4.2 Tonal neutralization in Standard Mandarin
4.2.2 Neutralization at lexical vs. postlexical levels
Put aside the underlying vs. derived distinction between inherent and non-inherent neutral tones, it seems that standard Mandarin is a parallel case to that of Taiwanese ― the threshold of tonal neutralization is set at the category with degree of grammaticalization equal to 3, as shown in (55).
(55) General distribution of tonal neutralization
NUM PHA
One may jump into conclusion that such a scenario suggests ranking ALIGN-R(ω, Fnc≤2) above SP-MAX-X0. This analysis, however, does not capture the picture of Standard Mandarin. As I shall argue next, the analysis of the grammatical words in Standard Mandarin would not be complete if we do not take into consideration the distinction between the two types of neutral tones.
4.2.2 Neutralization at lexical vs. postlexical levels
The fact that inherent can be distinguished from non-inherent neutral tones by the notion of underlying tonelessness in mainland Standard Mandarin is reminiscent of the lexical vs. postlexical distinction. This distinction plays an important role in generative phonology, where (morpho)phonological rules are postulated to apply at multiple stages
― lexical phonological rules are confined to lexicon and are thus confined to morphemes and single words, sensitive to sublexical structure, while postlexical rules, referred to as phrasal rules as well, operate across word boundaries and have access to b. Zhuo1shang4 zhi3you3=san1=ping2 jiu3
Table.on there.be.only=three=CL beer
‘There are only three bottles of beer on the table.’
‧
phrasal or syntactic structure. The difference in their domains of application implies that lexical rules must apply prior to postlexical rules and that there are distinct characteristics attributed to these two types of phonological rules, as shown in the following table (Kiparsky 1985).
(56) Properties of lexical vs. postlexical rules
Lexical Rules Postlexical Rules
a. Word-bounded Not word-bounded
b. Access to word-internal structure assigned at the same level
Access to phrase structures only
c. Cyclic Non-cyclic
d. Apply in derived environments Apply across the board
e. Structure-preserving Not (necessarily) structure-preserving f. Apply to lexical categories only Apply to all categories
g. may have exceptions Automatic
h. Semi-productive Fully productive
i. Categorical output May have gradient output
j. Obligatory Optional and may be sensitive to
rhythmic factors such as rate, register and pause
While the reliability of these properties has been challenged by many attested
applications that share characteristics of both lexical and postlexical types of rules, as a diagnostics they can still help us accounts for the two types of tonal neutralization in question, especially the properties (56g), (56i) and (56j). Based on Chao’s (1968) description, as well as my own observation, the occurrence of inherent neutral tones are obligatory, produce categorical outputs, and have exceptions ― all being properties of lexical rules. By contrast, in mainland Standard Mandarin neutralization of the non-inherent type is optional, has gradient output, and automatically applies ― all characteristic of postlexical rules. Note that the obligatoriness vs. optionality distinction is applicable only in some varieties, such as Taiwan Mandarin, a prevalent variety of Standard Mandarin spoken in Taiwan. According to authoritative teaching materials
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
and reference books published in Taiwan, degree 3 grammatical words (i.e. directional complements and object pronouns) are described to be optionally neutral-toned, and even if they are neutralized, the extent of the neutralization is also reported to be incomplete and/or variable.
We can also address the lexical vs. postlexical distinction with the aid of the famous T3 Sandhi. Details aside, this rule can be simply stated as follows: T3 changes into T2 before another T3, as formulated in (57).
(57) T3 Sandhi in Standard Mandarin T3 → T2 / _ T3
It is widely argued that this rule is both lexical and postlexical, since it applies and is sensitive to both word-internal/morphological and phrasal/syntactic structure (Shih 1986). As illustrated in (58),in the compounding word shui3guo3 ‘fruit,’ the first member of the compound shui3 becomes T2 for being followed by the second member, guo3, which is also a T3 morpheme, and this morpheme itself in turn changes into T2 due to the following T3 adjective hao3 ‘good,’ which is another word serving as the predicate/comment of the compound. This case shows not only that T3 Sandhi applies both within and across words, but that the application in lexical domain is prior to the application in postlexical domain.
