”053$?" .

‘ I A) H?)

o
:‘q" ‘J

‘8’;VA-.o

A

v V '7-’ W}. “‘0 -"‘I:4J]'-‘
' .

   
  
   

 
  
 

 

       

    
  
   

 
    
 

    

  

 
  

 
  
  
  
  
  

 
   
  
 

    

 
  

a},
.1 . ‘
3::
.. ' E:
. ~ ‘3“ . 2:3?” ‘h
(is? w“ 73' firm. 1- ”3!. "' ‘ . 5 g“
" . ‘ Eu ‘5' . .4. ~. - q- 0' 5 5 .c {a -. 3"- ‘
1 J'I'gvgr O,ogu;i‘;r , ‘fi. L n“. .0; ‘ . .' :‘{ ..‘l‘. '5‘“_-. 9...)? 0"”. «‘41....‘lwé‘!‘%:;“y=;.“;.- $3353: 1‘“
”02;: 7"_'r1‘):':_ 4 I t .-‘z 5“ -' :31). :‘u'C‘J-a. - .\:;'iQ~.h:) —.'B—4 ;‘§:.‘. .k‘1321 ; ‘5‘ ' V I a‘ - a. n
J“. 79‘. _- n a - ' | ~ I
r-' . .f'. ’o‘vuynfiu r-ri' ...} 3.3 7"‘3‘ '33.,»5“ ‘ ’i‘. ...-V ‘3. “’5"
, h": O‘r"~§vm‘ .‘ " ‘hyg‘
. J "Fit. ‘0- ‘ “:N'a. ’u‘t-‘f p01-
. . - ' . ~ V v._7 - ' . 4 1A
‘ , ‘ . .- I “5‘ " \
#‘VW . - ' ‘... . . kt.» 1“}. o 6"
wfix‘ijh 71 I); ‘9!!!“ , 9* who -° " .-x.:'§;; ‘3‘". 3“. I. 4. ’-
—,W . , r -. < . - w»- ; '.
:5“ ' ‘ 3‘- ., ~ "g t‘ ”331.32%”. I“ I ~ \
K .. . #5” so 5 , , _ . ' A; .
- rm!" 'n‘-"a- \45n’h l -
‘Hvt>*.la:r:v.q""l m" Vl.r‘ “‘17..
3:“ . 7" .1' a ' .—
-‘l- o 1
{ Nil-A“ §

    

   

.w‘s'vaJ- ,.. .T u
“R: V

mud? {$3372

 

«In
LV_’{%¢?\£ (‘73::

 
    

 

 

a .
3, “:9

 

  

  
   
 

 

. " ”a.“ f
:2“ ?. .aéauv .fi'fl‘. h

P. o v’ . I I‘. 4 ‘ "n’.30.‘::_" . 2£; 3“.
«$1.3 3‘ “' £3 3“ '41-”;

  
    
 
 
    
    
   

 
    
 
 

   
       

' V‘
""‘.‘-§ Aboay'p‘Q‘, , A. '
hkkififfl. 52.1%" ., ' , _ :
mf:.¢:sz'-...u.-"3.,- xamaaa ‘t‘ .
“‘RV 1-\ 5" ' ‘ ' ' ’

 
 
  
   

  
   
  
  
 

 

a'- 3
s“ I

fight: rm
. '9“ * $1.”! ¥4kl00313
49.31”;- {53"3-‘ ' .341" 3'31““. h“ «V «.7 - "an

‘1:ng mg; gag;
- (1.“. ‘1 ~48"? Vans”. ‘4“ -I aid-MN
'0- {£91 Y '- 5 ’ 1 . o4. . ‘zf , GE?“ ‘. what- ‘o ‘1‘ “;;,':Ar\'59.‘0:§.brh “I
PM" wgm 'k ‘ .. .. r ! :5 " ‘ -- ' ' ',,-- .a-wfkia-uE mfiww‘fisr—v "-*=‘-'°%<“a~¢r
g: “3,3‘ : ‘f “v." . . .‘u, , ' - ‘ f. ‘ ‘ ”’ h . .. W
32‘55' é} at“): .31; ‘5 "g {313:2 4:."1' :‘u ‘ ;w “‘2?

   
    
 

   

  
 
 
  
 

 
 

 

 

 
 

 

ward? mt”: «gh‘r‘kl:
’5’“ $V;L.‘

‘3)" a“ £2 .fihrr
”3’44 ’1;st “pi? £4;
wv ‘1..qu -:?'O.F§o’ any" ’3'"

 
 
 
 
 
  
 

 
 
 
    

 

 

 
 

 

 

 

 

J::o:io;“: .'
p" - r) 4:3 7:3:
{at} kt‘.‘ . . . 9.“ ..

was! a. g" ‘ v.3" ..

 

 
     

 

 

 

 

 

 

   

   
  
 

 

 

,gu' . - '\ w
2;;‘, $34” '. u“' i. .i::‘;; d
, , ' .’ ' V .

.tr‘ . ’ s -t.~?‘~fl’ra}1’2.""‘““
L‘shp 1 -,~ “Jr-...!‘m' v3 . . .

"““.‘w‘4 ‘ ‘ ...? 43‘” Mcnmt’v-r

" 1-). :3 ‘ \Vvfil;~“?';:.if? 3511mm...
chi-"IN, .. . ,5
dwvga can. '1 -.‘. I?" ;-: . .- ‘y’wngg £303“

 
 

«1 fig.“ ~lo§3§x

_ ‘ A: ; ,.‘ Lit?"
_ ' ‘3‘?! 0‘ 5. '_ v Q n’fiv 30' M F:.
' $5.: smfl’sxwé: M ’3 war

k m “ y I” L- ‘

  

 

.
" (0- ~
3' , p . ‘05! .
Emu: ’I'Dx'}. 1": {firm ' "(71¢

‘
Q’ngw'ga t ......"ia' " ‘55 v.“
a». an m. ‘4'?_
"J ‘ '

  
   

 
 
 

 
 
  
      

 

 
 
   
  
 

 
  
  

   
   
 

  
    
   

 
  
      

 
   
  
 

.5 .; :CL._ - - _A . ~
7' ”tr' .{v (q
IVV‘b'YO
’1';- ‘ 3:73-5:51 am 9‘}; ,_ '
3"” 22' 331+." - .~
- L' .‘o [rug-... .
5 . “... ‘ ‘n. - I
~
9 I
v3 1* 3:1
> ' O
' , : ' ‘ nu! o a . C2 . ‘ , .
. ‘ 91065.3? . 'f'”'¢£¢"- “:3. {‘3‘ «a iz.?‘$wj J52 6‘ . 3.“; ‘I'k.:' ’ .‘l' ‘ L' t 3’
th: flux-4U. -.J- ’ vi cw- .. 5r" " “A; . ‘ 6"“ 'a/ "S "3' ' w.» ' -“- -’
- 4 m n'l' rcycw n-.~L.¢.- - f" :' J“ .""!_.I«':. ' hilt-'1 \..'~' "“9"“? ”‘3’“ ‘ '3”:
. "(31 (4' «Zr-4'94"" 9.. ring: up -!,;“'+'-._ 1'"! ~ ""“"~r :‘3' ‘- " ‘4‘ , “we” . ' “
a :3!de 39" h»..- ‘ o s “"169”; 3.3,”. 1‘3" ",‘A‘gymcfio‘y ‘ ‘
5. r‘ .; ":~‘.~.'~’-‘:‘:"“‘ ‘14. tum-"r ' r-‘U‘s :fifi't'fia'm .
c 3:! ..f ‘-:'f. :".‘.‘":._,- ‘9‘}: "92:; .-"u"_,. (‘50.!1.’ . ‘.§'# a. h .’
. -"‘ ‘.‘l r r Y; ‘ 'u' z .
M"; (:17- n 0 we};

_ I ‘
_ arr; 1qlb-vac 1-. .. . ,-, 3",
r, ‘u . .a .‘ '. .' .. . § -_ ‘ E . ' - . Pk‘i'.
‘ 29$ '4-‘535‘ “2333a, ”“43 ’1'?“ . - 1‘ ”I . an . . .... 5'2“}? W“. 'v. , Wan r~‘:‘- ‘
‘1... .., L'!‘;.$?','.:i§:..w +52, 35,-; 5: Ln: ":1 . . gm .- - . ,
4 ‘J’_ “(i-*3? ‘1“! ,

§.':9¢2}'fi..*' .n‘r F'.n::;_' :.",' , ..r‘fi‘

 
 
 
 
 
 

 

 

  

 
  
  
 

 
 

PO}.

  
    
   

 

J; ‘
.31 .

 
  

     

I .1
:'°;‘.':.
.. ... ....
4.32:;
2 ‘5...
, o 2'. u v
.c... .b'£_ I‘. “.4,a:4.‘1
3- ..., m» "'16) ’-. .-w.~.r: m ' - .
ni-W z u . \r.‘.,..‘.;,‘_" ’95” . ‘ .
l

  
      
  
 

 

  

 
  
 
  

 

   

 
   
 
 

   
   
   
 
    

   
  
  

 

        

      
   
     
 
  
   

 
  
   
 

   

    
     

