Duanmu, San (2000) The Phonology of Standard Chinese, Oxford University Press, hardback, xv, 300pp., in 'The Phonology of the World's Languages' series, edited by Jacques Durand, ISBN: 0-19-829987-7.
Jie Zhang, UCLA and Harvard University.
PURPOSE AND CONTENTS OF THE BOOK
Duanmu's book is a comprehensive study of the phonology of Standard Chinese (or Mandarin). Its target readers are anyone with an interest in the synchronic or diachronic phonology of Chinese, or just Chinese languages or phonology in general. It strives to provide both accurate factual descriptions and theoretical analyses, at the same time avoiding esoteric jargon specific to historical Chinese philology or theoretical phonology. As the book succeeds in all accounts, it should appeal to a wide range of readership.
The book contains twelve chapters. It starts by introducing the history of Chinese languages and their speakers (Chapter 1), then moves on to discuss the sound inventory (Chapter 2), co-occurrence restrictions of sounds (Chapter 3), and syllable structure (Chapter 4) of Standard Chinese (henceforth SC). Chapters 5 to 8 are a detailed study of wordhood in Chinese, in particular, how the issues of word length and word order can be addressed in the light of a theory of stress in Chinese proposed by the author. Chapter 9 discusses the phonology related to the diminutive suffix [r] in SC. The next two chapters address the issue of tone in SC: Chapter 10 focuses on the basic properties of tone, such as its phonetic correlates, its featural representation, the concept of tone-bearing unit, and the tonal inventory of SC; Chapter 11 focuses on the infamous third tone sandhi of SC. Chapter 12 raises further issues of Chinese phonology not addressed in the book.
The introductory chapter of the book briefly reviews the historical development of Chinese languages and the establishment and preservation of a standard language. I see four interesting points conveyed by the author. First, although SC (or Mandarin, Putonghua 'Common Speech', Hanyu 'Chinese Language', Guoyu 'National Language') has only been the official language of China for a few decades, it follows from the tradition of having a standardized form of the language, from Yayan 'Refined Speech' in the Chunqiu period (722-482BC) to Guanhua 'The Official Language' in the Ming Dynasty (1368-1644). Second, unlike some standard European languages, such as RP British English, SC does not carry a superior social prestige. Third, there had been a great discrepancy between written and spoken Chinese throughout history, and only recently (since the beginning of the 20th century) has this discrepancy been significantly reduced. And fourth, there are tremendous differences among Chinese dialects, many of which are mutually unintelligible. These properties of the Chinese language determine that the study of spoken SC is necessarily different in nature than the study of many other languages.
Chapter 2 discusses the sound inventory of SC. Upon providing theoretical backgrounds in phonological representation of sounds, phonemic analysis, and syllable structure, the author proposes that there are three levels in the analysis of SC sounds: underlying, syllabic, and phonetic. On the underlying level, SC has 19 consonants and five vowels: p, ph, f, m, t, th, ts, tsh, s, n, l, voiceless unaspirated retroflex affricate, voiceless aspirated retroflex affricate, voiceless retroflex fricative, r, k, kh, x, ng; i, y, u, E (=schwa), a. On the syllabic level, there are also three glides that correspond to high vowels i, y, and u in the onset position, 29 consonant-glide combinations that share one onset slot, 2 syllabic consonants z and r that are derived from empty rhyme slots filled by the onset consonant, and a zero onset, which represents an empty onset slot. The phonetic level includes all other allophones, such as allophones of E--o, e, E, and the mid back unrounded vowel--in various environments. The most innovative proposal here is to treat the palatal affricates and fricative as dental+j combinations tsj, tshj, and sj underlyingly. The author's main argument comes from the fact that there is a group of SC speakers that realizes the palatals as dental+j combinations.
Chapter 3 is a detailed study of the co-occurrence restrictions and surface variations of SC sounds. The author starts by observing that of all the possible 2,280 syllables, only about 400 actually occur. E.g., [wau], [fje], [pjEu] (E=schwa) are bad forms in SC. He then proposes an analysis that accounts for the missing GVX forms by referring to a harmony constraint that requires the rhyme elements to agree in [back] and [round] features, and a dissimilation constraint that bans adjacent palatal sounds and adjacent rounded sounds. The author also proposes an Optimality-theoretic account for the surface realizations of the underlying forms using harmony constraints and constraints on featural combinations.
