Chapter 2

Copy and paste this link to your website, so they can see this document directly without any plugins.


that, syllable, which, with, will, syllables, vowel, Chinese, there, vowels, also, have, only, initial, discussed, glottal, surface, consonants, nasal, stop, voiced, examples, syllabic, chapter, this, This, because, from, voiceless, other


2 The Consonants and Vowels of Shaoxing:
Surface Representations
2.1 Introduction
This chapter provides an analytic description of the surface consonants
and vowels of Shaoxing Chinese (henceforth SX). Before analyzing the
distinct phonemes and their phonological features in chapter 3, I will
describe the overall numbers and forms of possible initials and finals of
SX. I use the terms “initials” and “finals”, which correspond to “shengmu” and “yunmu” in traditional Chinese accounts of syllable structure.
Wang & Smith (1997: 7) introduce the traditional way of representing
Chinese syllable structure, given here in the tree diagram in (1a):
(1) a. syllable b. σ
shengmu yunmu Initial Final
yuntou yunfu yunwei Glide Nucleus Coda
The diagram in (1a) shows that the Chinese syllable structure has two
main constituents: “shengmu”, corresponding to what current phonological theory would call onset, and “yunmu” (which literally means
‘rhyme’), referring to all that follows “shengmu”. “Yunmu”, in turn,
consists of three parts: “yuntou”, which is the position reserved for
prenuclear glides, “yunfu”, which is the nucleus, and “yunwei”, which is
the coda. The diagram in (1b) expresses the traditional representation of
Chinese syllable structure in Western terminology. In traditional Chinese
phonology,1 “yunmu” in SX (as well as in Mandarin and all other Chinese
dialects) is not equivalent to the syllable rhyme, because the Chinese
“yunmu” does not count in the poetic rhyming system, which only
includes “yunfu” and “yunwei”, i.e. it excludes “yuntou”. Thus, we cannot
adopt the term “rhyme” for “yunmu” in the Chinese syllable terminology.
1 Traditional Chinese phonology can be traced back to as early as the ‘Qieyun’ rhyme
table (AD 601), which deals with the pronunciation of Ancient Chinese (Karlgren 19151926; Wang 1963; Chao 1968; Xue 1986; Yang 1996).
To avoid confusion, many modern Chinese linguists (Chan 1997; Wang &
Smith 1997; Chen 2000; and many others) adopt the terms “initials” and
“finals” for the traditional terms “shengmu” and “yunmu”, instead of
general phonological terms such as “onset” and “rhyme”. In Chapter 4, I
will present an analysis of SX syllable structure and also seek to cast some
light on the syllable structure of Mandarin.
In this chapter, I also use the terms of “Initials” and “finals”,
referring to “shengmu” and “yunmu”, i.e. the constituents into which the
SX syllables can be divided and within which all the surface consonants
and vowels of SX may occur.
As was mentioned in chapter 1, both the constraint-based theory (e.g.
Prince & Smolensky 1993) and the rule-based theory (e.g. Chomsky &
Halle 1968) share the notion of an underlying form, or input, and produce
outputs, either having the advantage of the other in explaining the
language, so that both theories are applied to the analysis in this chapter.
2.2 Initials
I will refer to syllable-initial consonants as “initials” throughout this
chapter. The most remarkable characteristic of the SX initial consonants
(compared with Mandarin and other Chinese dialects) is the fact that SX
still retains the historical voiced obstruents, just as Middle Chinese did2
(Chao 1928; Yip 1980; Zhan 1991). In the following sub-sections, I will
discuss all five classes of possible SX initials: oral stops, the glottal stop,
affricates, fricatives, and sonorant consonants.
2.2.1 Stops
In this sub-section, I will discuss voiceless unaspirated stops, voiceless
aspirated stops and voiced unaspirated stops, all of which can appear as
distinctive initials in SX. I list the nine stops of SX in (2):
(2) stops
voiceless unaspirated p t k
voiceless aspirated ph th kh
voiced unaspirated b d g
2 Middle Chinese is the language of the Sui, Tang, and Song dynasties (7th–10th centuries
SX has voiceless unaspirated, voiceless aspirated and voiced
unaspirated bilabial, alveolar and velar stops. The voiced stops are typical
of the Wu dialects. Modern Chinese (Mandarin) and the other five
Chinese language families have lost all the voiced stops and voiced
affricates (Campbell 2003). SX has retained all the voiced stops and
affricates that were present in Middle Chinese. The nine stops in SX are
very commonly used as the onset of the syllable. Some examples are
presented in (3):
a b c
(3) [pIN33] ‘soldier’ [phIN33] ‘marry’ [bIN22] ‘ill’
[tIN33] ‘book’ [thIN33] ‘listen’ [dIN22] ‘decide’
[koN33] ‘supply’ [khoN33] ‘free time’ [goN22] ‘total’
In the examples above, the syllables in column a all have voiceless
unaspirated stops; those in column b have voiceless aspirated stops;
column c has voiced unaspirated stops. These examples show that the
tones of the words in column c are lower3 than those of the words in
columns a and b. In Middle Chinese tonogenesis, the syllables with
voiced obstruents in the initial had a lower tone than those with voiceless
obstruents (Chao 1928; Yip 1980, 1989; Bao 1999; Duanmu 2000b; and
many others), which is consistent with the articulatory and acoustic
properties of voiced and voiceless initial obstruents on the one hand and
pitch on the other (see also below). The current tonal structure of SX
therefore still reflects the tonal system of Middle Chinese. The Chinese
tones have been philologically and linguistically classified as “yin” and
“yang” registers since the 7th century (Bao 1999). The yin register is also
referred to as high register, corresponding to high-pitched tones, while the
yang register is also referred to as low register, corresponding to lowpitched tones. In SX, there are altogether eight tones. Among these tones,
[52], [35], [33] and [5] are in the yin register, which mainly occur in
voiceless-obstruent-initial syllables, and [31], [13], [22] and [3] are in the
yang register, which occur only in voiced-initial syllables, including
voiced obstruents and sonorants.
In SX, we can predict from the tone of the syllable whether the initial
obstruent is voiced or voiceless. The reverse, however, is not the case,
3 The pitch of the different tones is marked on a five-point pitch scale, in the same way
as in Mandarin. The highest pitch is marked 5 and the lowest is marked 1 (cf. also ch. 1).
See chapter 5 for a further exploration and discussion of SX tones.
because there are four low-register tones and four high-register tones with
voiced-initial syllables or voiceless-initial syllables, respectively (see
examples in (5) below). We can only tell from the voiced or voiceless
initials that the tone of the syllable falls within the yang (low) register or
yin (high) register – we cannot predict the individual tone. Moreover, it is
completely unpredictable whether voiceless stops in yin-register syllables
are aspirated or unaspirated, as the examples in (4) show:
(4) [pu33] ‘cloth’ [phu33] ‘berth’
[tIN33] ‘book’ [thIN33] ‘listen’
[kE/5] ‘cut’ [khE/5] ‘block’
but [bu22] ‘step’ *[bu33]
[dIN22] ‘decide’ *[dIN33
[gE/3] ‘squeeze’ *[gE/5]
Since the syllables with a voiced stop in the initial have a lower tone
than those with a voiceless stop, [b], [d] and [g] cannot form exact minimal pairs with syllables with voiceless aspirated or unaspirated initial
stops, as is shown in (4). Thus, the question arises whether the voiceless
stops and voiced stops in SX are allophones of each other. I claim that the
voiceless unaspirated and aspirated stops and voiced unaspirated stops in
SX are not allophones of each other, i.e. that they are distinctive
phonemes. There are two main points that bear on this issue. First,
allophones are variants of a distinctive phoneme which are usually in
complementary distribution. For two phones to be classified as allophones
of a single phoneme, they must exhibit phonetic similarity, and they must
not be in contrastive distribution (e.g. Trask 1996). For example, [k] and
[kh] are allophones of the same phoneme /k/ in English.
In SX, however, the tones on the syllable are not predictable from
the syllable-initial consonant. There are four tones that can occur in each
case. Take the four tones of a syllable with a voiced initial stop in SX.
Rhymes ending in a glottal stop have tone [3]. The other three tones, [22],
[13] and [31], are unpredictable. For example:
(5) a. [bIN22] ‘ill’ [bIN13] ‘judge’ [bIN31] ‘level’
b. [doN22] ‘cave’ [doN13] ‘pail’ [doN31] ‘same’
c. [gAÅ22] ‘a tool’ [gAÅ13] ‘confuse’ [gAÅ31] ‘finish’
The examples in (5) show that the tone pattern cannot be predicted from
the initial. However, the voicing quality of the initial consonant could still
be predictable from the tone registers. For articulatory and acoustic
reasons, a syllable beginning with a voiced obstruent intrinsically has a
lower tone than one beginning with a voiceless obstruent (Haudricourt
1954; Lehiste 1970; Matisoff 1973). This relation between voiceless
obstruents and high register, and between voiced obstruents and low
register, is widely attested in natural languages (Yip 2002). The consonant-tone correlation is also well documented in the tonogenesis
literature cross-linguistically (Hombert 1978; Hyman 1978; Hombert,
Ohala & Ewan 1979).
From an acoustic point of view, if the initials of the syllables in
column a and column c or those in column b and column c in (3) were in
complementary distribution, the tones on these syllables could be
analyzed as ‘allotones’, rather than allophones. Yip (1980: 138) also
realizes that in such cases the voiced stops are always accompanied by
low-tone allotones. I will assume that tones in high register and those in
low register occur both in underlying forms in SX. However, the exact
distribution of tones and their relations with initial consonants will be left
for more detailed discussion in chapter 5.
2.2.2 Glottal stop
Besides the bilabial, alveolar and velar stops discussed above, there is
another stop in SX, viz. the glottal stop [/]. Although about 46% of the
world’s languages have a glottal stop, according to Maddieson (1984a), in
many languages glottal stops serve to demarcate the boundaries of
phrases or other prosodic units (Ladefoged & Maddieson 1996). However,
in some languages like Arabic, Thai, Hebrew and Hawaiian, the glottal
stop is a contrastive consonant. In (6) some examples from Hawaiian are
given; the form in (a) comes from Ladefoged & Maddieson (1996) and
those in (b) and (c) from Gussenhoven & Jacobs (1998):
(6) a. [/a/a] ‘dare’
b. [aa] ‘jaw’
c. [/aa] ‘fiery’
The examples in (6) suggest that the glottal stop in Hawaiian is a
phonemic stop, although perhaps the issue is not decided conclusively,
lacking more data and a formal analysis. In Middle Chinese, the glottal
stop [/] served as an initial and was used before [u], [a], [i] and any of the
three glides, [j], [w] and [Á]4, when there was no other consonantal onset
in the syllable (Chao 1928). The glottal stop in SX seems to play the same
role in the syllable as it did in Middle Chinese. Some examples from SX
are listed in (7):
(7) [/e33] ‘love’ [/E/5] ‘duck’
[/jẽ52] ‘smoke’ [/Á∏̃52] ‘complain’
The examples in (7) show that in SX [/] can serve as an onset preceding a
vowel or glide in the syllable. Here the question arises whether the glottal
stop in SX is a phonemic stop, as in Hawaiian, or only a phonetic form in
words where no other consonant is present, as in German (see (8)) or
The status of the initial glottal stop in SX as well as in all other Wu
dialects has been a topic of debate, also with respect to Middle Chinese
(Karlgren 1915–1926; Chao 1928; Dragunov 1930; Hope 1953; and many
others). Dragunov (1930, following Karlgren) states that “in the ancient
Chinese languages, as is well known, words with zero initials were
differentiated from words beginning with a glottal stop. We discover this
distinction again, in certain Northern Chinese dialects of the 13th and 14th
centuries, as is shown by the hPhags-pa script, and in nearly all the
modern Wu dialects” (cited by Hope 1953).5 Hope (1953: 2) strongly
argues against the existence of phonemic initial glottal stop in Middle
Chinese and argues that “Karlgren presents little or no evidence; all he
really does is to make an assumption for the purpose of filling a
psychologically created lacuna”. Hope (1953: 13) further claims that “in
all modern Chinese dialects without exception the glottal stop, where it
exists, is not only of no phonemic significance but is not even heard by
the speakers of the language”. It is true that Karlgren (1915–1926) only
reconstructed an initial glottal stop for ancient Chinese (because the
phonetics of Middle Chinese are uncertain). Chao (1928) did fieldwork on
4 In traditional Chinese phonology, the three glides were transcribed as [i], [u] and [y],
i.e. using the same symbols as the counterpart vowels (Chao 1928; Wang 1963, 2003,
among many others). Through my dissertation, I transcribe them as [j], [w] and [Á],
respectively, following the general linguistic transcription. I will claim that the
prenuclear glides in SX are not in the onset. This will be discussed in detail in chapter 4.