I m»;
‘(.-'-_."" ‘ -
- z - u
I!“ --- I -,. u
z ...” .{ ”...' __~
' V ' a .- :w :3... .a ..|.-
, -'. ‘ . .V.’ ' ...:m‘m'. :’- .'.L I. l . .
‘ ' C" ‘0' "I"! '6 W' ’ C"I‘I?l' '.‘ "’ V I 'o
. "{.~ "..“_ J."‘. J'P'.rl’.|pm . :,';.°.',..."' '.-.‘:,‘. ,,_,.l.'_.-‘,." ..- .’
cob.‘ ,‘~,-0\DI-" o-gi 0' PV"' ". ,, . ‘ ' V' _'-"‘- "'f‘-V'-" " v-‘ P I
'm 9‘ 4 :vl'n' 5 ... u“, .r’Vl, .'., .i : _ ' ’_ ° ‘rv ‘0‘ "g “c '"Q. -. '0‘” . .‘:‘ '0 o ~r".. r - u , . -
I '5'?“ .6_'::.‘ .f" :I ‘Vfi'llo'u I' ~ . v ‘ ”-63 an o - .. '7‘", -' '.' ‘ “ -" "".,’ "'f.’1'."‘ ~."
“-1, a!" . ,| .". f "-v" Ha" ' '4 . '- _ . I‘ . . ‘ . N ' . . . O ' x ' -Q. . _ - , . - .; “.Px'. . .l-"';.‘..-" 05". ‘0 ' ‘.|,A 0'
o . ”V — - ‘ O .- , - n , . 4 . ' ‘ - ‘ .' 0V . 4 _ .. . ‘ -_ . ' , -"'me ~ I 01-. .oo
.:cn'-I”."l:"" 3‘ 3"..1‘ TO‘I'xf." 3.325233”! ._'.‘} .' " . . ‘,’~”. J , 9' ~. ..M‘“ ‘. o u- .0:.f..‘. or} .4 -..9... - .,A.‘ 3'“. .r&. .
‘.. - . ‘l ,4. 'i. _, .' ’. . -- ‘ - ' -.. ' ' _.. - 4. 5.. , ' P ' ' ° 0 ‘ 4 - '..r'. O - '1’ o
' ‘! 'W‘r ...”. ' $.14. .‘-. .4 n n'- °.. .. '- ° - . .. .r‘ ' ' ‘ ' "‘*"' ' ""‘""""{f'.'$'l"‘ '-
C '3“ ‘1"‘39‘.’ .. 51':th - wupd v. .. . ' ' ' ‘ '"9'1"" ‘ '. I 0']f"
q... ‘1”. .‘.,,.' _. , ‘ 9' O'. ' 0° .1 ' ,fif‘l'par "v 'vuwrr . . v.
m... ....' ’ ....r,. .V'. Q . 3.....7' _‘,. ". ..}...L 0 (f0 '. (V o..-.,. .
‘m‘.’-"'fl""l‘-"‘ . 9‘ :.'q(‘--o"oo"'afl.~:.‘ ’m’ 1..., .... .1.-
‘v~.'-’ '1’.- ......‘ “Q" -’!' ‘-(r" o' ‘ O I II to-.. o
'"a"? '. 0.; ‘. ' 1 : ...” Offiyv‘ ~-( 9 o- u, H. , I
R, . 1'; u: “,.I'Y' :_4" O." t:..:,Ioo-o .-. .’ .- .‘-- ~¢ .. o'. q ‘. ‘7. ’- ". _
'- ¢ ..u .. 1. ""0"! ' 0'0 00- -r rd» ”.0- -.‘; W 9" .c a
13" '3: ”as" ,1; \. l " "’“'7‘;~" 1‘33'3133-15surl- ”'1' . - - . -: >' . ~v ‘ .3 V‘.
I ‘1‘ '(y' ""'3'1' 1"? I” 3‘"? '1"*‘-'" ' f'z "- f' g ‘9. "
1 —-o-.o. ya 'fl-“' "~r'N"O.'J‘ ‘1‘“ .."H‘ """:--°‘v~.'° H
1' u... '04- 1 s w,-.,,. ,, " " 9 V D Or°',-"" '- "rr‘I' ’.' ' v; P} -- — -
.. ... v ' ,. OJ ”"?-:‘o:‘v;!m -0, 0’ - . - :- v~ \ r0! .... ..
u an -o"o'“'vvm vogJu301- . r" .. . ' ""' ."’ "'.‘ l""'?"v“'rr'"“’“ '?'~ 0' ‘ u. ‘ r9 "' '“ -
"(r-4 "r guy“. .... . ...... - '..'. t”-.' 'f‘“M"“s:J‘n v- -.., -- - ‘- -- '
0’. ',.-;to(r- "‘.(Z' .... J... _ ‘l"l“' . t g ‘ .. '.' 1;! -n,'°v_P.(._r'!~-¢;~ - a i ::f!’1. ~o-_o ’1' o '- 3, -u A": '.—.~. :-
" -"" .'d"""" «VV' ""1 J: '4:'& ”It..Y.-‘f:.'~':5!c"- “ H a" ’ ‘ do ...”: '::.’ " ... '3. H‘ ‘0-'""-3 ‘ ° ‘ A‘ ' "' " 2"” "" 'V‘I
. ' . .. . . _' ' ‘ - " r ‘ ‘ u’ d ' ' . :x-* .-| o—-~~v - . I — -~..1—. . . .
::;.|..'.’ 4".0}: '3’.‘ _" ..z’ . ......" . If- {- ‘1 g 0' '1“.' 3"“,‘. --.. .- .. ...?r 'f‘“ 2.. ',_‘.
'I‘O",('-l J.-a. ‘ u: "‘ 4""°0:"'r>-"v~'oo | 5w- «v - .‘v-Iv '-~- -.
:Ir‘wuu'v'w "0-. -.-»' 4" ‘1' ‘ "“'“' '3 no: v" - -r --' "-0 'x‘ .. -- ..' .
«gun-... —.‘..' '1 ,‘ ' t}:9-'(-§22'> vb a} -r-—,o.-r - I ,.~-‘--r
’,. ‘r"..a-.,):.g, , ’,".~ , . . . v -' ”:4, r; 'fr.on-Iu . 3 Q .- .07: . . . !,. .,.
‘ “""‘"'10‘-,oav0v.r'.'0 ’ u "' ‘3' " ‘°"‘"‘ \"-‘“a'*j;{zv": y'v-m ’-.'g"'vr‘- ..
.00 '9‘! ,.‘....‘ . ’., '7. ~‘F”'\." '. ”0': l 5’ 'Q’q 0. IT”! '.. u‘.... .
no "30'“. '-‘ 0“. . h-tr '~:,: .3" "'l""‘u'::""‘ -Q‘T, r’ ‘. ‘x - . .‘ :0 u". .
y ' Qp'1_.'.t4|-.' a ...)- , - ‘-" ' ‘ ' s'|--- 'y- ‘ m. .- ' r'flp, .
' ....g...' V"'I"‘d ,.;- fl' 'vvl.-o"a~y‘,q - v L-» 3“ .~‘,.-'-.r ”...-..- . ,.
0... fi..~V1.-o-" .1 , , 'l J, " ’ " -‘0".O.0"‘\ '0' - -O0-r“" I'f...‘,v-_l' H,‘
t' " ." ' "‘l." ( fa ‘ - Q - "l0_nn . .,.o}‘. '{g' r"r ‘1""'-I ~Vu"’no.-.‘s .. - n. - . v ”4" ' ,
.i’. ' 9'4‘lvn. 'o \ 1' 1.. - . ‘ "' ' ‘~‘ ‘7""‘" ' ,
-o -- ~:" 3' . ,. follxr'lzxy’gg't ’IO 5‘;'t30' d‘. >r.. '9":7”;‘:"."“’ r,
- an. ..aaq .v ,‘ . l 0 0'0 1 O'-— u-o-u - ,' ' _. _.,.. .’ ){j‘Of-ZU' .
w ”'0 . )co O V... _ , - . ’ - . - ).. . o . . 'O-y‘r a, y
04- O ..." c r... . . ’0'. "..: . .‘:l ."l I I~ I'. r. . ’ I Q - - ‘.' . :2 _, ‘.';'I.‘:-l.“.;‘ "P. ' . . ‘f if Q Q. 5 . ... ' .
’;Vuofl-I,' ‘ Jo. O‘rn- - ,' _ . -, . ':0 '5' Mil. v ' . ... ‘ ": 0"‘l ..rv-o
"""'f" ' Odp f‘rl - ‘Do‘ u~r-o'..-.‘. ‘-" - """H'I . 70
r-'—.u.-..a"u .. . '3 . '_,“. ‘l 4. ‘ 1 . ‘y L. “deny,
.J-'."3‘J"I"-""" ‘ , ._ -. 'I.‘.. I ‘. -'rJ- . to V53..-
‘ i o - ‘ ‘ ..‘1 l . o- (... J...“.. . ' ‘ ‘ :99 n <-. r. -. 5‘-
": 0‘1... '04.. ..¢. 1-500; v-, ".I“' - '
V ‘ ‘ _ l o . .
0. '1. ’ ' ‘ '4' _" g. 0 ' 9 ".- ""Io I’.’I 0‘ 3 . ¢'¢’;-q
O - ‘vo '7'.‘ f‘ . o . 1-0 5’! "~o~.,‘n
." ' 9|? ‘ 5- NU". O‘-‘o.,., o...
' 4 .0“. .'. 0- ~ .".'~ul ..ol'rg-
' . - M‘ I -'~ $5:Of H on n
- ' .. w far-r: '1 -‘
H" "d" ' "w: ”T ""‘"".‘r"- ft»
.--... - f“(’ o.- 1"" .. oo-o. ,..‘ . ", - ..' ' ' " J. ‘ a
. . 4 . . 9"." 'oer “v. o. ¢~- fil.‘c‘0.'tl - -
‘V-‘4 “V'f lhur'to ‘Q. . .... r 'H- . 4 I 00‘ {0. o.
0—. u . . I 0': . 9-, A b g o . J. ,., ....JI '. '1”. ..r ," T‘I. 'fVQ'I- oG‘VJ. 0 V ... . -. . g
. . ._ q' ...... 0'“... 3:.“V'U'! " . ' Ily‘uaov v.7.fi' g V. n» -0, .. ... -.. ‘1 ..‘u. - O
... ,,' .'| ”J. ., .'.,..“‘. . . . .‘fi ,. ‘ '_'. ‘Fcboo'v , “F." .5-5- 9 0-9 ug-Vv 5. .1... . ..,.. "'9‘. J. . -
-l,"0¢'v'o "0"" fl ." "g""‘, ""“'f.' ‘ ' ‘ 7‘0. ‘fco,-uao§ £y~s G
01.1.: .-v—.......o’. ... "' f'*""‘°.'.d' "’ I "' ‘ "V‘ “' ‘ boo 1> g
' r“.ll.vu- ' . *IOOO..:.-.‘,- ,., '7" .‘1..f".. '¢" ""‘gon'lw .90.)...4 . ‘,.....fl‘. 7 v o"
y t’ 1‘00. - o '..-9 ""‘°‘ '. O ' "f ...: ‘ ‘ ' 7‘ L, 9' 3 ‘ g‘ - .... yv'é-CQD f- . .... "”"P.‘“ '7‘ a... 9"? a, "Q ~. ~"'-'" '
‘ ...“ a ,. , Jt _ . -. "1‘:"II::.'-.'O'!‘ :‘ v2. ,q-, ,I. . . . "I ‘1' _ 0‘ ' ...‘r, N'f I 4. f If l.--. '-' VV..-":....':“ .f— g." . ’. .“ ‘. _ on
. , u ’ O . . I . - . . .._' . " "‘ ' " p: . u \ p.» w u . _. .
-.3. . _ , ,. .... ',“ ... . to v . 4 ...-"......f ,. .. .... ...“: 25'1"; 9"2'" ",*.(.~.:¢‘ 0.”. 14$.l’4o‘n-v-uw up! Jan-yo "P" 1 .... ...-nu -‘,-r—' .‘.§o'o-(“: u... . ‘ 3.:
a- .- ¢ .O‘L':I’"“l‘ “' "a “ " I In 0-,3. ...... a' . ‘ “.. ‘ 0"“..":".' a L‘u‘."’ .':.’ ..--.." r’u "" H I 0.
..I.. "a’o ..., coo. ....‘.. -..., . a~"’,.' '5, ,..' ."-.-..'/- i ,O"‘ .‘ ‘ I '- 0- 0-1- .‘, ‘ ."'
‘ ‘0. .— .... n. _ , ,. _ ... .’_ ...! OIIO‘ -ou‘qu..'.‘; .‘ I 0‘ 0. ‘Il' -
."" ‘F -0."o O ' I I vp;f.':'..' !'- o'wvr' m»..... '7‘. v. .c , ...
- o‘l'll'°_-vy' "om." ’l'..' ‘ 0' . ,,‘ _ .
9°~0 '°"‘O‘. ‘ 0- ' 1-0. you JOWC‘MOO“.
9‘ . '5 —-Q '.'.:....' ”‘.',.',~....,"q_. ' .- - .l'a>o(V‘., ha a
..(5'.v"4 .- - . ' '. ‘_ . . ' . ..- I. _| I ’3’, ' ' ' ¢~—' r 5-; ' 'x o r.-';..~'.o'v r.. ..V‘ . .
. " - 0" —.¢ .'-'v "or. a-.- ‘ O l - '9 O. ' . 1 _-
I. . 00‘ ‘IO- 0- Ca ‘ "E. o ‘0'“ v, "‘ - I," . II ‘ l r" '0‘. . - u ' . 'mn I ~;..“.-. ...“‘ ... "" "‘"l‘ 'l- r ' ‘. '. :::::' " .‘L 'L‘"' C ' ..-
~ . . —. .. _ r. 3.19;"- nu. —. . .. m. a J a. .... .\ ‘;‘;’.2‘.’4.*'.;' :"..::z." Mv - -. '- ~ «H ”1"”. L
v' Q".' '0 ’ , ... . I J. 2’ :" . 0.... . . ..‘v .... .‘ I0 . *‘~ .' ’ f.’0'— n‘! ., . NO“! .7". - .
' .. - o - I - or u : - . 0 On - n a .4.- - o _
.: OJ :1. ~ ' "‘.1 ‘ -‘ ' a I . ' 1"“, '1" ‘ l '1':.'.‘.'.' 3- ‘ ' . ' I "' - ‘ J“ Y. I" v- HIM“- , .rl) try:€.o~r.—p 90 :15} 'o cov‘a’o.0.o' o:
. ' r .0 ~- Ion-g Y- ‘7 l 7., ‘, 0 ‘ ans-Hv-vcuovoc- .- .’
C "' l'--.~- -—l ')|- '.. ‘n..,{.!.."._., v “‘9- or
. . ,‘-. u.‘ P... ‘ .... ', . ‘ r.‘: .
" 0" ...-’9‘. ‘ u- . l. . .o . -‘ .J , .O'o'l _.'._ . . nv¢.,— :00...:~
. ¢ ' o n . .. . _7‘ . ”0". va‘
0 l’0 - '1”: "o .T..:. so I0 'u -'o .1000; ... . 1.2%... - ' 1" " ‘ "U M" :O'.*' """" ’1-kf' 0".
3‘. -' u 4",“ -. .. ..-,J.. .. - . . - -, If}: ...".u ' .' '3:- H»?- gr» “‘9'? , -
" - a.‘ ....._ ..o '0' r' . -oo 0 r -. - 1 . _ . _- on ‘ WWW-”O ~ t—Mpo.
' ’t’ :00 n Y '0"Q t g . ’. o 9' v r ...... J"‘.Ill 'C‘t' ‘ -- ,O-dwt "s 6'6— ..v , ..J .';:'. ':.::. .‘.‘.':'r. ' '. 1’9..." 05-}. — -.:J: £2, I-fi:-o.v.: o:v:vo-v oL-ot .v v-
5 0.. 5 -J‘ . . o . .-.. '.. . ”I". ,.r.. ,. ,n..‘ "“""| ... '_..,.l.,,..~.. M~~.: D ..Joum4.” ..- .."..~ .k.- 'o-¢j. _
. I‘D! . 0 ’ '. - run. .. ... ., '. ""0:n '03-: odnovlv-r50vo...¢'.. . . 1- ' fl_..,"' :"I‘Prw'fn OW,'--:..;:Of.0\.o¢-1m ”00“” —.-'I.O..I'-~
'J‘ 0 .‘9' "H '- ' " ‘0" u'o o' . ' I‘"" "' “‘ " 0"?"v~ r'l".""~Afi1-Oo- oo‘ """"""“" "" "" "‘°‘ ‘1‘! '«Mrnuv¢*'r""' o
‘ ' ~.¢‘ " .. . 0' ......g. . . . ..‘ ‘-. -c-.~ q..51~.-r-~-.§.w.‘.r".‘ ._ ". ’". ...- ”I“ On-"w'~ ‘dl'flfij’.fuo .'.”~‘.OW' ......M '1‘,“_._”.
." .- ..,'l ." ‘t “ ou 0.0o‘AW-y I . (I. "" ."" .‘. I- ' °€“'.:-"’ 0-. .0 ‘ " V09 ..'.- .,~"’.4 “‘-'""‘".‘ ";"0.7-0'- "fi-v
'Ii- 0' ' ' — n,~o".- ... .I 3.,“3... "" C C-O-oa—».o¢°’ "II ‘h—oy.¢‘.‘¢ .,.. -. a. '4. ""‘"‘f‘ '-""'0“ 4 'I‘."o ’0 'OQOVOIO'H.
" ‘-"" "’v- ' " " " 3 u.,~.'.‘.. "'9' '~ " ‘ ~, 0- . v ~~.—....,. 3?... §",",""l_' ’g"""* 'r‘vooo'N-ov-m- «pun-coo.
..‘ol'.".'° ' 1'. ‘u -" 'J'J . ‘ -..' "" "” ’ "°"‘*‘° "' o 'rmv Y”. J.“ .'l""..':’..c;§ T:::"»_‘"", 'N"“'1"-0~70
o. - a O I . ... . .~~ . "a 0 ° ‘I I V..‘ ‘. M' ‘ . . O. Cc. ‘."d
. . . - . .. . . 1 v - .0. o J "Q h I! '~' 7 run w
? I"..o. ~. 0 - ‘ - . . :0 ¢_ . ...“ ‘5‘. . , pr: . -' ;’;; _' _ I. 951'. 'J'IJ? .\' .;.- I . '4‘.“ " ..‘g 3, OJ? .’.i. t:":':’: :::OI.O.¢..‘ .‘c‘ " V5.0; ’0.0
. ' l. ' '. q u» -— [a y ‘ ... '0 O ‘ Jr"""' .' ""‘ ’ "- '~--' '0..- .... J u If':" m".!:'.’:':“ ""
"' I. I ""‘ " H -o y"o°~‘ H ' ".°. '1 2"". ' “I f ' “ "'1 1 Iiod"..to' “' (Ex-I”, Wt; ... :v‘J-FJ” q”.
:',‘ r- y. .... '0 ..r. v .: .. . 4 ac ' ' l‘t’ . l. ' ”Q "I 0 9 0 o 0 -Q. .' ..I .oé €}‘.. .r 1-!" J, r ’ " J‘ "(‘0 ...." ’0’: ho 39“ .W ,. ~} '." 90<r.a.d.'
f '4 ‘ ““ '0 -- 'o ‘ ‘,' ... .' ' ' "1.;‘559 ‘l .‘o -.. ‘. ' “21.9.9.0:f00 D... obtréd‘ - o '.
' ’ . ‘~"'- ‘ ~. ' ‘..‘. ‘9', I"’ “..-. «I o. -' 1:. "J l"".' 0 010“' -‘~---‘
. . - . .! ..f. '0 , -' - _. , O . '. .‘. * . o - . .. _. .. ' _.. - . .30.. c _‘f" __. -
' "Ivo . -... I. ... , .-,. "~'C"~O I 0005. A‘
I ’ "‘ "-‘ " no 'r.‘ .. o " x-Vo.
"‘ ‘ ‘ ' ‘ —— . . ---
, .- ~»..-..32'n-4-.+

