CHAPTER 1 INTRODUCTION
1. Introduction
國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
1
CHAPTER 1
INTRODUCTION
1. Introduction
This study investigates the Mandarin focal tone sandhi under the theoretical framework of Optimality Theory (OT). Mandarin third tone sandhi has intrigued many scholars in the previous literature (Cheng, 1973; Shih, 1986; Zhang 1988; Hung 1989, among others). Third tone sandhi is a process in Mandarin Chinese that changes a third tone into a second tone when the third tone is immediately followed by another third tone. However, the interaction between focus and third tone sandhi has not attracted much attention. Shih (1990) and Hsiao (1991, 1995) propose that in Mandarin, the focal boundary is placed at the left edge of the focal element and that the foot formation ends at this boundary. Nevertheless, these studies lack a thorough investigation of the focal tone sandhi in different parts of speech and syntactic structures.
The term focus in this study is used to indicate the phonological and semantic contrast occurring in syntactic elements. In other words, focus indicates interconnections between syntax, phonology, and semantics. Specifically, a focus is contrastive when it explicitly contradicts a set of stated or predicted alternatives.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
When a syntactic element is under focus, the tone sandhi domain and the tone sandhi pattern are different from the regular reading. In other words, the prosodic domain is restructured so that the tone sandhi pattern is changed. Prosodic restructuring due to focus is common among languages and often displays an interface between syntax, phonology, and semantics (Condoravdi, 1990, Kanerva, 1990, Nagahara, 1994).
This thesis examines Mandarin focal tone sandhi based on following questions.
First, since the original studies of Mandarin focal tone sandhi were done roughly two decades ago, there is a question as to whether the findings from studies of the intuitions of youngsters who were born in the 1990s are the same or different from the claims of the operation of Mandarin focal tone sandhi in the previous literature?
Second, what is the focal phrasing pattern in terms of different syntactic structures and parts of speech? Third, how is the focused element aligned in relation to focal phrase? More specifically, is the focused element located at the right or left edge of focal phrase when the focus is in different syntactic structures or on different parts of speech? Fourth, what is the tone sandhi pattern of the focused element? Does the focused element undergo tone sandhi or retain the citation tone? Fifth, under the framework of OT, what are the prosodic constraints, tonal constraints, and the rankings that determine the output form of Mandarin focal tone sandhi?
The organization of this thesis is as follows. Chapter 1 introduces the general
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
background of Mandarin tone sandhi and the definition of focus. Chapter 2 reviews some previous rule-based studies, and some related theoretical background. Chapter 3 surveys the data from Taiwan youngsters and proposes certain generalizations about Mandarin focal tone sandhi. Chapter 4 adopts Optimality Theory and posits prosodic constraints and tonal constraints to account for Mandarin focal tone sandhi. Chapter 5 is the summary of this thesis.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
4
CHAPTER 2
LITERATURE REVIEW
This chapter will review some related theoretical frameworks dealing with the syntax-phonology interface and some recent studies in relation to Mandarin focal tone sandhi. I will also discuss the basic concepts of Optimality Theory, Correspondence Theory, Generalized Alignment, and Cophonology Theory.
2.1 Tonotactics of Mandarin
Mandarin is a tone language. Different tones will change the meaning of the words. There are four tones in the Mandarin tonal system: high level (H), rising (LH), low (L), and falling (HL). In particular, the low tone (L) is treated as the phonological representation of MLH (Hsiao 1991; Lin 2000). The Mandarin third tone sandhi rule and the application of the rule are illustrated below:
(1) Mandarin Tone Sandhi Rule: L → LH / __ L lao ban ‘bosses’
L L base tone
LH L surface tone sandhi pattern
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
This example shows that the first syllable lao, which carries a L tone, changes to a LH tone, when it is followed by another L tone. The last syllable ban does not undergo tone sandhi and retains its base tone in surface.
2.2 Syntax-Phonology Interface
In the literature, there are two main approaches to the mapping between syntax and phonology: the Direct Reference Hypothesis and the Indirect Reference Hypothesis. The Direct Reference Hypothesis claims that phonological rules are sensitive to syntactic structures, and that phonological rule domains are determined by syntactic surface structures (Kaisse, 1985; Odden 1987, among others). There is no intermediate level between syntax and phonology.
