On the nature of Mandarin tone and tone sandhi

Hua Lin

Bachelor of A rts, Lanzhou U niversity, 1982 M aster of E ducation, U niversity of V ictoria, 1987 A DISSERTATION SUBMITTED IN PARTIAL FULFILLMENT

OF THE REQUIREMENTS FOR THE DEGREE OF

A C C L

T K D

/■ ^ / , ] V TY We a c c e p t this d isse rta tio n as conform ing r /--- --- /.—J to th e req u ire d sta n d a rd

Dr. B arry ^ C j r l s o n , Supeipvisor (D ep artm en t o f Linguistics)

ThomasFT. Hukari^ e p a r ^ t m i S ^ J c ^ e m b e r (D epartm en t o f L inguistics) Dr.

Dr. Ja m e s A rthu rs, D e p a r t ( D e p a r t m e n t of L inguistics)

Dr. D aniel J . B r y ^ t , Outslde~TVlernber (Dgpagtxn~gnt of P a c ific and A sian S tudies)

Dr. William s W . V s lo ^ j^ x tè rn a l E :^ 4 m i^ r (^ n iv e rs j^ ÿ ^ f C alifo rn ia a t B erkeley) ® Hua Lin, 1992

UNIVERSITY OF VICTORIA 1992

All rig h ts reserv ed . This d is s e rta tio n may not be rep ro d u ced in whole or in p a rt, by m im eograph o r o th e r m eans,

ABSTRACT

T rad ilio n al rep re se n ta tio n s of M andarin tones have provided valuab le d e sc rip ­ tions o.' M andarin tone sandhi pro cesses. H ow ever, none of them has been able to to a s s o c ia te th e se processes in a p rincip led way, or to explain why th e y o ccu r a t all. In c o n tra s t, I have proposed in th is d isse rta tio n a unified an aly sis of M andarin to n e s and to ne sandhi, w ith an em phasis on th e rev elatio n of th e n a tu re o f th e se to n es and p rocesses. S pecifically, I have found th a t M andarin to n e s a re m ost o p ti­ m ally re p re s e n te d as follows: T o ne 1 Tone 2 Tone 3 To ne 4 To ne 0 a 1 0 1 0 1 a a 1 1 T 1 T 1 T T 1 T t o n a l t i e r / 1 \ / 1 \ / l \ / l \ 1 H H H M H H L L M H M L L t o n e m e t i e r

U nder such a re p re se n ta tio n , all M andarin tone sandhi p ro c e sse s (i.e. th e s e c ­ ond, th ird , and fo u rth to n e sandhi p ro cesses, and th e n e u tra l to n e sandhi p ro cess th a t has been uncovered in th is study) can be uniform ly a c c o u n te d fo r by th e fo l­ lowing Tone R eduction P rinciple:

( 0 . 2 ) . T o n e R e d u c t i o n P r i n c i p l e C l a u s e A: I n n o r m a l s p e e c h , r e d u c e a t o n e by o n e t o n e m e i f f i t i s i m m e d i a t e l y f o l l o w e d b y a n o t h e r t o n e w i t h i n t h e same p r o s o d i c f o o t . C l a u s e B: I n f a s t s p e e c h , r e d u c e a t o n e by o n e t o n e m e i f f i t i s i m m e d i a t e l y p r e c e d e d b y a n o t h e r t o n e , and a t t h e same t i m e i m m e d i a t e l y f o l l o w e d by a n o t h e r t o n e w i t h i n t h e same p r o s o d i c f o o t . n

-follow ing im p lem en tatio n rules: ( 0 . 3 ) . T T T / l \ + / h + / \ / T t ml tm2 tm3 tml tm2 tm3 t ml tm2 T T T / \ + / \ l / T _____T t ml tm2 tml tm2 tm2

With th e se tw o rules, M andarin to ne sandhi processes can be d escrib e d by the follow ing d eriv atio n s:

( 0 . 4 ) . a . The f o u r t h t o n e s a n d h i ( 4TS) ip ip T T T T / K + / l \ + / \ H M L H M L HM b . The t h i r d t o n e s a n d h i : A ( 3 T S ( A ) ) ip fp T T T T

/ l \

4. /IH

4.

/ \

/ l \

/ l \

♦ /IH

/ 1\

+ / \

/ | \

/ l \

+

I

/ l \

I

+ /IX

I

+ / \

I

While th e im p lem en tatio n ru les in (0.3) produce g ra m m a tic a l re s u lts in all o th ­ e r c a se s, th e y y ield o u tp u ts, in th e ca se s of 3TS(B) and OTS, th a t v io la te th e fo l­ low ing O C ? r e la te d , Mandarin sp e c ific WFC:

-l \

r\

X X

w h e r e % = H, M, o r L

T h ere fo re , th e se tw o outputs o b ligatorily undergo th e follow ing O CP re p a irs ((a) fo r STS and (b) for OTS):

( 0 . 6 ) . a T T

I X

r\

I X

y | x

In b rief, all th e M andarin tone sandhi p ro cesses a re fu n d am e n tally to n e re d u c ­ tion p ro cesses, th e re su lts of w hich may be su b je c te d to f u r th e r m o d ificatio n should th ey tu rn o u t to be violations o f c e rta in WFCs. In add ition to th e ru le s and d eriv atio n s p re se n te d above, th e analysis proposed also co n tain s a th e o ry of to n em es as tim in g u n its, a th eo ry o f M andarin syllable w eigh t and q u a n tity , and a th e o ry of th e diach ron ic im plications of th e analysis.

Dr. Bmyy F. C arlson, Supervisor (D ep artm en t of Linguistics)

D r. Thom as E. H u k a m ,D e p a rtm e n ta l M ember (D ep artm en t of L inguistics)

Dr. Ja m e s A rth u rs, D ep S rtm en tal Member (D ep artm en t of Linguistics)

Dr. D anieLJi'Bryant;-0utSTde-M «rai}er (D ep artm en t of P acific and Asian Studies)

Dr. William S-Y. Wang, E x tern al Exam iner (U niversity of C nlifornia a t Berkeley)

-A b s t r a c t ... "

T able of C o n t e n t s ... vi

A c k n o w le d g e m e n ts ... viii

D ed ication ... % Chapter I; In tro d u ctio n ...1

1.1 The Chinese L a n g u a g e s... 5

1.2 M andarin as S tandard C h in e s e ... 8

1.3 The Minimal Domain ... 9

1.4 D ata T r a n s c r ip tio n ... 10

Chapter II: A Review o f Previous Studies on T o n e ... 12

2.1 In tr o d u c tio n ... 12 2.2 Pike (1948) ... 16 2.3 Chao (1930, 1 9 6 8 )... 17 2.4 Wang ( 1 9 6 7 )... 19 2.5 Woo (1 9 7 2 )... 21 2.6 G oldsm ith ( 1 9 7 9 ) ... 27 2.7 Yip (1980a) ... 32 2.8 Yip (1 9 8 9 a ,b ) ... 37 2.8.1 Evidence from D a n y a n g ... 39 2.8.2 Evidence from T i a n j i n ... 41