 

'4‘
Eg.

CHI IGAN TSATE

ll illlll IUIIJIJMHMHI'IIJIlHllllHl

31293 00889 7948

IHIUH

 

 

 

 

 

 

 

 

 

 

LIBRARY
Michigan State
UnIversIty

 

 

 

This is to certify that the

dissertation entitled

On the Perception and Production of Mandarin Tones
by Adult English-Speaking Learners

presented by

Jya-Mei Lu

has been accepted towards fulfillment
of the requirements for

Ph. D. degree in Linguistics

 

 

WW

Major professor

Date 37 77M f/fyl
7

MSUis an Affirmative Action/Equal Opportunity Institution 0-12771

 

PLACE IN RETURN BOX to remove this checkout from your record.
TO AVOID FINES return on or before date due.

DATE DUE

   

DATE DUE DATE DUE

 

M: 132$! 17

 

 

 

 

 

NW 17 win

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

3t

 

 

 

 

 

ll

 

 

 

 

 

 

 

 

 

"‘1 1i

 

 

 

MSU Is An Affirmative Action/Equal Opportunity Institution

emcee-9.1

 

 

 

 

ON THE PERCEPTION AND PRODUCTION OF NANDARIN TONES
BY ADULT ENGLISH-SPEAKING LEARNERS

BY
Jya-Mei Lu

A DISSERTATION

Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

DOCTOR OF PHILOSOPHY
Department of Linguistics and Germanic,
Slavic, Asian and African Languages

1992

ABSTRACT

ON THE PERCEPTION AND PRODUCTION OF MANDARIN TONES
BY ADULT ENGLISH-SPEAKING LEARNERS

BY
Jya-Mei Lu

The learning of tone is difficult for adult learners of
non-tone languages learning a tone language. This study
investigated how Mandarin tones are perceived and produced
by adult English speakers in the first months of their
learning of Mandarin.

American university students taking Elementary Chinese
were tested (at the beginning and the end of the first term)
on their perception and production of Mandarin tones in
isolated syllables, 2-syllable words, and short sentences.
The production data were judged by native Mandarin speakers.
Scores of the tests were analyzed statistically.

Analyses of the results show that the perception and
production of Mandarin tones were affected by tonal context
and task. Tone perception and production were easier in
isolated syllables than in z-syllable words and short
sentences. The production of tone was better in imitating
than reading from a list.

Significant correlations between the perception of tone

in isolated syllables and in sequences (2-syllable words and

short sentences) were found. However, the correlations
between the production of tone in isolated syllables and in
Sequences were insignificant.

Significant correlations were found between the
perception and the production (reading) of Mandarin tones in
isolated syllables and 2-syllable words. However, the
correlation between tone perception and production in short
sentences familiar to the subjects was weak at the beginning
and became insignificant at the end of a term.

Significant improvement in tone perception was found
between the beginning of Mandarin learning and after a term
of learning. However, no significant improvement in tone
production was found.

Analyses of the tonal errors made by subjects show that
the difficulty order of the four Mandarin tones as measured
by error rates was different in different contexts at
different times. The patterns of misperception and
misproduction of Mandarin tones also differ depending on
context.

Some suggestions of teaching Mandarin tones are

proposed based on these findings.