5 “HPhags-pa” or the “hPhags-pa script” is the language created by hPhags-pa (1235–
1280) in the Yuan Dynasty (1207–1367). It is sometimes referred to as New Mongolian,
which had over 40 letters used to spell Mongolian, Chinese and Tibetan.
the phonetic realization of the initial glottal stop in modern Wu dialects,
and states (in his Table 3) that SX has a stable [/] as an initial preceding
[u], [e], [jẽ] and [Á∏̃]. I also claim that the initial glottal stop in SX exists,
but its status is phonetic rather than phonological. There are two main
reasons for this: First, in SX there is no minimal pair which would show
that [/] is a contrastive phoneme as was the case for Hawaiian in (6). The
syllables in (7) can optionally be pronounced without the initial glottal
stop and with the same tones and would then have the same lexical
meaning: thus, the glottal stop plays no contrastive role.
It is true that syllables in SX are strictly classified into two categories:
high-register syllables with high-pitch tones and low-register syllables
with low-pitch tones, which is determined by the voicing status of the
initial obstruents (c.f. above). As expected, the glottal stop [/] as the
initial of a syllable is accompanied by a high-pitch vowel so that the
initial glottal stop [/] is predictable in SX. However, underlyingly, there is
no initial glottal stop for the syllables in (7). A similar phonetic
phenomenon occurs in German as Hall (1992:58) describes “the glottal
stop is completely predictable in German, since it only occurs before a
vowel-initial stressed syllable and then only optionally”. For example
(Hall 1992:58):
(8) arm [/a !Rm] [a !Rm] ‘poor’
elf [/E!lf] [E!lf] ‘eleven’
oft [/O!ft] [O!ft] ‘often’
Uhr [/u!:ø8] [u!:ø8] ‘clock’
Hall formulates this phenomenon in German by way of a Glottal Stop
Insertion rule, which is presented in (9):
(9) Glottal Stop Insertion (GSI: optional)
P → [/] / F[ __ [−cons]
Jongenburger & van Heuven (1991) carried out a phonetic experiment in Dutch with the example of [dAt ´n /A:ntAl] ‘that a number’ and
found that there is a striking acoustic difference with a smooth vowel
onset in [´n] and an abrupt onset (glottal stop) in [/A:ntAl]. They claim
that the phonetic glottal stop is inserted only in hiatus position, and
nowhere else in Dutch (Jongenburger & van Heuven 1991, 1993). The
only difference between German and Dutch and SX is that the GSI rule in
German cannot apply to the syllable which is not foot-initial, while in
Dutch it only operates in hiatus position whilst GSI in SX applies to every
vowel-initial syllable with a high-register tone, due to the fact that SX is a
monosyllabic language. However, neither the initial glottal stop in
German or Dutch nor that in SX is an underlying phoneme. To briefly
summarize, SX, like many other languages, inserts a glottal stop in
syllables beginning with a vowel. Does this mean that there is no
phonemic /// in SX? I assume that there is a phonemic glottal stop /// in
SX, but it is in the final rather than in the initial. I will discuss the
phonological behavior of final /// in SX in §2.4.
2.2.3 Affricates
Phonetically, affricates are consonants whose articulation involves a
complete oral closure followed by a comparatively slow release, yielding
perceptible friction noise which is clearly longer than the noise burst of a
plosive. Affricates are not uncommon phonemes across languages
(Maddieson 1984a) in terms of their occurrence in languages. In SX, there
are six affricates, including three dental affricates and three alveolopalatal affricates, as shown in (10):
(10) dental alveolo-palatal
voiceless unaspirated ts t˛
voiceless aspirated tsh t˛h
voiced unaspirated dz d¸
Like the SX stops in (2), there are voiceless unaspirated, voiceless
aspirated and historically voiced unaspirated affricates in SX6. Phonetically and phonologically, these affricates are phonemic consonants,
although voiced affricates and voiceless affricates are accompanied by
different tones in the syllable, because of the association between stiff and
slack vocal cords and high and low tone register, respectively, as was
discussed above. For example:
6 Historically speaking, Middle Chinese had voiceless unaspirated, voiceless aspirated
and voiced unaspirated stops and affricates (Karlgren 1954; Wang 1985; Baxter 1992),
which are claimed to still retain in the Wu Chinese, like SX (Wang 1959; Zhan 1991;
Cao 2002) in spite of the fact that some phonetic experiments show that the historical
voiced stops and affricates in the Wu Chinese do not have VOT (Shryock 1995).
(11) [tsoN52] ‘end’ [t˛i33] ‘post’
[tshoN52] ‘dash’ [t˛hi33] ‘air’
[dzoN31] ‘worm’ [d¸i31] ‘ride’
Middle Chinese had nine affricates in initial position, including [ts],
[tsh], [dz], [tS], [tSh], [dZ], [t˛], [t˛h] and [d¸], according to the Guangyun7
(Ding 1984). Of these, SX has retained six. SX does not distinguish
between alveolar and palatal affricates, lacking post-alveolar and retroflex
affricates such as /tS/, /tSh/, /dZ/, /tß/, /tßh/, and /dΩ/.
The status of affricates as one or two segments has been a topic of
some debate. Some have argued that affricates are combinations of stops
and fricatives and should be treated as two segments (Brooks 1965;
Szigetvári 1997; and others). Durand (1990) cites some examples of
segment sequencing in English, which show that in English either /S/ or /t/
can be followed by /r/; but /tS/ cannot. For example:
(12) a. [Srimp] ‘shrimp’
b. [trai] ‘try’
c. *[tSr-]
The examples in (12) show that (c) is ruled out, possibly because
there are already two consonants before /r/ and the maximal onset cluster
is CC (except [s]+CC) in English. This kind of evidence can be used to
suggest that affricates form a consonant cluster in a particular language.
Harris (1994) claims that affricates are qualitatively complex but
quantitatively simple. Theoretically speaking, affricates are combinations
of the features of the two constituent phonemes rather than two separate
phonemes (see also Clements 1985; Anderson & Ewen 1987; McCarthy
1988; van de Weijer 1996 for discussion).
Evidence from SX strongly suggests that affricates are single
segments. SX is a language that has no complex onsets. 8 No
combinations of two consonants preceding a glide or a vowel are possible
in SX (*[sp], *[sh], *[sl], *[pl], *[kl], *[sn], *[ph], etc). The SX syllable
structure maximally has a CGVC, or CGVV pattern (where C is a
consonant, G is a glide, and V is a vowel). For example:
7 Guangyun (1008) is an ancient book of traditional Chinese phonology that describes
the pronunciation of the Chinese syllables and rhyming system.
8 It is also my claim that the prenuclear glides are not part of the onset in SX. This will
be discussed in detail in chapter 4 about the syllable structure.
(13) [pjAÅ35] ‘watch’
[˛jAÅ35] ‘small’
[kwÅN52] ‘light’
[HwoN31] ‘red’
There are no *CCV, *CCGV, or *CCGVC syllables in SX. However,
affricates may fill the C position in the syllable template; this is in fact
very common in SX. For example:
(14) [d¸jAÅ31] ‘bridge’ [tswo33] ‘do’
[t˛hjo/5] ‘lack’ [tshwo33] ‘wrong’
[d¸joN31] ‘poor’ [dzwo31] ‘tea’
The syllables in (14) are well-formed and acceptable. This is consistent
with the assumption that affricates are single segments, such that the SX
maximal syllable template of CGVC/CGVV can be maintained.
Another piece of evidence for the idea that affricates are single
segments in this language comes from SX loanwords. SX does not allow
complex onsets, as discussed above. This could be captured by an
Optimality-theoretic constraint on the syllable onset in SX:
(15) *COMPLEX(ONS) (Itô 1986; Blevins 1995)
*[σ CC: Onsets must be simple.
The constraint *COMPLEX(ONS) is highly ranked in SX, so that
complex onsets are not acceptable in SX, even in loanwords (Zhang
2003). For example, the English noun [khloun] ‘clone’ will be realized as
[khØloN], without the original CC cluster in the onset; rather, an
epenthetic vowel is inserted. SX has a very restricted coda. Only two
consonants, [N] and [/] are allowed in the coda position, which can be
stipulated by a coda-condition constraint, as stated below:
(16) CODA-COND (Ito$ 1989; Zhang 2003)
Coda can only be [/] or [N]9.
9Coda condition in SX will be discussed in the next section and in more detail in chapter
Another important phenomenon in loanword phonology of SX is the
tendency of disyllabification for a minimal word, which is also well
documented in many other Chinese languages (Yip 1993; Chen 2000;
Zhang (2003). There is a Minimal-word constraint as follows:
(17) MINWD (Yip 1993)
A loanword is minimally disyllabic.
Among the three constraints explained above, *COMPLEX(ONS) and
CODA-COND are inviolable in SX while there are occasionally some
exceptions for MINWD.10 For example, the English clone and tank are
[khØloN] and [thE‡khØ], respectively, in SX loan words. The tableau deriving
the optimal candidate is given in (18):
(18) input khloun *COMPLEX(ONS) CODA-COND MINWD
a. khloun *! * *
b. khloN *! *
c. khØloN
d. khoN *!
e. loN *!
In tableau (18), candidate (a) is the closest to the input, but violates all the
three constraints, so it is the worst candidate. Candidate (b) is also ruled
out first because it violates *COMPLEX(ONS). Candidates (d) and (e)
violate MINWD and are also ruled out. Candidate (c) is the winner because
it does not violate any constraints in (18) (it violates another constraint,
DEP-IO, which militates against insertion of material, so this constraint
must be lower-ranked).
If we compare a loanword with a cluster to a loanword with an initial
affricate, it turns out that (English) affricates are well-formed onsets in
SX. For example, English [dZi:p] ‘jeep’ is adapted as [t˛I/pu] in SX. The
selection of this candidate is represented in the following OT tableau:
10 Some technical terms may be still monosyllabic, e.g. chemical elements. This involves
more constraints, which falls outside of the scope of my discussion.
(19) input /dZi:p/ *COMPLEX(ONS) CODA-COND MINWD
a. t˛I/p *! *
b. t˛I/pu
c. t˛I/ *!
In tableau (19), candidates (a) and (c) are ruled out because they violate
CODA-COND and/or MINWD. Candidate (b) does not violate any constraint
and is the winner. This shows that affricates behave as single segments
and are acceptable in SX phonology: a candidate with epenthesis would
be treated as violate the constraint DEP-IO mentioned above. In short, for
any foreign word with a CC or CCC cluster in the onset, the SX loanword
phonology exceptionlessly has to insert a vowel between CC or two
vowels between CCC of the output loanwords. More examples
(disregarding tones) are given in (20):
(20) English Loanword in SX
a. [sprIN] spring [sbI/lIN] ‘spring lock’
b. [brœndi] brandy [pa/lẼdi] ‘Brandy wine’
c. [frœns] France [fa/lẼi] ‘France’
d. [tSOk´lit] chocolate [t˛hjAÅkh´/lI/] ‘chocolate’
e. [tS´…tSIl] Churchill [t˛hjØt˛I/] (person’s name)
In the examples in (20), vowels are inserted between the consonants in the
initial clusters of the words in (a), (b) and (c) to form an acceptable SX
syllable (other constraints in SX loanword phonology will be left aside
until later: see chapter 4). In the words for chocolate and Churchill,
however, the affricates are treated as single segments, so that no vowel
needs to be inserted to break up affricates.
2.2.4 Fricatives
Fricatives are sounds that are produced by a narrow approximation of two
articulators so as to produce a turbulent airstream (Ladefoged 1971: 46).