However, the Indirect Reference Hypothesis argues that phonological rules do not operate directly on syntactic structures. Instead, they operate on the intermediate level between syntax and phonology, namely, prosodic structures (Selkirk, 1984;
Nespor and Vogel, 1986; Shih, 1986; Hung, 1987; Hsiao, 1991, 1995, among others).
Prosodic structures are constructed based on syntactic information and serve as the domains for phonological rules, as illustrated in (2).
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
(2) Indirect Reference Hypothesis
There is a separate prosodic hierarchy, parallelling syntactic structure. Its main components include utterance, intonational phrase, phonological phrase, clitic group, phonological word, foot, syllable, and mora, as in (3).
(3) Prosodic Hierarchy U (utterance)
I (intonational phrase) φ (phonological phrase) C (clitic group)
ω (phonological word)
F (foot) σ (syllable)
µ (mora)
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
The hierarchy is governed by the Strict Layer Hypothesis. The Strict Layer Hypothesis determines the geometry of this constituent structure, as follows:
(4) Strict Layer Hypothesis (Selkirk 1984; Nespor & Vogel 1986)
a. A given non-terminal unit is composed of one or more units of the immediately lower category.
b. A unit of a given level is exhaustively contained in the superordinate unit of which it is a part.
2.3 Previous Studies
2.3.1 Zhang’s (1988) and Shih’s (1990) Analyses
Shih (1986) convincingly demonstrated that foot is not formed by surface syntactic structure, but it is a mediating prosodic structure partially conditioned by syntactic structure. Shih (1986) modified Chen’s (1984) rules, which were originally designed for scanning Chinese poetry and verse:
(5) Foot Formation Rules
a. Immediate Constituency (IC): Link immediate constituents into disyllabic feet.
b. Duple Meter (DM): Scanning from left to right, string together unpaired
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
syllables into binary feet, unless they branch in the opposite direction.
c. Superfoot (f’): Join any leftover monosyllable to a neighboring binary foot according to the direction of the syntactic branching.
Zhang (1988), based on Shih’s Foot Formation Rules, posits that foot formation starts with this accented syllable so that restructured foot domains would derive tone patterns different from the non-focal renditions. Example (6) is given for illustration.
(6) ‘I want to buy small cats’
wo xiang mai xiao hua mao I wish buy small flower cat IC DM Superfoot
L L L L H H base tone pattern L LH L L H H sandhi tone pattern
In (6), the DM starts scanning from the accented syllable xiang rightward, in order to string xiang and mai, instead of wo and xiang, into a foot. Then, wo joins the existing foot to form a trisyllabic foot to which tone sandhi is applied, and hence the low tone of wo remains invariant, unlike the non-focal reading is (LH.(LH.L))(L.(H.H)).
‧
Shih (1990), on the other hand, posits an ‘emphatic boundary’ before the focal element, suggesting that foot formation ends at the boundary ($ = emphatic boundary).
In (7a), zhi is the focal element, and thus the preceding syllables lao and li form a foot.
‧
In addition, xiang in (7b) is the focal element, and an emphatic boundary is placed before it. As a result, zhi must join lao li to form a foot, and retains its base tone.
However, the parameters proposed by Zhang (1988) and Shih (1990) fail to make a correct prediction of focal phrasing in flat structures. Consider the example of the flat structure in (8) where the focus is on the central syllable.
(8) ‘55955’
As in (8), Zhang and Shih would decide that the foot formation starts at the focused element jui. As a consequence, the focused element jui inevitable carries a sandhi tone instead of a base tone, but the output is ill-formed. In fact, in a flat structure, the focused element stands at the right edge of the focal phrase, and thereby the foot formation ends at the focused element. To account for this problem, I will propose a right-edge constraint in chapter 4 to explain the phrasing of the focal tone sandhi in a flat structure.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
2.3.2 Hsiao (1991, 1995)
In order to capture the psychological reality of the native intuition, Hsiao (1991, 1995) proposes a beat-counting device, showing that the structure of the foot should be built upon metrical beats, which mediate between the foot and the syllable. Lexical syllables and functor syllables are distinguished in his beat-counting theory:
(9) Beat Assignment
a. Lexical Beat-Assignment (LB): Every lexical syllable is assigned a metrical beat.
b. Functor Beat-Assignment (FB): A functor syllable is assigned a beat in normal or slow speech, behaving like a lexical syllable, and is left-adjoined to the nearest beat in fast speech.
c. FB takes place after lexical syllables have been made into ICFs or ABFs.
