posited.
(43) IDENT-High register (abbr. ID-Hr)
Assign one violation mark for the Hr tone which is not identical to its corresponding output tone.
ID-Hr has to be ranked lower than *Rise in order not to rule out the optimal output of HM-MH. In this case, the conjoined constraint is posited in (44).
(44) [OCP-Rise & ID-Hr]ADJ:
Assign one violation mark for adjacent tones which violate both OCP-Rise and ID-Hr.
‧
[OCP-Rise & ID-Hr]ADJ penalizes two adjacent rising tones and these tones are required to be identical to their input high register tones. This constraint needs to dominate [*IsRise & *T]ADJ to solve the problem in (41). The tableau in (45) illustrates how [OCP-Rise & ID-Hr]ADJ works.
(45) [OCP-Rise & ID-Hr]ADJ rules out LM-MH and LM-LM LM-HM → LM-HM
[lien tau] ‘sickle’
LM-HM [Lr, lh]-[Hr, hl]
[OCP-Rise & ID-Hr]ADJ OCP-Hr *Concave ID-T-L(Lr) [*IsRise & *T]ADJ [OCP-Reg & *Rise]ADJ ID-MH *Hr *IsRise *Rise *T OCP-Rise ID-Hr
☞a.LM-HM
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
optimal output consists of two tones in different registers whereas candidate (c) consists of two tones in the same register. This difference can be captured by OCP-Reg; however, the low ranking of OCP-Reg in this dialect limits its power. The conjoined constraint in (46) is needed.
(46) [OCP-Reg & OCP- ]ADJ:
Assign one violation mark for adjacent tones which violate both OCP-Reg and OCP- .
[OCP-Reg & OCP- ]ADJ penalizes adjacent tones which have the same register and have the same tone melody in the intersyllabic position. [OCP-Reg &
OCP- ]ADJ has to be ranked lower than ID-T-L(Lr) in order not to rule out tonal pairs, such as LM-ML. The OCP- restricts the power of OCP-Reg. It has to be ranked at the same level with *Fall so that MH-ML, the optimal output of MH-HM will not be ruled out. The tableau in (47) illustrates how [OCP-Reg & OCP- ]ADJ rules out candidate (c) in (41) and (45).
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
(47) [OCP-Reg & OCP- ]ADJ rules out LM-ML LM-HM → LM-HM
lien tau ‘sickle’
LM-HM [Lr, lh]-[Hr, hl]
[OCP-Rise & ID-Hr]ADJ OCP-Hr *Concave ID-T-L(Lr) [OCP-Reg & OCP-]ADJ [*IsRise & *T]ADJ [OCP-Reg & *Rise]ADJ ID-MH *Hr *IsRise *Rise *T OCP- OCP-Rise ID-Hr OCP-Reg
☞a.LM-HM
[Lr, lh]-[Hr, hl] * * * * *
* b.LM-MH
[Lr, lh]-[Hr, lh] *! * *
* * * *
c.LM-ML
[Lr, lh]-[Lr, hl] *! * * * * *
d.LM-LM
[Lr, lh]-[Lr, lh] *! * *
* * * * *
In (47), [OCP-Reg & OCP- ]ADJ successfully rules out LM-ML because LM and ML are in the same register and there are adjacent tone melodies. The ranking is revised and illustrated as follows:
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
(48) Temporal ranking 5
ID-T-R(Lr) *Concave
ID-T-L(Lr) [OCP-Rise & ID-Hr]ADJ
[OCP-Reg & OCP- ]ADJ
OCP-Hr [*IsRise & *T]ADJ
[OCP-Reg & *Rise]ADJ
*Hr ID-MH
*IsRise [ID-T-L(Hr) & *Fall]SEG *Rise
OCP- *T ID-T-L(Hr) *Fall OCP-Rise OCP-Reg ID-T-R(Hr) ID-Hr
In addition to LM-HM, ML-MH is the other tonal combination that will be incorrectly ruled out by [*IsRise & *T]ADJ in the temporal ranking. The constraint ranking in (48) still fails to predict ML-MH, as in (49).