2.8.3 Evidence fro m W u x i... 41

2.9 Bao ( 1 9 9 0 ) ... 43

2.10 C o n c lu s io n ... 48

2.11 N otes to C h a p te r 2 ... 49

Chapter III: The R egister Feature S y s te m ... 52

3.1 M andarin C ase 1: th e N eu tral T o n e ...54

3.2 M andarin C ase 2: Tone on P ost-V erb S y l l a b l e s ... 60

3.3 M andarin C ase 3: th e "Conclusion" I n to n a tio n ...65

3.4 M andarin C ase 4: "Intonation P a r t i c l e s " ... 70

3.5 Yin and Yang Tonal Split ...75

3.6 The C o nstraining of th e G e n erativ e P o w e r ... 80

3.6.1 P ack ard 's (1989) S t u d y ... 82

3.6.2 The F u nction o f a F e a tu re System ... 85

3.7 Physiological C o n s id e ra tio n s ... 88

3.8 N otes to C h a p te r 3 ... 89

-4.1 P r e lim in a r ie s ... 92

4.1.1 A Working D efinition for th e Term " T o n e " ...93

4.2 The R ep resen tatio n of M andarin Tones ...96

4.2.1 The N atu re of th e T o n e m e ... 99

4.2.2 The D uration of Tone 3 ... 101

4.2.2.1 Howie's S t u d y ... 104

4.2.2.2 Woo's S t u d y ... 109

4.2.3 The Tonem ic R ep re se n ta tio n o f a T o n e ... 112

4.2.4 Mandarin E vidence fo r T re a tm e n t of C ontours as U n i t s ...115

4.2.5 A Theory of M andarin Tim ing U nits ...117

4.3 N o tes to C h ap ter 4 ... 119

Chapter V: The Mandarin Syllable ... 123

5.1 P r e lim in a r ie s ... 123

5.2 The Q u an tity of th e M andarin S y lla b le ... 126

5.3 The S y llab ification of th e M andarin S y lla b le ... 130

5.4 The Tim ing Function of th e M andarin S y l l a b l e ... 136

5.5 T he Double F a c e ts of th e M andarin S y lla b le ... 136

5.6 N otes to C h ap ter 5 ... 138

C lis^ter VI: Mandarin Tone S a n d h i... 141

6.1 Which is U nderlying? ... 142

6.2 T he N atu re of M andarin Tone S a n d h i ...143

6.2.1 The F ourth-T one Sandhi (4 T S )... 144

6.2.2 The T hird-T one Sandhi: A (3TS(A)) ... 149

6.2.3 The T hird-T one Sandhi: B (3TS(B))... 152

6.2.4 The Second-Tone Sandhi (2TS) ... 162

6.2.5 The N eutral-T on e Sandhi (O T S )... 170

6.3 Sum m ary and Conclusion ...177

6.3.1 Tow ard A G en eralized T heory ... 182

6.4 N o tes to C h ap ter 6 ... 185

Chapter VII: Further E v id e n c e ... 190

7.1 Tem po and Domain Size; C heng's S t u d y ... 190

7.2 D iachronic Evidence: C hen's Study ... 192

7.3 D iachronic Evidence: Shih's S t u d y ... 194

7.3.1 Two Types of T onal S y s te m s ... 195

7.3.2 The H isto ric al D e v e lo p m e n t... 197

7.3.3 Evidence from X ianyou ...202

7.3.4 Evidence from Z h a n g p in g ... 205

7.3.5 Evidence from F u z h o u ...206

7.3.6 Evidence from S u z h o u ...208

7.3.7 Evidence from D a n y a n g ... 210

7.3.8 Evidence from W u x i... 214

7.3.9 C o n c lu s io n ... 215

I am g ra te fu l to many people h ere in V ictoria who have g enero usly given me tim e and ad v ice, and shown me kindness, u nd erstanding and p a tie n c e during my d o cto ra l stu d ie s.

I e x te n d th e d eep e st g ra titu d e to my c o m m itte e m em bers, p a rtic u la rly to my superv isor, Dr. B arry C arlson. I enjoyed v ery much Dr. C arlson's s ty le o f su p e rv i­ sion. He did n ot s e t too many rules fo r me to follow , but alw ays show ed his c o n fi­ dence in m e, and w as alw ays th e re when I needed help. Also, I g re a tly a p p re c ia te his v ery c a re fu l p eru sal of d ra fts of th is d is s e rta tio n . His ex ten siv e and d e ta ile d co m m en ts on th e se d ra fts led to many sig n ific a n t revisions of th is d is s e rta tio n . Dr. Tom H ukari, a n o th e r m em ber of th e c o m m itte e , also m ade m ajo r c o n trib u ­ tions to th is d is s e rta tio n . His " relen tless" co m m en ts on an e arly d r a f t p u t me on gu ard and m ade me to think tw ice b efo re m aking any conclusion. A lthough I found those co m m en ts hard to swallow a t th e tim e (but n ot b ecau se I w as n o t co n ­ vinced), th e y proved to be ex tre m ely insigh tfu l and b en eficial to my la t e r re v i­ sions. The c o m m itte e m em ber D r. Ja m e s A rth u rs was also th e in s tru c to r of my f ir s t lin g u istic s course. Thus, I am in d e b t to him also b eca u se it w as his q u a lita ­ tiv e p re s e n ta tio n of th e basics in th is fie ld , p u n c tu a te d w ith hum orous a n e c d o te s ab ou t lan g u ag es, th a t led me into th e fa s c in a tin g w orld o f lin g u istic s. As w ell, my thanks go to D r. D aniel B ryant, th e o u tsid e m em ber, who, to o, g a v e v alu ab le co m ­ m ents on th e fin al d r a ft of th e d is s e rta tio n , and to D r. R ich ard King, th e "ex "-o u tsid e m em ber, who had given m e e x c e lle n t advice and g u id an ce b e fo re h av ­ ing to le av e th e c o m m itte e a t th e fin al sta g e to go on sa b b a tic a l.

-The e x p e rtise Dr. Wang provided in th e a re a s of C hinese phonology and h isto rical phonology was of p a rtic u la r im p o rtan ce in th e fo rm u latio n of a num ber of ideas in th e d is s e rta tio n . A lthough he was an e x tre m e ly busy person, he was alw ays a v a il­ ab le, and alth ou gh he tra v e le d co n stan tly , he was alw ays reac h ab le. I still re m e m ­ b e r "bugging" him through e le c tro n ic m ail fo r m a te ria ls, advice and in fo rm atio n , when he was aw ay on in te rn a tio n a l trip s.

I am v ery g ra te fu l to th e D e p a rtm e n t of L inguistics and to its C hair, Dr. Josep h F. K ess, fo r th e fin an c ial support both in th e form of th e fellow ship and th e te a c h in g /re s e a rc h eissistantship, and to Dr. R ichard King fo r hiring me as a T eaching A ssista n t fo r C hinese. W ithout such fin an cial a ssista n c e , th e re would no t have been th is d isse rta tio n .

The follow ing people have also helped me in various ways during my d o cto ra l stu d ie s: P ro fesso rs Tom H ess, Leslie Saxon, John Esling, B arbara H arris, th e g en ­ e r a l o ffic e s e c r e ta r ie s D arlene and G retch e n , and my friend s Bob, Robynne and th e la te N ancy Sw ailes. The Swailes w elcom ed me to V ictoria and very kindly helped me th ro u g h th e c ritic a l period o f c u ltu ra l shock.

L ast b u t n o t le a s t, I thank my fam ily : my husband, my son, my m o th er, my b ro th e r and s is te rs . My son alw ays showed me his un derstanding when 1 was busy w ith th e d is s e rta tio n and could no t have tim e fo r him , even though he did n o t r e a l­ ly u n d erstan d why mom had to work on "th e book".

I assum e fu ll resp o n sib ility fo r any m istak es in th is d is s e rta tio n .

-INTRODUCTION

The m ost w ell-know n fe a tu re of th e C hinese language is probably th a t it is a lan gu age w ith phonem ic to n es. A lthough Chinese tones (or to n es in languages in g en eral) have su ffe re d re la tiv e n e g le c t in th e h isto ry of g e n e ra tiv e phonology, the la s t d eca d e has w itnessed a stead y g row th o f in te re s t in th e ir stu d y . Largely inspired by th e r e c e n t th e o re tic a l adv an ces made in g e n e ra tiv e phonology, in p a r­ tic u la r , by th e ad v en t of n o n -lin ear phonological th e o ries such as a u to seg m en ta l phonology (G oldsm ith 1979, 1990, L ieber 1987), m e trica l phonology (L iberm an and P rin c e 1977), and prosodic and se n te n c e phonology (Selkirk 1978, 1984, 1986; K aisse 1985, K aisse and Zwicky 1987, M cC arthy 1982, M cC arthy and P rin ce 1986, N espor and Vogel 1983, 1986, Vogel 1984, Zwicky 1985), th is in c re ased in te re s t has re s u lte d in a sizab le num ber of stu d ie s being conducted w ith an exclu siv e focus on

C hinese to n e s.

W ithin th e field s of prosodic and se n te n c e phonology, fo r in sta n c e , C hinese to n e s and th e ir w ell-no ted sandhi p ro cesses have been re -in v e s tig a te d , this tim e , fro m a m ore global p e rsp e c tiv e . E arlier, Cheng (1973) had m ade th e bew ildering o b se rv a tio n th a t although M andarin tone sandhi processes alw ays o ccu r w ithin an is o la te d s trin g of tw o syllables, th e ir ap plication becom es so m eth in g of a puzzle w hen m ore sy llab les a re added to th e strin g . S p ecifically, when th e strin g becom es lo ng er, th e a p p lic a tio n of th e ton e sandhi rules becom es som ehow c o n tro lled by th e s u rfa c e s y n ta c tic s tru c tu re . D espite the intriguing n a tu re o f C heng’s o b serv a­ tio n , th e s y n ta x -to n e re la tio n has rem ain ed elusive and in tra c ta b le fo r tw o d e c ­

ades, due ap p are n tly to the lack of a viable tool for its study. This situ a tio n , how­ e v e r, has g re a tly changed since th e th e o ries of prosodic and s e n te n c e phonology cam e into being. A rm ed with th e se th e o ries, many studies have been co n d u cted , probing into th e sy n tax -to n e re la tio n no t only in M andarin b u t in o th e r C hinese languages as well (Chen 1990, Cheng 1987, Chung 1992, Hung 1987, Lin 1990 and 1991, Shen 1986, Shen 1988, Shih 1986, Selkirk and Shen 1988 and 1990, W right 1983, and Zhang 1988). These stu d ies have added much to th e c u rre n t u n d e rsta n d ­ ing of th e in te ra c tio n in question. For in sta n ce, one m ajor u n d erstan d in g th a t has been adv an ced by th ese stud ies is th a t C hinese ton e sandhi ru le s do n ot e x a c tly apply on s y n ta c tic b rack e ts, as th e y had been tho ug ht to, bu t r a th e r , on som e kind o f an in te rm e d ia te prosodic s tru c tu re com p letely or p a rtia lly p ro je c te d fro m th e s y n ta c tic s tru c tu re .