To my parents and my husband, for their love and support.

iv

ACKNOWLEDGMENTS

I wish to thank my guidance committee: Dr. Grover
Hudson, Dr. Ruth M. Brend, Dr. Julia 8. Falk, Prof. James P.
Wang, and Dr. Yen-Hwei Lin, for their assistance and
valuable comments. Thanks to Dr. Patricia Lunn for serving
as the college reader and giving valuable comments. Special
thanks to my committee chair Dr. Grover Hudson for his
guidance and assistance throughout my doctoral program.

I am also indebted to Dr. Carl Lee of the Department of
Mathematics at Central Michigan University for his valuable
advice and assistance in the statistical analyses, to Dr.
Hiou-Ming Wei and T. S. Liu for spending long hours serving
as judges, and to Mrs. Betty Briggs for her assistance in
recording test materials and editing production data with
the language laboratory facilities.

Lastly, I am grateful to Dr. Yen-Hwei Lin and her
Chinese teaching assistants for helping in conducting the
tone tests in their classes, to the students of Chinese who
agreed to have their tests studied, and to the Chinese

speakers who served as controls.

TABLE OP CONTENTS

LIST OF TABLES O O O O O O O O O O O O O O O O O O O Xii

LIST OF FIGURES O I O O O O O O O O O O O O O O O O O Xiii

1. INTRODUCTION . . . . . . . . . . . . . . . . . . . . l
1.1 Introduction . . . . . . . . . . . . . . . . . 1
1.2 Scope of investigation . . . . . . . . . . . . 2
1.3 Outline of thesis . . . . . . . . . . . . . . 3
2. TONE AND INTONATION IN MANDARIN AND ENGLISH . . . . 6
2.1 Tone in Mandarin . . . . . . . . . . . . . . . 6

2.1.1 Tone in isolated syllables (citation

forms) . . . . . . . . . . . . . . . . . 6
2.1.2 Tone in 2-syllable words . . . . . . . 11
2.1.3 Tone in connected speech . . . . . . . 13

2.2 Word stress and sentence intonation in

English . . . . . . . . . . . . . . . . . . . 30
2.2.1 English word stress . . . . . . . . . . 31
2.2.2 English intonation . . . . . . . . . . 33

2.3 Comparison of the two systems . . . . . . . . 39

vi

3. PERCEPTION AND PRODUCTION OF SOME ASPECTS OF L2
PHONOLOGY . . . . . . . . . . . . . . . . . . . .
3.1 Perception and production of L2 phonology by
adults . . . . . . . . . . . . . . . . . . .

3.1.1 Variables affecting L2 phonology

learning . . . . . . . . . . . . . .
3.1.1.1 Phonological context . . . .
3.1.1.2 Task . . . . . . . . . . . .
3.1.1.3 Sex . . . . . . . . . . . .

3.1.1.4 L2 experience . . . . . . .

3.1.2 Relationship between the perception

production of L2 sounds in isolation
in sequence . . . . . . . . . . . .
3.1.2.1 Perception . . . . . . . . .

3 O 1 O 2 O 2 Praduction O O O O O O O O 0

3.1.3 Relationship between perception and

production of L2

phonology . . . . . . . . . . . . .

3.1.4.1 Perception . . . . . . . . .

3.1.4.2 Production . . . . . . . . .

3.1.4 Acquisition of L2 phonology over time

47

49

51

54

56

57

57

62

64

73

73

77

3.2 Perception and production of Mandarin tones by
adults L2
learners . . . . . . . . . . . . . . . . . .
3.2.1 Perception . . . . . . . . . . . . . .

3.2.2 Production . . . . . . . . . . . . . .

vii

4. ISSUES AND HYPOTHESES . . . . . .

4.1

4.2 Relationship between the perception and

Effects of some variables on
perception and production
4.1.1 Tonal context . . . .
4.1.2 Task . . . . . . . .
4.1.3 Sex . . . . . . . .

4.1.4 L2 experience . . . .

production of tone in isolated syllables and

in sequences . . . . . . .

4.3 Relationship between L2 tone perception and

4.4 Improvement of tone perception and production

tone production . . . . . .

4.5 Order of difficulty among the Mandarin

4.6

Error patterns . . . . . . .

5. METHODS AND PROCEDURE . . . . . .

5.1

5.2

5.3

5.4

5.5

5.6

Tone Perception Test . . . .
Tone Production Test . . . .
Test Material . . . . . . .
Subjects . . . . . . . . . .
Judges . . . . . . . . . . .
Procedures . . . . . . . . .

5.6.1 Testing . . . . . . .

5.6.2 Evaluation . . . . .

Computing the scores . . . .

viii

tones

95

98
99
100
101

102

103

105
107
110

111

113
113
116
118
121
122
122
123
126

129

5.7.1 Perception test .
5.7.2 Production test .

5.8 Statistical Analyses . .

5.8.1 Analyses of the effects of some

variables on tone perception and tone

production . . . .

5.8.2 Analysis of correlations between the

perception and production of tone in

isolated syllables and in sequences .

5.8.3 Analysis of correlations between tone

perception and tone production . . . .

5.8.4 Analysis of improvement in tone

perception and tone production . . . .

5.8.5 Analysis of order of difficulty among

the Mandarin tones

6. RESULTS AND DISCUSSION . . . .

6.1 Test validity and reliability . . . . . . .

6.2 Variables affecting tone
production . . . . . .
6.2.1 Tonal context . .
6.2.2 Task . . . . . .
6.2.3 Sex . . . . . . .
6.2.4 L2 experience . .

6.3 Relationship between the

production of tone in isolated syllables and

ix

perception and

perception and

129

129

131

132

133

133

134

134

136

136

137

138

141

143

146

in sequences. . . . . . . . . . . . . . . . 150

6.3.1 Perception . . . . . . . . . . . . . 150
6.3.2 Production . . . . . . . . . . . . . 154
6.3.2.1 Reading task . . . . . . . . . 154
6.3.2.2 Imitation task . . . . . . . . 155

6.4 Relationship between tone perception and tone

production . . . . . . . . . . . . . . . . 156
6.4.1 Tone in isolated syllables . . . . . 156
6.4.2 Tone in 2-syllable words . . . . . . 158
6.4.3 Tone in short sentences . . . . . . . 158

6.5 Improvement over one term of learning . . . 161
6.5.1 Perception . . . . . . . . . . . . . 161
6.5.2 Production . . . . . . . . . . . . . 163

6.6 Order of difficulty of the four tones . . . 166
6.6.1 Perception . . . . . . . . . . . . . 168
6.6.1.1 Tone in isolated syllables . . 168

6.6.1.2 Tone in 2-syllable words . . . 169

6.6.1.3 Tone in short sentences . . . 171

6.6.2 Reading . . . . . . . . . . . . . . . 175
6.6.2.1 Tone in isolated syllables . . 175

6.6.2.2 Tone in 2-syllable words . . . 175

6.6.2.3 Tone in short sentences . . . 178

6.6.3 Imitation . . . . . . . . . . . . . . 182

6.7 Error Patterns . . . . . . . . . . . . . . 185
6.7.1 Perception . . . . . . . . . . . . . 186
6.7.1.1 Tone in isolated syllables . . 186

6.7.1.2 Tone in 2-syllable words .

6.7.1.3 Tone in short sentences .

6.7.2 Reading . . . . .

6.7.2.1 Tone in isolated syllables

6.7.2.2 Tone in 2-syllable words .

6.7.2.3 Tone in short sentences .

6 O 7 O 3 Imitation O O O O

7. CONCLUSIONS . . . . . . . . .
7.1 Summary of findings . .

7.2 Suggestions for teaching

BI BLI OGMPHY O O O O O O C O O O

APPENDI x A O O I O O O O O O O 0
Tone Perception . . . . . .

Pronunciation . . . . . . .

APPENDIX B . . . . . . . . . . .
Perception Scores . . . . .
Reading Scores . . . . . . .

Imitation Scores . . . . . .

xi

189

192

194

194

195

199

204

207

207

211

215

228

228

229

230

230

231

232

LIST OF TABLES

Table 1. Correlation between actual and predicted tonal
contours . . . . . . . . . . . . . . . . . . . . . 23

Table 2. Group Voice Onset Time Means for Production . 81

xii

LIST OP FIGURES

Figure 1. Tonal variations in sentence final position . 21

Figure 2. Basic tunes of Mandarin . . . . . . . . . . . 22

Figure 3. Median scores of two pronunciation tasks

1. INTRODUCTION

1.1 Introduction

It is generally agreed that adults learning a second
language rarely, if ever, achieve a native-like accent
(Scovel 1988). This seems especially true for
suprasegmentals since L2 learners who speak with only a
slight accent are often found to have suprasegmental rather
than segmental errors (Ioup and Tansomboon 1987:344). Even
in the acquisition of non-tone languages, intonation
contributes to the foreign accent by which a native speaker
identifies the nationality of a non-native speaker (Van Els
and De Bot 1987). Not surprisingly, in learning a tone
language, non-tone language learners are likely to have
problems with tone. One study on the acquisition of Thai
(Ioup and Tansomboon, 1987) has shown that tone is one of
the last aspects of Thai grammar to be mastered by adults.

For English speakers learning Mandarin as a second
language, the difficulty of tone is well attested to by both
students and teachers (Kiriloff 1969, Shen 1989). Incorrect
production of tone is a very apparent feature of foreign
accent to native Mandarin speakers. If a Mandarin speaker
is asked to speak with a foreign accent, he/she is most
likely to do it by making the tones wrong. Because tone is
an integral part of almost every morpheme in Chinese,

incorrect production of tone is more than just a problem of

2
accent. Mispronunciation and misperception of tone very
often cause confusion and failure in communication (Chiang
1979, W.C.J. Lin 1985).

As stated by Ioup and Weinberger (1987:319), "Research
on the acquisition of suprasegmentals is a recent
development in studies of L2 phonology". Up until now,
there have been relatively few studies on the acquisition of
tones or other suprasegmentals by L2 learners. Thus, a
study on the perception and production of Mandarin tones by
L2 learners can definitely contribute to our understanding f”

of L2 phonology.

1.2 Scope of investigation

The purpose of this study is to see how Mandarin tones
are perceived and produced by English-speaking adult
learners (university students) in a foreign language
environment (Chinese class in the United States) in the
first months of their learning. More specifically, this
study will try to answer the following questions:

(1) Is tone perception and production affected by
variables of tonal context, task, sex, and L2 learning
experience?

(2) Is there a relationship between the perception of
tone in isolated syllables and in sequences, and the

production of tone in isolated syllables and in sequences?

3

(3) Is there a relationship between tone perception and
tone production?

(4) Do tone perception and tone production improve over
one term of learning?

(5) Among the four Mandarin tones, is there a
difficulty order for English speaking learners? If there is,
what can account for the order of difficulty?

(6) Are there general patterns in errors of tone made
by English-speaking adults learning Mandarin? If such
patterns can be found, what accounts for them?

Because of their important implications for theories of
L2 acquisition and for L2 teaching, similar questions have
been asked in other areas of L2 phonology concerning L2
learners of other languages. A more detailed account of
these questions and hypotheses concerning some of the

questions will be given in Chapter 4.