(10) Foot Formation Revisited
a. Immediate Constituent Foot (ICF): Any adjacent beats which are assigned to ICs form an ICF.
b. Adjacent Beat Foot (ABF): Any two adjacent beats which are not assigned to
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
ICs are paired into an ABF.
c. Jumbo Foot (JF): Any unpaired single beat is recruited by a neighboring foot to from a Jumbo Foot if the beat c-commands the adjacent beat contained in the foot.
d. Minifoot (MF): The leftmost single beat constitutes a Minifoot iff it is followed by an intonational phrase boundary %.
(11) Application Criteria
a. When all conditions are met, the ICF prevails over other footing processes.
b. Scanning starts from left to right and stops as soon as the environmental requirements for either MF, ABF or JF are met, and the principle triggered applies to the whole line.
c. Footing must not cross any intonational phrase boundary.
To account for tone sandhi in relation to focus, Hsiao proposes the principle of focal phrasing:
(12) Focal Phrasing (FB): A focal boundary $ is placed before a focused lexical syllable iff it is outside of an ICF or a JF.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
The general assumption is that foot construction cannot operate across the focal boundary $, which, however, can not be placed within an ICF or a JF. That is, a functor syllable does not have the ability to take the focal boundary $. ABF does not resist a focal boundary as ICF and JF do. This is because ABF usually renders a metrical tension, and hence is more breakable. The following example shows the application of focal phrasing.
(13) ‘A small horse is walking.’
xiao $ma zou lu
x x x x Lexical Beat ICF
*MF
*JF
L L L HL base tone pattern L LH L HL *sandhi tone pattern
When ma in (13) is focused, $ is placed before it, forcing xiao to emerge as a Minifoot. The focal ma then joins the ICF to form a Jumbo Foot. As a consequence,
ma inevitably carries a LH tone, and the derived reading is ill-formed. The placement
‧
of a focal boundary $ is prevented after the first ICF, as dictated by the Focal Phrasing Principle.
Nevertheless, when focal phrasing occurs in the marked reading of Type C1, Hsiao’s focal phrasing is unable to make a correct prediction, as illustrated in (14).
(14) ‘Dogs bite Small Bao.’
The focal boundary is required to be placed before the focused element guo since it is a lexical syllable. As a result, guo undergoes tone sandhi and changes to a LH tone.
Yet, this is not the correct pattern because in a marked reading, the focused element usually retains the base tone. Therefore, this study will propose a right-edge parameter
1 As Chapter 3 will show, there are two alternative focal phrasing patterns among youngsters when focal phrasing occurs in Type C. One is common and unmarked while the other is less frequent and marked.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
to account for the focal phraing patterns in Mandarin.
The foot formation should be stopped before the boundary. The boundary itself does not reveal any phonetic cues, such as the change of pitch or duration. By using the term focal phrase, there is no need to further point out the phonetic properties once the phrase is under focus.
2.3.3 Cho (1990) & Kanerva (1990)
Cho (1990) examines the relationship of syntax and phrasing in Korean. She discovered that the semantic notion of focus, together with its phonological realization as high pitch, plays a crucial role in phrasing. The focus phrasing data in the study show that the determination of prosodic structure is not entirely dependent on syntactic notions. Consider how the presence of focus affects phrasing in (15).
(15) a. nuka ‘who(Nom)’ [+high]
[núga gayo] ‘Who is going?’
who goes
b. [Súni-ga gayo] ‘Suni is going.’ (in answer to (a))
c. [Suni-ga] [kayo] ‘Suni is going.’ (normal declarative sentence) Suni-Nom goes
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
As shown in (15b), the accented subject is phrased with the non-branching verb phrase. It is necessary to include one more clause in phonological phrasing of Korean.
