2. Literature Review
2.2 Derivation of Tone Sandhi Patterns
國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
2.2 Derivation of Tone Sandhi Patterns
To investigate the derivation of TS patterns, it is necessary to define the domain to which TS applies in the first place. Shih (1986) proposes a set of FFR, which is
captured in (5) below.
(5) Foot Formation Rule (Shih, 1986, p.110) I. Foot (f) Construction
a. IC (Immediate Constituency): Link immediate constituents into disyllabic feet.
b. DM (Duple Meter): Scanning from left to right, string together unpaired syllables into binary feet, unless they branch to the opposite direction.
II. Super-foot (f’) Construction
Join any leftover monosyllable to a neighboring binary foot according to the direction of syntactic branching.
Her foot-based definition of the TS domain leads the following researchers of Mandarin TS to a new direction of analysis that does not regard syntax as the sole reference for the formation of TS domain. However, her analysis lacks in explaining the foot formation of certain phrases and function words. First of all, for a sequence of third tones with a hierarchical syntactic structure like (6), foot formation seems to be a post-lexical rule, since the foot boundary cannot sit between gou3 “dog” and bing3 gan1 “cookies”; otherwise, it will affect the internal structure of the lexical item gou3
bing3 gan1 “dog’s cookies” and renders a different meaning as in (6a). On the other
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
hand, for a sequence of third tones with a flat structure like (7), foot formation can either apply within a lexical item, rendering it a pre-lexical process.
(6) “Soft dog’s cookie”
ruan3 gou3 bing3 gan1 soft dog cookie
a. *(2 3) (3 1)
b. (3 (2 3 1))
c. (2 2 3 1)
(7) “Somalia”
suo3 ma3 li3 ya3 (2 3) (2 3)
Shih (1986) deems it unnecessary to distinguish pre- from post-lexical foot formation in this matter because FFR applies in the same way, only that sentences with hierarchical syntactic structures should be restricted by the Lexical Integrity Principle, so that foot formation cannot parse (6) into two feet. In (7), since the phrase
is considered as having a flat syntactic structure, the syllables are formed into disyllabic feet according to DM.
Nevertheless, Hsiao (1991) reveals the problem of the reliance on the Lexical Integrity Principle by giving hierarchically structured PhPs that can be parsed into
two feet, as in (8). He proposes the Phonological Integrity Constraint that forbids foot
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
formation from applying to PhPs with uniformly left- or right-branching structures as in (6).
(8) “excellent character”
you1 liang2 pin3 de2
(1 2) (3 2)
In fact, another solution to the problem that arose from the contrast of foot parsing between (6) and (8) is to abandon these stipulations of the integrity principles and treat bing3 gan1 “cookies”, you1 liang2 “excellent” and pin3 de2 “character” as PhWds, and to hypothesize a PhWd level on which TS applies, prior to the PhP and the IP level, which will be introduced in 4.1.1.
As mentioned briefly before the integrity problems, Shih (1986) observes that foot formation can be said to operate both in the pre-lexical system and the
post-lexical system. This reflects the foot’s different nature from prosodic categories like the PhWd, the PhP and the IP, which only exist in the post-lexical system. Hsiao (1991) in fact treats the foot and the syllable as metrical categories, as opposed to the other categories like the PhP and the IP, which are termed prosodic categories. In doing so, Hsiao (1991) hypothesizes that the metrical foot and the syllable, the PhP and the IP should respectively belong to the metrical hierarchy and the prosodic
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
hierarchy; and that TS “applies to the domain of the foot, which is cyclically formed on prosodic constituents”, namely, the PhP, the IP etc. (p.97). As it turns out in 4.1.4,
this view proves to be crucial for the definition of the intonational phrase.
Another problem in Shih (1986)’s analysis is the foot formation of prepositions (e.g. bi3 “than”, wang3 “toward”) and classifiers (e.g. dian3 “some”). For sentences like (9), Shih (1986) suggests not to apply FFR-IC to the prepositional phrase (bi3 gou3 “than dogs”) with a full noun phrase (gou3 “dog”), so that the sentence can be
parsed into two feet by FFR-DM.