(59) Lexical/Postlexical application of T3 Sandhi
[Shui3guo3N] [hao3A] ‘Fruit is good.’
Fruit good
Lexical domain (( T2 T3 )
Postlexical domain (( T2 T2 ) T3 )
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
Let us turn back to the neutral tones. There is an asymmetry between inherent and non-inherent neutral tones with regards to the tone sandhi rule: T3 sandhi applies before a non-inherent neutral tone that is originally T3, but fails to do so before an inherent neutral tone grammaticalized from a T3 verb, namely liao3 ‘to end.’
(60) Interaction between T3 Sandhi and neutral tones a. Inherent neutral tones:
b. Non-inherent neutral tones
The non-application of T3 Sandhi before inherent neutral tones suggests that T3 Sandhi applies after the neutralization, meaning that inherent neutral tones come into being at the stage earlier than lexical T3 Sandhi. Here I assume that the occurrence of inherent neutral tone is in the domain of stem level, while the lexical T3 Sandhi is a word-level application. Accordingly, the processing of non-inherent neutral tone must be postlexical. This is illustrated in (61).
(61) Rule application in different morphosyntactic domains
To summarize, inherent neutral tones, a stem-level application which is obligatory, and [Da3=le0VP] [ta1D] ‘have hit him/her’
To.hit=PFV him/her ( T3 T0 )
[Da3=wo3VP] ‘to hit me’
To.hit=ME
( T2 T0 )
Inherent Neutral tone T3 Sandhi Non-inherent neutral tone Stem-level
Word-level
Phrase-level
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
categorical, occurs in highly grammaticalized words (i.e. sentence-final particles, modifier markers and aspect markers), while non-inherent neutral tones, a phrase-level application which may be optional and/or gradient, occurs in medium grammaticalized words (i.e. object pronouns and directional complements). This is schematized in (62).
(62)
NUM PHA/CL DIR/PRO ASP MOD SFP
Fnc=1 Fnc=2 Fnc=3 Fnc=4 Fnc=5 Fnc=6
Unneutralized Non-inherent Neutral Tones (Phrase-level)
Inherent Neutral tones (Stem-level)
This derivational analysis can be recast under the framework of stratal OT that assumes a classical modular feedforward architecture of grammar (Bermúdez-Otero 2012, 2015).
The model consists of multiple grammars referring to stem-, word- and phrase-level, each may involve different constraint ranking that evaluates in a parallel way as the classic version of OT, yet the grammars are feedforward: the output of stem-level grammar serves as the input of word-level grammar, and the output of word-level grammar feeds the input of phrase-level grammar. Adopting this model, I assume different rankings wih respect to ALIGN-R(ω, Fnc≤2) and SP-MAX-X0 for stem- and word-level grammar. Specifically, in stem-level grammar, ALIGN-R(ω, Fnc≤2) ranks below SP-MAX-X0, hence no inherent neutralization for degree 3 grammatical words.
In phrase level, on the other hand, the ranking of ALIGN-R(ω, Fnc≤2) with respect to SP-MAX-X0 is variable, and thus there is postlexical, optional, non-inherent neutralization for degree 3 grammatical words. By contrast, ALIGN-R(ω, Fnc≤3) outranks SP-MAX-X0 at every stage of grammar, which leads to inherent neutralization
for highly grammaticalized function words. This analysis is shown in (63) and (64).
‧
(64) (One possible) Phrase-level grammar
Constraint ranking: ALIGN-R(ω, Fnc≤3) ≫ SP-MAX-X0 ≫ ALIGN-R(ω, Fnc≤2)
This chapter addresses grammaticalized function words in Taiwanese and Standard Mandarin. Both languages contain a large range of non-prominent function words in
[Da3=le0VP] [ta1D]
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
tonal neutralization, with the threshold set on categories with degree of grammaticalization equal to 3, hence the constraint ranking ALIGN-R(ω, Fnc≤2).≫ SP-MAX-X0. In Taiwanese, non-prominent words undergo tonal neutralization and prosodic invisibility only at phrase-final positions. This is because of the interaction between EQUALSISTERS, ALIGN-R(ω, Fnc≤2) and ALIGN-R(ϕ, Fnc≤2). In Standard Mandarin, non-prominent function words can be further divided into two groups according to the presence/absence of optionality for tonal neutralization ― relatively lower grammaticalized function words optionally bears neutral tone, while higher grammaticalized ones always appear with neutral tone. This division is parallel to the point in tone they are grammaticalized.