That is, phrase any focused word with the next word, unless that word is already phrased.
Kanerva (1990) investigated the focus in relation to phonological phrasing in Chichewa, a Bantu language spoken in East Central Africa. The major finding in his study is that Chichewa phonological phrases are strongly dependent on focus, and are generally larger and more variable with respect to syntax than what is reported for phonological phrases in other languages. In light of this, the Chichewa domains could represent a truly distinct pattern of prosodic organization since no phonological phrase algorithms hold up under scrutiny of focus. Four focal phrasings of VP are given below. Focus is indicated by bold typeface.
(16) a. (VP) A-namenya nyumba ndi mwala.
‘pro-hit the house with a rock’
b. (VP) (Anaményá nyumbá ndí mwáála) c. (V OBJ OBL) (Anaményá nyumbá ndí mwáála) d. (V OBJ) (OBL) (Anaményá nyumbá) (ndí mwáála) e. (V) (OBJ) (OBL) (Anaményá) (nyumbá) (ndí mwáála)
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
The phrasing pattern above is that when there is a focus inside the VP, a domain starts at the verb and ends at the focused constituent and any non-focus constituents each form their own domain.
Collectively, the phonological phrase algorithms of Nespor and Vogel (1982), Mchugh (1986), Selkirk (1986), and Chen (1987) expectedly allow (16d) and (16e).
However, they fail to derive (16b) and (16c). Therefore, Kanerva proposes that Chichewa domains instantiate a previously unrecognized level of prosodic structure, Focal Phrase (FP), and that this level is intermediate between the phonological phrase and the intonational phrase.
In summary, the studies of Cho (1990) and Kanerva (1990) have shown a problem that the focal phrase of Chichewa and Korean is generally larger than the phonological phrase but smaller than the intonational phrase. This problem could be resolved by two ways. One way is to include one more focus rule in the phonological phrase formation, as Cho (1990) suggests. The other is to posit a new distinct level between phonological phrase and intonational phrase, as Kanerva (1990) proposes.
This thesis will examine the size of the Mandarin focal phrase as to whether it matches the phonological phrase, intonational phrase, or whether it is a truly distinctive pattern of prosodic organization.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
2.4 Theoretical Background
2.4.1 Optimality Theory
Optimality Theory (OT) is a constraint-based framework which is proposed by Prince and Smolensky (1993/2004). Unlike traditional transformations, OT disallows serial derivations. The operation in OT mainly involves two systems: Generator (GEN) and Evaluator (EVAL). An OT grammar can be schematically represented as in (17).
For every possible input, the generator (GEN) produces a candidate set. Inputs are in principle unconstrained linguistic objects such as lexical items in word phonology.
(17) OT Schema
A candidate set contains output structures. These structures are possible analyses of the input. According to the principle of inclusiveness, GEN produces all of the analyses of the input that ‘are admitted by very general considerations of structural
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
well-formedness’ (McCarthy and Prince 1993), which could include universal properties. The evaluator (EVAL) evaluates candidate sets with respect to particular rankings of the constraint inventory Con. It is often assumed that Con is universal. Its members are simple and conflicting statements about the form of the output or the relation between the output and the input. Because of the conflict between constraints, all conceivable linguistic structures will be assumed to violate at least some of the constraints.
(18) Constraint A >> Constraint B >> {Constraint C, Constraint D}
/Input/ Constraint A Constraint B Constraint C Constraint D
☞Candidate (a) *
Candidate (b) *! *
Candidate (c) *!
The operation of OT can be demonstrated with the tableau as shown in (18). In the left column, the input is placed at the top left cell and the other columns show the candidates generated from the input by GEN. In the right columns, the constraints of interest are each named in separate columns. The ranking constraints are ranked lower from left to right. Thus, constraint A is ranked higher than constraint B, followed by
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
constraint C in a certain language. The finger marker indicates the optimal output of the evaluation while the reverse one indicates the wrong prediction of the optimal output. The solid line distinguishes the priority accorded to one constraint over another. The dotted line between two constraints shows that the ranking is unknown.