(9) “Cats are smaller than dogs.” (NA refers to non-application; UR refers to underlying tones; ST refers to surface tones.)
mao1 bi3 gou3 xiao3
UR 1 3 3 3
FFR-IC: NA 1 3 3 3
FFR-DM (1 3) (2 3)
ST 1 3 2 3
On the other hand, Hung (1987) views Mandarin prepositions (e.g. wang3
“toward” and bi3 “than”) as co-verbs, treating them as lexical words like verbs,
which may take arguments or modifiers. This view is also supported by Lin (2006), who holds that prepositions in Chinese are developed from the grammaticalization of verbs, so they still retain verbal characteristics such as taking aspectual markers and
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
functioning as the main predicate of the sentence. Under this assumption, Hung (1987) adopts Selkirk (1984)’s Sense Unit Condition (SUC), which is specified in (10), and he suggests a modification of Shih (1986)’s definition of FFR as shown in (11), in order to derive all the possible TS patterns with the prepositional phrases in (12) and (13). The round brackets indicate the boundaries of sense units. The prepositions are boldfaced.
(10) Sense Unit Condition (Selkirk, 1984, p.291)
Two constituents Ci, Cj form a sense unit if (a) or (b) is true of the semantic interpretation of the sentence:
a. Ci modifies Cj (a head).
b. Ci is an argument of Cj (a head).
(11) Hung (1987)’s revision of FFR
(a) Form binary-branching feet by linking syllables into disyllabic sense units wherever possible;
(b) For all remaining syllables, form trisyllabic or quadrisyllabic feet by linking them to the above to form larger sense units.
(12) “Cats are smaller than dogs.”
mao1 bi2 gou3 xiao3
UR 1 3 3 3
a. ST1 (1 ((2 2) 3)) b. ST2 ((1 3) (2 3))
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
(13) “Dogs walk toward the north.”
gou3 wang3 bei3 zou3
UR 3 3 3 3
a. ST1 (3 ((2 2) 3)) b. ST2 ((2 3) (2 3))
In (12) and (13), since prepositions (i.e. bi3 “than” and wang3 “toward”) are treated as co-verbs, they can form sense units with either their subject arguments (i.e.
mao1 “cats” in (12) or gou3 “dogs” in (13)) or their complements (i.e. gou3 “dogs” in
(12) and bei3 “north” in (13)) based on their head-argument relations required by the SUC. In (12a) and (13a), quadrisyllabic sense units are formed after the disyllabic sense units (i.e. bi3 gou3 “than dogs” and wang3 bei3 “toward the north”) are formed.
However, in (12b) and (13b), the last two syllables gou3 xiao3 “dogs, small” and bei3 zou3 “north, walk” still form respective sense units even though how these
constituents satisfy the SUC is questionable. Gou3 “dogs” and xiao3 “small” in (12b) and bei3 “north” and zou3 “walk” in (13b) do not share full modifier-modified
relations, because, strictly speaking, xiao3 “small” in (12) is modified by the entire PP (bi3 gou3 “than dogs”), rather than just the noun (gou3 “dogs”); and zou3 “walk” in (13) is modified by the entire PP (wang3 bei3 “toward the north”), not just the noun (bei3 “north”). Therefore, whether combining the SUC and the FFR is enough to explain the formation of TS domain remains debatable.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
The classifier problem left by Shih (1986) also remains unsolved in Hung (1987), because the SUC cannot explain the adjoining of classifiers with the verbs on the left, instead of the noun phrases they specify. Furthermore, this approach alone cannot account for the foot formation of multisyllabic words as in (7), because each lexical item is a sense unit and it cannot be analyzed into smaller sense units to serve as the TS domain. As a result, we find that the formation of D cannot fully rely upon the SUC, although it may describe the D formation of prepositional phrases in Mandarin.
For sentences with classifiers as in (14), on the one hand, Shih (1986) resorts to cliticization and treats the verb and the classifier as one clitic word to form a foot
first, and the remaining syllable shui3 “water” forms a super-foot with gei3 dian3
“give some” afterwards.