Chapter 4 is the analysis of syllable structure in SC. Traditionally, the structure of SC syllables is considered variable from a minimum of one segment, V or C, to a maximum of four segments, CGVC (coda C = n or ng) or CGVG. The author here, following his dissertation (Duanmu 1990), proposes a novel approach in which all SC syllables are either full or weak; full syllables have the structure CVX, with one onset slot and two rhyme slots; and weak syllables have the structure CV, with one onset slot and only one rhyme slot. A number of arguments is presented to support this approach, the most convincing of which come from phonetics: the prenuclear glide, when present, is phonetically realized as the secondary articulation of the onset consonant; the rhyme portion of full syllables has comparable duration--an open full syllable has a long vowel and a closed full syllable has a short vowel; weak syllables are phonetically CV syllables with a short vowel. When a full syllable with a diphthong or a nasal coda is reduced to a weak syllable, the diphthong is monophthongized and the nasal coda is realized as vowel nasalization. These proposed syllable structures will also turn out to be useful in the account of SC tonal distribution (Chapter 10).
Chapter 5 marks the beginning of the discussion of word- level phonology in SC. The author first calls attention to the importance of the compound/phrase distinction in Chinese by pointing out that they might have different tone sandhi patterns. E.g., in Shanghai, [tsho ve], when used as a compound meaning 'fried rice', has the tonal pattern L-H; while when used as a phrase meaning 'to fry rice', has the tonal pattern LH-LH. He then reviews the tests that have been previously proposed to distinguish compounds from phrases, and points out the conflicts among them. The tests that the author eventually adopts are the following: Conjunction Reduction, Adverbial Modification, XP Substitution, and Productivity. He also argues that the tests Freedom of Parts, Semantic Composition, and Exocentric Structure can detect compounds when failed, but cannot guarantee that the structures which pass them are phrases, and therefore should be adopted with limitations. The consequence of applying the adopted tests is that in Chinese, a modifier-noun [M N] without the particle 'de' is a compound, so are its derivatives [M [M N]], [[M N] N], [[M N] [M N]] etc.
Chapter 6 discusses the author's innovative approach to Chinese stress. The main proposals are the following. First, contra traditional assumptions, SC has word stress. Second, the basic stress unit is a syllabic trochee. Third, a simple di- or polysyllabic word forms trochees left to right; a compound or phrase assigns stress cyclically according to the Nonhead Stress rule (Duanmu 1990, Cinque 1993). The advantages of this approach in accounting for the disyllabic requirement, the restrictions on word length, and the restrictions on word order in SC are also briefly touched upon in this chapter, and are spelled out in greater details in later chapters. The author also addresses the obvious question to stress in Chinese: why do native speakers lack good intuitions about where the stress falls? He argues that this is due to the fact that Chinese is a tone language, thus one of the major stress indicators, f0, cannot be freely altered to indicate stress, since it is being used to mark lexical contrasts.
Chapter 7 elaborates on the author's approach to the word length problem in SC under his theory of SC stress. The basic data patterns to be explained are the following. For modifier-noun [M N] constructions, a [1 2] (the numbers indicate the numbers of syllables in each syntactic category) structure, such as *mei shang-dian 'coal store', is bad, while [2 2], [2 1], and [1 1] structures are good. But for verb-object [V O] constructions, [2 1] (*zhong-zhi suan 'plant garlic') is bad, while [2 2], [1 2], and [1 1] are good. The author argues that the reason for the difference between [M N] and [V O] is metrical. In [M N], M is the nonhead and should receive stress. If it is [1 2], N can form a binary foot, but M cannot. Under the assumption that the main stress must be in a binary foot, the structure is ungrammatical. For [V O] however, O is the nonhead and receives stress, therefore [1 2] is good, but [2 1] is less optimal, since the object can only form a binary foot with the following empty beat, thus the main stress falls in a weak foot. The author also dismisses two widely held myths about Chinese words, the first being most Chinese words are monosyllabic, the second being Chinese developed many disyllabic words recently due to sound changes that resulted in a decrease in the size of its syllable inventory. According to two corpus studies (ZWGW 1959, He and Li 1987), the author argues that the majority of the modern vocabulary is in fact disyllabic. As for the increase in disyllabic vocabulary, he argues that it was primarily caused by the disyllabic or longer borrowings from Japanese and English, not by homophone avoidance.