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
(49) Yinqu ML-Yangqu MH in temporal ranking 5 ML-MH → ML-MH
[kau tʰiau] ‘entomb’
ML-MH [Lr,-hl]-[Hr, lh]
[OCP-Rise & ID-Hr]ADJ OCP-Hr *Concave ID-T-L(Lr) [OCP-Reg & OCP-]ADJ [*IsRise & *T]ADJ [OCP-Reg & *Rise]ADJ ID-MH *Hr *IsRise *Rise *T OCP- OCP-Rise ID-Hr OCP-Reg
a. ML-HM
[Lr, hl]-[Hr, hl] *! * * * * *
☞b.ML-MH
[Lr, hl]-[Hr, lh] *? * * * *
* c.ML-ML
[Lr, hl]-[Lr, hl] *! * * * * *
?d.ML-LM
[Lr, hl]-[Lr, lh] * * * * * *
Candidates (a) and (c) are ruled out by [*IsRise & *T]ADJ because this ML is a citation tone and there is a rising pitch in the intersyllabic position. The optimal output should be candidate (b); however, it is ruled out by [*IsRise & *T]ADJ as well.
Candidate (d) is then wrongly selected as the optimal output. In order to rule out ML-LM, OCP- is posited.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
(50) OCP- :
Assign one violation mark for adjacent tones which have the same tone melody in the intersyllabic position.
However, this constraint may penalize the optimal output as well. Besides, it has to be dominated by *Rise in order not to rule out the unchanged tonal pair, such as HM-LM. It cannot solve the problem in (49) alone. [OCP- & ID-MH]ADJ is then posited to rule out candidate (c), ML-LM. Its definition is exhibited in (51).
(51) [OCP- & ID-MH]ADJ:
Assign one violation mark for adjacent tones which violate both OCP- and ID-MH.
The OCP in (50) targets the -melody instead of the general tone melody in order not to rule out MH-ML, the optimal output of MH-MH. The ranking of [OCP- &
ID-MH]ADJ should be higher than [*IsRise & *T]ADJ to preserve the optimal output, ML-MH. Besides, it has to be ranked lower than ID-T-L(Lr) to prevent ML from changing. The tableau in (52) shows how [OCP- & ID-MH]ADJ works.
‧
[OCP-Rise & ID-Hr]ADJ OCP-Hr *Concave ID-T-L(Lr) [OCP- & ID-MH]ADJ [OCP-Reg & OCP-]ADJ [*IsRise & *T]ADJ [OCP-Reg & *Rise]ADJ ID-MH *Hr *IsRise *Rise *T OCP- OCP- OCP-Rise ID-Hr OCP-Reg
a. ML-HM
[OCP- & ID-MH]ADJ successfully rules out candidate (d) because there are adjacent tone melody in the intersyllabic position and the input rising tone is not preserved.
Nevertheless, the optimal output still loses to candidate (c). Notice that candidate (c) consists of two ML’s. OCP-ML is posited in (53).
(53) OCP-ML:
Assign one violation mark for adjacent ML.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
However, the tonal pair, ML-ML, does exist and this constraint cannot rule out the optimal output of tonal pairs, such as HM-MH. Therefore, this constraint has to be dominated by *Rise. Due to the low ranking of OCP-ML, it cannot solve the problem in (52). In this case, the constraint in (54) is needed.
(54) [OCP-ML & *T]ADJ:
Assign one violation mark for adjacent tone which violate both OCP-ML and *T.
[OCP-ML & *T]ADJ has to dominate ID-T-L(Lr) so that the unchanged tonal combination, ML-ML, will not be wrongly changed. Besides, it has to dominate *Rise in order to rule out candidate (c) in (52). The OCP in (54) target the ML instead of the general falling tone in order not to eliminate another unchanged tonal combination, HM-ML. Moreover, the OCP that targets a specific tone category is not an idiosyncratic phenomenon in this language. In Mandarin Chinese tone sandhi, there is a requirement avoiding two adjacent low tones. It can be regarded as an OCP effect.