B esides being approached ex te rn a lly in te rm s of th e dom ains of th e ir sandhi b eh av io r, C hinese ton es have also been ta ck led from w ithin. The ag e-o ld q u estion of how C hinese ton es should be m ost o ptim ally re p re s e n te d by a phonological f e a - tu ra l sy stem is raised again in ligh t o f th e new ly-developed a u to se g m e n ta l phonol­ ogy. N ot e x a c tly surprisingly, th e inquiry in th is d irectio n has led to th e conclusion th a t C hinese to n es a re essen tially a u to seg m en ta l (Yip 1980a). A c o n c e p t new ly develo ped by G oldsm ith (1979), th e te rm "au to seg m en tal" r e fe r s to th e s ta t e of a phonological fe a tu r e no t behaving co m p lete ly in a c c o rd an ce v nth th e r e s t o f th e f e a tu r e s , b ut showing independent or autonom ous tr a its . Since Yip's p io n e e r work in 1980 (i.e. 1980a), th e re have been a num ber of stu d ies on th e in te rn a l s tr u c tu re o f C hinese to n e s (Yip 1989a,b, Bao 1990, Shen 1985, P ack ard 1989, Zee 1991).

In s p ite of th e in creased in te re s t and th e larg e num ber of stu d ie s th a t have e m erg ed as a re su lt of it, a review of th e lite r a tu r e still re v e a ls som e p rob lem s (or gaps) in th e stud y of C hinese to n es. G enerally speaking, to n e s can be ap p roached

from a t le a s t tw o d iffe re n t p e rsp e c tiv e s. They can be stu d ied in te rm s o f th e ir in te rn a l re p re s e n ta tio n , or they can be ta ck led in term s of th e ir sandhi pro cesses. The sandhi p ro cesses can them selv es be approached in tw o fashions. They can be analysed to i.nsw er the question of w here th ey actu ally occur in a p h rase o r sen te n c e . A lte rn a tiv e ly , th ey can be exam ined in tern ally to see w hat e x a c tly h ap ­ pens when th e y ac tu a lly occur; nam ely, w hat the ex act n atu re of th e se sandhi p ro ­ cesses is, or why ton e sandhi p rocesses occu r a t all w ithin th o se sp e c ifia b le dom ains.

H ow ever, fo r som e reason o r o th e r, th e overw helm ing m ajority of the re c e n t stu d ie s on C hinese ton es focus exclusively on th e problem of the dom ains o f C hi­ nese to n e sand hi, co nd ucted w ithin th e prosodic and sen ten c e phonology. R e la tiv e ­ ly few ex am in e th e in tern al re p re s e n ta tio n of Chinese to nes. In ad d itio n , none of th e se stu d ie s has a tte m p te d to explain why C hinese tone-sandhi p ro cesses o ccu r. A lthough it m ay be tru e th a t th e stu d ie s of th e tone-sandhi dom ains m ay, to a la rg e e x te n t, ignore th e in te rn a l s tr u c tu re of th e p e rtin e n t to ne, th e e x a c t shape o f th e to n e sandhi rule fo r w hich th e dom ains of ap plicatio n are sought c ru c ia lly depends on an o p tim al re p re s e n ta tio n of th e p e rtin e n t to n e. Such a re la tio n b etw ee n re p re s e n ta tio n and p ro cesses is due to an understanding re c e n tly re a c h e d in n o n -lin e a r g e n e ra tiv e phonology. A ccording to th is understanding, phonological re p re s e n ta tio n is m ore fu n dam ental than phonological p rocesses in th e sen se th a t c o n fig u ra tio n o f ad eq u a te ru les of phonological processes (such as to ne sandhi rules) c ru c ia lly depend on an o ptim al re p re s e n ta tio n of th e e le m e n ts (such a r tone) involved in th e phonological p rocesses.

A n o th er g en era l problem in th e stu d ie s of Chinese to n es is th a t m ost of th e stu d ie s c o v e r in th e o ry all of the to n al sy stem s of th e C hinese languages. A lthough such a co v erag e w ithin a single work of th e varied (though closely r e l a t ­

ed) languages may not in itse lf p re se n t a problem , it does p oint to a lack of serio u s e f f o r ts in m ore focused, in-d epth stu dies of ju s t one of th e languages c o v e re d under th e proposed th e o rie s. One m ajor c h a ra c te r is tic of th e se r e c e n t stu d ie s is th a t th ey are all theo ry -d riv en , which is not surprising given th a t th e y a re all inspired by th e re la tiv e ly re c e n t av ailab ility of n o n -lin ear phonological th e o rie s . Being th e o ry -d riv en is again not qu estion ab le in lin gu istic stu d ies. H ow ever, when a study is both th e o ry -d riv en and a t th e sam e tim e ta rg e te d a t a la rg e a r r a y of various (though clo sely -related ) languages, th e v alid ity of th e th e o ry em erg in g from it may need to be fu rth e r v e rifie d . One n a tu ra l w ay to v e rify such a th e o ry is, of cou rse, to con du ct in-depth stu d ies o f th e individual languages c o v e re d by it.

Tne g e n e ra l purpose of th is d isse rta tio n is defin ed according ly . It is to co n d u ct an in -d ep th study of one of th e C hinese languages: M andarin. S p ecifically , th e p u r­ pose is tw ofold: to in v e stig a te (1) th e n a tu re o f M andarin to nes, nam ely, th e r e p ­ re s e n ta tio n a l a s p e c t of M andarin to n es, as w ell as (2) th e n a tu re o f M andarin to n e sandhi pro cesses.

As im plied in th e above discussion, th e re p re s e n ta tio n a l a s p e c t o f M andarin to n e s has not re c e iv e d su b sta n tia l, clo se-u p stu d y in g e n e ra tiv e phonology. F u r­ th e rm o re , no stu d ie s have d e a lt w ith th e e x a c t n a tu re o f th e M andarin to n e sandhi pro cesses, though much has been th e o riz e d w ith re g a rd s to w h ere, given a strin g of syllables, th e y will o ccu r. H ow ever, th e f a c t th a t M andarin to n e sandhi v a rie s in n a tu re from to n e sandhi in o th e r C hinese languages has been n o te d in s e v e ra l stu d ie s (Chen 1990, Zhang 1989 am ong o th e rs). But w h at e x a c tly is m e a n t w hen Chen rem a rk s th a t M andarin tone sandhi p ro cesses a re basically p h o n e tic in n a tu re w hile th o se in o th e r C hinese languages a re n o t? To answ er th is q u e stio n , it is obviously n ece ssary to know f irs t of a ll th e tru e n a tu re of M andarin to n e s and to n e sandhi p ro cesses.

The o rg an izatio n of this d isse rta tio n is as follows: th e rem ain ing sectio n s of th is c h a p te r provide som e prelim in ary inform ation about th e focus of this stu dy and th e d a ta -tra n s c rip tio n conventions adopted in it. In C h ap ter 2, a g en era l su r­ vey o f th e lite r a tu r e on ton al re p re se n ta tio n is co n du cted. One study b riefly rev iew ed (see § 2.6) has sp ecial im po rtan ce to th e p re se n t study. It is G oldsm ith's (1979) A u to se g m e n ta l Phonology. In fa c t, the review of G oldsm ith's work is also m e an t to in tro d u ce th e basic th e o re tic a l fram ew ork used in th e analysis proposed in th is d is s e rta tio n . Within th e lite r a tu r e on C hinese tonal re p re s e n ta tio n , Yip's (1980a) work stan d s o u t as v irtu ally th e only one th a t is both com prehensive and co n d u cted w ithin th e c u rre n t n o n -lin ear phonology. H ow ever, it is d e m o n stra te d in C h a p te r 3 th a t Yip's fo u r-le v e l, re g is te r - fe a tu r e to nal system lacks m o tiv atio n fro m M andarin. F u rth e r, in C h a p te r 4, 5 and 6, it is shown th a t M andarin to nes a re b e s t an alysed in te rm s of a th re e -le v e l sy stem , w ith w hich a unified analysis o f th e M andarin to n e sandhi ru les becom es fo r th e f ir s t tim e possible. The th r e e - le v el system is sp elled o u t in C h a p te r 4. C h ap ter 5 discusses an im p o rta n t th e o ­ r e tic a l c o n c e p t, a co n c e p t co ncerning seg m ent tim in g. This co n c e p t c o n s titu te s an in te g ra l p a r t o f th e th e o ry proposed in C h ap ter 4. C h a p te rs 6 probes in to th e n a tu re o f th e M andarin to n e sandhi p ro cesses, and a u nified acc o u n t of them is proposed in it. F u rth e r su p p o rt fo r th e proposed M andarin to ne sandhi analysis is d raw n from elsew h ere, including h isto rical eviden ce th a t involves o th e r C hinese lan gu ages. The discussion of th is fu rth e r evidence is provided in th e la s t c h a p te r.