Chapter 8 continues the discussion of word-level phonology and elaborates on the author's approach to the word order problem in SC, again in the light of his theory of SC stress. The basic patterns are as follows. For [V-O N] compounds, when V and O are both monosyllabic, [V O N] is the only possible order, e.g., qie cai dao 'cut vegetable knife'; but when V and O are both disyllabic, [O V N] is the only possible order, e.g., luo-bo jia-gong dao 'turnip process knife'. For [X Y N] compounds where X and Y are modifiers of N, when X and Y are both disyllabic, there is a fixed order between the two modifiers, e.g., da-xing han- yu ci-dian 'large-scale Chinese dictionary'; but when one of them is monosyllabic, the preferred pattern is to have the monosyllabic modifier in the Y position, e.g, han-yu da ci-dian 'Chinese large dictionary'. The full analysis of these patterns involves elaborate constraints and arguments, but the basic idea still follows from the theory of SC stress: first, word order variation is triggered by metrical requirement; second, the mechanism for word movement is Nonhead Fronting. The difference between [V-O N] and [X Y N] compounds comes from the difference in the position of the head.
Chapter 9 discusses the diminutive suffix [r] in SC. Traditionally, there have been three mysteries to the realization of stem syllables with the diminutive suffix. First, why does the [r] suffix replace some portion of the rhyme in certain rhymes, but is added to the rhyme in others? E.g., pai+r --> par 'board', pau+r --> paur 'robe'. Second, why is a schwa added between the rhyme and the [r] suffix when the rhyme has a high vowel? E.g., i+r - > iEr 'clothes'. Third, after the [r] suffix is added, why is the alveolar nasal coda lost, while the velar nasal coda is preserved as vowel nasalization? E.g., pan+r --> par 'plate', pang+r --> pa~r 'side'. The author answers these questions as follows. First, when a sound is articulatorily incompatible with [r], the sound is replaced. Otherwise [r] is added to the sound. Assuming that the phonetic realization of pang+r is pangr, this accounts for both the first and the third mysteries. Second, a high nuclear vowel spreads to the onset and thus leaves the nucleus slot empty; the default height value of the nucleus is then filled in, which is mid.
Chapter 10 lays out the basic properties of tones in SC. The discussion starts from the phonetic correlates of tone and the different systems of tonal transcription used in Chinese linguistics including IPA, number marking, and Chao letters. The author then moves on to discuss the phonological representation of tone and proposes to represent tone levels with two features--Register and Pitch. The two values for Register are stiff vocal cords and slack vocal cords, which indicate non-murmured and murmured quality of the vowel respectively. The two values for Pitch are thin vocal cords and thick vocal cords, which indicate high and low pitch respectively. The cross- classification of these two features yields four possible levels of tonal contrasts. The author then takes on the issues of contour tone representation and tone-bearing unit (TBU). He argues that contour tones are combinations of level tones, and the TBU is not the syllable or the rhyme, but the moraic segment. This approach accounts for the fact that a weak syllable in SC can carry only level tones, a full syllable can carry simple contour tones, and complex contour tones are usually restricted to final position only, where the syllable is lengthened. In this system, the four Mandarin tones on nonfinal full syllables--55, 35, 21, and 51 in Chao letters--are represented as [-mur, H], [-mur, LH], [+mur, L], and [-mur, HL] respectively, each associated with two TBUs. On monosyllables or final position of di- or polysyllabic words, the third tone 21 lengthens the syllable to trimoraic and is realized as 214. Various phonetic variations of these tones in different contexts are discussed in the light of these representations. The most interesting one also relates to the stress theory of SC proposed earlier: there is a difference in the realization of a final third tone between [M N] and [V O]. For 'sai ma', when it is a [M N] structure, which means 'a race-horse', the third-toned 'ma' is more likely to be realized as 21; but when it is a [V O] structure, which means 'to race horses', the third-toned 'ma' must be realized as 214. The account of these facts is again Nonhead Stress. In [M N], N is the head and thus not stressed, while in [V O], O is the nonhead and thus stressed. It is more likely for 214 to surface on a more heavily stressed syllable.