While it is true that there are fricatives in nearly all of the world’s
languages, there have been relatively few studies of their precise
distribution or of the patterns of occurrence which they show, according
to Maddieson (1984a). However, some research has been done on the
description of the articulation, classification of fricative sounds and their
acoustic properties cross-linguistically (e.g. Jakobson, Fant & Halle 1952;
Nartey 1982; Ladefoged 1983). Nartey (1982) presents a cross-linguistic
phonetic analysis of fricatives in 14 languages11, measuring the phonetic
differences within and between languages. In this sub-section, I will
examine the behaviour of fricatives in SX. I provide the surface fricatives
of SX in (21):
(21) labiodental dental alveolo-palatal glottal
voiceless f s ˛ h
voiced v z ¸ H
SX has four pairs of voiceless and voiced fricatives. In most
languages, there is a tendency to prefer voiceless fricatives and to avoid
voiced and voiceless pairs of fricatives at the same place of articulation
(Maddieson 1984a). Although this tendency holds for obstruents in
general, fricatives appear to be more asymmetric. Most Southeast Asian
languages have relatively few fricatives; Mandarin has five fricatives,
which are all voiceless; Cantonese has four fricatives; Thai, Korean and
Taba12 (Abramson 1999) have only three fricatives. Indo-European languages usually have eight or more fricatives. The SX voiceless and
voiced fricatives are contrastive phonemes in spite of their being
accompanied by different tones in the syllable, which follows the same
register division as with the stops and affricates (cf. above). For example:
(22) voiceless voiced
[fu33] ‘pay’ [vu22] ‘attach /enclose’
[sAÅ35] ‘little’ [zAÅ13] ‘make/invent’
[˛y33] ‘need’ [¸y22] ‘tree’
[ho52] ‘shrimp’ [˙o31] ‘river’
Most languages in the world have no contrast between voiced and
voiceless glottal fricatives (/h/ vs. /˙/). In fact, Maddieson (1984a) lists
only two languages in the world which have a contrast between /h/ and /˙/,
and one of these is from the Wu language family, just like SX. Although
there has been some disagreement on the classification of /h/ and /˙/ (e.g.
whether they are fricatives or laryngeals, vowels or approximants; see
Maddieson 1984a), /h/ and /˙/ in SX are undoubtedly glottal fricatives
11 Nartey’s (1982) analysis includes Amharic, Arabic, English, Hebrew, Hopi, Japanese,
Korean, Navajo, Papago, Pima, Polish, Swedish, Yoruba and Zuni.
12 Taba is an Austronesian language spoken in the northern Maluku province, Indonesia.
and both are phonemic consonants which may occur in the syllable onset.
For example:
(23) voiceless voiced
[ho52] ‘shrimp’ [˙o31] ‘river’
[hE/5] ‘blind’ [˙E/3] ‘narrow’
According to Maddieson (1984a), about 63.7% of the languages in
UPSID 13 have voiceless /h/, while only 4.1% of the languages have
voiced /˙/. In SX, voiced /˙/ is much more frequent than voiceless /h/ in
syllable onsets. Besides the words mentioned above, I list some more
examples in (24):
(24) Voiced Voiceless
[˙AÅ31] ‘roar’ [hAÅ52] ‘spend (time)’
[˙´N31] ‘stable’ [h´N52] ‘groan’
[˙o/3] ‘learn’ [ho/5] (surname)
[˙wo22] ‘speech’ [hwo33] ‘spend’
[˙we31] ‘return’ [hwe52] ‘ash’
[˙wẼ31] ‘play’ [hwẼ52] ‘well-behaved’
[˙wÅN31] ‘king’ [hwÅN52] ‘nervous’
[˙u22] ‘unclear’ [hu33] ‘call’
[˙ÅN31] ‘line’ [hAN52] ‘ram’
[˙wa22] ‘bad’ ? (undecided possibility)
[˙jẽ22] ‘hate’ * (systematic
[˙jØ31] ‘oil’ * impossibility)
[˙jo/3] ‘moon’ *
[˙ja22] ‘night’ *
[˙Á∏̃31] ‘cloud’ *
The examples in (24) show that where there is an acceptable syllable
beginning with voiceless [h] in SX, there is also a word with [˙] followed
by an identical final vowel or combination. The reverse is not true,
however. In short, [˙] is more often found in this position in SX than [h].
In fact, [h] plus a front high vowel is systematically ruled out in SX
(*[h][+high, -back]), while [˙] can be followed by all different vowels, as
13 The UCLA Phonological Segment Inventory Database.
shown in (24). There are no constraints on the distribution of [H] as an
onset. This will be discussed in detail in chapter 4.
In some languages [h] and [H] have been described as voiceless
versus breathy-voiced counterparts of the vowels that follow them
(Ladefoged 1971). In SX, vowel-initial syllables invariably receive a [H]
if the tone is low-register, but [/] appears (not [h]) if it is high-register, as
discussed in §2.2.2 above. More examples are given in (25):
(25) [˙]: [/]: [h]:
[˙i31] ‘move’ [/i52] ‘clothes’ *[hi52]
[˙jØ31] ‘oil’ [/jØ52] ‘low voice’ *[hjØ52]
[˙jẽ31] ‘salt’ [/jẽ33] ‘smoke’ *[hjẽ33]
[˙Á∏̃31] ‘cloud’ [/Á∏̃33] ‘complain’ *[hÁ∏̃33]
Generally speaking, there are more voiceless obstruents in a language
than voiced ones; if a language has a voiced obstruent, it usually also has
the voiceless counterpart (Maddieson 1984a). However, the asymmetric
distribution of [h] and [˙] in SX is inconsistent with this general situation.
I assume that the wider distribution of [˙] compared to [h] in SX is caused
by the tonal system, which has a voiceless-obstruent/high-register and
voiced-obstruent/low-register correlation (see further discussion about it
in chapter 5). This means that it is impossible for a syllable to have a lowregister tone and a voiceless initial obstruent or a high-register tone and a
voiced initial obstruent. There is cross-linguistic evidence that the glottal
stop [/] is often inserted before a vowel-initial syllable, as discussed in
§2.2.2. The situation in SX is similar in that [/] is always inserted before a
high-register syllable when there is no other initial consonant. To satisfy
the tonal system of SX, a voiced glottal obstruent is required before a
low-pitched rhyme when there is no other voiced consonant (the tonal
system will be discussed in detail in chapter 5). But there is no voiced
counterpart of the glottal stop [/] in the consonant inventory of the
world’s languages. However, [/] and [˙] form a natural class in certain
aspects. Both can be regarded as laryngeals and the laryngeals are always
considered as placeless-component, lacking the complexity when
compared with other consonants (Harris & Lindsey 1995; Humbert 1995;
Botma 2004). Many Chinese scholars assume that [h] is a phonemic
consonant while [/] and [˙] are a pair of phonetic onsets in the Wu
Chinese (Xu & Tang 1988). The optimal choice of [/] and [˙] for the
onset of a high-register syllable and low-register syllable, respectively,
when there is no other onset consonant, can be stipulated by the following
four onset-condition constraints:
(26) Onset-condition Constraints:
a. ONSET(Itô 1989)
Syllables must have an onset.
b. Voiced-L
Voiced initial obstruents must have low-register tones on the
following vowels.
c. Voiceless-H
Voiceless initial obstruents must have high-register tones on the
following vowels.
The constraint ONSET in (26a) will rule out full-tone syllables without
an onset such as *[e52], *[a33], *[u31] and *[o22] in surface representation
and the constraints of (26b) and (26c) will also rule out any syllable that
has voiced initial obstruent for high register or voiceless initial obstruent
for low register. Such a consonant-tone correlation is well explained by
Halle & Stevens’ (1971) laryngeal feature specifications ([stiff] & [slack],
which will be discussed in chapter 5).
In short, I assume that the onset [˙] in SX is both a phonological
onset when it is contrastive with [h], as shown in (24), and a phonetic
onset just like [/] in surface representation, as shown in (25). Such is the
case decided by the onset-condition constraints in (26). This is why [˙] is
more frequent than [h] in the formation of syllables in SX.
2.2.5 The sonorants
Besides the 24 obstruents discussed above, there are also five sonorant
consonants that can appear in syllable-initial position in SX. In this
subsection I present a brief analysis of the five sonorant initials of SX. In
(27), I list the four nasal initials:
(27) bilabial alveolar alveolo-palatal velar
m n  N
SX has a bilabial, alveolar, alveolo-palatal and velar nasal in surface
representation, all of which can appear in the initial position of the
syllable, but not all are in contrastive distribution. Consider the three
groups of syllables in (28):
(28) [o] [i] [u]
[m]: [mo22] ‘grind’ [mi13] ‘rice’ [mu22] ‘grave’
[n]: [no22] ‘glutinous’ *[ni] — [nu22] ‘anger’
[]: *[o] — [i22] ‘two’ *[u] —
[N]: [No22] ‘hungry’ *[Ni] — *[Nu] —
The examples in (28) show some syllables with the four initial nasals and
three different vowels, [o], [i] and [u], and also show that not all the four
nasals can surface with the same vowel. This means that the four nasals
cannot be all contrastive with each other. In fact, only [m], [n] and [N] can
all occur before [o]. I assume that [m], [n] and [N] contrast with each
other, as attested by more examples such as [me22] ‘younger sister’, [ne22]
‘patient’ and [Ne31] ‘fool’, and [ma13] ‘buy’, [na13] ‘milk’ and [Na13] ‘we,
us’. Thus, [m], [n] and [N] are phonemic nasals in SX. The velar nasal [N]
cannot occur before the high vowels, [i] and [u], in SX. This can be
formalized by a constraint:
(29) *[N][+high]
[N] cannot occur before any [+high] (semi-)vowel.
The constraint *[N][+high] in (29) forbids */Ni/, */Nja/, */Nu/ and */Nwa/
in SX. The alveolar nasal [n] cannot appear before the front high vowel [i]
or the glide [j] while the alveolo-palatal nasal [] can only appear before
[i]/[j].14 This indicates that [] is in complementary distribution with [n].
Thus, [] is an allophone of the distinctive nasal /n/, which can be
expressed by a nasal palatalization rule as in (30) or a constraint as in (31):

__ +high –back
or (31) *[n] +high
The rule in (30) formalizes that in SX the coronal nasal becomes alveolopalatal [] when preceding a front high (semi-)vowel, e.g. [i], [I] or [j],
14 Actually in SX syllables, [m] can appear before all vowels; [n] and [N] can appear all
except high front (semi-)vowels. The phonotactics of the SX segment sequences will be
discussed in chapter 4.
and the constraint in (31) rules out any syllable where [n] is followed by a
front high (semi-)vowel. More examples that involve this nasal
palatalization rule are given in (32):
(32) [i22] ‘two’ *[ni]
[jØ31] ‘cow/ox’ *[njØ]
[IN31] ‘silver’ *[nIN]
[I/3] ‘hot’ *[nI/]
From the analysis above, we have observed that in SX, there are six
alveolo-palatal consonants: [tÇ], [tÇh], [dÛ], [Ç], [Û] and []. All the
alveolo-palatal consonants have the same contribution that they can only
precede high front (semi-)vowels. Some scholars (e.g. Duanmu 1999)
assume that all the alveolo-palatal consonants are only allophonic
segments. The distribution of all the consonants will be discussed in
chapter 4.
However, the nasal palatalization rule does not apply to the bilabial
nasal in SX, presumably because the coronal in general shifts to the
alveolo-palatal place of articulation more easily than the bilabial one.
However, nasal asymmetric behavior is common cross-linguistically,
though they are of a natural class (e.g. Bhat 1978; Botma 2004). For
example, the formation of compounds in Dutch has optional place
assimilation of /n/, but not of /m N/, which strongly suggests asymmetric
behaviour in the class of the nasals (Botma 2004), as shown in (33):
(33) steen+bok stee[mb]ok ‘Capricorn’ (*[nb])
tram+kaart tra[mk]aart ‘tram ticket’
meng+paneel me[Np]aneel ‘mixing panel’
Finally, SX also permits the lateral /l/ in the initial position of a
syllable, but it has no /r/. /l/ is a common onset in SX syllables. For
(34) [li13] ‘inside’ [l´N13] ‘cold’
[le31] ‘come’ [laN13] ‘cool’
[la52] ‘pull’ [loN31] ‘dragon’
[lAÅ13] ‘old’ [lo/3] ‘green’
[ljØ31] ‘flow’ [ljaN22] ‘bright’
Altogether, SX has 29 consonants in the surface representation, all
of which can be used as the initial of a syllable. Of these, only the glottal
stop /// and the velar nasal /N/ can appear in postnuclear position (cf. § The constraints that formalize this will be discussed in more
detail in the next section. I list all 29 consonants of SX in (35):
(35) SX consonant inventory:
-asp p t k /
+asp ph th kh
+voice b d g
-asp ts tÇ
+asp tsh tÇh
africate +voice dz dÛ
-voice f s Ç h fric. +voice v z Û H
nasal m n  N
lateral l
Next, I will present the feature specifications for the 29 consonants
in SX in (35) for the sake of further phonological analysis of their exact
distribution, which will be discussed in chapter 4. To specify the 29 SX
consonants, I use ten distinctive features, which include:
Laryngeal features: [stiff], [slack], [spread];
Place features: [anterior], [coronal], [dorsal];
Stricture features: [nasal], [continuant], [strident], [sonorant].
Among the ten features above, following Halle & Stevens (1971), I
use [stiff] and [slack] instead of [voice] to specify obstruents, [+stiff] for
[–voice] and [+slack] for [+voice], although one of them is redundant for
the specifications of obstruents. Sonorants have default voicing so that
they are not specified for [voice] in (36). The feature [dorsal] has the
same value as [back] for consonants. To distinguish dental/alveolar from
alveolo-palatal consonants, I use the feature [anterior]. In the text, I also
use [apical] as a feature (as assumed by Williamson (1977)), when
discussing the distribution of the apical vowel []; sometimes, I use [back]
instead of [dors] for consistency with vowel, and sometimes I also use
other more features such as [labial] for both [p] and [u]. In table (36), I
only mark “+”, leaving “–” blank for simplicity.