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y 5. Grammaticalized Function Words in
Shanghainese
5.0 Introduction
Wu Chinese is probably the richest in neutral tones of all the major families of Sinitic languages. In a large majority of languages in Wu Chinese, especially in the Northern branch, underlying tonal contrast is maximally neutralized in the sense that the application is across categories. On the one hand, tonal neutralization applying in the functional categories tend to be irrespective of the distinction between classes with different degrees of grammaticalization. Even low grammaticalized function words that are prosodically prominent in other Sinitic families are normally neutralized in Wu Chinese. On the other hand, neutralization of tonal contrast occurs in lexical categories as normal as in functional categories. A vast majority of Northern Wu languages have general rules of tone sandhi that replace the full lexical tones carried by non-initial syllables in a lexical word with the pitch contour either by spreading from the initial syllable or by epenthesis of some unmarked pitch target. In consequence, tonal neutralization in Wu Chinese involves conflation between not only distinct classes on the grammaticalization scale, but also the functional categories and lexical categories.
A representative example of this dual conflation that I address in this chapter is Shanghainese, where only numerals denoting specific quantity preserve their distinct lexical tones, and the preservation is confined to the initial syllable as it is in the lexical words.
Besides tonal neutralization, Shanghainese also instantiates the invisibility of the
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
high grammaticalized function words. Only modifier markers and sentence-final particles are invisible to the normal application of spreading from the preceding tone.
This phenomenon of extrametricality, as it were, serves to distinguish between classes with high degree of grammaticalization and classes with non-high degree of grammaticalization. Table (1) summarize the two types of tonal attrition in Shanghainese.
(1) Tonal attrition of functional categories in Shanghainese
NUM PHA/CL DIR/PRO ASP MOD SFP
Unneutralized Neutralized
Visible Invisible
Discussions of the remaining sections mainly concerns the two types of tonal attrition in Shanghainese, including tonal neutralization (§5.1) and the prosodic invisibility to melody redistribution (§5.2). The conclusion follows in §5.3. The data in each section is primarily from my own observation, personal communication and consultation with informants. Part of data are adapted from online disctionaries and previous studies (e.g.
Qian 1997, T.-C Huang 2015).
5.1 Tonal attrition in Shanghainese
Shanghainese is perhaps the most recognized dialect of Northern Wu Chinese. As stated in the outset, grammatical words in this language are characteristic of both maximal neutralization and invisibility. This section examines the first one, with a general introduction to tonal neutralization of grammatical words first (§5.1.1), followed by the discussion of the distinction between neutralization in lexical and functional categories (§5.1.2).
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
5.1.1 Tonal neutralization of grammatical words
Shanghainese has five citation tones, three of which being long while the other two being short, or checked. Like many other Sinitic languages, long tones and short tones are in complementary distribution conditioned by the type of syllables that bear the tone, and therefore they can be treated as allophonic variants. However, as our discussion in the following subsections focuses on the long tones, the distinction between long tones and short tones would be maintained for the sake of clarity. Table (2) shows the five lexical tones in Shanghainese, including T1, a fall, T2, a small rise, T3, a big rise, T4, a short high, and T5, a short low rising. Note particularly that the long tones are all contour tones.
(2) Citation tones in Shanghainese
Categories T1 T2 T3 T4 T5
Long 52 34 13
Checked 5 12
The marking of citation tone for grammatical words in Shanghainese is an issue of mess.
Some treat part of the grammatical words as having a fixed citation tone as lexical words do, while others leaving unspecified the function words in most of the classes.
My conjecture is that most of the function words in Shanghainese underlyingly carry a pitch contour lexicalized from a variety of sources that we have addressed in the cases of Cantonese and Hakka, such as default unmarked form (i.e. a checked tone), boundary tones, and intonation. Later on, this lexicalized pitch contour is attributed by the speakers to one of the five tonal categories with which it shares the most perceptual similarities, and yet this categorization is far from stable. Accordingly, I assume that most of the function words in Shanghainese do carry one of the five full tones in citation, as illustrated in (3b-g). However, sentence-final particles, especially those that Paul
‧
(2014) argues are situated in Attitude C, could be underlying toneless, given their more variable pitch realization than that of the other grammatical words. This status is shown in (3a) by marking these particles as T0 in citation.