The tableau in (55) illustrates how [OCP-ML & *T]ADJ predicts ML-MH as the optimal output.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
(55) [OCP-ML & *T]ADJ rules out ML-ML ML-MH → ML-MH
[kau tʰiau] ‘entomb’
ML-MH [Lr,-hl]-[Hr, lh]
[OCP-Rise & ID-Hr]ADJ OCP-Hr *Concave ID-T-L(Lr) [OCP-ML & *T]ADJ [OCP- & ID-MH]ADJ [OCP-Reg & OCP-]ADJ [*IsRise & *T]ADJ [OCP-Reg & *Rise]ADJ ID-MH *Hr *IsRise *Rise OCP-ML OCP- OCP- OCP-Rise ID-Hr *T OCP-Reg
☞b.ML-MH
[Lr, hl]-[Hr, lh] * * * * * *
* c.ML-ML
[Lr, hl]-[Lr, hl]
*! * * * * * * *
Candidate (c) in (55) is ruled out by [OCP-ML & *T]ADJ because there are adjacent ML and the first ML is a citation tone. Therefore, candidate (b) is successfully selected as the optimal output. The constraint ranking illustrated in (48) will be revised in (56).
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
(56) Temporal ranking 6
ID-T-R(Lr) *Concave
[OCP-Rise & ID-Hr]ADJ ID-T-L(Lr)
[OCP-Reg & OCP- ]ADJ [OCP- & ID-MH]ADJ
[OCP-ML & *T]ADJ
[*IsRise & *T]ADJ OCP-Hr [OCP-Reg & *Rise]ADJ
*Hr ID-MH
*IsRise [ID-T-L(Hr) & *Fall]SEG *Rise
OCP- *T ID-T-L(Hr) *Fall OCP-ML OCP- OCP-Rise OCP-Reg ID-T-R(Hr) ID-Hr
The ranking in (56) successfully accounts for the non- ru tone sandhi. The ru tone sandhi will be discussed in the following section.
4.3.2.5 Ru tone sandhi
In terms of ru tones, it follows the argument in the last chapter. Owing to different syllable structures, limited tones could be realized in the ru tone syllable.
Also, the tone sandhi patterns differ from non-ru tones. In this dialect, the trigger of
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
the ru tone sandhi is contextual. Adjacent high ru tones are prohibited. Besides, when ru tones are next to non-ru tones, the constraints targeting the interaction between
syllables still work, such as [*IsRise & *T]ADJ and [OCP-Reg & *Rise]ADJ. However, the preservation of ru tones is different from the preservation of non-ru tones. Ru tones tend to stay unchanged, except for the high register ru tone in the second syllable. In this case, the ranking of ID-T-L(Hr) may be different in dealing with ru tones and non-ru tones; otherwise, the example in (57) will be wrongly predicted.
‧
(57) High register ru tones in the first syllable H-HM → H-ML
nat kɔŋ ‘moonlight’
H-HM [Hr,-h]-[Hr, hl]
[OCP-Rise & ID-Hr]ADJ OCP-Hr ID-T-R(Lr) *Concave ID-T-L(Lr) [OCP-ML & *T]ADJ [OCP- & ID-MH]ADJ [OCP-Reg & OCP-]ADJ [*IsRise & *T]ADJ [OCP-Reg & *Rise]ADJ ID-MH *Hr *IsRise [ID-T-L(Hr) & *Fall]SEG *Rise OCP-ML OCP- OCP- OCP-Rise ID-Hr *T OCP-Reg *Fall ID-T-R(Hr) ID-T-L(Hr)
a. L-HM because of H in the first syllable. Candidate (c), L-ML, is wrongly predicted. Since H
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
in the first syllable never changes, the constraint for the preservation for H in the first syllable should be undominated. Therefore, ID-T-L(Hr)ru is posited. The problem in (57) is solved in (58).
(58) ID-T-L(Hr)ru rules out L-LM H-HM → H-ML
nat kɔŋ ‘moonlight’
H-HM [Hr,-h]-[Hr, hl]
ID-T-L(Hr)ru [OCP-Rise & ID-Hr]ADJ OCP-Hr ID-T-R(Lr) *Concave ID-T-L(Lr) [OCP-ML & *T]ADJ [OCP- & ID-MH]ADJ [OCP-Reg & OCP-]ADJ [*IsRise & *T]ADJ [OCP-Reg & *Rise]ADJ ID-MH *Hr *IsRise [ID-T-L(Hr) & *Fall]SEG *Rise OCP-ML OCP- OCP- OCP-Rise ID-Hr *T OCP-Reg *Fall ID-T-R(Hr) ID-T-L(Hr)
c. L-ML [Lr, l]-[Lr, hl]
*
! * * * * * *
☞h. H-ML
[Hr, h]-[Lr, hl] * * * * * *
In (58), candidate (c) does not preserve the input H in the first syllable so it is ruled out by ID-T-L(Hr)ru. H-ML is then selected as the optimal output. This is the case which H is in the first syllable. As for the cases which H is in the second syllable, unchanged tonal pairs, such as LM-H, ML-H and L-H, may be wrongly ruled out by [*IsRise & *T]ADJ. The example is given in (59).