1.1

The Chinese Languages

T h ere has been m uch c o n tro v e rsy concerning w h eth er to speak of the C hinese lan gu age in th e sin g u lar or th e Chinese languages in th e p lu ral. It is g en era lly known th a t C hinese has eig h t bran ch es. The c o n tro v e rsy is hence over w h eth er to

re g a rd th e eig h t branches as d ia le c ts o f a single language o r langu ages th e m selv es. The m ajor arg u m e n t for the l a t t e r position is based on th e C rite rio n of M utual In te llig ib ility (Steinbergs 1987). By th is c rite rio n , tw o ling uistic e n titie s a re d ia­ le c ts if th e y a re m utually understandab le; o th erw ise, they a re lan g u ag es. Since th e eig h t b ran ch es of Chinese a re m utually u nintelligible, they should th en be re g a rd ­ ed as languages.

An a rg u m e n t fo r th e opposite po sition is th a t th ese b ran ch e s of C hinese a re m u tu ally unintelligible prim arily b ecause th ey have sig n ifican tly d iff e r e n t phono­ log ical s tr u c tu re s . When it com es to m orphology and syntax, r e la tiv e ly li ttl e d if­ fe re n c e is found am ong them . O th e r arg u m e n ts fo r th e o p p o site p o sitio n a re draw n on lan g u ag e-ex te rn al f a c to rs such as th a t all th e se b ran ch e s s h a re th e sam e w ritte n fo rm , and th a t all th e sp eak ers of C hinese a re of th e sa m e e th n ic origin, sh arin g th e sam e history and c u ltu re (cf. B ax ter 1992, D e F ra n c ic 1984, R am sey

1987).

I do n o t wish to join th is a rg u m e n t h e re . In th is p a rtic u la r stu d y , I will use "th e C hinese languages" in th e p lu ral fo rm . This choice is m ade in lig h t o f th e co n sid e ra tio n s th a t th e p re se n t study is a phonological study (and n o t a m orpholog­ ic al o r s y n ta c tic one), and th e f a c t th a t, as ju s t m entioned, th e s e b ra n c h e s o f Chi­ nese a re sig n ific a n tly d iffe re n t phonologically. The choice o f th e p lu ra l fo rm also r e f le c ts a p r a c tic a l con sid eratio n; 1 can th en r e f e r to th e su b -b ra n c h e s o f one of th e se C hinese languages as d ia le c ts, th us avoiding u n n ecessary confusio n th a t is likely to re s u lt if th e te rm " d ia le c t" is used fo r item s a t b o th le v e ls. The follow ­ ing illu s tra te s my use of th e te rm s "d ia le c t" and "language".

L a n g u a g e s D i a l e c t s 1. M a n d a r i n ( N o r t h e r n D i a l e c t ) Mandar i n ( S t a n d a r d D i a l e c t ) M a n d a r i n ( B e i j i n g ) O t h e r s Wu S h a n g h a i O t h e r s

Gan Nanc hang

O t h e r s

X i a n g C h a n g s h a

O t h e r s

5 . H a k k a ( K e j i a ) M e i x i a n

O t h e r s

S o u t h e r n Min Xiamen( Amoy)

O t h e r s

N o r t h e r n Min Fuzhou

O t h e r s

Yue G u a n g z h o u ( C a n t o n e s e )

1.2 Mandarin as Standard C hinese

The list of th e Chinese languages and d ia le c ts given in th e previous se c tio n rev ea ls one problem . T h at is, th e te rm Mandarin has been used sim u ltan e o u sly to mean th re e things: M andarin as one of th e C hinese languages, M andarin as th e S tandard D ialect, and M andarin as any individual d ia le c t (such as Beijing) o f th e M andarin language. A question thus arises h ere as to w hat I m ean by claim in g th a t th e p re se n t study focuses on M andarin. A lthough th e d e m a rc a tio n m ay n o t be te rrib ly necessary when a m orphological or a s y n ta c tic stu d y is co n ce rn ed , it is r a th e r im p o rtan t in a phonological study. E specially in te rm s of to n al c a te g o rie s , tonal values and tone-sandhi phenom ena. M andarin as a language can be q u ite d if­ f e re n t from o th e r C hinese languages. F or in sta n c e . M andarin has c a te g o ric a lly fo u r ton es, bu t C an to n ese, a Yue d ia le c t, has as m any as nine. M andarin to n e sa n ­ dhi p ro cesses have been observed to be b asica lly p h o n etic in n a tu re , w hile to n e sandhi p ro cesses in o th e r Chinese languages m ay r e la te to m orphology. Even th e d ia le c ts w ithin th e M andarin language can d iffe r g re a tly in th e ir to n a l sh ap es and ton e sandhi p ro cesses. A lthough to n e sandhi p ro cesses in th e lang uag e o f M anda­ rin m ay sh are th e com m on fe a tu re of all being p h o n etic in n a tu re , th e y n e v e rth e ­ less may be q u ite d iffe re n t from d ia le c t to d ia le c t.

In th is study , th e fo cu s is tru ly on M andarin as th e S tan d ard D ia le c t (MSD). Taken as th e sta n d a rd C hinese in both th e P eop le's R epublic of C hina as w ell as th e R epublic o f China (i.e. Taiw an), MSD has been o ffic ia lly d efin ed in th e P e o ­ ple's R epublic of C hina as a d ia le c t whose p ro n u n ciatio n is b ased on th a t o f th e Beijing d ia le c t (though n o t id e n tica l to it), and whose m orphology and s y n ta x a r e based on th e M andarin language as a whole. This sta n d a rd d ia le c t is th e d ia le c t used by th e variou s b ro ad ca stin g m edia, and is w idely in use ev ery w h ere in C hina. In th e non-M andarin a re a s, it is norm ally used as a "second lang uage". In th is d

is-s e rta tio n , I will uis-se th e te rm M andarin is-sp ecifically for MSD, unleis-sis-s o th erw iis-se n o t­ ed.

1.3 T h e M inim al D omain

As I m entioned e a rlie r, linguists have found th a t M andarin tone-sandhi p ro ­ c e sse s do n ot ju s t o ccu r w h erever th e ir s tru c tu ra l d escrip tion s a re m et; ra th e r, th e y o c c u r in som e d esig n ated fashion on a prosodic s tru c tu re bu ilt co m p letely or p a rtia lly on th e s u rfa c e s y n ta c tic s tru c tu re . One a re a of study on M andarin tones is th u s to d e te rm in e e x a c tly how this in te rm e d ia te s tru c tu re is built on a given s trin g of syllab les, so as to find o u t w here ex a c tly on th is strin g th e se to n e sandhi pro cesses a c tu a lly o ccu r. E arlier, 1 have also m entioned th a t one of th e sp e c ific purposes o f th is d is s e rta tio n is to explore th e n a tu re of M andarin to n e sandhi. Such a fo cu s o f in v e stig atio n m eans th a t I am in te re s te d in why M andarin to n e - sandhi p ro cesses happen r a th e r th an w here th ey happen. In o th e r words, th e p re s e n t in v e stig atio n is lim ited to th e m inim al dom ain o f M andarin to n e sandhi.

W hat th e n is th e m inim al dom ain of M andarin to n e sandhi? Can one c h a r a c te r ­ iz e it? And how? T hese q uestions a re easily an sw ered if th e strin g s under scru tin y co n ta in only (iso lated ) tw o -syllab le words or p hrases (or th re e -sy lla b le w ords o r ph rases in th e case o f th e seco n d -to n e sandhi to be discussed in § 6.2.4), sin ce th e s e strin g s co n sist of th e m inim al dom ains th e m selv es. How then can one d efin e th e m inim al dom ain on lo ng er strin g s? F o rtu n a te ly , this definition has been p ro ­ vided by stu d ies w ithin prosodic and se n te n c e phonology. In th e se stu d ie s (Shih

1986 m any o th e rs), th is m inim al dom ain fo r M andarin to n e sandhi has been d efined as a prosodic fo o t.

A ccording to Shih, th e prosodic fo o t is bu ilt by th e following th re e ru les (p.