(3) Prescriptively marked tones of function words in Shanghainese a. Sentence-final particles in Attitude C
a0 Smooth-alert, adhortative ya0 Smooth-alert, adhortative o0 Smooth-alert, adhortative le0 Smooth-alert, assertion
la0 Follow-up question, with adhortative touch noa0 Smooth-alert, warning
b. Sentence-final particles outside Attitude C leh5 Realization of state
geh5 Neutral assertion of relevance va3 Used in yes or no questions
c. Modifier marker
geh5 (Adjectival) modifier marker, possessive marker d. Aspect markers
leh5he0 Continuous marker leh5 Perfective marker
ku2 Experiential perfect marker c. Directional complements
le3 ‘towards the speaker’
qi2 ‘away from the speaker’
zang3 ‘upward’
‧
hoa2 ‘properly done, well set’
theh4 ‘realization of lost or eroded’
toa2 ‘at, to, arrived’
f. Classifiers
geh5 generic classifier
po2 for something long and has a hilt or handle
pe1 for a flight, train or bus at a particular time; for a group of people
bu3 for vehicle
khue2 for a clump or blob of thing that usually has some weight dioa3 for long piece of thing
g. Numerals
In connected speech, all classes of grammatical words ― with numerals as the only exception ― surface with no tonal contrast. This is shown in the examples (4-11), where the neutral-toned syllable is marked with a superscripted 0. Details of the phonetic manifestation of the neutral tones are addressed in the next subsection.
‧
(4) Neutralization in sentence-final particles
(5) Neutralization in modifier markers
(6) Neutralization in aspect markers
a. Geh5=tsah0 ng3 ghe3=zy0 sang1=geh0 This=CL fish still=COP raw=MOD
‘The fish remains undercooked.’ ‘That is his/her favorite dress.’
a. geh5=dji0 yi1zang0 tsah4-lah0he0, tse2 veh5=we0 sang1fong0 This=CL clothes put.on-CONT, just NEG=will to.catch.a.cold ‘Put the clothes on, so you will not catch a cold.’
b. Qieh4-tsy0 ve3, nong3 zhieu3 qi2 doh5sy0 Eat=PFV meal, you then to.proceed studying
‘You should study once you finish your dinner.’
c. Geh5=peng0 sy1 ngu3 koe2-ku0 hoa2ji0-thang0 This=CL book I read-EXP several-times ‘I have read the book several times.’
‧
(7) Neutralization in directional complements
(8) Neutralization in object pronouns
(9) Neutralization in phasal complements
I just NEG=want take.notice.of=him/her=SFP
‘I’m just not going to take any notice of him/her.’
b. Ngu3 tseng1.geh0 loa3 xiang2=nong0=geh0 hurry.up OM=you=MOD homework write-PHA-SFP
‘Finish your homework right away.’
c. Coa2ku2 ze3-zah0=ti0 xioa2ce0-dong3di0 Speculate.in.stocks earn-PHA=little appetizer.money ‘He made a little money in the stock market.’
a. Geh5=bu0 qi1co0=geh0 ying1djing0 zy3=jing2kheu0=ge4 This=CL car=MOD engine COP=imported=SFP
‘The engine of this car was imported.’
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
(11) No neutralization for Neumerals
To sum up, in Shanghainese function words in the vast majority of grammatical categories have their distinct tones neutralized, except numerals, the only type of function words that is prosodically prominent enough to preserve the contrast. Given the grammaticalization scale established in Chapter 2 (as repeated in (12), where the number denotes the degree of grammaticalization, in the sense that the greater the number, the more grammaticalized the category), the prominence asymmetry between numerals and the other classes of grammatical words in Shanghainese can be translated as the distinction between the least grammaticalized function words (i.e. Fnc≤1) and the function words that are relatively higher grammaticalized (i.e. Fnc>1). In other words, the distinction between the categories with grammaticalization degree higher than 1 are all conflated. That is, with respect to tonal neutralization, there is no distinction between degree 2 and degree 3, or between degree 3 and degree 4, and the like, as depicted in (13). The picture forms a mirror image of the case in Cantonese and Miaoli Sixian Hakka.