H + + +
h + +
/ +
N + + +
g + +
kh + + +
k + +
 + + +
Û + + + +
Ç + + + +
+ + +
h + + + +
+ + +
l + + +
n + + + +
d + + +
th + + + +
t + + +
z + + + + +
s + + + + +
dz + + + +
ts h + + + + +
ts + + + +
v + + + +
f + + + +
m + + +
b + +
ph + + +
p + +
ea tu re sp ec ifi
ca tio
ns fo r t
he 2
co ns on an ts in S
ff sl ac k sp r an t co r do rs na s co nt st rid
In the feature matrix in (36), the 29 consonants in SX are all
distinguished from one another, using ten features. Since SX still retains
historically voiced obstruents, I still apply [±voice] to the specifications
for obstruents as well as sonorants throughout this dissertation when [stiff]
and [slack] are applied particularly to the analysis of consonant-tone
interaction. There are controversial specifications for the feature [strident].
I assume that all the anterior and coronal fricatives and affricates in SX
are [+strident], while in some languages, [±strident] is used to distinguish
[f] ([+strid]) and [F] ([–strid]), [s] ([+strid]) and [T] ([–strid]), and [S]
([+strid]) and [ç] ([–strid]) (see Trask 1996). All the analyses of the SX
consonants are based on the feature specifications in (36) in this
The initial consonants have been relatively stable through time and
there is not much disagreement on the number or realizations of the initial
consonants among field workers or researchers, with the exception of the
initial glottal stop [/]. Some researchers (e.g. Yang & Yang 2000;
Campbell 2003) prefer the term “zero onset” with respect to syllables
beginning with vowels or glides. The syllabic status of glides will be
discussed in detail in chapter 4. The behaviour of vowels and glides in
Finals will be discussed in the next section.
2.3 Finals
At the beginning of this chapter, I explained that “finals” in Chinese
syllable structure refer to all material that follows syllable-initial consonants. However, “finals” in SX are not equivalent to “rhymes”. The
syllable structure of SX will be discussed in chapter 4. In this section, I
will present a phonologically analytic description of all the possible
surface segments in the finals of SX.
The finals in SX are much more complicated and unstable than the
initials. Different native speakers may realize different phonetic vowels in
the same lexical form. Another remarkable phenomenon of SX is that
quite a number of lexical syllables are different in pronunciation when
comparing literary and colloquial style.15 For example:
15 In SX, literary style refers to the syllables for the written forms, which are mostly
borrowed from Mandarin or phonetically influenced by Mandarin and are always more
formal; colloquial style refers to the syllables for the oral forms, which are less formal.
(37) Literary style Colloquial style
[vi31] [bi31] ‘grease’
[vi22] [bi22] ‘taste’
[tÇjAÅ52] [kAÅ52] ‘hand in’
[da22] [do22] ‘big’
[z´N22] [≠IN22] ‘recognize’
[tso35] [tÇja33] ‘left’
The difference between literary style and colloquial style in SX can be
found in the onset, the rhyme, or the whole syllable, as shown in (37).
This difference is caused mainly by the influence of Mandarin, so that the
literary style is close to the Mandarin pronunciation. These two styles are
equally frequent in modern SX, each usually occurring with fixed lexical
collocations. Such differences are not so common in other Chinese
dialects. Researchers disagree on the phonetic transcription or even the
number of finals that exist in SX. In this subsection, I will first discuss the
Final inventory of SX because all syllables in SX are usually split into
two parts: Initial and Final, and all the surface vowels occur in fixed
combinations in Finals.
2.3.1 Final inventory
The exact inventory of Finals in SX is a controversy issue, not only with
regard to the number of Finals but also with regard to vowel qualities of
some final combinations. In this subsection, I will introduce some
different versions of the Final inventory of SX and present my proposal
after comparison with other proposals. Chao (1928) made an investigation
of Wu dialects and recorded all the phonetic vowels occurring in the
Finals16 of SX, which can be summarized in the following table:
16 In traditional Chinese phonology, Finals were referred to as ‘yunmu’, classified into
four categories (call ‘sihu’), viz. Kaikouhu, Qichihu, Hekouhu and Cuokouhu.. Kaikouhu
includes those ‘yunmu’ with simple vowels as the rhyme or beginning with a vowel;
Qichihu are those with i as the rhyme or beginning with i; Hekouhu are those with u as
the rhyme or beginning with u; Cuokouhu are those with y as the rhyme or beginning
with y. In the following tables of Final inventory, all ‘yunmu’ are categorized into sihu.
(38) Chao’s Finals in SX17
Kaikouhu Qichihu Hekouhu Cuokouhu
 i, ij u yj, y
e ie ue
a ia ua
Ø, , ∏ iØ, IØ
o io uo yo
œ,  uœ
In open
en, ,  I,  u∏n, u∏
a ia u
Å iÅ uÅ
With nasal
u iu
œ) iœ uœ 
∏ u∏ y∏
vowels e  I  ue 
consonants m, N, l 
Chao’s surface Final inventory in (38) contains altogether 53 different
Finals in the form of simple vowels, syllabic consonants, and
combinations of GV, VV, VG, VC, and GVC. Chao (1928) recorded
these Finals according to individually investigated subjects. His recording
represents the phonetic realization of Finals in SX, so that it is not
phonologically systematic. Besides, the pronunciation of some vowels
can be different from subject to subject. Some of his different forms of
Finals are just free variations, such as [u∏], [u∏n] and [u∏], [i] and [ij],
[en] and [e ], all groups or pairs of which could be used to pronounce the
same lexical syllable, differently from speaker to speaker. In Chao’s
recording, there are VG combinations such as [yj] and [yÁ], which, I
claim, are unacceptable in the SX surface representation because they
violate the OCP.
Campbell (2003) presents his inventory of 47 Finals in SX as follows:
17 This is my summary of Chao’s (1928) Table of Yunmu (in Wu dialects), in which the
prenuclear glides were originally transcribed in [i] and [u], which stay in the same
symbols as the original in (38), (39) and (40). However, in my analysis through the
dissertation, I present the two pre-nuclear glides as [j] and [w].
(39) Campbell’s 47 Finals in SX18
Kaikouhu Qichihu Hekouhu Cuokouhu
 i u
e ie ue yÁ
a ia ua
o io uo
Ø iØ
In open
 I
Å uÅ
a ia ua
With nasal
u iu uu
œ  iœ uœ 
∏ u∏ y∏
vowels I 
/ I/
/ i/ u/
o/ uo/ yo/
With stop
e/ Syllabic
consonants m n N l 
Campbell’s (2003) 47 Finals in SX as shown above are more systematic
than Chao’s (1928) except some combinations such as [] and [uu]
which are rather uncommon in Chinese dialects and can be hardly
manifested by the present data. In Campbell’s version of the SX Finals,
there is one VG combination [yÁ] which sounds unusual. Yang & Yang
(2000) present a similar Final inventory of SX to Campbell’s in (40):
18 Campbell’s (2003) 47 Finals in SX are presented on a website:, which may be upgraded every year. The three glides
are transcribed as [i], [u] and [y], following the original version.
(40) Yang & Yang’s Finals in SX19
Kaikouhu Qichihu Hekouhu Cuokouhu
 i u y
Å iÅ
a ia ua
o io uo
Ø iØ
In open
e ie ue
 i
Å iÅ uÅ
a ia ua
With nasal
o io uo
E iE uE
Ø  uØ  yØ 
vowels e  ie 
´/ i´/
a/ ia/
o/ io/ uo/
With stop
E/ uE/
consonants m  N l 
In (40), Yang & Yang present 49 Finals in the SX surface representation.
The Final inventory in (40) is phonologically more systematic than both
Chao’s (38) and Compbell’s (39), leaving out free variations as well as
the VG combinations in the 49 Finals in SX and not much difference in
phonetic realization of the surface vowels from the data. However, some
Finals in (40) remain problematic from either a phonetic or a
phonological viewpoint, e.g. with regard to the question whether [Å] alone
can be the rhyme of a syllable or whether the combination of [iN] in the
surface SX exists. This will be discussed in the following sub-sections.
The main difference between Campbell’s analysis and Yang & Yang’s
analysis concerns different vowel qualities. For example, [/] and [æ̃] in
(39) refer to [a/] and [Ẽ] in (40), respectively. Based on the different
versions of the Final inventory of SX, including Chao’s (1928), Yang &
Yang’s (2000), Campbell’s (2003) and others, and also through Yang’s
19 In (40), Yang and Yang (2000) also transcribed the three glides in SX as [i], [u] and
[y], instead of [j], [w] and [Á] which are used through out in this dissertation.
(2000), Campbell’s (2003) and others, and also through exhaustive
consultation with the SX native speakers and according to the data and
my native intuition of SX, I will present my proposal of the Final
inventory of SX, as shown in (41):
(41) Finals in SX20
Kaikouhu Qichihu Hekouhu Cuokouhu
 i u y
e je we
a ja wa
o jo wo
Ø jØ
In open
 I
Å wÅ
a ja wa
With nasal
o jo wo
E jE wE
∏ w∏ Á∏
vowels e  je 
´/ I/
a/ ja/ wa/
o/ jo/ wo/
With stop
consonants m n N l 
The Final inventory of SX I present in (41) contains 48 different forms of
Finals, including 17 in open syllables, ten with a nasal coda, eight with
nasalized vowels as the rhyme, and nine with final stops. The 48 Finals in
(41) can also be re-classified as V, C, 21 GV, VV, VC, and GVC, in
general linguistic terms. The difference in the Final inventory of SX
between Yang & Yang’s (2000) (indicated by ‘Yangs’’ below) and my
proposal (indicated by ‘Zhang’s’ below) can be summarized as follows:
20 In (41), I transcribe the three glides in SX in [j], [w] and [Á] which are equivalent to [i],
[u] and [y], respectively, in traditional Chinese transcription mentioned in this chapter.
21 Here ‘C’ refers to syllabic consonants.
(42) Yangs’  Ø  Å i iÅ ia uØ  yØ  i/ uE/
Zhang’s n ∏ AÅ I ja w∏ Á∏ I/ wa/
The difference between the Yangs’ and Zhang’s representations in the
Final inventory of SX, as shown in (42), concern either phonological
behavior or phonetic realization of some vowels, which will be discussed
in the following sub-sections and next chapter. However, in my analysis,
SX has 48 Finals in three different forms, including syllabic consonants,
vowels (monophthongs) and complex finals (combinations). In this
section, I will discuss consonant syllabicity, vowel nasalization,
phonological behaviours of vowels in the syllable, and different patterns
and combinations of complex Finals in the SX syllables, based on the
Final inventory in (41).
2.3.2 Syllabic consonants Rules of syllabicity
In this subsection, I will examine the syllabic consonants and analyze the
phonological rules and constraints concerning the syllabicity of
consonants in SX. A syllabic consonant is a segment which has the
phonetic characteristics of a consonant but which, in a particular case,
functions as a syllabic nucleus (e.g. Trask 1996). SX has four sonorant
consonants that can be syllabic, which are listed in (43):
(43) mÆ nÆ N lÆ
The consonants in SX that can be syllabic are the bilabial nasal, the
alveolar nasal, the velar nasal, and the lateral liquid. These are all
distinctive sonorants in the SX initial position, which means that all the
phonemic sonorant consonants can be syllabic in SX. Syllabicity of
consonants is found in many languages. In English, for example, syllabic
sonorants appear after an obstruent in word final position, as shown in
(44) [»str√g] struggle
σ σ
µ µ µ
str √ g 
The example in (44) shows that in English, the liquid /l/, when in wordfinal position and preceded by an obstruent, is syllabic, making struggle a
disyllabic word [»str√g], in which the syllabic lateral [] is the nucleus of
the second syllable.