1.3 Outline of thesis

The following is a brief description of the rest of the
chapters in this dissertation:

Chapter 2 gives descriptions of the Mandarin tone
system. Studies on acoustic analysis and native speakers
perception of the Mandarin tones are reviewed. The use of
pitch in words and sentence in English are also briefly
reviewed. Differences in the use of pitch in the two

languages are compared.

4

Chapter 3 reviews studies on the perception and
production of L2 phonology by adult L2 learners in general
and also the perception and production of Mandarin tones by
English speakers. The former part is focused on some
aspects of L2 phonological learning that are of interest for
the learning of tones in this study. These include the
effect of some variables (phonetic context, task, sex and L2
learning experience) on L2 phonological learning, the
relationship between the learning of L2 phonology in
isolated words and in connected speech, the relationship
between the perception and production of L2 phonology, and
the improvement of L2 phonology over time. The latter part
concerns the specific difficulties encountered and the
errors made by English speakers learning Mandarin tones.

Based on the literature reviewed in Chapter 3, Chapter
4 addresses each question raised in this chapter in more
detail and proposes specific hypotheses in the form of
expected answers to most of the questions.

Chapter 5 describes the methods and procedures of
collecting the data on the hypotheses and how the data are
to be analyzed.

Chapter 6 discusses the results in regard to the
hypotheses proposed. The results are also compared to the
learning of other aspects of L2 phonology and to other

studies on the learning of Mandarin tones.

5
Chapter 7 summarizes the results discussed in Chapter 6
and provides suggestions for teaching Mandarin tones to L2

learners.

2. TONE AND INTONATION IN NANDARIN AND ENGLISH

2.1 Tone in Mandarin

The term Mandarin in this study is used in a narrow
sense to mean the standard Mandarin Chinese based on the
Beijing (Peiking) dialect. It is the language most likely
to be taught in a Chinese class in a foreign language
environment. In this section the tonal system of Mandarin
will be described. The Pinyin romanization system is used

to spell Mandarin words.

2.1.1 Tone in isolated syllables (citation forms)

The basic tone-carrying unit in Mandarin is the
syllable. Almost every syllable in Mandarin (as written in
the form of a character) is a morpheme/word and can be read
in isolation (citation form) with a distinctive tone.
Syllables consisting of the same consonant and vowel but
differing in tones will be perceived by a native Mandarin
speaker as different morphemes/words. (There are also many
homophones in Mandarin Chinese that are written in different
characters.)

1 There are four distinctive tones in Mandarin.

They are usually numbered as Tones 1, 2, 3, and 4 (or lst,
2nd, 3rd, and 4th tones). The characteristic pitch features
of the four tones in citation forms are: Tone 1 starts high

and stays high, Tone 2 rises from mid to high, Tone 3 starts

7
at mid-low, dips to low and rises to mid-high, and Tone 4
falls from high to low. This can be summarized in the
following table using Chao's five-point tone-letter system

(Chao 1930, 1948):

Tone Description Pitch Tone-letters
Tone 1 high-level 55 1
Tone 2 high-rising 35 1
Tone 3 low-dipping 214 v1
Tone 4 high-falling 51 V

The tone-letters without the vertical reference line ( -, /,
v0 and \) are used as tone marks when writing Mandarin in
some romanization systems, including Pinyin. They are
placed above the syllabic vowel of a syllable, e.g., tan,
tan, tin, and tan. There are a limited number of morphemes
(e.g., sentence particles, grammatical morphemes, and
suffixes) which do not have a distinctive tone and are
pronounced in a neutral tone whose pitch is generally
determined by the preceding tone (see 2.1.2 below). These
morphemes are never spoken in isolation in natural speech.
The actual pitch range and interval of the tonal
reference points described by Chao are relative rather than
absolute, according to the sex and voice of the individual,

and the mood of the moment (Chao, 1948:25). Since the

8
description and the notation first used by Chao were based
on auditory analysis, subsequent acoustic studies (e.g.,
Rumjancev 1972 (according to Lyovin 1978), Howie 1976, Ho
1976a, and Tseng 1981) have tried to find the acoustic
correlates of the four tones in citation forms. The
acoustic studies usually measure the absolute values of the
sound waves (for example, the fundamental frequency (F0)
correlates to the absolute pitch values). Not surprisingly,
different figures were obtained from different individuals.
Variation has also been found for different syllable
structures (Howie 1976, Ho 1976a). One of the variations
found is a dip located at the beginning (about 15% of the
duration of the syllable nucleus) of some Tone 2 syllables
(Ho 1976a). Generally, however, the acoustic studies confirm
the characteristic pitch contours of the four tones observed
by Chao.

Experiments on tone perception in citation forms by
native speakers usually result in near perfect
identification. For example, in Tseng's (1981) study the
syllable /i/ spoken with the four Mandarin tones and
produced by a native speaker as citation forms (along with
tokens obtained by editing these recorded syllables) were
presented four times each in random order to four native
speakers. They were correctly identified 100% of the time.
Experiments on the dimensions of tone perception (e.g.,

vowel quality, duration, loudness) by native speakers have

9
also been performed by manipulating the acoustical signals.
For example, in a perception test conducted by Howie (1976),
the stimulus sets consisted of synthetic tone contours
(based on average fundamental frequency contours obtained
from native speakers) imposed on the real-speech (level)
Tone 1 syllable bao, yielding a set of four tonally
differentiated words. Chinese listeners identified the
tones correctly (in agreement with the synthetic tonal
contour) 95% of the time. This experiment and others have
shown that pitch is the primary factor in tonal perception;
other concomitant features of tone, such as vowel quality,
duration, and loudness are only of marginal importance. In
another test conducted by Howie, the original fundamental
frequency contours in syllables differing only in tone
(ying, ying, ying, ying) were replaced with a constant
fundamental frequency of 128 Hz. He found that Chinese
listeners did poorly in identifying the original tones under
this condition.

Tseng's (1981) study investigated the role that
fundamental frequency contour and duration play in
production and perception of Mandarin tones. She found that
fundamental frequency contour plays the primary role in both
tone production and perception. Duration is not a primary
phonetic parameter since the intrinsic vowel duration in the
citation form is not maintained in spontaneous speech, and

the tone of a syllable (/i/ in her experiment) can be

10
correctly identified with 25%, 50%, even 75% edited out from
the offset if the basic tone feature is left intact (e.g.,
Tone 1, high level tone). She concluded that, when
misidentification of the tones occur as a result of cutting
part of the syllable, it is "due to the absence of the
particular fundamental frequency contour, not purely due to
the vowel length" (p.47).

Rumjancev (according to Lyovin 1978) also conducted
perception experiments which involved erasing part of a
syllable produced in isolation. Although his major concern
was to find which part of the syllable carries the tonal
features, he also found that when the final nonsyllabic
elements (e.g., nasal in yan or o in yao) were erased, the
tones were still correctly perceived by native listeners,
provided that the total duration of the nonerased portion of
the syllable was between 180 and 200 ms (p.132).

Massaro, et a1. (1985) investigated the evaluation and
integration of pitch height and pitch contour in the
perception of Mandarin tones in isolated syllables. They
found that "[b]oth the F0 contour and the F0 height
influenced the perceptual recognition of tone. The cues
were about equally effective. As in the perception of many
other contrasts, lexical tone appears to be recognized on
the basis of, at least, two cues, and the influence of one
cue is the largest when the other cue is relatively

ambiguous" (p.267).

11
2.1.2 Tone in 2-syllable words

Historically, most 2-syllable words in Mandarin have
evolved from the combination of two monosyllabic morphemes
or words. However, 2-syllable words now occur almost as
frequently as l-syllable words in Mandarin vocabulary
(M-Y. Chen, 1981). One and two syllable words are the basic
units in Mandarin speech. It has been proposed that,
phonologically speaking, tones in all longer words and
utterances are based on the combination of tones of one and
two syllable words (Wu 1982).

Besides the 16 tone combinations that can occur in two
syllable words (4x4), the second syllable of a 2-syllable
word can also be a neutral tone. The neutral tone occurs in
a completely unstressed syllable, where the distinctive tone
features disappear and the pitch is generally determined by
the tone of the preceding syllable. The neutral tone is not
marked in Pinyin romanization and is represented as a dot to
the left of the reference line in Chao's five-point tone
letter system. Although the actual pitch contours of
neutral tone can be seen from a spectrogram (falling when
following Tones 1, 2, and 4, and level or rising when
following Tone 3), they are usually too short to be
perceived clearly. The average pitch height of the neutral
tone, according to a recent acoustic study (Gao, 1980), is

shown below:

12

after Tone 1 43
after Tone 2 13
after Tone 3 14
after Tone 4 J2

There are some minimal pairs of 2-syllable words (or
phrases) which have exactly the same phonological structure
and tones except for the stress on the second syllable: in
one the second syllable is normally stressed and bears a
basic tone and in the other the second syllable is
unstressed and bears a neutral tone (e.g., déngxi "East and
West" and déngxi "things". See Lin and Yan, 1980). However,
except for these cases and a small number of suffixes, which
are lexically marked neutral tone (toneless), if the second
syllable of a 2-syllable word can be pronounced unstressed
with a neutral tone, it can also be pronounced with normal
stress with no difference in meaning.

When two syllables form a 2-syllable word (or close-
knit phrase), some tone sandhi rules apply. The most
important one is that Tone 3 is realized as Tone 2 when
followed by another Tone 3. This rule makes the tone
sequence 3-3 phonetically indistinguishable from 2-3 (Wang
and Li, 1967), and therefore only 15 tonal patterns exist
for 2-syllable words. Another rule of Tone 3 involves only
a phonetic change: Tone 3 is realized as low falling or low

level (without the rising part which occurs in citation

13
forms), sometimes called half Tone 3, when followed by a
syllable bearing one of the other tones. This variant of
Tone 3 in fact occurs more often (83%) than the citation
form in actual speech (M-Y. Chen 1983).

Although there can be some other changes because of the
shorter duration of the syllables in 2-syllable words and
the anticipation of the following syllable in pronouncing
the first syllable (e.g., when followed by another Tone 4,
Tone 4 in the first syllable does not fall as much as in a
citation form), the changes usually do not change the basic
features of the tones. The pitch contours in 2-syllable
words, except for cases where tone sandhi occurs, are
basically a succession of the basic tones of the two

syllables.

2.1.3 Tone in connected speech

It is generally agreed that "many of the functions of
intonation in non-tone languages are fulfilled in Mandarin
by other devices than intonation, such as the use of
particles" (Chao 1933:130). According to Lyovin (1978:143),
Rumjancev believes that in tone languages "the chief
acoustic features of intonation are the amplitude and
relative durations of various segments". Chao used pitch,
duration (tempo), and loudness in describing the intonation

patterns in Mandarin (Chao 1968:40-44).