( 1 . 2 ) . F o o t F o r m a t i o n R u l e s a . I m m e d i a t e C o n s t i t u e n c y : L i n k i m m e d i a t e c o n s t i t u e n t s i n t o d i s s y l l a b i c f e e t . b . D u p l e M e t e r : S c a n n i n g from l e f t t o r i g h t , s t r i n g t o g e t h e r u n p a i r e d s y l l a b l e s i n t o b i n a r y f e e t , u n l e s s t h e y b r a n c h t o t h e o p p o s i t e d i r e c t i o n . c . T r i p l e M e t e r : J o i n any l e f t o v e r m o n o s y l l a b l e t o a n e i g h b o r i n g b i n a r y f o o t a c c o r d i n g t o t h e d i r e c t i o n o f s y n t a c t i c b r a n c h i n g t o f or m a s u p e r f o o t .

A lthough co n tro v ersy s till ex ists in th e e x a c t d e ta ils of th e fo rm a tio n o f a prosod­ ic fo o t, th is co ntro versy need not co ncern us h ere. The p re s e n t stu d y will be lim ­ ite d to th e exclusive dom ain of a prosodic fo o t, re g a rd le ss of how ex a c tly such a fo o t is d efin ed , but assum ing sim ply th a t it is a lre ad y o p tim ally d efin e d in pro sod­ ic and se n te n c e phonology.

T h ere is one m ore c la rific a tio n th a t needs to be m ade h e re reg ard in g th e m in­ im al dom ain of a fo o t. As C heng (1973) has observed, th e dom ain o f M andarin to n e sandhi becom es la rg e r as th e tem po of speech in c re a se s (cf. § 7.1). T his m eans th a t given enough speed, a ll th e syllables (no m a tte r how many) in a m u lti­ sy lla b ic se n te n c e may be com pressed in to one single dom ain. In o rd e r to k eep th e p re s e n t stu d y w ithin its p ro p er lim its, I will consider p rim arily th e m inim al dom ain o f a fo o t under slow o r norm al sp eech r a te .

1.4

Data Tranxription

U nless oth erw ise n o ted , th e M andarin d a ta p re se n te d in th is d is s e rta tio n w ill be tra n s c rib e d in th e w ritin g system o f Hanyu pinyin (h e n c e fo rth pinyin), th e s ta n ­ d ard R om anized w riting system o ffic ia lly adopted in 1958 in th e P eo ple's R epublic o f C hina. F or th e la s t te n o r so y ears, this system has b eco m e q u ite pop ular am ong ling uists of M andarin. In f a c t, it has v irtu ally b eco m e th e only sy stem o f

tra n s c rip tio n in M andarin ling uistics. In this d isse rta tio n , n o t only will my own M andarin d a ta be p resen ted in this sy stem , M andarin d a ta c ite d from e a rlie r stu d ­ ies p re se n te d in o th e r system s (e.g. Chao, 1968) will also be p re s e n te d h ere co n­ v e rte d to pinyin - unless o th erw ise noted . The In tern atio n al P h o n etic A lphabet sym bols w ill also be used alongside of pinyin, w here th e ir 'e b eco m es n ecessary. Follow ing tra d itio n , th ese sym bols w ill, how ever, be included in slashes / / (or sq u are b ra c k e ts D).

A REVIEW OF PREVIOUS STUDIES ON TONE

2.1 In tro d u ctio n

One o f th e m ost im p o rtan t th e o re tic a l adv ances m ade in r e c e n t g e n e ra tiv e phonology lies in its s h ift of focus from rules to re p re s e n ta tio n . The idea is th a t only if th e re p re s e n ta tio n is c o rre c t can op tim al ru les be fo rm alize d ; th e re v e rs e is, how ever, not tru e . The lite r a tu r e review con du cted in th is c h a p te r fo cu ses acco rd in gly on th e phonological re p re s e n ta tio n of to n es, e sp e c ia lly , of C hinese ton es.

A num ber o f re le v a n t issues have been explored in th e lite r a tu r e on to n a l r e p ­ re s e n ta tio n . Among them is, f ir s t of all, w hat c o n s titu te s a to n a l fe a tu r e . S p ecif­ ically, th e q uestion concerns: should th e u niversal s e t of to n a l fe a tu r e s c o n ta in u n itary co n to u r fe a tu re s such as [rising] and [falling], or should it c o n ta in only re g ­ is te r (or level) fe a tu re s such as [high] and [low]? In th e c u rre n t n o n -lin ear phonol­ ogy, th is is a q uestion of w h eth er th e co n to u r to n es should be re g a rd e d as indivisi­ ble un its d escrib a b le in te rm s o f u n itary co n to u r fe a tu re s , o r be seen as being com posed of a m ere sequence of lev el to n em es d escrib ab le only in te rm s o f re g is ­ te r fe a tu r e s . These tw o positions a re illu s tra te d below using th e M andarin risin g to n e as an exam ple. ( 2 . 1 ) . Example mà Representation a . LH b . [ + r i s i n g ] 12

-T rad itio n ally , th e (b) typ e of approach was widely ado pted (e.g. Pike 1948, EH- m eleeh 1974, Schuh 1978), p artic u la rly w ith reg ard s to C hinese tones (Wang 1967). H ow ever, th is approach cam e under serious a tta c k in 1969 by Woo, who proposed th e (a) typ e o f ap pro ach in stead , basing h er proposal on ev idence from C h inese. The c e n tr a l claim of th e (b) tr e a tm e n t was fu rth e r challenged by G oldsm ith in his A u to se g m e n ta l Phonology (1979),^ w ith evidence from A frican tones. Yip (1980a), follow ing G oldsm ith, has developed th e fir s t n on-linear re p re s e n ta tio n of C hinese to n e s, and ju s t as in G oldsm ith, her re p re se n ta tio n tr e a ts th e co n to u r to n es as being com posed of lev el to n em es.

In r e c e n t y e a rs, how ever, a th ird approach has been proposed by Yip (1989a,b) in an a tte m p t to solve th e p ro b lem atic cases th a t can n o t be reaso nably handled by a le v el-p rim itiv e-o n ly analysis. The th ird ap proach is, in a sense, a com prom ise: it is less stro n g th a n e ith e r of th e tw o fo rm er positions, w ith th e basic claim th a t alth ou gh a t one le v el, co n to u r to n es a re b e s t acc o u n ted for in te rm s of seq u en ce s o f lev el to n e s, th e y may be describ ab le a t an o th er level only if tr e a te d as p rim i­ tiv e u n its.

B esides th e issue of th e co n to u r to ne tr e a tm e n t, an o th e r im p o rta n t focus o f s tu d ie s in th e a re a of to n a l re p re s e n ta tio n concerns th e question of how to n a l f e a ­ tu r e s can b e s t be co nfig ured in th e phonological re p re s e n ta tio n . This is th e q ues­ tio n w hich e v en tu ally led to G oldsm ith's (1979) rev o lu tio n ary m o dificatio n of th e g e n e ra tiv e ap p ro ach to phonological fe a tu re re p re s e n ta tio n . The m ajor q u estio n add ressed in G oldsm ith is w h eth er th e to n al fe a tu re s in som e languages (or r a th e r c e r ta in f e a tu r e s , n o t n ece ssarily to n al, in som e languages) should be g ra n te d autonom ous s ta tu s and re p re s e n te d s e p a ra te ly from th e r e s t of th e f e a tu r e s . T he raisin g of th is q uestio n is in its e lf q u ite unconventional, as it im plies th a t f e a tu r e s a f te r all m ay n o t n ecessarily alw ays be bundled to g e th e r w ithin th e sam e seq u en ce of f e a tu r e m a tric e s.