The asterisk denotes a violation and the exclamation mark represents a fatal violation that rules out a losing candidate. Shaded areas in the tableau indicate that the constraints in the shaded areas should no longer be considered since a higher ranked constraint has already been violated.
As can be seen in (18), Candidate (b) is ruled out because Candidate (b) violates Constraint A, the highest ranked constraint. Candidate (c) violates Constraint B, and Candidate (a) violates Constraint C. In this tableau, Constraint B is ranked higher than Constraint C, so Candidate (c) is ruled out. Candidate (a) is the optimal output which is indicated by a finger marker ☞.
2.4.2 Correspondence Theory
McCarthy and Prince (1995) enriched the OT grammar with Correspondence Theory, which defines faithfulness constraints as evaluating the individual elements of the input and the output so that constraints must assess the correspondence and identity of the correspondent elements. Elaborated in many subsequent researches (Ito,
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
Kitagawa and Mester 1996, Kager 1999, Nelson 2003), this theory posits the correspondence between the output base and the reduplicant. In Benua (1997), the Basic Model was developed into Transderivational Correspondence Theory, in which identity relation holds between two surface words. The correspondence model can be summarized as (19).
(19) Output-to-output Correspondence
Input
IO-correspondence
Base Output
OO-correspondenceEach variable dimension of the representation is governed separately, by a separate faithfulness constraint. The constraints in (20) demand complete and exclusive correspondence between strings.
(20) a. MAX: Every segment in S1 has a correspondent in S2.
b. DEP: Every segment in S2 has a correspondent in S1.
c. IDENT[F]: Correspondent segments are identical with respect to feature F.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
In this study, the IDENT constraint will be adopted to constrain not only the correspondence between input-to-output tones but also the correspondence between base-to-output tones in Mandarin focal tone sandhi.
2.4.3 Prosodic Correspondence in Tone Sandhi
Drawing evidence from tone sandhi in Mandarin, Lin (2005) argues that the correspondence relation can be extended to stand between outputs that are related in the prosodic structures. Consider the following examples.
(21) a. mai ba san ‘buy an umbrella’ a’ {σ{σσ}}
Input: L.L.L → Output: LH.LH.L ((LH.LH).L) b. ya er ma ‘Armagh’ b’ {σ{σσ}}
Input: L.L.L → Output: LH.LH.L ((LH.LH).L)
In (21a) and (21b), while the morphosyntactic structure is right branching {σ{σσ}}, the tonal domain is left branching ((σσ)σ). In the above examples, the tone sandhi
domains cannot be morphosyntactically defined. Thus, Lin proposes a correspondence model for tone sandhi, as in (22).
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
(22) Output-to-output Correspondence for Tone Sandhi
Input Tone Input Tone
Tb.Tc Ta.Tb.Tc
IO-Faith
(Tb’.Tc’) (Ta”.(Tb”.Tc”
Base Tone Output Tone
))BOT-IDENTITY (…) = prosodic structure
In this correspondence model, two correspondence relations are involved, an input-to-output relation and a base-to-output relation. Unlike the transderivational model proposed in Benua (1997), the two tonal outputs are related by the prosodic structures rather than by the morphosyntactic structures. Two freestanding tonal outputs that are compositionally related are governed by the base-tone-to-output-tone correspondence. For prosodic structures such as ((σσ)1σ)2 and (σ(σσ)1)2, the tonal information in prosodic constituent 1 would serve as the base for constituent 2. The bi-tonal sequence in the inner domain of the tonal output is evaluated with a bi-tonal base with which it shares the same underlying tones. In (22), the base is Tb’.Tc’ and the reference output is Tb”.Tc”. The tonal sequence Tb”.Tc” within the inner prosodic constituent of (Ta”.(Tb”.Tc”)) is evaluated with Tb’.Tc’ for correspondence.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
2.4.4 Generalized Alignment
Developed from the edge theory of Selkirk (1986), and Inkelas (1989), McCarthy and Prince (1993) propose the idea of Generalized Alignment, in which the phonological representation of a sentence is constructed through the mapping of the
Developed from the edge theory of Selkirk (1986), and Inkelas (1989), McCarthy and Prince (1993) propose the idea of Generalized Alignment, in which the phonological representation of a sentence is constructed through the mapping of the