However, sonorant consonants in SX are syllabic in a different
environment from English. In SX, a sonorant becomes syllabic when
standing alone. It is always assumed that syllabic [mÆ ], [nÆ ], [N] and [lÆ] in
SX are monosyllables, without any other consonant or a schwa preceding
or following, to form a lexical syllable by themselves. For example:
(45) Lexical syllable
a. [mÆ 22] ‘yes’
b. [nÆ 33]([~jØ35]) ‘not’ (only used together with ‘have’)
c. [N13] ‘five’
d. [lÆ22] ‘also’
I assume that underlyingly, the lexical items in (45) consist of a
single nasal or lateral for the whole lexical syllable. In (45a), the word for
yes is a single syllabic bilabial nasal. The lexical syllable [mÆ ] does not
combine with any other syllable to form a phrase: it is a lexical word in
itself. In SX, the syllabicity of sonorants can be formulated as a
phonological rule, as in (46):
(46) [+son] → [+syll] / $ _ $
Rule (46) says that any sonorant becomes syllabic when it occurs on its
own between syllable boundaries underlyingly. This rule also includes
Like many other Chinese dialects, SX has no contrast between long and
short vowels. Thus, a feature like [± long] or an underlying mora
distinction between long and short vowels (or consonants) (cf. ch.1) plays
no role. Surface syllable weight is not decided at the phonemic level or by
the syllable structure (e.g. CV, CVC, or CGVC), but by stress. Stressed
syllables (as opposed to unstressed ones) may be signaled acoustically
with higher pitch, greater intensity, longer duration or some combination
of these (Selkirk 1984). However, stress in tone languages such as SX and
Mandarin plays a quite different role from that in stress languages such as
English and French. In tone languages, it is often difficult or impossible
for someone who is not a native speaker of the language to identify stress
functioning separately from tone: syllables may sound stronger or weaker
according to the tone they bear. Generally speaking, in tone languages
pitch is divorced from stress and prominence, which means that various
combinations of H and L (and sometimes also M=mid) tones may occur
in a single word. SX is a tonal monosyllabic language. Almost every
syllable is a word and every word has a stress to demarcate it as a lexical
unit. Underlyingly, every syllable is bimoraic in SX, but the weight of the
syllable is realized by stress. A lexical syllable must be prominent
underlyjngly in SX and its bimoraic status has to be realized by way of a
full tone, which means that only a bimoraic syllable is a stressed syllable.
Thus, stress in tone languages plays a role of realizing syllable weight in
the form of tone pitches. Accordingly, if a syllable bears a neutral or zero
tone, it has no stress. In SX, as in many other languages, all stressed
syllables are heavy and heavy syllables must be stressed (Prince 1990;
Prince & Smolensky 1993). In SX, not only are all stressed syllables
heavy, but also all unstressed syllables are light, and light syllables cannot
bear a full tone. In terms of mora structure, all heavy syllables are
bimoraic and all light syllables are monomoraic (Duanmu 1999, 2000;
Wang 1999). The correlation between syllable weight and stress in SX is
stated in (47):
(47) Weight-by-stress
Stressed syllables must be heavy and bimoraic; unstressed
syllables must be light and monomoraic.
The weight-by-stress principle is largely consistent with constraints
(STRESSTOWEIGHTPRINCIPLE) 23 (Prince 1983; Myers 1987; Prince &
Smolensky 1993). Neither WSP nor SWP, however, constrains unstressed
syllables. Weight-by-stress in (47) is a typical characteristic of moraic
syllable structure of SX. Weight-by-stress also differs greatly from
Weight-by-position (Hayes 1989). Weight-by-stress only allows stressed
syllables to be bimoraic no matter whether they are CV, CVV, V or even
just syllabic nasals, as in (45). In SX as well as in Mandarin, most of the
lexical syllables are stressed. Only some particles for grammatical
functions are always unstressed, such as [lF] to mark ‘past tense’ in
Mandarin and [go/] to mark the status of ‘adjective’ in SX. For example,
in [hAÅ35go/] ‘good’, the second syllable is unstressed because it is not
lexically meaningful. Their moraic syllable structure can be represented
as follows:
(48) σ σ
µ µ µ
[hAÅ35go/] ‘good’
The representation in (48) shows that the first syllable in [hAÅ35go/] is
stressed and is heavy and the second syllable is only an adjective marker
grammatically and is unstressed. The unstressed syllable is short and
always toneless or has a neutral tone because it is monomoraic. Thus, the
unstressed syllable is always a non-TBU in Chinese (Pulleyblank 1986;
Duanmu 2000a). The length of a vowel in SX is a morphophonological
matter rather than a phonetic matter. If the same syllable plays a different
lexical role or a sentential role, it has a different weight. For instance,
[go/3] in [go/3lAÅ31] ‘standstill’ is stressed and bimoraic, because it is a
lexical word. The moraic structure of the stressed [go/] in [go/3lAÅ31] is
represented as follows:
22 WSP (WEIGHTTOSTRESSPRINCIPLE): Heavy syllables must be stressed.
23 SWP (STRESSTOWEIGHTPRINCIPLE): Stressed syllables must be heavy.
(49) σ
µ µ
g o/ in [go/3lAÅ13]
According to Weight-by-stress, the syllabic sonorants in (45a), (45c)
and (45d) are all independent lexical words and are all stressed and
bimoraic because they are all full-tone syllables. For example, the moraic
syllable structure of [mÆ 22] ‘yes’ can be illustrated in (50):
(50) σ
µ µ
[mÆ 22]
Duanmu (1999, 2000a) claims that all heavy syllables (CVV,24 CVC,
CVG) of Chinese dialects are bimoraic and that all monomoraic syllables
are light (CV). This means that a syllable like CVC or CVG has to be
stressed, which is not the case with SX. This issue will be discussed in
chapter 4.
It should be noted that in the list of the syllabic sonorants in (45), all
syllables have low-register tones except [nÆ 33], which has a high-register
tone. Is it possible that a sonorant initial has a high-register tone in the
syllable? According to the onset-condition constraints in (26), no highregister vowel is preceded by a voiced obstruent onset. However, a voiced
initial sonorant can be a high-register syllable. Yip (2002) points outs that
in sonorant consonants the rate of vibration of the vocal folds is
controlled by a number of factors. Rotation of the thyroid and cricoid
cartilages with respect to each other can be deformed in several ways, and
as a result the vocal cords may or may not be stiff. So, articulatorily, a
24 ‘VV’ here refers to either a two-vowel sequence, or a long vowel, or a stressed
bimoraic vowel (Duanmu 1999).
syllable with a sonorant initial may have a low-register tone or a highregister tone. There are some examples in SX listed in (51):
(51) a. [lØ22] ‘leak’
b. [lØ52] ‘hollow out’
c. [mI/3] ‘extinguish’
d. [mI/5] ‘screw’
In the examples in (51), the syllables of (a) and (c) have low-pitch tones
and those of (b) and (d) have high-register tones.25 This fact is also captured by the configurations of Halle & Stevens’ (1971) laryngeal feature
specifications that sonorants are specified as [–stiff, –slack] and high
register is specified as [+stiff] or [–slack] and low register is as [+slack]
or [–stiff]. The consonant-tone correlation will be discussed in more detail
in chapter 5. The example in (45b) also shows that the syllable [nÆ 33] for
‘not’ is syllabic only when uttered in the combination [nÆ jØ35], which is
actually a disyllabic lexical item, meaning ‘not have/haven’t’. A question
may arise: is [ nÆ jØ35] a CGV monosyllabic unit? If not, what is its
phonological representation? In fact, the prosodic word [nÆ jØ35] deserves a
comment. I assume that [nÆ ] in [nÆ jØ35] is a clitic, which is a toneless
syllable by itself and is phonetically and phonologically fused with the
host syllable [jØ]. As a syllabic nasal, [nÆ ] is both a lexical syllable by itself
and the onset of the second syllable. The moraic structure of [nÆ jØ35] can
be represented in (at least) four possible ways, given in (52):
(52) a. σ σ
µ µ µ
n j Ø
b. σ σ
µ µ µ
n j Ø
c. σ σ
µ µ µ
n j Ø
d. σ σ
µ µ µ
n j Ø
The possible moraic syllable structures in (52) all show that the syllabic
nasal is the nucleus of the first syllable and also the onset of the second
syllable. The only difference is the location of the prenuclear glide [j]. It
can be assumed that [j] has one mora independently as [Ø] does, as in
25 According to the tonal structure of SX, the high register carries high-pitch tones of 52,
35, 33, 5 and the low register carries low-pitch tones of 31, 13, 22 and 3. The details of
the tonal system of SX will be discussed in chapter 5.
(52a), suggesting that [j] is in the Nucleus; or [j] is moraic but shares one
mora with [Ø], as in (52b), suggesting [j] is in the Nucleus but cannot bear
a tone by itself; or [j] is non-moraic and is in the Onset, as in (52c); or [j]
is non-moraic and is neither in Onset nor in Nucleus, as in (52d). I claim
that the status of the prenuclear glide [j] in [nÆ jØ35] is like in (52d), which
will be discussed in detail in chapter 4. However, there still remain some
mysterious questions as to how [nÆ ] becomes a clitic; why [nÆ ] in [nÆ jØ35] is
not palatalized as [] according to the nasal palatalization rule in (30);
why [nÆ jØ35] is disyllabic; and what the phonological motivation for the
syllable structure in (52) is. All these issues will be discussed in chapter 5
(see also Zhang 2005: 69-79).
In (45c), the syllabic velar nasal [N] has three different meanings
with different tones, representing three different lexical words, as listed in
(53) a. [N31] ‘fish’
b. [N13] ‘five’
c. [N22] ([t∏̃33~22]) ‘a dragon boat festival’
In (53c), the lexical syllable [N22] only exists in a fixed phrase for the
meaning of ‘dragon boat festival’, a traditional Chinese festival. All the
three forms of the syllabic velar nasal in (53) are stressed syllables
because they all bear full tones, so each can constitute a bimoraic syllable
by itself, according to Weight-by-stress, as illustrated in (54):
(54) σ
µ µ
[N31] ‘fish’
Nasals can be syllabic in many languages. However, it is not very
common for nasals to be syllabic in a monosyllable and by themselves as
in SX.
The liquid [lÆ] in (45d) is syllabic in much the same way as it is in
English. When it is syllabically articulated, the middle of the tongue is
excessively lowered so that the sound is pronounced with no intervening
vocoid. There is a big difference in articulation and acoustics between the
syllabic lateral and the non-syllabic lateral. In SX, the initial [l] and
syllabic [lÆ ] cannot appear in the same world because SX is a monosyllabic
language and has no coda [l]. Like English, the initial [l] and syllabic [lÆ]
in SX are allophones of the phoneme /l/ because the syllabic [lÆ] is only
possible when it occurs alone. It is in complementary distribution with the
initial [l]. For example:
(55) [la52] ‘pull’
[lAÅ13] ‘old’
[lu22] ‘road’
[ljaN22] ‘two (people)’
[lo/3] ‘green’
[lÆ22] ‘and/also’
The examples from (43) to (55) all show that sonorant consonants in
SX can be syllabic, as is stipulated by the rule in (46). Actually, all
consonants can be syllabic in one way or another, but only if there is no
other better peak available (Laver 1994; Ladefoged & Maddieson 1996;
among others) because every syllable must have a peak. I propose the
following peak principle:
(56) Peak Principle:
Segment α can be the syllable peak iff α is the most sonorant
segment in the syllable.
The Peak Principle in (56) says that any segment can be the peak of a
syllable in a certain environment. Vowels are always good peaks because
they are more sonorant than any consonant; sonorant consonants are likely
to be the peak because they are [+son]; obstruents can also be syllabic, but
very rarely so because they are the least sonorant. However, some
languages, such as Imdlawn Tashlhiyt Berber (Dell & Elmedlaoui 1985)
and Bella Coola (Bagemihl 1991), have been described as having syllabic
obstruents (see also Botma 2004: 263).
2.3.3 Vowels
As was mentioned at the beginning of this section, there are three
categories of “Finals”, one of which is the category of single vowels. In
this sub-section, I will present an analytic description of all the single
vowels in the surface representation of SX. I use the term ‘single vowel’
instead of monophthong to avoid any association with diphthongs, for
there has been a discussion in the literature whether there are diphthongs
and triphthongs in Mandarin and other dialects (Zhan 1991; Chan 1997;
Wiese 1997; and others). I claim that there are no triphthongs in Mandarin
and that there are no triphthongs or diphthongs in SX (see the next
subsection). Vowels involved in a combination will also be discussed in
the next subsection.
In this subsection I will discuss how single vowels are used as
rhymes in SX in surface representation, focusing on the surface vowels,
vowel nasalization, and their different distributions. In SX, there are ten
single vowels in the surface representation, used as rhymes in the
syllables. They are presented in the vowel diagram in (57):
(57) Ten single vowels in SX:
Front Back
High  i y u
e ∏ Ø o
Low a
In (57), the circled vowels are rounded. The vowel diagram in (57) offers
a clear picture of what single vowels26 may occur in the rhyme of SX
syllables and where they are located. SX has more front vowels than back
vowels and more unrounded vowels than rounded vowels, which is a
natural arrangement because front vowels are naturally unrounded. We
can easily acoustically locate these vowels in the above vowel diagram,
except [], which is a very remarkable phone in SX as well as in Mandarin.
[] is usually regarded as an apical vowel which exists in many Chinese
dialects. There has been a discussion of the phonetic and phonological
status of the apical vowel [] in Chinese (Karlgren 1915–1926; Chao 1968;
26 The ten vowels in (57) are all single vowels which occur alone as the Rhyme in the SX
syllables, excluding vowels which only occur in combinations such as [I], [´], [A], and
Kratochvil 1968; Ladefoged & Maddieson 1990; Wiese 1997). Wiese
(1997) assumes that [] in Chinese is not a vowel, but a syllabic fricative.