‧
(59) High register ru tones in the second syllable L-H → L-H
tsat sat ‘tightly weaved (basket)’
L-H [Hr,-h]-[Hr, hl]
ID-T-L(Hr)ru [OCP-Rise & ID-Hr]ADJ OCP-Hr ID-T-R(Lr) *Concave ID-T-L(Lr) [OCP-ML & *T]ADJ [OCP- & ID-MH]ADJ [OCP-Reg & OCP-]ADJ [*IsRise & *T]ADJ [OCP-Reg & *Rise]ADJ ID-MH *Hr *IsRise [ID-T-L(Hr) & *Fall]SEG *Rise OCP-ML OCP- OCP- OCP-Rise ID-Hr *T OCP-Reg *Fall ID-T-R(Hr) ID-T-L(Hr)
☞a. L-H OCP-Hr. Candidate (d) violates ID-T-L(Lr) because the input L in the first syllable is not preserved. Candidate (a) should have been the optimal output. However, it is incorrectly ruled out by [*IsRise & *T]ADJ and candidate (b) is selected as the wrong optimal output. To fix this wrong prediction, the preservation of H in the second syllable is considered. ID-T-R(Hr) is indexed as ID-T-R(Hr)ru which has to be ranked at the same level with [*IsRise & *T]ADJ so that the optimal output of MH-H, ML-H, MLM-H and MH-H will not be wrongly ruled out. Nevertheless, ID-T-R(Hr)ru is not
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
sufficient to account for the problem in (59). The tableau in (60) illustrates the situation which ID-T-R(Hr)ru is not enough.
(60) ID-T-R(Hr)ru is not enough L-H → L-H
tsat sat ‘tightly weaved (basket)’
L-H [Hr,-h]-[Hr, hl]
ID-T-L(Hr)ru [OCP-Rise & ID-Hr]ADJ OCP-Hr ID-T-R(Lr) *Concave ID-T-L(Lr) [OCP-ML & *T]ADJ [OCP- & ID-MH]ADJ [OCP-Reg & OCP-]ADJ ID-T-R(Hr)ru [*IsRise & *T]ADJ [OCP-Reg & *Rise]ADJ ID-MH *Hr *IsRise [ID-T-L(Hr) & *Fall]SEG *Rise OCP-ML OCP- OCP- OCP-Rise ID-Hr *T OCP-Reg *Fall ID-T-R(Hr) ID-T-L(Hr)
☞a. L-H
[Lr, l]-[Hr, h] * *
? * *
*
?b. L-L
[Lr, l]-[Lr, l] * * * * * *
In (60), ID-T-R(Hr)ru is included. Candidate (b) violates it because the input H in the second syllable is not retained. However, it is not enough to prevent candidate (a) from being eliminated. In (60), candidate (a) is still wrongly ruled out by *Hr due to the retention of the H. In this case, the conjoined constraint, [OCP-Reg & *T]ADJ is posited.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
(61) [OCP-Reg & *T]ADJ:
Assign one violation mark for adjacent tone which violate both OCP-Reg and *T.
[OCP-Reg & *T]ADJ has to dominate *Hr to solve the wrong prediction in (60).
Besides, it has to be dominated by [*IsRise & *T]ADJ so that L-LM, the optimal output of L-HM, will not lose to candidates such as L-ML. The tableau in (62) illustrates how [OCP-Reg & *T]ADJ works.