14

However, among pitch, loudness, and duration, "pitch is
the prosodic feature most centrally involved in intonation"
(Cruttenden 1986:3). Since pitch is used to distinguish
lexical meanings in a tone language, it can be inferred that
the use of pitch to express other information would be
restricted in order not to interfere with the distinctive
function of lexical tones. Ross, et al. (1986) compared the
affective functions of pitch movement in tone (Taiwanese,
Mandarin, and Thai) and non-tone (English) languages and
found that the local movement of pitch in tone languages is
greatly restricted compared with English (p.283):

The presence of tone in a language significantly
inhibits the unrestricted manipulation of three
acoustical measures of prosody which are involved
in producing local pitch changes in the
fundamental frequency contour during affective
signalling.

However, the use of pitch in sentence intonation does
exist in Mandarin. Chao wrote about the existence of
expressive intonation in Chinese (1933:130):

Owing to the fact that Western learners of Chinese
tend to use intonation in a purely expressive
role, writers on and teachers of Chinese have gone
to the other extreme of regarding Chinese
intonation as entirely fixed by word-tones. It is
true that many of the functions of intonation in
other languages are fulfilled in Chinese by other
devices than intonation, such as the use of
particles. But there still remains a good deal of
pitch movement in Chinese speech which expresses
moods and attitudes of the speaker in addition to
the modulation of pitch due to the etymological
word-tones and the more or less regular forms of
neutral speech intonation when tones are joined
together.

15

As to the question of how intonation and tone interact
in terms of pitch movement, Chao compared syllabic tone and
sentence intonation to small ripples riding on large waves.
"The actual result is an algebraic sum of the two kinds of
waves. Where two pluses concur, the result will be more
plus, when a plus meets a minus, the algebraic addition will
be an arithmetical subtraction" (Chao 1968:39). He also
states that there are at least two types of intonation
additions: simultaneous and successive (Chao 1933:131).

Simultaneous addition occurs when one of the following
four features is added simultaneously to tones to form the
resultant speech melodies.

(a) general raised level of pitch,

(b) general lowered level of pitch,

(c) widening of range,

(d) narrowing of range.
All of these may affect either the whole or a part of the
intonation group. (The general pitch level has also been
referred to as register by some authors; see Cruttenden
1986:129-130 for a discussion.) Successive addition is the
intonational rising or falling endings added after the pitch
contours of the final lexical tone. Chao later (1968)
treated these two intonational endings of phrases and
sentences as particles. The following table shows the
modification of tones (adapted from Chao 1933:133 and Chao

1968:812, 814: numbers represent pitch height).

16

rising endings falling endings
Tone 1 55 = 556 55 - 552
Tone 2 35 I 356 35 8 352
Tone 3 214 3 2145 214 - 2141
Tone 4 51 = 513 51 I 5121

In contrast to "expressive intonation", there is
neutral or normal intonation in "plain matter of fact
statements", which is the regular succession of tones (with
their regular tonal changes) in connected speech (Chao
1933:127). Chao observed that for normal intonation "in
short sentences of three or four syllables, there is no
special intonational modification. In longer sentences,
there is a slight tendency for the pitch to trail off to a
lower key toward the end" (Chao 1968:40).

After Chao's studies, there have also been some
acoustic studies on Mandarin tones in connected speech.
Rumjancev's study (1972, as reviewed by Lyovin 1978) found
that "syllable duration is generally shorter in connected
speech than in isolated citation forms. Also, the longer the
duration of a given syllable in connected speech, the
greater the likelihood that the rise or fall of F0 will be
more pronounced" (Lyovin 1978:134).

Regarding the interaction of Mandarin tones and
intonation,

Rumjancev briefly enumerates various factors
which affect various parameters of tone in

17

connected speech (phrase intonation, logical
emphasis, emotional nuances, syntagmatic and
actual segmentation of sentences, and speech
tempo) and then discusses some of them in detail.

In discussing phrase intonation and its
effects on tone he mentions what is already known
from previous studies done by others (mainly his
Soviet colleagues); e.g., he mentions what happens
when the pitch features of intonation clash with
the pitch features of lexical tones: for the most
part, the basic contour features of lexical tones
are left intact, but register features are
significantly affected. In this connection, he
also discusses what he calls 'supplementary post-
tonal stretches of phonation'--a phenomenon which
has been observed by some earlier investigators.
Thus, for example, Tone 4 syllables falling within
a zone of rising pitch intonation sometimes
exhibit their usual falling pitch contour, plus a
terminal rise in pitch. (Using Y.R. Chao's
notation, we can say that in such cases Tone 4 is
realized as 513, and that the 13 stretch is a
'supplementary post-tonal stretch of phonation'.)
(Lyovin 1978:143-144)

What Rumjancev calls "supplementary post-tonal
stretches of phonation" appears to be equivalent to Chao's
"successive addition".

' According to Rumjancev's findings, these
'extra' stretches are entirely conditioned by
intonational factors, but are not always present
where they may be expected. Even in those cases
where they are present (and their presence seems
to be conditioned by emotional factors), the basic
intonational contrast is signalled primarily by
the pitch register of what precedes such
stretches. In other words, such stretches are
largely redundant, except perhaps in those cases
where intonational contrasts involve emotional
nuances. (Lyovin 1978:144)

Rumjancev also says that intonational factors have a
very significant effect on various acoustic parameters of

neutral tone syllables (Lyovin 1978:145). In some cases the

18
pitch level of neutral tone (e.g., sentence-final particles)
is determined by intonation, not the context of the
preceding lexical tone.
Rumjancev examined the acoustic features which signal
the contrast between interrogative and declarative
intonation:

According to Rumjancev's findings, the
primary differentiation of interrogative and
declarative intonation is carried out by means of
pitch features....The basic contour features of
lexical tones are not significantly deformed by
intonation: primarily it is the register features,
the steepness of the rise and fall of F0, that
signal the intonational contrast. (Lyovin
1978:148)

However, the most important intonational differences
are in the last word (Lyovin 1978:149):

To be sure, intonational differences are
found as a rule in all parts of the sentence from
its beginning to its end (though the acoustic
parameters involved may differ from section to
section). However, the critical intonational
differences are localized only in the last word.
Elsewhere, all intonational differences are
secondary and redundant. This is the conclusion
reached by Rumjancev on the basis of experiments
which involved the 'transplanting' of the final
word from a recorded interrogative sentence into
the context of a recorded declarative sentence
from which the final word had been erased and vice
versa: the resulting sentences were judged by the
native auditors to be either declarative or
interrogative solely on the basis of the
transplanted last word in spite of the fact that
intonational cues in the rest of the sentence
clashed with the cues in the last word. Moreover,
such 'prepared' sentences were not felt to be in
the least 'unnatural'.

19

Ho (1976a) investigated the acoustic correlates of the
four Mandarin tones in isolation, and variations in these
parameters in relation to six different sentence
environments: (1) before a question marker ma in a sentence
with rising intonation (X__Q#); (2) at the end of a sentence
with falling intonation (x__#): (3) at the end of a sentence
with falling intonation with a preceding emphatic modifier
(emphX__#); (4) at the end of a clause in a sentence with
falling intonation (X__,): (5) immediately following a
sentence-initial demonstrative in a sentence with falling
intonation (dem__X), and (6) in the initial position ofpa
sentence with falling intonation (#__X). She found that
sentence environment affects the realization of the tone,
and the influence of sentence environment is greater than
that of the vowel and the initial consonant of the syllable.
However, the four tones maintain their basic shapes.

Ho (1976b) found that in declarative sentences, tone
contours in sentence-initial position are higher on the
frequency scale than those in sentence medial position (at
the end of a clause), and tone contours in sentence-medial
position are higher on the frequency scale than those in
sentence-final position. She also compared the influence of
falling and rising intonations on tone contours in sentence-
final position (actually, in rising intonation it is the
position before the question marker ma): all four tones

start and end in higher frequency in environment X__Q# than

20
in environment x__# except that Tone 3 ends higher in the
latter environment than in the former environment (possibly
due to the occurrence of the following question marker ma).
She also observed the influence of an emphatic modifier or a
demonstrative on tone contours. Again, the four tones
maintain their basic shapes.

Ho (1977) investigated the acoustic correlates of
intonation for declarative, interrogative and exclamatory
expressions in a frame sentence of five words: Zhei ge zi
shi nian . ("This word is pronounced .") Results
showed that interrogative and exclamatory intonation have a
higher fundamental frequency throughout the whole sentence
and also have a rising contour in the sentence-final
position, while declarative intonation has a distinctive
lower fundamental frequency and a falling contour in
sentence-final position. Ho also studied and compared the
influence of intonation and word position on some acoustical
parameters (duration, amplitude, and fundamental frequency)
of individual syllables (Ho 1977:454):

To summarize the findings of this study, the
duration of syllable nuclei in a sentence varies
according to word position rather than intonation,
except that in the sentence-final position. Both
intonation and word position have influenced the
fundamental frequency, tone range, and peak
amplitude of syllable nuclei. As for the four
tones in sentence-final position, the duration,
fundamental frequency and peak amplitude are
modified by intonation. However, the basic
characteristic of the shape of the four tones

remains very close to that of tones in isolated
words.

According to Ho's statistics:

21

...word position has more influence on duration,
fundamental frequency at the beginning and middle
points, peak amplitude and tone range, but
intonation has more influence on fundamental
frequency at the end point. Furthermore, it has
been found that intonation accounted for more
variation on fundamental frequency at the end

(Ho 1977:454)

point in words 5 and 6.

The duration and fundamental frequency of sentence-final

syllables by tone and intonation are shown in Figure 1 from

Ho (1977:453).

 

 

-——4m«mwflnimnflu1
-—-menmemMHfin
~—-Jhwmwakmndut
“meniscufim
Ema
I 2 3 L
Mmm{::;u- ,// // §f\\‘
ZNAW a“ "\..‘- ‘
.tn~“ ‘4‘ " \\
i so-
I r I 7 I 7 I I j I 7 T I 1 T T fi—
0102030010203040000040 0030
Dumfiomcsec

Figure l. Tonal variations in sentence final position

(Ho 1977:453)

Shen (1990) studied sentences and other utterances

consisting of syllables of the same tone (e.g., all

syllables in a sentence are Tone 1) spoken in different

prompted contexts. She tried to find the basic intonation

22
patterns of Mandarin and the results from her instrumental
analysis lead to the following conclusion (p.78):

The intonation minimum...consists of three
tunes: Tune I, Tune II, and Tune III. They
function basically as statement (Tune I) and
question (Tunes II and III). The clue to
differentiating them involves their starting
point: Tune II and Tune III start higher than Tune
I regardless of their ending point, which may be
in the high key (Tune II) or the low key (Tune
III). Thus, the intonations are layered: an upper
layer for interrogative and a lower layer for
assertive.

The three tunes are shown in Figure 2 from Shen

(1990:26).

 

Figure 2. Basic tunes of Mandarin (Shen 1990:26)

Like Rumjancev and Ho, Shen found that tone contours
are affected by intonation, especially for Tone 1, which is
transformed from level into a slope. However, "[r]egardless
of these modifications, the basic features of the four tones

are left intact. The perception test carried out by Tseng

23
(1981) revealed that the shallow slope of Tone 1 singled out
by careful instrumental analysis is still perceived by
listeners as a level tone" (p.38).