This qu estio n, how ever, has been considered reso lved a f te r G oldsm ith's insightful work and many sub sequent stud ies w ithin th e fram ew o rk developed by him . T hese stu d ies have argued co g en tly and re p e a te d ly w ith ev id en ce from v a ri­ ous languages th a t tonal fe a tu re s (as well as som e o th e r n o n -to n al fe a tu re s ) o fte n need to be re p re se n te d se p a ra te ly from th e re s t of th e phonological fe a tu re s . A lthough th e d eb ate over th e possible autonom ous s ta tu s of th e to n a l fe a tu re s in th e re p re s e n ta tio n is largely s e ttle d , th e p o ssibilities brought ab o u t by such a posi­ tion lead to new questions to be an sw ered. One such question is; if a to n e is a u to - s e g m e n ta l in a language, w hat th e n is th e ton e b earing u n it (TBU) in th a t la n ­ guag e? Is it, say, th e vowel or th e syllable? A gain, le t m e ta k e th e M andarin seco n d -to n e d m orphem e m â "hem p" fo r an exam ple to illu s tr a te th e d iffe re n c e (The arro w s o f th e shape <- a re used sim ply to highlight th e f e a tu r e s re le v a n t in th e p re s e n t discussion.): C o n t o u r T r e a t m e n t ( i ) . (2.2) . a . seg m en tal tr e a tm e n t L e v e l T r e a t m e n t ( i i ) . _ m _ _ a _ _ m _ _ a _ |+ c o n s | 1-c o n s 1 1+ c o n s 1 |- c o n s 1 l+son 1 1- h i 1 i +son j j - h i 1 l+nas 1 1+ lo 1 l+nas j l+ lo j j+ la b 1 1+bk 1 j+ la b j j+bk j 1 -c o n tl 1 j - c o n t 1 1 : 1 + r i s e | < - j+ hi j <— f a l l l < -_ : -_ i | —lo 1 <— a u to se g m e n ta l tr e a tm e n t : vowel a s TBU _ m _ _ a _ _ m _ _ a __ 1+ c o n s 1 -c o n s 1 1 + c o n s1 |- c o n s ! m e lo d ic l+son ! - h i j j +son j j - h i j l+nas 1 + lo 1 j +nas 1 j+ lo i j+ la b I +bk 1 j+ la b j 1+bk 1 1- c e n t I 1 . 1 : 11 j- c o n t j 1 1 1 : 1 I I - : _ 1 _ l_ : - I I— • / \ + r i s e 1 - f a l l l |+ h i~ 1 | - h i 1 to n a l _ • _ 1

_{L _1}

_L

a u t o s e g m e n t a l t r e a t m e n t : s y l l a b l e as T B U _ m _ _ a _ m _ a l + c o n s l | - co n s| l + c o n s l 1- c o n s 1 l + s o n 1 j-hi ! l + s o n 1 1- hi 1 l + n a s 1 l + l o 1 l + n a s 1 1+ lo 1 1 + l a b 1 j+bk 1 1 + l a b 1 1+ bk 1 j - c o n t l 1 : I 1 . 1 1 1 j - c o n t l 1 1 - 1 1 : 1 1 _ • 1 1 _ • — ' \ / \ 0 a _ 1 _ _ /.. _ \ _ 1 + r i s e 1 1+hi 1 - h i 1 1- f a l l l _{L _ l} _ _! m e l u d i c 11 PI t o n a l t i e r

As illu s tra te d above, th e re a re in th e o ry a t le a st fou r possible tre a tm e n ts w ithin th e n o n -lin ear approach (b and c). If one re je c ts the tr e a tm e n t of co n to u r to n e s as co n sistin g of level to nes only, th e re still rem ain tw o p o ssib ilities (i.e. bii and cii).

O th e r issues concern ing to n a l re p re s e n ta tio n include how m any lev els should be d istin g u ish ed in th e to n a l sy stem of a given language or languages. It seem s th a t in m o st o f th e A frican and N o rth A m erica to n al languages, tw o is th e answ er. F or C h in ese-lik e o rie n ta l languages, how ever, sev eral th e o ries have been pro­ posed, c o v erin g a wide ran g e fro m th r e e lev els in H alle and S tevens (1971), fo ur le v e ls in Yip (1980a) and G ruber (1964), to five levels in Sampson (1969), M addie- son (1972),^ Wang (1967), and Woo (1972).^ F or a given individual language, th e hy p o th esis can be q u ite diverse as w ell. For in sta n ce. M andarin to n e s have been a n aly zed as c o n tra stin g th re e levels (Woo 1972, Zee 1991), fou r lev els (Yip 1980a), and fiv e le v e ls (Chao 1968).

In th e r e s t of th is c h a p te r, 1 will b riefly review th e m ajor stu d ie s on tones, p a rtic u la rly w ith re s p e c t to th e ir positions in th e issues m en tion ed above, it should be n o te d , how ever, th a t all stu d ie s do not b ear on all th e issues, which m eans th a t, in th e course of th e review , 1 will focus on one o r a n o th e r of the issues w here it becom es p e rtin e n t fo r discussion.

2.2 P ike (1948)

One m ajor c h a ra c te ris tic of Pike's study of tones th a t is of sp ecial co ncern h e re is th a t P il'e is probably th e e a rlie s t proponent of th e tr e a tm e n t o f co n to u r to n e s as un its. In his now classic work on tone, Pike c le a rly distin g u ish es tw o ty p e s o f to n al languages, le v el-to n e languages such as M ixteco (a language of so u th ern Mexico), and co n to u r-to n e languages such as C hinese. Pike arg u es th a t w hile th e co n to u r ton es in le v e l-to n e languages may be seq u en ce s of level to n e m es, in th e description of th e co n to u r to nes of th e c o n to u r-to n e lan gu ages, " th e basic to n em ic u nit is gliding in ste ad of level (p. 8)." Pike gives f u r th e r com ­ p ariso n b etw een a co n to u r-to n e language and a re g is te r-to n e lan g u ag e, and o b serv es th a t in a co n to u r-to n e language (Pike 1948, p.8),

1. The u n i t a r y c o n to u r g l i d e s c a n n o t be i n t e r r u p t e d by morpheme b o u n d a rie s a s can th e nonphonem ic compounded ty p e s o f a r e g i s t e r sy ste m .

2 . The b e g in n in g and ending p o i n t s o f th e g l i d e s o f a c o n to u r s y s ­ tem c a n n o t be e q u a te d w ith l e v e l tonemes in th e same s y ste m , w hereas a l l g l i d e s o f a r e g i s t e r system a r e to be i n t e r p r e t e d p h o n e m ic a lly in term s o f t h e i r end p o i n t s .

3. In th e p r in te d m a t e r i a l exam ined, c o n to u r sy stem s had o n ly one toneme p e r s y l l a b l e , w hereas some of th e r e g i s t e r - t o n e la n g u a g ­ e s ...m a y have two or more tonem es p e r s y l l a b l e .

B ased on th e se o b serv atio ns, Pike concludes th a t

" in a p u re c o n to u r sy ste m , th e g l i d e s a r e p h o n o lo g ic a lly u n i t a r y , m o rp h o lo g ic a lly s im p le , and n o t s t r u c t u r a l l y r e l a t ­ ed to a sy stem o f l e v e l tonem es; th e g l i d e s a r e minimum s t r u c t u r a l u n i t s o f le n g th in words and s y l l a b l e s . " ( p .8 )

One f e a tu r e in Pike th a t is w orthy o f re m a rk h e re is his use o f th e te rm to n e m e , w hich is used to re fe r to a m inim al com ponent of a sy llab le to n e . His uni­ ta ry tr e a tm e n t of co n to u r to nes is also re f le c te d in his use o f th is te rm ; unlike

th e m inim al to n al com ponent defined to be alw ays level in c u rre n t g e n e ra tiv e pho­ nology, his to n em e may be an ato m ic co n to u r in itse lf. Thus, acc o rd in g to him , a to n e of a sy llab le typ ically co n tain s one level ton em e in a r e g is te r-to n e language, b u t a single co n to u r tonem e in a c o n to u r-to n e language. H ow ever, ju s t as a tone in a r e g is te r-to n e language may occasionally co ntain two level to n e m es, re su ltin g in a sim ple glide, so may a co n to u r tone in a c o n to u r-to n e language co n tain tw o co n to u r to n em es, resu ltin g in a com plex co ntou r tone (i.e. a con cav e or convex tone).

F o r som e reaso n or o th e r. Pike's te rm to n e m e as a m inim al com p on ent of a to n e has n o t been ado pted am ong la te r to no lo gists, although it c le a rly allow s for a u seful (and necessary) d istin ctio n to be m ade m ore ex p licit b etw ee n th e to n e a t th e sy llab le lev el and th a t a t th e sub-syllable level. N orm ally, tono lo gists sim ply use to n e fo r u nits of both lev els. The re lu c ta n c e to use th e te rm "ton em e" is probably due to th e g r e a t co m p lex ity in th e possible in tern al s tr u c tu r e o f w hat m ay be c a lle d a to n e in th e v ariou s to n a l sy stem s in hum an languages.

2.3

Chao (1930^ 1968)

It is p robably s a fe to say th a t Yuen Ren Chao is th e f o r e fa th e r of co n tem p o ­ ra ry lin g u istic stu d ies of C hinese. One p a rtic u la rly ou tstan d in g co n trib u tio n he has

m ade to th e s e stu d ies is th e sc a le of five p itc h levels he developed fo r to n a l r e p ­ re s e n ta tio n . An in te re s tin g phenom enon one will no t fail to n o tice in read in g lite r a tu re on C hinese to n e s is th a t ph eno logists o f C hinese may each arg u e fo r a q u ite d iffe re n t th e o ry of Chinese to n al analysis, bu t when it com es to th e in tro d u ctio n of to n a l d a ta a t th e o u ts e t of th e ir stu dy , th ey alm o st invariably choose Chao's fiv e -le v e l sch em e fo r th e ta sk , no m a tte r how much m ore adv an ced and insightful th e ir th e o rie s may be. In w hat follow s, I will illu s tra te Chao's sch em e using th e M andarin to n e s as an exam ple.