Chapter 11 focuses on the infamous third tone sandhi rule in SC. The exact application of this sandhi rule, especially in longer strings, is dependent on many factors, such as the syntactic branching, syntactic categories involved in the structure, and emphasis. When there are multiple ways to apply the sandhi, it sometimes gives various alternative surface patterns. And sometimes the sandhi rule is optional. The author proposes an analysis that accounts for all the facts. The analysis again relies on the syllabic trochee foot structure and Nonhead Stress and considers that the sandhi rule applies cyclically starting from each foot, and it is optional between two cyclic branches. The advantages of this metrical approach over previous approaches are also discussed.
The final chapter of the book, Chapter 12, raises further issues in SC phonology that the author has not covered. They include various connected speech phenomena such as consonant reduction, rhyme reduction, vowel devoicing, and syllable merger; phonological processes in other Chinese dialects, such as tone sandhi in Wu and Min dialects and rhyme changes under affixation in various dialects; and properties of Taiwanese accented SC.
CRITICAL EVALUATION
As I have said at the beginning of the review, this book is a comprehensive study of the phonology of Standard Chinese. Not only is it rich in detailed and amazingly accurate factual description, it also proposes elegant theoretical solutions to many long-standing problems in Chinese phonology, such as word length variation, word order, and the application of the third tone sandhi. Another great strength of the book is that in every chapter, the generative literature on related issues is carefully reviewed. Therefore it can also serve as a great reference book for the past advances in Chinese generative phonology. Moreover, the book is written in a down-to-earth fashion and is very approachable by anyone with the slightest interest in Chinese languages or phonology but relatively little training in either area.
As for the specific issues discussed in the book, Chapters 5 to 8 are probably my personal favorites. I find the author's arguments for the metrical structure of Chinese extremely cogent. The stress, word length, and word order problems have been long-standing in Chinese phonology, and the author's approach is the most systematic and comprehensive I have seen on these subjects. It is hard not to be deterred by the numerous phonetic studies that show the lack of consistent acoustic correlates and speaker intuition for SC stress, as many researchers are, but the author convincingly shows that stress must be phonologically relevant for SC, since otherwise many phenomena would go unexplained. Moreover, the author also shows that this stance is not necessarily in conflict with the phonetic studies, since unlike English, SC is a language with tonal contrasts, therefore its pitch cannot be freely used to mark stress.
I also like the author's position on the syllable structure of SC. Distinguishing SC syllables into two categories CVX and CV is phonetically accurate and phonologically beneficial, considering, for example, rhyming, which does not take into account the prenuclear glide; and tonal distribution, which clearly displays the distinction between full and weak syllables--full syllables can carry contrastive tonal contours, while weak syllables cannot.
But there are also issues on which I do not completely agree with the author.
In his discussion of the sound inventory of SC, the author does not seem to have a clear stand on the status of phonemic economy in phonemic analysis. At one point, he questions the importance of this concept (p.17), but later on, when proposing the treatment of palatals as dental+j combinations, he takes phonemic economy as one of the arguments for it (p.87). This reader would have liked to see more detailed discussion on the psycholinguistic importance of phonemic economy if it was going to be taken as an argument for an analysis, since the concept seems to have followed more from the tradition of treasuring elegance and symmetry than from any such importance.
I am also not sure about the arguments presented in the book for treating palatals as dental+j combinations. The author's main argument comes from the palatal-fronting speakers in Beijing, who pronounce palatal consonants as dental+j combinations. The author argues that the rule of making the Dorsal as a major articulator as well as the Coronal, which would create the palatals phonetically, does not apply for the palatal-fronting speakers; and only by treating the palatals this way can the relation between the two types of speakers be captured. But the author has noted that palatal-fronting speakers are mostly children and young women. In fact, it is associated with high feminimity. Therefore, even if the palatals are just palatals underlyingly, when the speaker on the one hand values feminimity highly and thus prefers high frication noise, but on the other hand only allows a minimum deviation from the underlying form, palatals will surface as dental+j combinations. There is no need to consider them to have come from the same underlying source. Moreover, this accounts for why palatal-fronting speakers are mostly children and young women. But Duanmu's approach here provides no explanation as to why this particular group of speakers is more prone to dropping the rule that makes Dorsal a major articulator.