Tseng (1981), on the other hand, has quite different
findings analyzing data from spontaneous speech. She found
that only 36.13% of the tones in her spontaneous speech data
correspond to tonal contours predicted by phonological
rules. "[TJhe predictions were made on the basis of word
and phrase levels only, i.e., on the rules that would
predict word stress, modification of neutral tone suffixes,
as well as neutralization of certain tones" (p.67). With a
further break-down of the syllables into unreduced, reduced
and neutral tone according to phonological prediction, Tseng
reported the following percentages of correlation between

actual and predicted tonal contours (p.73):

Table 1. Correlation between actual and predicted tonal
contours

phonological prediction percentage of correlation

syllable whose lexical tones 40.17%
should be present in production

syllables that are 29.55%
neutral-tone suffixes

syllables whose lexical tones 16.67%
are reduced to neutral tone

The low correlation is in sharp contrast to the
findings of Rumjancev, Ho and Shen, who found that tonal

contours are not greatly affected in connected speech. This

24
suggests that read or monitored speech is quite different
from spontaneous natural speech in respect to tone
production. In spontaneous running speech, "the speakers'
production is the consequence of a complicated interaction
among lower level phonetics, phonology, and higher levels,
syntax, semantics, and moreover their knowledge of the
language as well as intention toward such an activity.
...This can be one interpretation of the low phonological-
phonetic correlation of the production study" (Tseng 1981:
103).

Nevertheless, in accordance with Chao's observation of
normal intonation in sentences of three or four syllables,
Tseng found "...cases where the sentence is rather short in
duration, about 700-900 msec. The lexical tones are clearly
preduced, leaving no trace of intonation....In this case,
each lexical tone is clearly identifiable..." (p.162).

Wu (1982) reached the following conclusions concerning
Mandarin intonation, based on his acoustic study of
intonation in (prepared) statements, questions and emphatic
sentences, in both normal and accelerated speed:

(1) Statements tend to have falling contour at
sentence-final position.

(2) Questions without the question particle can have
rising intonation at the end: questions with the question

particle usually have the same ending as statements.

25

(3) In emphatic sentences, generally the pitch range is
raised or the duration lengthened. Questions can also be
emphasized and modified the same way.

(4) In accelerated statements, the pitch ranges of
"non-critical" morphemes or words tend to be narrowed or
even leveled. In very fast speech, these morphemes or words
become the transition between the preceding and following
tones and lose their original tone contours.

Besides doing acoustic analyses on the production of
tone in connected speech, some researchers have also
performed experiments on the perception of tone extracted
from connected speech. Rumjancev (Lyovin 1978) has found
that there exists a ”duration threshold" for tone
perception: when syllables shorter than 100-120 ms are taken
out of the context of connected speech, the tone
characteristics are leveled or neutralized and native
listeners have difficulties identifying the tone correctly.
According to Lyovin, Rumjancev found that

...in syllables whose tone contours are fully

manifested, the said contours are sufficient for

correct discrimination among the tones on the part

of the auditors, and the latter do not need to pay

much attention to register, which is a highly

relative component of tone (since it is dependent

on the individual speaker's voice). On the other

hand, in syllables whose duration is close to

minimal (as in allegro speech), contour features

tend to be neutralized (leveled), and tones

therefore are recognized mostly on the basis of

register features, especially the register of

their starting point. In such cases, difficulties

arise since two...tones begin in the high register

(Tone 1 and Tone 4) and the neutral tone may begin
in any register (depending on the tone of the

26

preceding syllable). This means that out of

context these three tones tend to be confused by

the auditors. (Lyovin 1978:136-137)

However, Rumjancev's results indicate a high percentage
of correct identification of tones out of context (Lyovin
1978:140):

Tone perception experiments involving syllables

removed from their original contexts in connected

speech indicate that even in the case of allegro
speech the percentage of correct tone

identification solely on the basis of the acoustic

properties of tone is still relatively high (about

65%). In normal tempo this percentage, is, of

course, significantly higher (80-85%).

On the other hand, Tseng's (1981) perception test
showed that selected syllables taken out of the context of
spontaneous running speech are correctly identified only
40.5% of the time by native speakers. For unreduced tokens,
55.66% were correctly identified. For the reduced tokens,
25.4% were correctly identified. If scoring is based on
actual phonetic output, i.e., the tone of a syllable is
determined by its actual fundamental frequency pattern
rather than by the predicted tone according to lexical and
phonological rules, the identification rate was not much
different: 58.58% for unreduced tokens and 24.83% for
reduced tokens. (The correlation of phonological and
phonetic tone was 90% for unreduced tokens, 53% for reduced
tokens). This identification rate is much lower than that

found in Rumjancev's study. The reason for this low

identification rate is not clear. The duration threshold

27
for tone perception found in Rumjancev's study may account
for some of the difficulties since in Tseng's study the
duration of each syllable varies greatly (108-618 ms for
unreduced tokens p.92). The greatly modified pitch register
in long running speech may be another reason for the low
identification out of the context. The fact that the
listeners were given six choices, including neutral tone and
nothing recognizable in addition to the four lexical tones,
may also have affected the results. However, data provided
by Tseng were not sufficient for us to make a firm
conclusion.

Leather (1983) studied speaker normalization in the
perception of lexical tone. Two sets of synthetic Mandarin
tone stimuli based on the syllable /y/ (lip-rounded high-
front vowel) were created. They were then embedded in
natural speech utterances of two native speakers with
different but overlapping voice ranges, and presented to
native listeners for lexical tone identification. All
synthetic stimuli in the two sets had the sameduration,
formant frequencies, amplitudes and bandwidths, but the two
sets differed in being adapted to the vowels of the two
speakers. Two stimuli in each set had "good" Tone 1 and
Tone 2 F0 contours, other stimuli had starting F0 or rate of
rise between "good" Tone 1 and Tone 2, and four of them had
F0 contours identical to both sets. While individual

listeners differed somewhat in the acoustic criteria by

28
which they apparently decided lexical identity (Tone 1 /y/
"muddy" or Tone 2 /y/ ”fish"), all listeners significantly
assigned "ambiguous" stimuli with identical absolute F0
contours to different lexical categories depending on which
of the speakers was heard to ”produce” them. Leather
concluded that the experiment provided support for the
supposition that inferred F0 range of a speaker is a
determining factor in the categorization of lexical tones.
If the pitch range of a speaker can be significantly
affected in spontaneous running speech, then listeners
likewise need to "normalize" the pitch range in that context
to perceive the tone correctly, especially when the contour
features are not clear.

Connell, et al. (1983) conducted an experiment designed
to determine how far the shape of a tone in Mandarin Chinese
can be changed before it is consistently recognized as a
different tone. The end point of pitch contour of naturally
produced monosyllabic words was raised or lowered to
simulate the effect of intonation on sentence-final
syllables. These were then presented to a group of native
Chinese speakers for tone identification. Results indicate
changes in perception of tone. These changes correspond
generally with their expectations based primarily on pitch
contours. For example:

[T]one 1 lowered by a slope of approximately 40 Hz

or more is largely recognized as tone 4. On the

other hand, when raised even by as much as 80 Hz,
tone 1 was only very rarely recognized as tone

29

2...[which] lends some credence to Rumjancev's

(1972) belief that the consonant to vowel

transition contains sufficient information for

tone recognition, since this portion of the

stimuli was least affected by our pitch

modifications. (Connell et al. 1983:349-350)
The recognition of tone appears to be stable over a wide
range of pitch contour changes. The authors concluded
(Connell et al. 1983:350):

It is apparent that lexical tones...can withstand

a large degree of perturbation before the risk of

ambiguity arises. This range of stability is

crucial, as it permits intonation to function in

the same manner in Chinese as it does in non-tonal

languages. Were the range of stability smaller,

the use of pitch as the main cue for indicating

such linguistic functions as interrogative and

indicative, and paralinguistic functions such as

attitude and emotion, would be severely

restricted.

Garding, et al. (1986) showed that the production of
Tone 4 (mai "sell") and Tone 3 (m5! "buy"), both falling in
non-final position, are realized in different ways by
different speakers in different speech acts, but some
features are constant. Tone 3 generally has a low pitch
level throughout the second half of the vowel and Tone 4
with a gradual fall over the main part of the vocalic
segment. These observations were tested in a series of
manipulations of pitch movements over mai from Tone 4 to
Tone 3 in the sentence Song Yan mai niurbu ("Song Yan
sells/buys beef"). The manipulated sentences were presented

in a test, in which listeners were asked if they heard mai

("buy") or mai ("sell"). The result confirmed the observed

30
constant features. In addition, when the pitch contours
lacked a clear reference (e.g., had conflicting cues or had
no distinctive tonal cues at all) the identification by
native speakers did not show a clear preference for Tone 3
or Tone 4.

Although the multiple use of pitch in Mandarin speech
has clear effects on the production and perception of
Mandarin tones, especially in spontaneous running speech, it
can be concluded that in stressed positions tonal features
generally are clearly produced in connected speech.
Although the identification by native speakers of tone
extracted from connected speech is far from perfect, the
degree of correct identification is far above chance, even

in fast reading and spontaneous speech.

2.2 Word stress and sentence intonation in English

In non-tone languages such as English, pitch is not an
integral part of words or morphemes, yet all utterances
still must be spoken with pitch. "Every sentence, every
word, every syllable, is given some pitch when it is spoken.
Even a sound in isolation is produced by vibrations whose
frequencies constitute its pitch” (K. Pike 1972:53).
Although English is probably the most studied of all
languages, its use of pitch and other related prosodic
features is still far from being well understood. There are

different theories, models and hypotheses in the literature;

31
none is agreed upon by most researchers. The purpose of
this section is not to give a full description of English
prosody but to give a brief account of what has been found
about English prosody which is relevant to this study,
namely what is relevant to the use of pitch in words and

sentences.

2.2.1 English word stress

English is often described as a stress-timed language.
Stress plays an important role in English rhythm: "a general
rule of English rhythm is that we take an equal amount of
time from one stressed syllable to the next" (Cruttenden
1986:24). Word-stress is defined by Lehiste (1970:150) in
relation to sentence stress and intonation. "Word-level
stress is the capacity of a syllable within a word to
receive sentence stress when the word is realized as part of
the sentence."

Word-stress in English is not predictable in a simple
way as in some other languages (e.g., French) which have
word-stress regularly in a certain position on almost all
words. Nevertheless, word-stress in a majority of English
words can be predicted by a set of complex morphological and
phonological rules (Chomsky and Balls 1968). For our
purpose, these rules of English word-stress will not be

described. What concerns us here are the acoustic

32
parameters used by native English speakers in the production
and perception of word-stress.