( 2 . 3 ) . To ne 1 Tone 2 Tone 3 Tone 4 • j 4 3 -2 -1 -/ \ / \ / / \ \ / \

\ /

\

Chao's schem e, as is illu s tra te d above, divides th e hum an vocal ran g e in to fiv e lev els, w ith th e f irs t and th e f if th m arking re sp e c tiv e ly th e lo w est and th e h ig h e st p itc h e s a sp e a k e r re a c h e s in norm al sp eech prod uctio n. H ere th e te rm "norm al" is c i t ' c a l since th e schem e does not deal w ith unusual p itc h lev els th a t, fo r in sta n c e , a re raised higher th an th e f ifth in singing or due to a f f e c tiv e f a c to r s such as an g er o r joy.

On C hao's sc a le , th e f irs t M andarin to n e (being th e f ir s t o r any o th e r n um b er m erely by convention) s ta r ts a t th e top of th e sp eak er's p itc h lev el (m ark ed as le v ­ el 5) and rem a in s so fo r a c e rta in period of tim e . The seco nd s t a r t s in th e m iddle (level 3) o f th e vocal range and rises sharply to th e top. The th ird is a m ore c o m ­ plex one. It f ir s t drops from le v el 2 to th e lo w est p itc h le v e l (level 1) and th e n r is ­ es again to th e middle point. Based on th e read in gs on th is s c a le , Chao r e f e r s to th e fo u r M andarin to n e s by th e num ber com po sites such as (55) fo r th e f ir s t to n e , (35) fo r th e second, (213) fo r th e th ird and (51) fo r th e fo u rth .

C hao's sc a le read ing s also h ave had im p a c t on th e kind o f d ia c ritic s t h a t hav e been ad o p ted to m ark M andarin to n es. F or in sta n c e , th e fo u r d ia c ritic s used in th e R om anized C hinese w riting system pinyin a re n o t to ta lly a r b itr a r y sym bols. A com [,arison b etw ee n th e se d ia c ritic s (shown below in colum n (a)) and C hao's sch em e re v e a ls th a t eac h d ia c ritic has th e sam e co n to u r sh ap e as th e sh ap e o f th e to n e re p re s e n te d on Chao's scale;

( 2 . 4 ) . ( a ) ( b ) ( c ) p i n y i n number g l o s s c o m p o s i t e s mâ ( 5 5 ) "mother" mâ ( 3 5 ) "hemp" mâ ( 2 1 3 ) " h o r s e " mà ( 5 1 ) " s c o l d "

One unique c h a r a c te r is tic of Chao's system is th a t it is a s c a la r ra th e r th an a b in a ry -fe a tu re sy stem . This c h a ra c te ris tic distinguishes C hao's from all la te r an aly ses of C hinese to nes. A nother c h a ra c te ristic o f C hao's th a t is w orth m en­ tio n in g is th a t it is b asically an analysis th a t recog nizes co n to u r ton es as u nits. Of c o u rse , w hen C hao's th e o ry was f ir s t developed (1930), th e re was not y e t an aw aren ess of a d istin ctio n betw een th e tw o app ro aches to th e re p re s e n ta tio n of co n to u r to n e s. In f a c t, no discussion o f th e issue was m ade in his work 36 y ears l a t e r (1968). N ev erth eless, C hao's th e o re tic a l o rie n ta tio n w ith re g a rd to th is issue can b e d escrib e d w ith f a ir am o u n t of c e rta in ty . As Bao (1990) has rem ark ed ,

" i n C h a o ' s n u m e r i c a l r e p r e s e n t a t i o n , t h e t o n e s a r e c o n c e i v e d t o b e s i n g l e , a t o m i c e n t i t i e s . 53 d o e s n o t i m p l y t h a t t h e h i g h f a l l i n g t o n e i s c o m p o s e d o f t h e h e i g h t p o i n t 5 f o l l o w e d b y t h e mid p o i n t 3 . 53 i s a u n i t a r y h i g h f a l l i n g t o n e , . . . a n d 3 1 , a u n i t a r y l o w f a l l i n g t o n e . " ( p . 2 1 )

2.4

Wang (1967)

If C hao's th e o ry is " p re -th e o re tic a l" , Wang is th e f ir s t to develop a to n al r e p ­ r e s e n ta tio n fo r th e C hinese to n es, couched in th e th en re la tiv e ly new model o f g e n e ra tiv e lin g u istics. In f a c t , Wang's study re p re s e n ts th e n ex t m ajor work on C hinese to n e s a f te r Chao, and it pioneers g e n e ra tiv e re s e a rc h on th e then linguis­ tic a lly l i ttl e known phenom enon called tone.

A d e p a rtu re Wang makes from C hao is th a t he is an e x p lic it p ro p o n en t of a u n ita ry an alysis of contour to n es. Within th e g en era tiv e school, Wang devises th e f ir s t s e t o f binary fe a tu re s sp e c ific a lly fo r ton al re p re s e n ta tio n . This s e t of f e a ­ tu re s co n tain s u n itary contour f e a tu r e s as well as re g is te r f e a tu r e s . In all, Wang develops seven binary tonal fe a tu re s , fo u r co n to u r fe a tu re s : [rising], [falling], [contour], and [convex], which a re to re p re s e n t th e contours o f v ario u s to n e s, and th r e e r e g is te r fe a tu re s : [high], [cen tral] and [mid], which a re to d e fin e fiv e levels o f to n e s. The fiv e levels of tones a re defined as follows:

( 2 . 5 ) . [ h i g h ] [ c e n t r a l ] [ m i d ] ( 5 5 ) + - -(4 4) + + — ( 3 3 ) _ + + (2 2 ) + ( I I ) - -

-In Wang’s f e a tu r e system , th e fo u r to n e s of M andarin can be r e p re s e n te d as in (2.6). It should be borne in mind in read in g (2.6) th a t Wang's sy ste m is in ten d ed not ju s t fo r M andarin, and th e re fo re , som e of th e fe a tu re s in (2.6) m ay n o t a p p e a r as

m o tiv a te d as th e y a ctu ally a re:

( 2 . 6 ) . T o n e 1 [ - r i s i n g , - f a l l i n g , - c o n v e x , - c o n t o u r , ( 5 5 ) + h i g h , - c e n t r a l , - m i d ] T o n e 2 [ + r i s i n g , - f a l l i n g , - c o n v e x , + c o n t o u r , ( 3 5 ) + h i g h , - c e n t r a l , - m i d ] T o n e 3 [ + r i s i n g , + f a l l i n g , - c o n v e x , + c o n t o u r , ( 2 1 3 ) - h i g h , - c e n t r a l , - m i d ] T o n e 4 [ ^ f a l l i n g , - r i s i n g , - c o n v e x , + c o n t o u r , ( 5 1 ) + h i g h , - c e n t r a l , - m i d ]

While th e m o tiv atio n behind th e use of th e fe a tu re s [rising], [falling], [contour] and [high] a p p e a rs q u ite stra ig h t-fo rw a rd , his po stu latio n o f th e r e s t o f th e f e a ­ tu re s n eed s a l i ttl e exp lanation. It seem s th a t in ord er to re p r e s e n t a com plex c o n to u r to n e (i.e. concave, convex, e tc .), w hich co n tain s b o th risin g and fallin g co m p o n en ts, Wang uses th e f e a tu r e [convex] to ensure th e d e sire d o rd e r o f th e ris ­

ing and fa llin g com ponents. Wang's use o f th e fe a tu re s [cen tral] and [mid] seem s a t f ir s t sig h t red u n d an t. Their c o n tra s t, how ever, is to distinguish th e to n e s 22, 33 and 44. A to n e w ith a value of 22, fo r in sta n ce, is [+ central, -m id , +high]; one w ith a value o f 33 is [+central, +mid, -high]; and one w ith a value of 44 is [+cen- tr a l, -m id , -high].

It seem s th a t Wang was n o t only th e f ir s t, b ut also th e la st to have developed and used co n to u r fe a tu re s ex p licitly and in a r a th e r sy ste m a tic way fo r to n al a n a l­ ysis. A lthough n o t long a f te r his w ork, his approach of tre a tin g co n to u r to n es as un its g av e w ay to a new approach which considers con to ur tones as sp ecia l form s o f lev el to n e s, no n eth eless, th e p la u sib ility of th e fo rm e r app roach has y e t to be judged. R e c e n t stu d ies by Yip (1989a,b), fo r in sta n ce, have found th a t a t le a s t a t som e le v el in som e Chinese languages, co n to u r tones should be tr e a te d as units. S im ilar findings a re also made in th e p re se n t study (cf. § 4.2.4.).