In the discussion of phonetic realizations of SC rhymes, Duanmu uses the rhyming groups as evidence for determining the surface forms, and in turn, underlying forms of the rhymes. E.g., since [in] and [yn] belong to the same rhyming group, he considers them to have to the same surface rhyme, i.e., [yn] is in fact [yin], and [yin] comes from underlying /win/ by spreading the high vowel to the onset. This practice is somewhat worrisome, since it is well known that semi-rhymes are commonly practiced in the poetry of many languages. So there is no guarantee that two rhymes are identical even though we know that they rhyme. Thus it is entirely possible that [in] and [yn] rhyme because they are phonetically similar, not because they are identical. Duanmu is in fact aware of this problem and discusses it in the 'Further Issues' section of Chapter 3. But without a clear argument against this approach, it is not clear to me why he opts for the alternative that makes more assumptions. This also brings me to a minor point: in the appendix, Duanmu lists all the full syllables in SC, with both underlying and surface forms. But a few of the listed items do not seem to agree with what is argued for in the text. E.g., in the text, all glides are underlyingly high vowels, but in the appendix (p.276), the underlying forms for [waa], [wai], [wan], etc. are listed to have glide onsets--/wa/, /wai/, /wan/, etc.; in the text, [yn] is argued to be [yin], which comes from underlying /win/, but in the appendix (p.277), the underlying form for the same syllable is written as /yn/.
Regarding the n-ng asymmetry in [r] suffixation, Duanmu argues that n is incompatible with [r] and must be replaced, and ng is compatible with [r], so the [+retroflex] feature is simply added onto the syllable with an ng ending. He also argues that the [CVngr] is phonetically similar to [CV~r], since Wang (1993)'s phonetic study shows that SC nasal codas are nasal glides without a velar closure. But Zhang (2000) shows that in the unsuffixed form, [CVng] has a clear dorsal raising movement, but in the suffixed form, there is no trace of dorsal raising left. This indicates that [CVngr] and [CV~r] are phonetically different. So if the analysis outputs [CVngr] instead of [CV~r], the pattern is not fully explained. Zhang (2000) offers a perceptual account couched in Optimality Theory to explain the n-ng asymmetry and testifies that the factorial typology of the proposed constraints only produces attested patterns.
I have some reservation about the use of binary features Register and Pitch to represent tonal levels. Phonetically, it seems unlikely that if a language has four level tones, it uses exactly the combinations of stiff/slack and thin/thick vocal cords to implement these pitch levels. Moreover, it is not clear to me what the author means by 'murmur'. On the one hand, he equates murmur with breathiness, and he states that it 'correlates with broader formant width and flatter spectral envelop' (p. 213), although a flatter spectral envelop is the spectral characteristic of creakiness. On the other hand, he claims that the third tone in SC has a [+murmur] quality, but it is well known that this tone is creaky and is unlikely to have slack vocal cords. Phonologically, it makes the typologically odd prediction that in a language with four level tones, the middle two tones, although they phonetically have the most similar pitches, will not behave as a natural class, since they have exactly the opposite feature values. This prediction has been successfully challenged by Tsay (1994) and her later works. The scalar representation proposed by Tsay seems to be a better characterization of phonological pitch levels.
