Combinations and Variation
3.5.3. Missing forms
Given four choices for G (three glides, plus one without G), fi ve for V, and fi ve for X (two high vowels, two nasals, one without X), there are 100 combinations of (G)V(X), of which only 35 actually occur, shown in (20). The top row indicates the choices for G, where [Ø−] means no glide.
The fi rst column indicates the choices for X, where [−Ø] means no X. The second column indicates the underlying forms of V(X), using [ə] for the mid vowel and [a] for the low vowel.
(20) Actual and missing (G)V(X) forms in phonetic symbols
Ø- j- w- form with (+) indicates one that does not contrast with another (indicated on the right-hand side) because of the constraints Merge and G-Spreading , to be discussed below.
There are two possible views on the missing forms. The fi rst is that they are due to historical accidents or arbitrary choices of a dialect, and there is no further explanation. Another view is that the missing forms indicate systematic constraints on possible syllable structures. To support the second view, one should show that there are indeed reasonably natural constraints for most of the missing forms. I offer such an analysis in the next section.
3.5. T H E D ATA 59
3 . 6 . R H Y M E - H A R M O N Y , M E R G E , A N D G - S P R E A D I N G
The analysis I offer makes use of three constraints to rule out most non-occurring forms. They are shown in (21), where α is any feature value and
−α is the opposite value.
(21) Rhyme-Harmony: VX cannot have opposite values in [round] or [back]:
*[+back][−back], *[−back][+back]
*[+round][−round], *[−round][+round]
Merge: Two tokens of the same feature merge into one long feature:
Fi Fi Fi
G-Spreading: A high nuclear vowel spreads to the onset C:
XX XX →
[Ci]→[Cji], [Cu]→[Cwu],[Cy]→[C y]h
Rhyme-Harmony rules out opposite values of [back], but if a sound is unspecifi ed for [back], it can freely combine with [+back] or [−back].
In other words, [+back][Ø back], [−back][Ø back], [Ø back][−back], [Ø back][+back] are allowed. The same is true for [round].
There is a likely physical reason for Rhyme-Harmony and Merge: it is hard for an articulator to move fast enough to execute two opposite gestures in a rhyme (such as [+back][−back]), or the same gesture twice in a rhyme (such as [−back][−back]). If so, one might fi nd similar constraints in other languages, a topic that is beyond the scope of the present study.
G-Spreading i s an anticipatory process. I will argue in Chapter 4 that there is only one onset slot in the syllable, which can be fi lled by C, G, or CG. G−Spreading does not apply when the onset is G or CG. For example, G−Spreading does not require [wi] to become [ɥi] or [wji], or [twi] to become [tɥi] or [twji], or [ ju] to become [ɥu] or [ jwu]. However, I will discuss some cases below where G−Spreading seems to occur when the onset already has a glide. Also, I will argue in Chapter 4 that the onset slot is optional, and G-Spreading does not apply when there is no onset. This explains why the English word [ist] east does not become [ jist], the latter being a different word, yeast. The reason is that yeast has an onset while east does not.
Of the 25 VX rows in the GVX table, Rhyme-Harmony and Merge r ule out nine, shown in (22), where I consider [n] to be [−back] in SC and [ŋ]
to be [+back]. Since SC has only one mid vowel and one low vowel, they are unspecifi ed for [back] or [round], and so they can combine with any glide or high vowel.
(22) VX rows ruled out by Rhyme-Harmony and Merge *row-[un] differ in frontness
*row-[iŋ] differ in frontness
*row-[yŋ] differ in frontness
*row-[ui] differ in rounding and frontness
*row-[yu] differ in frontness
*row-[yi] differ in rounding
*row-[iu] differ in rounding and frontness
*row-[ii] Merge, [ii] = [iː]
*row-[uu] Merge, [uu] = [uː]
Because all regular SC syllables are heavy, a rhyme with just [i] or [u] is in fact long [iː] and [uː] (see Chapter 4). Therefore, there is no contrast between [iː] and [ii], because the latter will change to [iː] under Merge.
Similarly, there is no contrast between [uː] and [uu].
It can be seen that most of the rhymes in (22) contain a pair of high vow-els. Y. Lin ( 1989: 59) suggests that SC has a constraint *[+high][+high]
that prohibits two high vowels from occurring next to each other. In the present analysis, Lin’s constraint is not necessary because the effect can be achieved by Rhyme-Harmony and Merge, which also rules out bad VN rhymes.
There remain 16 rows of rhymes, all of which satisfy Rhyme-Harmony.
They are shown in (23). The mid vowel [ə] and the low vowel [a] are unspecifi ed for rounding and frontness (see Chapter 2).
(23) Row Comment
[ai] [a] unspecifi ed for [back]
[əu] [ə] unspecifi ed for [back]
[au] [a] unspecifi ed for [back]
The rhyme [in] is [−back][−back], and so it will undergo Merge. This is shown in (24), where I use ‘−B’ for [−back].
(24) i n i n
−B−B −B
→
The result is still [in], which does not overlap with another rhyme.
Similarly, [yn] and [uŋ] will undergo Merge but will still remain separate rhymes.
The 16 rows contain 64 forms, listed in (25), where G-Spreading (G) accounts for six of them. The equivalent pairs under G-Spreading are list-ed on the right. For example, [u] and [wu] are equivalent, which means that [Cu] will become [Cwu].