2.5

Woo (1972)

The s tu d ie s I have review ed so f a r all sh are m ore or less th e com m on fe a tu re o f re p re s e n tin g co n to u r tones in a u n ita ry w ay. This assum ption is f ir s t ch allen ged by Woo (1972). Woo's fe a tu re sy stem fo r to n a l analysis assum es a to ta l o f th re e fe a tu re s , [high], [modify] and [low], all being re g is te r fe a tu re s . T hese fe a tu re s com bine to y ie ld fiv e level to n e s, ju s t as Wang's level fe a tu re s do. T hese fiv e lev ­ els a r e shown below (Wco, p.71):

( 2 . 7 ) . [ h i g h ] [ m o d i f y ) [ l o w ] ( 5 5 ) + - -(4 4 ) + + -( 3 3 ) - - -(2 2) _ + + (1 1) - - +

A lthough Woo's fe a tu re s are d efin e d in a way q u ite sim ilar to th e r e g is te r f e a ­ tu re s in Wang, h e r system d iffe rs d ra s tic a lly from Wang's in th a t it dispenses to ta lly w ith c o n to u r fe a tu re s such as [rising] and [falling]. A ccording to Woo,

"a s y s t e m u s i n g t h e f e a t u r e s [ r i s e ] and [ f a l l ] i s n e i t h e r m e c h a n i c a l l y n o r t h e o r e t i c a l l y a d e q u a t e t o d e s c r i b e t h e

t o n e s o f a l l l a n g u a g e s . We t h u s a b a n d o n t h i s a p p r o a c h and t u r n i n s t e a d t o a s y s t e m w h i c h d e s c r i b e s c o n t o u r t o n e s a s s e q u e n c e s o f p i t c h h e i g h t s . " ( p . 6 4 )

Woo goes a g a in st th e tra d itio n of her tim e w ith th e claim th a t in d eep s tr u c ­ tu r e , co n to u r to n es a re m ade up of level com po nen ts. . andarin to n e s, in p a r tic u ­ la r, becom e re p re s e n te d fo r th e f ir s t tim e in th e sam e way as som e o f th e A frican co n to u r to n es a re re p re s e n te d , th a t is, in te rm s o f a sequ ence o f le v e l to n e m e s. In p a rtic u la r, th e se M andarin to nes co n tain tw o to n e m es each (Woo, p.75);

( 2 . 8 ) . T o n e l T o n e 2 T o n e 3 T o n e 4

( 5 5 ) ( 3 5 ) ( 2 1 3 ) ( 5 1 )

[ h i g h ] + + — + - - +

-[ l o w ] — — — — + + — +

[ m o d i f y ] - - - - - -

-As shown above, th e fe a tu re [modify], w ith its n e g a tiv e value found fo r all th e to n em es, is n o t d is tin c tiv e fo r M andarin. This m eans th a t M andarin to n e s can be sp ecified using ju s t th e o th e r tw o fe a tu re s . Thus, th e above can also be s ta t e d in th e follow ing fashion (Woo, pp. 73-74),

( 2 . 9 ) . T o n e l T o n e 2 T o n e 3 To ne 4

[ + h i ] [ + h i ] I - h i I [ + h i ] [ + l o ] [ + l o ] [ + h i ] [ + l o ] I- l o i

o r using th e p opular ab b rev iated sym bols H, M and L;

( 2 . 1 0 ) . T o n e l T o n e 2 T o n e 3 To ne 4

HH MH LL HL

Many stu d ies a f te r Woo have d e m o n stra te d th a t th e r e p re s e n ta tio n o f c o n to u r to n e s as being com posed o f level to n em es is ab so lu tely n e c e ssa ry in ph onological

d escrip tio n , con firm in g th a t Woo's tonem ic tr e a tm e n t of co n to u r to n e s re p re s e n ts an im p o rta n t ad v an ce in th e phonological study of to n e s. H ow ever, Woo's th e o ry also has p roblem s, th e m ost serio us o f which re su lts from th e th e o re tic a l fr a m e ­ work h er analysis is couched in: it seem s h er le v el-to n e m e tr e a tm e n t o f co n to u r to n e s is in c o n flic t w ith th e th e o re tic a l fram ew o rk . Within th e fram ew o rk of s ta n d a rd g e n e ra tiv e phonology. Woo finds it im possible to re p re s e n t in th e sam e f e a tu r e m a trix th e sequ en ce of tw o lev el ton em es she proposes fo r th e C hinese to n e s. L et me illu s tra te th e problem w ith th e follow ing tw o s e ts o f fe a tu re m a tr i­ ces: (2.1 1) . ( a )

.

( b ) .

_ m

m

- h i

:

.

_ • — I L : - I

I

Illu s tra te d in th e above a re re p re s e n ta tio n s of th e M andarin secon d (35) to n e . In (a), a u n ita ry c o n to u r ap pro ach is used. U nder th is ap p ro ach , th e rising lo n e is r e p re s e n te d by p lacin g u n ita ry fe a tu re s such as [rising] ([rise] fo r sh o rt) in th e f e a ­ tu r e m a trix o f th e vow el. In c o n tra s t to (a), (b) re p re s e n ts th e sam e risin g to n e by using le v e l to n e m e s such as L and H. Both (a) and (b) a re couched in th e tra d itio n a l se g m e n ta l ap p ro ac h , w hich tr e a ts to nal f e a tu r e s as p ro p e rtie s of th e vow el (or th e to n e -b e a rin g so n oran t).

While th e use o f u n ita ry co n to u r fe a tu re s in (a) fits into th e tra d itio n a l s e g ­ m e n tal ap p ro ach , a problem arises in th e h y p o th e tica l (b) in w hich tw o f e a tu r e m a tric e s (a b b re v ia te d as M and H) in a d esig n ated o rd e r a re found. H ow ever, an im p o rta n t te n e t o f th e sta n d a rd model she follow s is th a t f e a tu re s in th e sam e m a trix a re u n -o rd ered . This m eans th a t she sim ply c a n n o t ensure th a t th e M f e a ­ tu re s te m p o ra lly p re c e d e th e H fe a tu re s .

Woo's solution to th e ord ering problem is to "enlarge" th e to n e -b e a rin g se g ­ m en t, in p a rtic u la r, to assum e th a t th e re are no sin g le-so n o ran t sy llables in th e lang uage. All of the open syllables tra d itio n a lly reg a rd e d as being com posed o f one single vow el co n tain tw o vowels in Woo, so th a t ma becom es m aa, di b ecom es dii, and so fo rth . With the addition of one vowel, th e sequence of le v el to n em es (tw o m axim um in Woo for underlying re p re s e n ta tio n of C hinese to n es) can th en be evenly d istrib u te d am ong th em , and th e case in (2.11b) can th e n be re p re s e n te d as follow s, w here th e aw kw ard situ a tio n seen th e re d isappears:

(2.1 2) . _ m _ _ a _ a _ + c o n s 1 1- c o n s 1- c o n s + s o n 1 i - h i 1 - h i + n n s 1 l + l o l + l o + l a b 1 l+bk 1 +bk - c o n t 1 1 : 1 : 1 M 1 H <

-In such an analysis. Woo a p p a re n tly assum es th e assum ption of sta n d a rd g e n e r­ a tiv e phonology th a t to n al fe a tu re s a re b u t fe a tu re s of th e vow els (or th e to n e b earin g son oran ts). In o th e r w ords, to n e m es n ece ssarily b e a r a s tr ic tly o n e -to -o n e re la tio n sh ip to vowels. This assum ption, how ever, has been q u ite convincingly r e je c te d by G oldsm ith's (1979) a u to se g m e n ta l phonology and Yip's (1980a) a u to se g ­

m e n tal a c c o u n t of C hinese to n a l re p re s e n ta tio n . In f a c t , it is due to th is assu m p ­ tio n o f th e sta n d a rd g e n e ra tiv e phonology, r a th e r th a n any th in g e lse , th a t Woo's n um ber of to n em es has to be lim ite d to no m ore th a n tw o fo r an y underlying C hi­ nese to n e . The sig nificance of th is tw o -to n em e lim ita tio n lies in th a t no convex o r co n cav e to n e s are reco gn ized in d eep s tru c tu re . L et us ta k e th e th ird M andarin to n e (213) as an exam ple. This to n e has th re e fo rm s, (213) in fin a l positio n o r in iso la tio n , (21) when follow ed by any o th e r to ne th an its e lf, and (35) when fo llow ed by its e lf . Woo's analysis fo r th is to n e is d iffe re n t fro m e a r lie r stu d ie s n o t only b ecau se she re p re s e n ts it w ith le v el to nem es b ut also b ecau se she assum es th e