(14) “give some water”
On the other hand, Hsiao (1991) provides a different solution to this problem. He posits a set of Beat Assignment and Foot Formation Rules represented in (15) to (17).
gei3 dian3 shui3
UT 3 3 3
((2 3) 3)
ST (2 2 3)
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
Different from Shih (1986), he assigns a metrical beat to each syllable before foot formation. In other words, in his study, foot formation groups beats, rather than syllables, into feet. Moreover, he distinguishes function words from content words by stipulating that the beat assignment and foot formation of function words (i.e. functor syllables) only take place after lexical beats are assigned and lexical feet are formed, as in (15c). He also specifies that immediate constituent foot (ICF) must be formed before adjacent beat foot (ABF), jumbo foot (JF) or minifoot (MF) are formed, as in (17a)
(15) Beat Assignment (Hsiao, 1991, p.37-38)
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 and ABFs.
(16) Foot Formation
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 ICs are paired into an ABF.
c. Jumbo Foot (JF): Any unpaired single beat is recruited by a neighboring foot to form a Jumbo foot if the beat c-commands the adjacent beat contained in the foot.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
d. Minifoot (MF): The leftmost single beat constitutes a Minifoot if and only if it is followed by an intonational phrase boundary %.
(17) 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.
In addition, Hsiao (1991) assumes that the metrical categories and the prosodic categories belong to separate hierarchies, and foot formation operates cyclically within the phonological phrase and the intonational phrase. He states that a
phonological phrase boundary is marked on the left edge of a branching Xmax. The intonational phrase is constructed by one or more phonological phrases that share either head-argument or modifier-modified relations with each other as required by the SUC.
By stipulating that functor beat assignment takes place after the foot formation of lexical syllables, he provides another solution to the classifier problem, as shown in (18), and the preposition problem, as shown in (19). Nonetheless, as will be explained in the following, the phonological phrasings of these two function words differ from each other in his analysis.
‧
In (18), through left-cliticization in phonological phrasing, the classifier dian3
“some” is left-adjoined to form a PhP with the verb mai3 “buy”. After that, since mai3
“buy” and dian3 “some” cannot form an IP by themselves due to the violation of the
SUC, they must join jiu3 “wine” to form an IP.
(18) “buy some wine”
Phonological Phrasing [mai3 dian3] [jiu3]
Intonational Phrasing [ ]
In contrast to the classifiers, which can only cliticize to the left, prepositions optionally undergoes left-cliticization. In (19a), wang3 “toward” cliticizes to its left to form a phonological phrase, but it does not undergo cliticization in (19b). This
produces different foot domains of TS and different tone patterns.
(19) “The dog runs to the north.”
a.
Phonological Phrasing [gou3 wang3] [bei3 pao3]
Possible Intonational Phrasing (I) [ ]
UT 3 3 3 3
‧
In the phonological phrasing of the classifier dian3 “some”, cliticization is left-adjoined to the neighboring phonological phrase, as in (18). In that of the
preposition, cliticization is, however, optional, as in (19). Hence, Hsiao (1991) makes a further stipulation, as in (20), to distinguish between the cliticization of these two function words.
(20) Cliticization (Hsiao, 1991, p.95)
a. A functor which is a non-head of an Xmax must be cliticized to the left of the preceding phonological boundary.
b. Cliticization is optional for Xhead functors.
However, as Chapter 4 develops, this stipulation seems to be dispensable if we consider prepositions to be co-verbs or lexical words as Hung (1987) and Lin (2006) do.
Despite the fact that Hsiao (1991) provides a possible solution to the preposition and the classifier problem, his model still lacks in predicting the variations collected
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
in the corpus of this thesis, as shown in (21). The boundaries of the PhP are marked on the left edges of each branching Xmax, and a possible IP break can sit between Lao3 Li3 “Old Li” and zhao3 “find”. Foot formation first forms ICFs (Lao3 Li3 “Old Li”,
zong3 tong3 “president”, pao3 bu4 “run”) and then ABFs (zhao3 Ma3 “find Ma”). If
TS does not operate across feet, the surface TS pattern will be ST1. If TS operates across feet, the surface TS pattern will be ST2. However, both predicted TS patterns do not identify with the data collected from two informants (abbreviated as inf.) in the corpus of the present thesis. Therefore, a model that is able to predict more variations is needed.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
(21) “Old Li finds President Ma and asks him for a run.”