(62) [OCP-Reg & *T]ADJ rules out L-L L-H → L-H
tsat sat ‘tightly weaved (basket)’
L-H [Hr,-h]-[Hr, hl]
ID-T-L(Hr)ru [OCP-Rise & ID-Hr]ADJ OCP-Hr ID-T-R(Lr) *Concave ID-T-L(Lr) [OCP-ML & *T]ADJ [OCP- & ID-MH]ADJ [OCP-Reg & OCP-]ADJ ID-T-R(Hr)ru [*IsRise & *T]ADJ [OCP-Reg & *T]ADJ [OCP-Reg & *Rise]ADJ ID-MH *Hr *IsRise [ID-T-L(Hr) & *Fall]SEG *Rise OCP-ML OCP- OCP- OCP-Rise ID-Hr *T OCP-Reg *Fall ID-T-R(Hr) ID-T-L(Hr)
☞a. L-H
[Lr, l]-[Hr, h] * * * *
* b. L-L
[Lr, l]-[Lr, l] * *
! * * * * *
In (62), [OCP-Reg & *T]ADJ successfully rules out candidate (b) because both tones in candidate (b) are in the same register and one of them is a citation tone. In
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
this case, the input L-H is finally preserved. The ranking will be revised and the final constraint ranking of construction is illustrated in (63).
(63) The constraint ranking of Construction ID-T-R(Lr) *Concave
ID-T-L(Hr)ru [OCP-Rise & ID-Hr]ADJ ID-T-L(Lr)
[OCP-Reg & OCP- ]ADJ [OCP- & ID-MH]ADJ
[OCP-ML & *T]ADJ
[*IsRise & *T]ADJ ID-T-R(Hr)ru OCP-Hr
[OCP-Reg & *T]ADJ [OCP-Reg & *Rise]ADJ
*Hr ID-MH
*IsRise [ID-T-L(Hr) & *Fall]SEG *Rise
OCP- *T ID-T-L(Hr) *Fall OCP-ML OCP- OCP-Rise OCP-Reg ID-T-R(Hr) ID-Hr
The motivation for tone sandhi in construction is not purely positional as that in construction and in Liujia Raoping Hakka. The relevant tone sandhi that selects optimal output forms is mainly ontextual. [*IsRise & *T]ADJ and OCP-Hr can be regarded as the contextual triggers while *Concave, *Rise and * Hr can be regarded
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
as positional triggers. Conjoined constraints, [OCP-Rise & ID-Hr]ADJ, [OCP-Reg &
OCP- ]ADJ, [OCP- & ID-MH]ADJ, [OCP-Reg & *T]ADJ and [OCP-ML & *T]ADJ are posited to solve the problem where unchanged tonal pairs might be wrongly ruled out.
[OCP-Reg & *Rise]ADJ and [ID-T-L(Hr) & *Fall]SEG are posited to prevent optimal outputs from losing to other candidates.
4.3.3 OT analysis of Tone sandhi in construction and
In sections 4.3.1 and 4.3.2, tone sandhi in both construction and are discussed under the framework of OT. To provide a comprehensive analysis for this language, it is necessary to investigate both tone sandhi mechanisms under the same constraint ranking. To account for different phonological processes resulting from the morphological effects, the and indexed approach is adopted.
In terms of construction , the tone in the second syllable never changes;
therefore, ID-T-R(Hr) and ID-T-R(Lr) are undominated. On the other hand, in construction , the tone in the second syllable is not always retained. Since ID-T-R(Lr) is undominated in both constructions, it needs no specific index and is undominated in the comprehensive ranking. However, ID-T-R(Hr) has different rankings in the two constructions. It needs to be indexed as ID-T-R(Hr) and ID-T-R(Hr) . ID-T-R(Hr) is undominated while ID-T-R(Hr) stays as the lowest ranked as in (63). For the first
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
syllable in both constructions, ID-T-L(Lr) is dominated by *Concave while ID-T-L(Hr) is ranked as the lowest ranked. Due to the same ranking of ID-T-L(Hr) and ID-T-L(Lr) in both constructions, they do not need to be indexed.
ID-t-L should be indexed as ID-t-L because this preservation occurs in construction only. As for markedness constraints, according to Pater (2007, 2010), they can be indexed. *Concave has the same ranking in both constructions. However, the ranking of *Fall is slightly different in both constructions, i.e. *Fall is ranked
ID-t-L should be indexed as ID-t-L because this preservation occurs in construction only. As for markedness constraints, according to Pater (2007, 2010), they can be indexed. *Concave has the same ranking in both constructions. However, the ranking of *Fall is slightly different in both constructions, i.e. *Fall is ranked