Regarding the representation of contour tones as sequences of level tones, there might be both psycholinguistic and phonetic evidence against it. Psycholinguistically, Wan (1999) and Wan and Jaeger (1999) show from the behavior of speech errors that SC tones are underlyingly unitary and are not made of level tone sequences, since no tone blends or tone spreading errors were observed. (On a related note, it is somewhat unfortunate that the works by Wan and Jaeger on speech errors were not cited as evidence for or against various proposals on the sound inventory and syllable structure of SC.) Phonetically, Xu (1998) shows from the consistency of tone-syllable alignment across different conditions that contour tones are probably implemented as dynamic targets rather than sequences of static targets. Zhang (2001), on the other hand, shows that even if contour tones are considered units, all is not lost regarding the relation between syllable duration and contour tone distribution. If all tones are associated with a tonal complexity index, and a complex contour tone has a higher tonal complexity than a simple contour tone, and a simple contour tone has a higher tonal complexity than a level tone, then the relation between tonal distribution and duration can still be successfully drawn. Moreover, Zhang (2001) shows that only an enriched representation like this can capture all the attested patterns of contour tone distribution.
Finally, when I was reading the book, although on the one hand, I was glad that it was generally presented in a theory-neutral fashion, since this would make it approachable to a wider audience and more room was left for the presentation of the data; on the other hand, I also secretly hoped that there was a consistent theoretical framework that the book held onto, so that more accurate predictions can be made using the general principles proposed by the author, and more accurate comparisons can be made with other proposals. As it is, Chapter 3 (sound combinations and variation) is presented in a somewhat odd version of Optimality Theory, which looks like it has restrictions on the input, but no faithfulness constraints. The rest of the chapters is presented in a combination of rule-based framework and constraint-based framework, which at times makes the proposals hard to evaluate.
In summary, although there are various points where my view diverges from the author's, overall I agree with the author much more than I do not. The book will prove to be an invaluable resource for anyone interested in the study of Chinese phonology, and it has set a high standard for researchers in Chinese phonology to follow. Somewhat unrelatedly, one particularly enjoyable moment of book comes during the preface, when the author compares the question 'which language do you study' to a linguist with the question 'which country do you study' to a geologist. At least now we know what to say when we encounter a question like that.
BIBLIOGRAPHY
Cinque, Guglielmo (1993) 'A Null Theory of Phrase and Compound Stress', Linguistic Inquiry 24: 239-297.
Duanmu, San (1990) A Formal Study of Syllable, Tone, Stress and Domain in Chinese Languages, Ph.D. dissertation, MIT.
He, Kekang and Dakui Li (1987) Xiandai Hanyu Sanqian Changyong Cibiao [Three Thousand Most Commonly Used Words in Modern Chinese], Beijing: Beijing Shifan Daxue Chubanshe [Beijing Normal University Press].
Tsay, Suhchuan Jane (1994). Phonological Pitch, Ph.D. dissertation, The University of Arizona.
Wan, I-Ping (1999) Mandarin Phonology: Evidence from Speech Errors, Ph.D. dissertation, SUNY at Buffalo.
Wan, I-Ping and Jerri Jaeger (1999) Speech Errors and the Representation of Tone in Mandarin Chinese. Phonology 15: 417-461.
Wang, Jenny Zhijie (1993) The Geometry of Segmental Features in Beijing Mandarin, Ph.D. dissertation, University of Delaware, Newark.
Xu, Yi (1998) 'Consistency of Tone-Syllable Alignment across Different Syllable Structure and Speaking Rates'. Phonetica 55: 179-203.
Zhang, Jie (2000) 'Non-Contrastive Features and Categorical Patterning in Chinese Diminutive Suffixation-- Max[F] or Ident[F]?' Phonology 17: 427-478.
Zhang, Jie (2001) The Effects of Duration and Sonority on Contour Tone Distribution--Typological Survey and Formal Analysis, Ph.D. dissertation, UCLA.
ZWGW (Zhongguo Wenzi Gaige Weiyuanhui Yanjiu Tuiguang Chu [Chinese Language Reform Committee Research and Popularization Office]) (1959) Putonghua Sanqian Changyong Cibiao [Three Thousand Commonly Used Words in Standard Chinese], preliminary edition, Beijing: Wenzi Gaige Chubanshe [Language Reform Press].
BIOLOGICAL SKETCH
Jie Zhang received his Ph.D. in linguistics at UCLA in June 2001 and will be a lecturer in the Dept. of Linguistics at Harvard University for the academic year 2001-2002. His research interests include phonology, phonetics, the effects of phonetics in phonological patterning, patterns of tones and nasals, Chinese languages, Athapaskan languages, and Otomanguean languages.
|