PhP [[Lao3 Li3 [[zhao3 Ma3 zong3 tong3 pao3 bu4
IP [ ] [ ]
ICF [(x x)] [x x (x x) (x x)]
ABF [(x x)] [(x x) (x x) (x x)]
ST1 2 3 2 3 2 3 3 4
Cross-feet TS 2 3 2 3 2 2 3 4
ST2 2 3 2 3 2 2 3 4
Inf. 1 *2 2 3 3 2 3 3 4
Inf. 2 *2 2 2 3 2 2 3 4
The findings and analysis in Chapter 4 are divided into two main parts. The first part is about the domain of TS application. As explained in this chapter, the formation of D under the IP level is still open to debate (the foot in Shih (1986) and Hsiao (1991); the phonological word, the phonological phrase and the intonational phrase in Cheng (1987) etc.), which makes it difficult to define the boundaries of IP. Hence, the following analysis examines another model of representation, the Prosodic Hierarchy, as developed by Selkirk (1984), Nespor & Vogel (1986), Inkelas (1989), Bickmore (1989) among others, and tries to provide a unified definition of the prosodic hierarchy and, most important of all, a clearer definition of the IP.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
The prosodic hierarchy has different levels of prosodic constituents that are organized into a hierarchy different from the syntactic one. Nespor & Vogel (1986) proposes a complete prosodic hierarchy as shown in (22). The levels in the prosodic hierarchy all meet the following two requirements: (a) Exhaustive Parsing and (b) Strict Layering. Simply put, exhaustive parsing refers to the regulation that every
element at each level must be parsed into some kind of prosodic domain; and strict layering designates that the prosodic domain at one level must contain at least one or more prosodic domains formed at its lower level.
(22) The prosodic hierarchy a. the syllable
b. the foot
c. the phonological word d. the clitic group
e. the phonological phrase f. the intonational phrase g. the phonological utterance
The present thesis investigates the levels involving the process of Taiwan Mandarin TS derivation and whether the definition of the prosodic categories in Nespor & Vogel (1986) can be the domain of TS. Through the observation of the TS patterns in the corpus, four levels of the prosodic hierarchy are assumed to be relevant:
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
the phonological word (PhWd), the clitic group (C), the phonological phrase (PhP) and the intonational phrase (IP). The primary goal of this thesis is to examine how TS cyclically applies to these levels and to adjust the construction and reconstruction rule of these prosodic categories in Nespor & Vogel (1986) whenever necessary.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
24 Chapter 3 Data Collection
3.1 Data Collection
As discussed in the previous chapter, the faster the speech rate, the wider the D.
In Shih (1986), it is presumed that D is the IP at the presto speech rate. Cheng (1987) and Hsiao (1991) also support the view that TS applies cyclically up to the IP, or simultaneously to the IP at a relatively fast speech rate. Therefore, this thesis constructed a corpus of third tone sequences recorded from four Taiwan Mandarin native speakers (two females and two males, three of them at the age of 25, one of them 27). The author presumed that TS is more likely to apply to the IP level at a speech rate faster that the presto, which, grounded on the musical tempo, is 168 to 200 beats per minute. Hence, the four informants were asked to read sentences made of sequences of third tones at a speech rate of above 200 beats (or words) per minute, which is termed as the prestissimo speech rate. The recording process depended on the tempo of a metronome to control the speech rate of the informants.
3.2 Introduction to the Corpus
The corpus comprises 324 sentence tokens (81 sentences*4 speakers) containing 6 to 24 syllables and 49 types of syntactic structures, which are shown in (23) and (24) (see Appendix for a full list of tokens for each sentence structure).