I argue that [] in SX as well as in Mandarin is an apical vowel and an
allophone of /i/. I will present my analysis of the apical vowel [] in the
following subsection. Apical vowel
Apical vowels are phonetically vowels, which are produced with the tip
of tongue touching the anterior portion of the palate. Thus, they are also
called fricative vowels (Ladefoged & Maddieson 1996). The contact
location is in the denti-alveolar. There are several apical vowels in
Chinese dialects such as [   ].27 In SX, there is only one apical vowel
[], as shown in (57). Its phonetic and phonological status and its
distribution in SX are very similar to that in Mandarin. Wiese (1997: 239)
claims, for Mandarin, that [] is a pseudo-sound that should not have any
place in either a phonological or a phonetic description. He regards [] as
a syllabic fricative, as shown in (58):
(58) Mandarin Wiese’s assumption
a. [s51]28 ‘four’ [sZ]
b. [Ω51] ‘day’ [ΩΩ ]
c. [s33] ‘four’
d. [z22] ‘word(s)’
Wiese argues that [] in (58a) and (58b) is “a syllabic consonant
identical in place and continuancy to the preceding fricatives”. However,
Wiese does not give any evidence for denying [] the status of a vowel.
He proposes a filter (1997: 242) to rule out high vowels preceded by
[+cor] consonants, as illustrated in (59):
(59) * + cons – cons
– back – back
+ cor + high
27 [] and [] are also apical vowels in some Chinese dialects. They are the rounded
counterparts of [] and [] respectively, differing from the rounded front glide [].
28 The syllables in (58a) and (58b) carry Mandarin tones, which are different from those
in SX.
Wiese’s filter in (59) attempts to state that after [+cor] consonants,
front high vowels are not acceptable so that [] may be a syllabic fricative.
However, he realizes that the filter cannot be correct as stated in (59)
because [t] and [th] are also [+cor] and [ti] and [thi] are well-formed
syllables in both Mandarin and SX. The main reason for Wiese’s
assumption is that [in] and [iN] are well-formed and wherever [i] is
acceptable, [in] and [iN] are also acceptable with the same initial
consonants in Mandarin, such as [pi55] ‘close’, [pin55] ‘guest’ and [piN55]
‘soldier’ while *[n] or *[N] is never possible in any case. However,
whether [i], [in] or [iN] can be preceded by the same consonant is simply
a matter of phonotactics. For example:
(60) Mandarin SX
a. [thi55] ‘shave’ [thi33] ‘shave’
*[thin] *[thIn]
[thiN55] ‘listen’ [thIN33]29 ‘listen’
b. *[si] *[si]
*[sin] *[sIn]
*[siN] *[sIN]
The examples in (60) show that /i/ and /iN/ can occur after /th/, but /in/
cannot, while /i/, /in/ or /iN/ cannot occur after /s/30 in either Mandarin or
SX, because of their phonotactics. Wiese claims that the nucleus preceded
by /s/ in (58) must be a syllabic fricative [Z ] because [sZ] is acceptable, as
shown in his argument in (58), while *[sZ ̀n] or *[sZ ̀N] is not found. This is
ill-formed only because of the Sonority Sequencing Principle (SSP).31
Wiese argues that if the nucleus in (58) were a vowel, [n] or [N] should
also be allowed in Mandarin. However, the examples in (60) prove that
[V], [Vn] and [VN] could have a different distribution. There are more
examples such as in (61):
29 [I] is an allophone of /i/ in SX, which will be discussed in chapter 3.
30 Here ‘//’ is used instead of ‘[ ]’ for the purpose of indicating that the same underlying
phonemes are involved in both Mandarin and SX.
31 SSP: Sonority increases towards the syllable peak and decreases towards the syllable
margins (see Clements 1990; Roca 1994; Morelli 1999).
(61) Mandarin SX
a. [mo51] ‘mill’ [mo22] ‘mill’
b. *[mon] *[mon]
c. *[moN] [moN] ‘dream’
The examples in (61) show that [o] is acceptable after [m], but [on] or [oN]
is not possible after the same consonant in Mandarin while in SX [on] is
not possible after [m]. However, it would be senseless to argue that [o] in
[mo] is not a vowel but a syllabic consonant because there is no *[mon]
or *[moN]. In SX, [Ø] can occur after different onset consonants, but [Ø]
can never be followed by any consonant, disallowing any combination of
*[ØC] in surface representation, such as *[ØN] and *[Ø/], while [VN] and
[V/] are well-formed combinations in SX. For example:
(62) [dØ31] ‘head’ *[dØN] *[dØ/]
[fØ35] ‘deny’ *[fØN] *[fØ/]
[lØ22] ‘leak’ *[lØN] *[lØ/]
[tsØ35] ‘walk’ *[tsØN] *[tsØ/]
[khØ33] ‘button’ *[khØN] *[khØ/]
However, there is no reason to doubt that [Ø] is a vowel. Neither
Wiese’s filter in (59) nor his argument for *[sz̀n] and *[szǸ] can support
his denying [] the status of a vowel. If, as Wiese assumes, [] is a syllabic
fricative identical in place and continuancy, it must be an allophone. Then
what is the underlying segment― a phonemic /z/ or /¸/? If a syllabic
fricative itself is a distinctive phoneme, what is its phonological property
as an underlying phoneme? The facts suggest the contrary, viz. [] is a
vowel and phonologically is in complementary distribution with /i/ in
both SX and Mandarin. Phonetically, [] has formant structure, according
to Howie (1976).
Some Chinese scholars (see Li, Yu, Chen & Wang 2004: 257–258)
present a comparative analysis of vowel formants between Standard
Chinese (SC) and Shanghai-Accented Standard Chinese (ASH),32 among
which the formants of [i] and [] pronounced by male and female are
shown in the following table:
32 SC refers to Mandarin and SAH refers to the standard Chinese spoken by Wu native
speakers, like SX natives.
(63) Average values of F1, F2, F3 of [i] and [] by male and female
Vowel F1 (Bark) F2 (Bark) F3 (Bark)
[i] 3.11352 2.96491 13.6248 14.04572 16.10078 16.20306
[] 4.05798 3.90083 9.92734 10.55437 15.19762 15.3489
[i] 3.35486 3.28039 15.21406 15.32132 17.0888 17.42927
[] 4.50622 4.1378 11.0423 12.19668 15.76756 16.05551
Table (63) shows the formants of [i] and [] for male and female speakers
of Mandarin and Wu, indicating that [] in both Wu and Mandarin is a
vowel in aspect of acoustics. [] is an allophonic vowel which is made
with the tip of tongue touching the upper articulator and is thus called
apical vowel. [] only occurs after dental fricatives and affricates in SX,
as shown in (64):
(64) [ts35] ‘paper’
[dz22] ‘late’
[tsh33] ‘wing’
[z22] ‘word
[s33] ‘try’
Following Bright (1978), Ladefoged and Maddieson (1996: 163)
regard “the dental sibilant as being apical”.34 Presumably, the reason that
[] only occurs after [ts tsh dz s z] in SX is that the apical vowel [] shares
with the apical sibilants the [+apical] feature, 35 as proposed by
Williamson (1977). Thus, the apical vowel is made with the tongue in
essentially the same position as in the corresponding fricatives or
affricates. In short, [] is a vowel, whether it is called fricative vowel or
apical vowel, but an allophonic vowel of the phonemic /i/ in SX as well as
33 BJ refers to Beijing speakers who speak SC; SH refers to Shanghai (Wu) speakers
who speak ASH.
34 According to Trask (1996), sibilants refer to fricatives and affricates and the feature
[+sibilant] is similar to [+strident] by nature. In SX, the dental sibilants include [ts tsh dz
s z].
35 In Williamson’s (1977) feature system, [±apical] is used as a feature to replace
[±distributed] in SPE. Neither [apical] nor [distributed] is used for the feature
specifications of the 29 consonants in SX, as shown in (36).
in other Chinese dialects. This can be expressed by a phonological rule, as
presented in (65):

/ +cons +apical
Rule (65) shows that the high front vowel /i/ becomes an apical vowel []
when following an apical consonant. The distribution of [] and [i] will be
discussed in chapter 3. Other vowels
Out of the ten single vowels, as shown in (57), I claim only six are
underlying phonemic vowels, viz. /i/, /u/, /e/, /Ø/, /o/ and /a/, which
contrast with each other, as shown in (66):
(66) [di22] ‘earth’
[du22] ‘ferry’
[de22] ‘pocket’
[dØ22] ‘bean’
[do22] ‘big’
[da22] ‘wash in pan’
The examples in (66) show that the six vowels, /i/, /u/, /e/, /Ø/, /o/ and /a/
all can occur alone as the rhyme in an open syllable contrasting with each
other. This suggests that they are all phonemic vowels. The underlying
vowel inventory will be discussed in detail in chapter 3. Another vowel [y]
can also occur as the rhyme of a syllable in surface representation,
although its distribution is almost as limited as []. It can only occur after
the lateral [l], the five alveolo-palatal affricates and fricatives and the two
phonetic onsets [] and [/], leaving many systematic impossibilities.
Some examples are presented in (67):
(67) [ly31] ‘donkey’
[tÇy33] ‘expensive’
[tÇhy35] ‘take’
[dÛy22] ‘live’
[Çy52] ‘book’
[Ûy22] ‘tree’
[/y33] ‘silted’
[Hy13] ‘rain’
I assume that the front high rounded vowel [y] is not a phonemic
vowel because of its limited distribution. This gives rise to the question as
to what the underlying vowel is in case of surface [y]. I will argue that the
underlying form of [y] is not a single phonemic vowel, but underlyingly a
GV combination /iu/ or /wi/, which merges into [y] in surface representation because of an OCP constraint. This issue will be discussed in the GV
sub-section (§, where more details of vowel distribution and the
phonological motivation for /iu/ or /wi/ to merge into [y] will be discussed (see chapter 3).
2.3.4 Vowel nasalization
Some vowels have to be, or can be, nasalized when occurring as finals in
SX. Among the vowels represented in (57), there are three vowels that
can also occur nasalized in the rhyme, viz. [ẽ], [Ẽ] and [∏̃]. Consider the
following examples:
(68) [khẽ33] ‘look at/watch’
[p∏̃33] ‘half’
[pẼ52] ‘class’
Of the three examples in (68), only [ẽ] has an oral counterpart [e], a
phonemic vowel that can appear alone as the rhyme. The other two
vowels, [Ẽ] and [∏̃], have no oral counterparts appearing alone as the
rhyme in surface representation. I assume that the three nasalized vowels
in (68) are not underlying vowels. Vowels usually become nasalized in a
phonetic environment that is conducive to nasalization. For example, in
English, [Q] is nasalized to [ œ̃ ] in [T œ̃ Nk] ‘thank’ because of the
following nasal. In most cases, vowel nasalization is predictable. Why do
the three vowels in (68) have to be nasalized in a syllable of SX?
Compare the cognates in (69) of Mandarin and SX, both of which
developed from Middle Chinese:
(69) Mandarin SX
[nan35] [nẼ31] ‘difficult’
[hwan35] [vwẼ31] ‘return’
[thwan55] [th∏̃33] ‘swallow’
[khun51] [khw∏̃33] ‘sleepy’
[pen35] [bẽ22] ‘pail’
[kan35] [kẽ35] ‘catch up’
The examples in (69) show that in Mandarin all these syllables end with
an alveolar nasal, whereas in SX the rhymes consist of nasalized vowels,
missing the final alveolar nasal in surface representation (which confirms
CODA-COND (16)). This phenomenon suggests that the underlying forms
of the syllables above in SX could also have a final alveolar nasal. Strong
evidence for this analysis comes from examples of nasal gemination
between a syllable ending in a nasal or a nasal vowel and a syllable
beginning with a vowel, as shown in (70):
(70) Monosyllable Disyllable
a. [tshoN52] ‘dash’ [tshoN.a]36 → [tshoN52.Na] ‘Dash!’
b. [nẼ31] ‘difficult’ [nẼ.a] → [nẼ] ‘Difficult!’
c. [thjẼ52] ‘heaven’ [thjẼ.a] → [thjẼ] ‘God!’