b. Yieh4=khue0 de3koa0 yoa2 tu1soa0 dong3di0=ne0 One=CL cake need how.much money=SFP
‘How much does a cake cost?’
a. Qieh4-tsy0 ng3=pe0 jieu2 tse2 qi2 zang3pe0 Have-ASP five=CL beer not.until go to.work ‘(I) had five glasses of beer before going to work.’
b. Geh5=tsah0 ghah5deu0-li3xiang0 yeu3teh0 se1=tsah0 keu2
This=CL box.inside have three=CL dog
‘There are three dogs in the box.’
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
(12) Grammaticalization scale
NUM PHA
CL
DIR
OP ASP MOD SFP
1 2 3 4 5 6
(13) Conflation of Fnc>1
The neutralization vs. non-neutralization distinction between grammaticalization degrees 1 and 2 can be captured in the currently proposed framework by ranking ALIGN -R(ω, Fnc≤1) above SP-MAX-X0, definition of these constraints being given below.
(14) ALIGN-R(ω, Fnc≤1):
Assign one violation mark for every prosodic word (ω) that is not right-aligned with a syntactic word with grammaticalization degree lower than or equal to 1 (i.e. numerals and lexical words).
(15) SP-MAX-X0
Assign one violation mark for every syntactic word (X0) in the input syntactic representation S that does not have a corresponding prosodic word (ω) in the output phonological representation.
This version of alignment constraint in (14) is highly stringent in that it allows an
Fnc=1 Fnc=2 Fnc=3 Fnc=4 Fnc=5 Fnc=6
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
extremely small subset of word types (i.e. numerals and lexical words only) to license prosodic words at its right edge. When this constraint outranks SP-MAX-X0, which requires that every syntactic word have a corresponding prosodic word, a wide range of types of function words fails to fall under the scope of the prosodic word status. The tableau in (16) gives an illustration, where various prosodizations of function words in categories with different degrees of grammaticalization are evaluated. Prosodizations from CAND 1 through CAND 5contain a variant amount of instantiation of Fnc>1 right-aligned with ω, and thus they fatally violate the high-ranked ALIGN-R(ω, Fnc≤1) at least once. The violation(s) on ALIGN-R(ω, Fnc≤1) is removed in the remaining two candidates: CAND 6 parses lexical words and numerals in ω, and CAND 7 parses lexical words only. The latter loses to the former as desired because leaving numerals unparsed would incur excess violation on SP-MAX-X0. This constraint ranking successfully ensures that only numerals, the lowest grammaticalized type of function words, are eligible for the status of ω, where they can be accented at syllabic level and thus are as
qualified as lexical words in the capability of preserving the tonal contrast.
(16) ALIGN-R(ω, Fnc≤1) ≫ SP-MAX-X0 INPUT:
CAND 1: (ω Lex=3) (ω Fnc=3) (ω Lex=2) (ω Fnc=5) (ω Fnc=1) (ω Fnc=2) (ω Lex=3) (ω Lex=3) (ω Fnc=4) (ω Fnc=6) (ω Fnc=6)
CAND 2: (ω Lex=3) (ω Fnc=3) (ω Lex=2) (ω Fnc=5) (ω Fnc=1) (ω Fnc=2) (ω Lex=3) (ω Lex=3) (ω Fnc=4) (ω Fnc=6) (ω Fnc=6)
CAND 3: (ω Lex=3) (ω Fnc=3) (ω Lex=2) (ω Fnc=5) (ω Fnc=1) (ω Fnc=2) (ω Lex=3) (ω Lex=3) (ω Fnc=4) (ω Fnc=6) (ω Fnc=6)
Meng3=nong0 jia2=geh0 se1-peng0 sy1 khoe2-ku0=leh0=ya0 Ask=you borrow=MOD three-CL book read-EXP-SFP-SFP
Meng3=nong0 jia2=geh0 se1-peng0 sy1 khoe2-ku0=leh0=ya0 Ask=you borrow=MOD three-CL book read-EXP-SFP-SFP