(25) Cells ruled out by G-Spreading (G)
Ø- j- w-
i + G − − [ ji] = [i]
u + − G − [wu] = [u]
y + − − G [ɥy] = [y]
ə + + + +
a + + + −
in + G − + [ jin] = [in]
yn − − − G [ɥyn] = [yn]
ən + − + − an + + + +
uŋ + + G − [wuŋ] = [uŋ]
əŋ + + + −
aŋ + + + −
əi + − + −
ai + + + −
əu + + − −
au + + − −
The 35 occurring forms are indicated by ‘+’. There are 23 non-occurring forms, indicated by ‘−’, which I list in (26). About half of them seem to be accountable in some ways, and I have made some tentative comments on why they do not occur independently. For each pair of variants, such as [wi]–[wəi] or [ɥuŋ]–[ juŋ], there seems to be no principled way to decide which variant is underlying and which is missing or derived.
(26) Forms Comments wi Variant of [wəi]
ju Variant of [ jəu]
win Same as [yin] if G-Spreading app lies to [win]
ɥuŋ Same as [ juŋ] if G-Spreading app lies to [ juŋ]
In SC [wi] is a variant of [wəi]; the variation is conditioned by tone, to be dis-cussed below. Similarly, [ ju] is a variant of [ jəu], also conditioned by tone. In [win], [ɥuŋ], [ jy], [wy], [ jyn], and [wyn], there is a prenuclear G and a high nuclear vowel. If G-Spreading app lies to the high nuclear vowel even if G is already occupied, these forms will become the same as some others (which can be occurring or non-occurring) and so not independently contrastive. In [ɥəu], [ɥau], [wəu], and [wau], both the onset and the coda are [+round], and there seems to be a dissimilation eff ect against that. It is interesting to note that [wau] is marginal in English, too, and only occurs in three words, powwow, bowwow, and wow. More frequent in English are quote, quota, swollen, woeful, and wont, etc., probably because some speakers use [wou]
or [wo] (with one long [+round]) and some use [wəɯ]. So there does not seem to be a good reason to rule out [wau] and [wəu] completely. The syllable [ jəi]
is similar to [ jei] in English, which only occurs in a marginal word yea and a few French borrowings such as soigné. Finally, it is interesting to note that 16 out of the 23 non-occurring forms, or 70 per cent, contain [y] or [ɥ].
It is possible that under G-Spreading [ɥu ŋ] and [iuŋ] are identical. If so, the underlying form could be either [ɥuŋ] or [iuŋ], although I have as-sumed that it is the latter. The ambiguity may explain some confusion in the literature. In traditional Chinese phonology, syllables are sometimes grouped according to the prenuclear glide. The SC word ‘use’ is some-times thought to be [ juŋ] and grouped with the [ j]-group (Chao 1968) and sometimes [ɥuŋ] and grouped with the [ɥ]-group (Hsueh 198 6). The pres-ent analysis suggests that there is no simple way to resolve the ambiguity.
Duanmu (20 03) proposes a rule called Triphthong Raising, accord-ing to which a form with three vowels [high][mid][high] is raised to [high][high][high]. The effect has been observed in SC (Zee 200 3) and can be understood in terms of articulatory effort reduction. The rule accounts 3.6. R H Y M E-H A R M O N Y, M E R G E, A N D G-S P R E A D I N G 63
for the alternation in the pairs [wəi]–[wi] and [ jəu]–[ ju]. It can also ac-count for the lack of [ɥəu], [wəu], [ɥəi], and [ jəi], all of which contain [high][mid][high]. However, Triphthong Raising is not universal, because English uses [wi] ‘we’ and [wei] ‘way’ contrastively. Also, English uses [wou] quite often, as just seen.
In (27) I summarize the analysis of GVX forms.
(27) Summary of the analysis of GVX for ms Reason Number of forms Rhyme-Harmony 28
Merge 8
G-Spreading 6
Other missing ones 23
Occurring 35
Total 100
65% of all missing forms are accounted for 77% of all forms are accounted for
3 . 7 . A L L O P H O N I C VA R I AT I O N S
Let us now consider surface variations of SC sounds, or allophonic varia-tions. I focus on the G-Spreading eff ect and surface variations in the mid and low vowels.
3.7.1. G-Spreading
G-Spreading req uires a high nuclear vowel to spread to the onset. As a result, there is a lack of contrast between many pairs of forms, such as [sin]
vs. [sjin] or [ɕin]. The process is graphically shown in (28), where O is the onset, N the nucleus, and C the coda. As will be discussed in Chapter 4, in a CGVX syllable both C and G are in the onset.
(28) [sin] [sjin] or [ in]
ONC ONC
s i n s i n
→
Because of G-Spreading, ther e is no surface [sin] but only [sjin], which is used by some speakers, and [ɕin], which has further undergone palataliza-tion and is used by most speakers (see Chapter 2).
G-Spreading may also be responsible for the lack of contrast between [win] and [wjin] or [ɥin], or between [ juŋ] and [ jwuŋ] or [ɥuŋ], if it can apply, at least sometimes, when the onset already has G. Assuming that [ɥ]
is a combination of [ j] and [w], [ juŋ] can be analysed as in (29).
[ juŋ] [ jwuŋ] or [Ëuŋ]
ONC ONC
j uŋ j uŋ
(29) →
Chinese also lacks the contrast between V and GV, or VC and GVC, where V is a high vowel and G is the corresponding glide, such as the pairs in (30).
(30) Contrasts not found in Chinese [i]–[ ji] [u]–[wu] [y]–[ɥy]
[in]–[ jin] [uŋ]–[wuŋ] etc.
One possibility is that every syllable has an onset, so that a high vowel is always preceded by G. For example, [in] is in fact [Øin], which has an empty onset Ø, and it becomes [ jin] under G-Spreading, as shown in (31).
(31) [in] [ jin]
ONC ONC
i n i n
→
Another possibility, to be proposed in Chapter 4, is that an onset is not required for every syllable and Chinese does not use the empty onset. If so, Chinese simply lacks syllables like [ ji], [wu], [ɥy], [ jin], etc., because there is no onset slot for the high nuclear vowel to spread to.