The syllables in (70) are subject to certain phonetic and phonological
environment where liaison between the syllables occurs. Usually liaison
does not occur between a consonant-final syllable and a vowel-initial
syllable if both syllables have full tones. I assume it is because a full-tone
syllable must have an onset which is required to assign the register feature,
as is stipulated by the onset-condition constraints in (26). Liaison or
gemination never occurs between two full-tone syllables in SX (the
details of the process of liaison and gemination will be discussed with
more examples in chapter 4). Liaison or gemination only occurs when the
second syllable is toneless and has no phonological onset. The syllable
[tshoN] in (70a) ends with a velar nasal. In the disyllabic form of (70a), the
first syllable-final velar nasal is followed by another syllable [a], which
has no real lexical meaning but only expresses a certain emphasis, so that
it is an unstressed syllable and is toneless. In this context, liaison triggers
gemination of the final nasal in the preceding syllable, or phonetically, the
36 The dot “.” here indicates a syllable boundary.
final nasal [N] becomes ambisyllabic, producing a phonetic onset of the
following syllable. In this way, the prosodic word [tshoN.a] is uttered with
a geminated velar nasal as [tshoN.Na]. In this form, [N] does not
autosyllabically become the onset of the second syllable because [N] is
still required to be the coda of the first syllable for its integral bimoraic
status as a TBU. As for the examples in (70b) and (70c), I assume that the
underlying forms of the surface nasalized vowels are VN combinations,
which, when followed by a toneless vowel-initial syllable, triggers the
phonetic restitution of the lost syllable-final nasal, which is realized as
nasal insertion, as shown in the disyllabic forms in (70b) and (70c). In
connected speech, there can be a nasal consonant in the onset of the
second syllables following SX words such as those in (70), which
strongly suggests that both [nẼ] in (70b) and [thjẼ] in (70c) also have a
final nasal in the underlying form.
There is cross-linguistic evidence that the surface syllable-final nasal
vowels are underlying sequences of an oral vowel + a nasal consonant.
For example, Portuguese nasal vowels have been analysed as
phonological vowel + nasal consonant sequences and the syllables ending
in a nasal vowel are treated as closed syllables (Parkinson 1983). More
details of the process of the phonological change of the vowel
nasalization will be discussed in chapter 4. However, the syllable
structure of the final should be VN([+cor]) underlyingly, viz. a vowel
followed by an alveolar nasal. The final alveolar nasal in SX was
debuccalized through historical attrition (nasal debuccalization will be
discussed in chapter 4) so that it drops off and the [+nasal] feature spreads
to the preceding vowel, producing vowel nasalization. This can be
expressed by the [+nasal] feature spreading (see van de Weijer 1994: 200),
as illustrated in (71):
(71) V C
Place [stop]
The feature-spreading structure in (71) can also be represented by a vowel
nasalization rule, as follows:
(72) V → Ṽ / /__ N/
The vowel nasalization rule in (72) shows that a vowel becomes nasalized
when followed by a nasal underlyingly, which suggests that the deep
structure of the surface [kh ẽ 33] ‘watch/look at’ is /khVn33/ in SX.
Nasalized vowels are just the surface forms. The phonological motivation
for vowel nasalization will be discussed in detail in chapters 3 and 4. In
chapter 3, I will present my analysis of the underlying vowels of the three
surface nasal vowels, and in chapter 4, I will discuss why vowel
nasalization only happens in the case of [+cor] nasals in the coda, and not
with a [+dors] nasal in SX.
2.3.5 Complex Finals
There are four kinds of complex finals in SX. They are GV, VC, VV and
GVC. SX has a limited number of rhymes, mainly because there are very
few postnuclear consonants in the coda position. In this section, I will
discuss all these four types of final combinations. I will introduce the
Diphthong Constraint and the OCP(H) constraint in SX. There has been
discussion in the literature whether there are diphthongs and triphthongs
in Mandarin and other dialects (Zhan 1991; Chan 1997; Wiese 1997; and
many others). I claim that there are no triphthongs in Mandarin and that
there are no triphthongs or diphthongs in SX. I will present my analysis of
these issues in this section. GV
GV is a combination of a glide and a vowel. In SX there are 12 such
combinations as surface syllable finals. They are: [jo], [ja], [je], [jØ], [wa],
[we], [wo], [jẽ], [jẼ], [wẼ], [w∏̃], and [Á∏̃], five of which have a nasalized
vowel. Of these 12 GV combinations, five are preceded by the glide [w],
six by [j], and one by [Á], which is the glided counterpart of the rounded
vowel [y]. In SX [Á] only occurs in combination with [∏̃], which is also a
[-back] rounded vowel. I assume that [Á] is in complementary distribution with [j] because [j] is more often used and mostly precedes
unrounded vowels such as [Ø], [a], [e], and [Ẽ], with the exception of back
rounded [o]. This phenomenon suggests a rule or an equivalent constraint,
as in (73) and (74), respectively:

__ +round
or (74) *[j] +round
The rule in (73) says that /j/ becomes rounded when followed by a [–back]
rounded vowel, and the constraint in (74) rules out such combinations as
*[jy] and *[j∏], permitting the existence of [jo] in SX. This roundedness
rule is well supported by the existence of [Á∏̃] in SX, which, however,
gives rise to the question why [j] in [jo] does not get rounded. There is
cross-linguistic evidence that [–back] roundedness is marked and [+back]
roundedness is unmarked, which suggests that marked features can
trigger assimilation more strongly than unmarked features.
The rule in (73) makes it possible that [Á ∏̃ ] exists in SX.
Undoubtedly, [Á] is an allophone of the underlying glide /j/ in [Á∏̃]. The
details of the distribution of glides will be discussed in chapter 3. All the
12 GV combinations are well-formed in lexical syllables when preceded
by an onset. For example:
(75) [Çjo33] ‘blood’ [˙we31] ‘return’
[˛ja33] ‘write’ [˙wo22] ‘speech’
[Á∏̃13] ‘soft’ [gwẼ22] ‘circle’
[t˛hjØ33] ‘autumn’ [hw∏̃33] ‘happy’
[pjẽ52] ‘edge’ [˙wa22] ‘bad’
The examples in (75) show that all these combinations begin with a glide
[w], [j] or [Á] in surface representation. The syllabic status of the prenuclear glides has been the topic of many Chinese linguistic studies. I claim
that the prenuclear glides in SX are not in the onset (which will be
discussed in detail in chapter 4). This gives rise to the question whether
these well-formed GV combinations are in the nucleus and are therefore
diphthongs. The topic of the representation of diphthongs is a
controversial one (cf. Laver 1994; Casali 1996; Trask 1996; Wiese 1997).
According to Trask (1996), a diphthong is “a single syllable nucleus
which begins with one vowel quality and changes more or less smoothly
to a second quality, as in [ju] and [aj]. Usually one of the two vocalic
elements is more prominent than the other, this other consisting only of a
preceding glide (an on-glide, as in [ju]), or a following glide (an off-glide,
as in [aj])”. I formalize Trask’s definition of a diphthong as follows:
i. Diphthongs form the nucleus of a syllable.
ii. Within the diphthong, the vowel quality changes (involving high,
low, back and front).
iii. Either part of the diphthong is a glide.
GV combinations in SX look like diphthongs in many other
languages, satisfying Trask’s properties (ii) and (iii). Traditional Chinese
phonology claims that there are not only diphthongs but also triphthongs
in Mandarin as well as in other Chinese dialects (Wang 1963, 1985; Chan
1985; Norman 1988; Wiese 1997; and others). This claim is mainly based
on the acceptable combinations in Mandarin, such as [ai], [au], [wa], [wo],
[ei], [ou], [wei], [wai], [jau], [jou],37 etc. The major motivation to assume
that Mandarin has diphthongs and triphthongs is to deal with the on-glides
in the syllable structure because every ‘diphthong’ is a combination of a
monophthong and either an on-glide or an off-glide (GV or VG) and
every ‘triphthong’ is a combination of a monophthong and both an onglide and an off-glide (GVG) in Chinese.
Wiese (1997) proposes an underspecification analysis of the Mandarin
vowel system and claims the existence of triphthongs in Mandarin,
regardless of their position in the syllable structure. On Wiese’s (1997:
219) assumption, there are at least four well-formed triphthongs in
Mandarin, viz. [jau], [wai], [jou], and [wei], all of which have a GVG
structure. Glides play a crucial role in forming a diphthong or triphthong.
It is essential to identify the phonological status of glides in the syllable
structure when determining whether the sequence is a diphthong or
triphthong. As widely claimed, a diphthong must be the single syllable
nucleus. However, as attested in all the Chinese literature, is that the
prenuclear glide in GV or GVG is never included in the Chinese poetic
rhyming. Consider the examples of Mandarin in (76):
(76) A B
a. [t˛ja55] ‘family’ [ma55] ‘mother’
[kwa55] ‘melon’ [la55] ‘pull’
37 All the off-glides in Chinese dialects are written as [i] and [u], rather than [j] and [w],
because I assume that VG in Chinese is a diphthong.
b. [xwai35] ‘chest’ [tshai35] ‘finance’
[kwai55] ‘well-behaved’ [khai55] ‘open’
c. [thjau55] ‘choose’ [thau55] ‘hollow out’
[phjau55] ‘flow’ [phau55] ‘throw’
In the examples in (76), all words in column A have prenuclear glides and
those in column B have no prenuclear glides. The A-B pairs of words in
groups (a), (b) and (c) rhyme very well with each other.38 Obviously, the
prenuclear glides are not in the rhyming unit in Mandarin, which is also
true in SX. Besides, the prenuclear glides are not TBUs since they are
non-moraic (Howie 1976), so that they are not in the nucleus constituent.39 The nucleus constituent must be fully counted towards the rhyme.
In a prosodic unit, a diphthong must be an integral constituent of Nucleus.
Thus, any combination of GVG in Mandarin can never be a phonological
triphthong; nor can GV phonologically be a diphthong. In all the GV
combinations of SX, G is never counted in the rhyme and there is no VG
combination in SX, as shown in (75). The evidence from the data strongly
suggests that there are no phonological triphthongs in Mandarin and that
there are neither triphthongs nor rising diphthongs in the SX surface
representation. I propose that only (falling) VG is a real diphthong and
that (rising) GV is not a diphthong in Mandarin or any other Chinese
dialect. VV
In SX, there is only one VV form in complex final combinations, which is
[AÅ]. It is very frequently found in syllables with an initial consonant, such
as the following:
(77) [sAÅ35] ‘little/few’
[pAÅ52] ‘wrap’
[˛jAÅ35] ‘small’
[d¸jAÅ31] ‘bridge’
[NAÅ31] ‘tolerate’
38 This is the Chinese poetic rhyming pattern.
39 The position of prenuclear glides in the syllable structure will be discussed in more
detail in chapter 4.
VV combinations as the nucleus of a syllable can be found in many
other languages such as [E´] in English (Heffner 1949) and [çå] and [eå]
in German (Rohler 1999, from International Phonetic Association,
Corporate Author International Phonetic 1999). In terms of diphthongs,
one element of the combination is always a glide in GV or VG, according
to Trask’s definition, as discussed in the previous sub-section. However,
VV combinations such as [E´], [çå] and [eå] are also regarded as
diphthongs, because there is a vowel quality change or tongue movement
from the first vowel to the second vowel and both vowels form a single
nucleus of the syllable in English and German, respectively. More
specifically, they are falling diphthongs since the second, schwa-like,
element has the weaker intensity. The consideration of [E´], [çå] and [eå]
as diphthongs challenges Trask’s diphthong definition (ii). Moreover,
whether [AÅ] in SX is a diphthong remains a question.
Before answering this question, I first introduce a well-attested
principle: the Obligatory Contour Principle (Leben 1973; Goldsmith 1976;
McCarthy 1986; Yip 1988; Wiese 1997):
(78) Obligatory Contour Principle (OCP)
Adjacent melodic autosegments cannot be identical.
The validity of the OCP in (78) has been confirmed in numerous studies,
not only for features of tone but also for segmental features (see (80)
below)). It is assumed that every feature occupies a tier of its own,
ultimately associated with a root node, R, which encodes, by bundling
temporally co-occurring segmental features, the notion of a segment and
is also situated on a tier of its own (Wiese 1997). If the features of the two
adjacent segments of a diphthong are the same, there will not be any
tongue movement, which should begin “with one vowel quality and
changes more or less smoothly to a second quality (Trask 1996).” [E´] in
English is a diphthong in which the tongue begins from front mid and
changes to central mid, as is illustrated in the vowel chart in (79):
(79) i y µ u
e O F o
E ø √ ç
a Ø A Å
Singh & Singh (1976) propose an important criterion for diphthong
formation in that “diphthongs involve an appreciable amount of tongue
body movement within the perimeter of one syllable”. However, [AÅ] in
SX is composed of [A] and [Å], which are identical in height and position
and only differ in [±round] so that they are regarded as a pair of vowels
(cf. (79)). There are generally eight pairs of such vowels, as shown in (79),
none of which is used or regarded as a diphthong in the world’s languages
according to Ladefoged and Maddieson (1996). Casali (1996:40) claims
that V1 and V2 in a diphthong must differ in at least two features of
dimension (height and position) and regards sequences like /ae/ and /ao/
as ill-formed because the two vowels are “not sufficiently distinct from
each other”. I argue against Casali’s claim that if there is a place
movement between V1 and V2 in a vowel combination and if the
combination is the integral nucleus of a syllable, it is a well-formed
diphthong. In my view, /ae/ and /ao/ are diphthongs if they occur as a
single nucleus in a language since there is a tongue movement in their
articulation. However, accepting /ae/ and /ao/ as diphthongs will
challenge Trask’s diphthong criterion (iii). With the [AÅ] combination, no
tongue movement is involved in its articulation. It is well attested in the
literature of the world’s languages that a well-formed diphthong is always
faithful to the OCP in terms of features of height and position and that
violation of the OCP always fails to produce a well-formed diphthong
(Casali 1996; Ladefoged & Maddieson 1996; Trask 1996; Wiese 1997).