‧
(23) Number of sentences of different lengths
Number of syllables Number of sentential tokens
6 32
(24) Syntactic structures and sentence tokens in the corpus
Types of sentential structures No. of
tokens
‧
‧
This thesis adopts the prosodic hierarchy proposed by Nespor & Vogel (1986) as in (22), assuming that TS cyclically applies to the Ds on the following four levels: the PhWd, the C, the PhP and the IP level. First of all, the phonological word here refers to the prosodic unit that is “coextensive with the terminal node of the syntactic tree”
(Nespor & Vogel, 1986, p.116). In other words, one PhWd occupies one terminal node of the syntactic tree, as exemplified in (25). They are marked as the lexical categories like N, V, ADJ, ADV and P etc. in the corpus. We can also say that a PhWd is a stem.
Therefore, affixes like -li3 “inside” and de (possessive suffix) etc. do not form independent PhWds. They must be attached to other stems to form PhWds. In (25), the PhWds are xiao3 “small”, lao3 shu3 “mouse”, yao3 “bite” and lao3 hu3 “tiger”.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
(25) “A small mouse bites a small tiger.”
xiao3 lao3 shu3 yao3 xiao3 lao3 hu3
The corpus data indicate that TS still tends to start applying cyclically from the PhWd level. This is observed in syntactic binary-branching phrases in which the left terminal nodes end with the third tone and the right terminal nodes begin with at least two successive third tone syllables, as in (26). If TS starts on the PhWd level, TS for
the final syllable on the left branch will be blocked, producing a tone pattern such as
“3 2 3”; if TS skips the PhWd level and applies to a higher level, a tone pattern as “2
2 3” will be produced. The total number of these binary branching structures is 624.
(26)
UT (…3) (3 3…)
Li3 xiao3 jie3 “Miss Li”
xiao3 lao3 shu3 “small mouse”
hao3 hao3 zou3 lu4 “walk properly”
mai3 bao3 yang3 pin3 “buy skin care products”
As for the clitic group (C) level, the clitics in the corpus include pronouns and classifiers. A clitic group is assumed to contain a nonclitic word plus, if there are any, one or more clitic words. It should be noted that this thesis assumes Hung (1987) and Lin (2006)’s hypothesis that prepositions are co-verbs, which can be assumed to be
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
lexical words rather than function words; thus, the phrasing of prepositions is not to be considered on the clitic group level.
The total number of clitic tokens in the corpus is 20, as shown in (27). On this level, the analysis examines how a clitic word cliticizes to a nonclitic word to form a D, and how forming a D on the clitic group level can account for the TS patterns in the corpus. For every sequence of a non-clitic word plus a clitic word, if the last syllable of the left adjacent non-clitic word, such as gei3 “give” or wu3 “five” in (28a) and (28b), carries a sandhi tone, the clitic shows left-cliticization. Right-cliticization is
shown (29), where the pronoun wo3 “I” is cliticized to the adjacent verb xiang3
“want” on its right. The clitics are boldfaced in (28) and (29).
(27) Clitic tokens in the corpus Pronouns Classifiers Total
12 8 20
(28) Left-cliticization of clitics a. “give some water”
gei3 dian3 shui3
UR 3 3 3
PhWd-D (3) (2) (3)
C-D (2 3) (3)
‧
b. “five kinds of antipruritic powder”
wu3 zhong3 zhi3 yang3 fen3
After the C level is the PhP level. A PhP, as defined by Nespor & Vogel (1986), includes a head X and all elements on its nonrecursive side within Xmax. In syntax, the recursive side refers to “the side of the head on which internal arguments or
complements in the X’-theoretic sense occur” (Giorgi and Longobardi, 1991, p.98).
The nonrecursive side refers to the opposite side of the recursive side. For instance, if the complement of a verb head is located on the right side of the head, the right side is the recursive side, and the left side is the nonrecursive side.
Furthermore, they propose a PhP reconstruction rule that says, besides the inclusion of all the Cs on its nonrecursive side within the maximal projection, the
Furthermore, they propose a PhP reconstruction rule that says, besides the inclusion of all the Cs on its nonrecursive side within the maximal projection, the