For example:
(80) a. [E´] R R
+front –front
b. [ai] R R
–high +high
c. *[iy] R R
+front +front
(or +high +high)
d. *[AÅ] R R
+back +back
(or +low +low)
The examples in (80) show that the OCP must be observed in the
formation of a diphthong. We conclude that [AÅ] is not a diphthong in SX,
but a true VV combination. What matters whether [AÅ] is a diphthong or
not is a matter of a vowel system of SX. The exclusion of [AÅ] as a
diphthong stipulates the *DIPHTHONG constraint in SX, which helps to
decide that the merged [y] is from [wi] or [ju] rather than [uj] or [iw].
The latter is a well-formed diphthong and does not violate OCP (H) (see
(83) below), so that they will not merge into [y], as it is in Mandarin. VC
In this subsection, I will give a brief introduction to all the surface VC
forms which constitute the rhyme of a syllable in SX. There are altogether
ten such complex finals in VC structure in SX. They are [´N], [IN], [aN],
[ÅN], [oN], [o/], [a/], [´/], [I/], [E/]. In these ten surface VC combinations,
the final C is always either one of the two consonants: the velar nasal /N/
or the glottal stop ///. This satisfies CODA-COND in (16) which says that
the coda in the SX syllables can only be [/] or [N]. I list some examples of
CV combinations in (81):
(81) [n´N31] ‘able’ [˙´/3] ‘join’
[t˛IN52] ‘gold’ [zI/3] ‘enter’
[NaN22] ‘hard’ [ba/3] ‘white’
[zÅN31] ‘taste’ [pE/5] ‘eight’
[loN31] ‘dragan’ [mo/3] ‘ink’
The examples in (81) show that the ten VC finals can be divided into two
groups: one with five VN combinations, and the other five with V/
combinations. The vowels of the two columns of VC combinations are
almost the same except for one case in each set, which is [Å] in [ÅN] and [E]
in [E/]. The distribution of vowels in combinations will be discussed in
chapter 3 and the details of the phonotactic constraints of segment
sequences in SX will be discussed in chapter 4. GVC
The maximal final combination structure in SX is GVC. This means that a
vowel is preceded by a glide and followed by a consonant. There are
altogether nine such complex finals: [jaN], [waN], [wÅN], [joN], [woN],
[wo/], [jo/], [ja/], [wE/]. As with the VC structure, the GVC structure
can only have [N] and [/] in the final C position. I list some examples in
(82) [˛jaN52] ‘scent’ [khwo/5] ‘wide’
[vwaN31] ‘horizontal’ [t˛hjo/5] ‘lack’
[kwÅN52] ‘light’ [t˛ja/5] ‘foot’
[d¸joN31] ‘poor’ [vwa/3] ‘slide’
[˙woN31] ‘red’
The examples in (82) show that there are five GVN and four GV/ finals in
SX. In § and §, I have discussed the structures GV and VC
in complex finals of SX. GVC could be either a combination of GV plus
C or G plus VC. On the basis of the data in (82), I assume that the GVC
structure in SX is G plus VC rather than GV plus C, because we have
more VC combinations with the same phonemes than combinations of
GV with the same phonemes as those in GVC. This assumption is also
well supported by the fact that GV in GVC is not an independent single
constituent in SX, while VC is the integral rhyming unit in GVC since G
is excluded from the nucleus constituent in SX. So, I may formalize:
GVC is G + VC. The status of the syllabic structure of G will be
discussed in detail in chapter 4.
The data in (82) also show that *[juN] and *[wuN] do not exist in SX
just as there is no *[wu], *[ju], *[wi], or *[ji] in GV combinations in SX. I
assume such GV combinations are not acceptable in the SX surface
representation because they violate OCP for height features. This suggests
a specific OCP constraint on height in SX, as stated in (83):
(83) OCP(H)
Two adjacent segments cannot be identical in height in GV
In (78), I introduced OCP as one of the criteria of diphthong
formation, which mainly involves two vowel parameters, height and
position. OCP(H) only involves height. OCP(H) is a surface constraint
which only applies to GV combination in surface representation, ruling
out *[ji], *[jy], *[ju], *[Ái], *[Áu], *[wi], *[wy] or *[wu] on the surface,
while underlyingly /ju/ and /wi/ could be possible since we do have /jo/,
/je/, /jØ/, /ja/, /we/, /wo/ and /wa/ in SX, formalized in a combination of
GV underlyingly. However, I assume that OCP in both height and
position (OCP(H&P)) when applied to GV combination is an underlying
constraint which rules out such combinations as */jy/, */ji/ and */wu/.
OCP(H) is a remarkable phonological constraint in SX, since [ju], [Áu]
and [wi] are well-formed combinations or sequences in surface
representation in some other Chinese dialects such as Mandarin. I list
some examples to show how the OCP(H) constraint works in SX for
words of the same lexical meaning in the same combination in Mandarin:
(84) Mandarin SX
[tÇju35] [tÇjØ33] ‘nine’
[ji35] [Hi31] ‘move’
[Áu214] [Hy13] ‘rain’
[khwi55] [khwe52] ‘loss’
The fact that SX has such syllables as in (84), different from
Mandarin, is not a coincidence. Because of the OCP(H) constraint in SX,
if there should be any GV combination identical in height underlyingly,
some adaptation should be made in surface representation to satisfy
OCP(H), as shown in the examples in (84).
There is another remarkable final combination in SX, [jAÅ], which
has remained undiscussed so far. In fact, [AÅ] in SX behaves like a long
vowel phonologically, though it is a combination of [A] and [Å]
phonetically, as discussed above, so that [jAÅ] behaves the same as a GV
combination in terms of syllabic constituency. This will be discussed in
chapter 3.
2.4 Phonemic [/] in SX
Although the initial glottal stop [/] in SX and other Wu dialects, as well
as in other languages in the world, has been a problematic segment in
terms of its phonological status in the syllable-initial position of the
languages that have it, there seems not to be much discussion over the
syllable final [/]. SX has the glottal stop [/] both in the initial and in the
final, possible even in the same syllable, as shown in (85):
(85) [/E/5] ‘duck’
[/I/5] ‘one’
As was discussed in §2.2.2, the initial glottal stop [/] in SX plays the
phonetic role of indicating the beginning of an underlying onsetless highregister syllable, rather than any phonological role in the syllable structure.
However, the status and properties of the syllable-final [/] differ
phonologically from that of the initial [/]: it is phonologically present as
the syllable final while it only has phonetic status as the syllable initial.
This can be supported by the fact that there are near minimal pairs with
and without the final [/] in SX. For example:
(86) a. [to/5] ‘supervise’ c. [ba/3] ‘white’
b. [to52] ‘many’ d. [ba31] ‘card’
The examples in (86) show that (a)-(b) and (c)-(d) are two near minimal
pairs (differing in tones). It is also true that syllables with different tones
have different lexical meanings in SX. Tones and segments are the
integral syllable constituents which decide the lexical meaning. The
syllables of (86a) and (86c) both have the tone feature [h] and have the
final glottal stop [/] which is dominated by one mora (the moraic
structure of syllables will be discussed in chapter 4). The syllables of (86b)
and (86d) both have the tone features [hl] with the tone pitch falling down.
It is obvious that the final glottal stop [/] plays a phonological role in
deciding the syllable tone feature thus deciding the lexical meaning of the
syllable. We therefore assume that it is underlyingly present.
According to Qieyun, a book of ancient Chinese phonology, Middle
Chinese had [p], [t], [k], [m], [n] and [N] in the coda. Syllables ending in
[p], [t] or [k] fell into one group and had entering tone,40 similar to that of
Modern SX. In this way, the syllables with the same tone but different
final stops had different lexical meanings, because the different stops in
the coda were distinctive phonemes. Some modern Chinese dialects still
have [p], [t] or [k] in the coda such as Xiang, Hakka, Min and Cantonese
(Zhan 1991). These three final stops – [p], [t] and [k] in Middle Chinese –
merged into the glottal stop [/] in the coda in modern Wu dialects through
historical attrition. However, it is believed that the disappearance of the
final stops had a significant influence on the tones in Wu dialects (Zhan
1991; Cao 1998, 2002, 2004). In one way or another, the final glottal stop
[/] still has a similar phonological function to that of the final [p], [t] or [k]
in Middle Chinese. It is widely agreed that Chinese underwent a process
by which the voicing distinction on initial consonants was transformed
into a tonal distinction, doubling the number of tones (Hombert 1978; Yip
2002, also see above). I assume that tones developed as a result of the loss
of some phonemes originally. Much cross-linguistic evidence shows that
a syllable-final consonant influences the tones on the preceding vowels
(Haudricourt 1954; Pulleyblank 1962; Wang 1963; Hombert 1978; Baxter
1992; Chen 2000). Hombert (1978: 92) also finds that the effect of a
glottal stop on the pitch of the preceding vowel is widely attested, e.g. in
Vietnamese, Burmese, and Middle Chinese.
On the basis of the diachronic and synchronic studies, I assume that
the syllable-final glottal stop [/] (which results from the debuccalization41
of the syllable final stops [p], [t] and [k] in SX as well as other Wu
dialects) has phonological status in the syllable structure, making [to/]
and [to] or [ba/] and [ba] a pair of different lexical items. The details of
the debuccalization of the syllable-final stops will be discussed in chapter
40 The term entering is one of the four tones, viz. even, rising, going and entering, which
are translated from the traditional Chinese terms, ping, shang, qu and ru, respectively.
The entering tone only occurs on the syllables ending with a stop in Chinese. The
Chinese tonology will be discussed in chapter 5.
41 Debuccalization, also called deoralization, is a phonological process in which a
consonant segment loses its oral articulation (see Humbert 1995; Trask 1996). The
debuccalization of the SX syllable-final stops and nasals will be discussed in chapter 4.
2.5 Summary
Through phonetic observation and phonological analysis of the
consonants and vowels of SX in surface representation, focusing on their
distribution, phonetic and phonological realizations, constraints and rules,
I presented the total of 29 surface consonants in SX, as shown in (35), and
presented the feature specifications for the 29 consonants in (36). I have
also presented a clear picture of all the 48 Finals, as shown in (41), in the
forms of syllabic C, V, VV, GV, VC, and GVC, which remarkably
excludes a combination of VG in SX. That is, there are no postnuclear
glides in SX syllable structure while VG is, as a matter of fact, a very
common segment sequence in many of the world’s languages, including
Mandarin and other Chinese dialects. This chapter also presents 14
surface vowels in SX, as shown in the vowel chart of SX in (87):
(87) The Vowel Chart of SX
Front Central Back
i y u

e P F o
id E
id O
a A Å
round +round roun
d +round roun
d +round
The Vowel Chart (87) shows that there are altogether 14 vowels in SX
which appear in surface representation in the rhyme of syllables either by
themselves or in combinations. Besides the 14 vowels, there are also three
glides: [j], [Á] and [w], in surface representation. However, among these
14 surface vowels, I assume that there are only six phonemic vowels
underlyingly. These are /i/, /e/, /a/, /Ø/, /o/ and /u/. The discussion of the
14 vowels, their phonetic and phonological behaviour, and why only
these six vowels are the underlying phonemes in the vowel inventory of
SX, will be the topic of the next chapter.

PDF Document reader online

This website is focused on providing document in readable format, online without need to install any type of software on your computer. If you are using thin client, or are not allowed to install document reader of particular type, this application may come in hand for you. Simply upload your document, and will transform it into readable format in a few seconds. Why choose

  1. Unlimited sharing - you can upload document of any size. If we are able to convert it into readable format, you have it here - saved for later or immediate reading
  2. Cross-platform - no compromised when reading your document. We support most of modern browers without the need of installing any of external plugins. If your device can oper a browser - then you can read any document on it
  3. Simple uploading - no need to register. Just enter your email, title of document and select the file, we do the rest. Once the document is ready for you, you will receive automatic email from us.

Previous 10

Next 10