As we will see below, in acoustic studies investigating
the relative importance of pitch (fundamental frequency F0),
loudness (intensity) and duration in the perception and
production of English word-stress, pitch has been found to
be the primary correlate while loudness and duration are
only secondary in the perception and production of stress.

Lieberman (1960) studied the acoustic correlates of
word-stress in American English. Twenty-five verb-noun
pairs (e.g., CONflict-conFLICT) were recorded by 16 speakers
of American English. The test words were presented in
context in lists of sentences but subjects read only the
test words aloud. Only utterances whose stressed syllables
can be clearly perceived were analyzed. He found that the
stressed syllable had a higher fundamental frequency than
the unstressed syllable of the same utterance in 90% of the
cases, a higher peak envelope amplitude in 87% cases, and a
longer duration in 66% instances.

Fry (1958), in his experiments in the perception of
stress, tested three physical dimensions--intensity,
duration, and fundamental frequency--as cues for the
perception of stress. Synthesized English noun-verb pairs
(e.g., SUBject-subJECT) varying systematically in certain
dimensions were judged by listeners. Duration ratio was

found to be more important than intensity ratio in the

33

judgement of stress, and the fundamental frequency cue may
outweigh the duration cue. Moreover, change in fundamental
frequency differs from change of duration and intensity in
that it tends to produce an all-or-none effect. "...so long
as the resulting pitch change is easily perceptible to the
listener, he tends to judge a higher syllable as more
stressed, but the magnitude of the pitch change makes little
contribution to his judgement" (p.144).

Morton and Jassem (1965) conducted another study on the
perception of word stress:

Morton and Jassem (1965) used synthetic nonsense

syllables of the form /sisi/, /3239/ and /sasa/

...The stimuli were expressly constructed to have

the same vowels in both syllables in order to

avoid the problem of different intrinsic

intensities. English listeners showed a high

degree of consistency in their responses.

Variation in fundamental frequency produced far

greater effects than variations in either

intensity or duration: a syllable was marked

stressed if it was different from the "context"

fundamental, and a raised fundamental was more

efficient than a lowered one. In general, the more

intense and longer syllables were more likely to

be marked as stressed. (Lehiste 1970:128)
2.2.2 English intonation

In intonation languages such as English, intonation
"involves the occurrence of recurring pitch patterns, each
of which is used with a set of relatively consistent
meanings, either on single words or on groups of words of
varying length" (Cruttenden 1986:9).

One important difference between Mandarin tones and

English intonation is that English intonation can occur with

34

utterances of varying length. As stated by K. Pike
(1972:59):

An intonation contour is not limited to specific

syllables or words, but may be spread out over as

many syllables and words as are colored by the

speaker's attitude. For example, an intonation

contour which begins low and rises slowly could be

spread over three syllables, as in (He's) doing

it? or the entire rising contour may occur on a

single syllable such as Tom?
When pitch patterns are associated with more syllables than
contour points (those points in the contour crucial to the
establishment of its characteristic rises and falls),
"[tJhese extra syllables can be pronounced with intermediate
pitches..., or with considerable variation in the amount of
drop from syllable to syllable" (K. Pike 1972:62). Thus the
same contour and the same words can be associated in
different ways. The following example shows how the same

pitch contour can be realized differently on the same words

(From K. Pike 1972:63):

telephone number!

2 3 + 3 4 + 4 (2 - high,
2 3 3 3 4 3 = mid,
2 3 4 4 4 4 = 16w.)

2 4 4 4 4

Another difference between tone and intonation is in

the way meaning is associated with them. Tone is an

35
integral part of a lexical item (morpheme or word) which has
lexical meaning, while intonation meaning is quite
different.

English words have basic, intrinsic meanings;

these lexical meanings are the ones found in the

dictionary.... The intonation meaning is quite the

opposite. Rather than being a stable inherent part

of words, it is a temporary addition to their

basic form and meaning....Rather than contributing

to the intrinsic meaning of a word, it is merely a

shade of meaning added to or superimposed upon

that intrinsic lexical meaning, according to the

attitude of the speaker. (K. Pike 1972:55)

One assumption behind the study of intonation is that
pitch variation in intonation is used in a consistent way.
As K. Pike stated (1972:53): "The changes of pitch which
occur within a sentence are not haphazard variation. The
pattern of variation, the rules of change, are highly
organized." However, the perception of the pitch patterns
in English intonation by English speakers is not distinctive
at the lexical level as is the perception of tone by
Mandarin speakers. As found in Lieberman's study (1965),
even linguists familiar with the same (Trager-Smith)
transcription system do not agree on the assignment of pitch
levels in their transcription of recorded English sentences
60 percent of the time.

To describe the rules of intonation, the first step is
to find the basic unit of intonation. However, there is much

disagreement among researchers about this task.

...the lack of consensus on what constitutes the
basic unit of intonation has led to a plethora of

36

competing terms: breath group, phonemic clause,

tone group, tone unit, tune, intonation group,

intonation contour, intonation chunk, etc. Whether

any or all of these refer to the same thing,

however, is a matter for careful consideration.

Depending on what defining criteria are set forth,

quite different units may result. (Couper-Kuhlen

1986:76)

Nevertheless, despite the different defining criteria, the
basic units of intonation in the majority of cases can be
found because the native speakers tend to produce them
consistently. For example, Crystal (1972) found that "when
native speakers were presented with the task of repeating an
utterance, there was maximum agreement (84.8 per cent) over
the location of tone-unit boundaries" (Crystal 1972:110-
111) .

Once utterances are divided into basic intonation
units, the internal structure of the basic intonation unit
is most commonly viewed as consisting of one obligatory
constituent, the nucleus, which is the most prominent
syllable in an intonation unit, and a number of optional
constituents. There are different views about how the
optional constituents should be divided or described. One
way is to describe the head (the pattern beginning on the
first stressed syllable), the pre-head (syllables before the
head) and the tail (syllables following the nuclear
syllable) (Crystal 1972). Thus the structure of the basic

intonation unit can be represented as follows:

(pre-head) (head) nucleus (tail)

37
The following example has all the possible constituents in a

intonation unit (White 1980:11 adapted from Lindstrom 1978):

rP1 run-"1 rN'I r ---------- '1' --------- '1
She won't be 'a; that school much longer.

(' = mid-fall)

In describing the pitch patterns of English intonation,
one way is to describe 'whole tunes', the contours produced
by joining all the accents together. However, this approach
is rejected by some researchers ”because it misses important
generalisations dependent on the occurrence of similar tones
starting from the nucleus" (Cruttenden 1986:56).

Since it is generally agreed that the pitch pattern
starting from the nucleus (often termed nuclear tone)
carries the most important intonation meaning (Cruttenden
1986, Crystal 1972, Lindstrom 1978), most intonation
analyses of English have been concerned with describing it.
The optional pre-nuclear constituents (head and prehead)
"only modify the meaning associated with the nuclear tone"
(Cruttenden 1986:63) and are less studied and less
understood.

Although most researchers agree that nuclear pitch
patterns carry major differences of intonation meaning, they
disagree in their description of nuclear pitch patterns.

This is an area where almost every analyst varies

in his judgement of what constitutes a 'major

difference of meaning' and hence in the number of

nuclear tones which are set up. Nor are the
arguments for any one particular set of nuclear

38

tones ever very convincing or indeed explicit: in

fact, given that intonational meanings are often

intangible and nebulous, it is difficult to see

how a wholly convincing case for any one set of

nuclear tones can be made. (Cruttenden 1986:58)

For example, according to Cruttenden (1986:58f), seven
nuclear pitch patterns can be set up for English: high-fall,
low-fall, rise-fall, low-rise, high-rise, fall-rise, and
mid-level. All the seven nuclear pitch patterns are shown to
carry some different meanings. For instance, ”The low-fall
is generally more uninterested, unexcited, and dispassionate
whereas the high-fall is more interested, more excited, more
involved..." (Cruttenden 1986:100). However, these nuclear
pitch patterns can be grouped into falls and rises and be
ascribed abstract meanings: "high-fall, low-fall, rise-fall
can be grouped semantically as falls, which involve a sense
of finality, completeness, definiteness and separateness,
particularly when used with declaratives" (Cruttenden
1986:100). On the other hand, low-rise, high-rise, fall-
rise, mid-level can be grouped as rises, and "...the use of
a rising tone signals dependency or non-finality”
(Cruttenden 1986:102).

A major reason that the meanings of intonation are so
difficult to grasp is that they vary in different contexts.
For example, Cutler (1977) provided some convincing evidence
that the so-called contradiction contour (an initial rise

followed by a fall followed by a final rise of some sort)

proposed by Liberman and Sag (1974) can be used to express

39

quite different meanings such as enlightenment or
impatience, and that there are alternative contours for
signaling contradiction (Scott 1987:99). The following are
some potential context effects identified by Cruttenden
(1986:114):

non-final v. final: declarative v. yes/no

interrogative v. wh interrogative v. imperative v.

exclamative: tense: agreement v. disagreement:

relationship to preceding tone: and general

context, including preceding utterances and

physical situation. Lurking behind many of these

particular factors is the more general one of

speaker/listener relationship.

From the literature cited above, we can see that
disagreement among analysts exists on virtually every aspect
of English intonation. This is very different from the

analysis of Mandarin tones, about which almost everyone

agrees.

2.3 Comparison of the two systems

It has been suggested that intonation is a universal
phenomenon and that its correlates may be relatedmore to
innate biological factors than to conventional grammatical
factors. Bolinger (1978:471) surveyed languages of the
world and found intonation universals:

The traits of intonation shared by the majority of
languages, not excepting tone languages, are both
formal and semantic, and cover the two main non-
tonal uses of pitch variation: to form closures
(descending lines, clause-final falls and non-
falls) and to form accents (obtrusions for
prominence, mainly upward). Terminals are almost
universally low or falling for finality and

40

assertion, and high or rising for the opposite,

including yes-no questions. Accents are generally

set off by contrasting pitch levels, and their

position in the sentence indicates both focus and

climax.
In section 2.1 above, it is shown that Mandarin also
exhibits these intonational universals. However, the
Mandarin lexical tones are usually modified without changing
their tonal characteristics. For example, a level tone may
be transformed into a shallow slope, and the slope of a
rising or falling tone may become steeper or shallower.

The most important difference between Mandarin and
English is that pitch is distinctive primarily at the
syllable (morpheme) level in Mandarin while in English the
pitch functions on the word level or above. This difference
(lack of lexical tone in English) is related to other
differences in the use of pitch in the two languages.

For example, the pitch patterns associated with English
word-stress are more flexible for native English speakers
and are more susceptible to the influence of intonation
(Chao 1956). On the other hand, the tonal patterns in
Mandarin words are fixed and generally cannot be robbed of
their basic characteristics by intonation. Thus, as Chao
(1956:42) noted, in the speech of Chinese learners of
English the pitch patterns associated with word-stress in
isolated English words tend to be maintained in phrases and

sentences.

41
Affective or attitudinal functions which in English are
signaled by local pitch change are not signaled in the same
way in Mandarin. A study by Ross, et al. (1986) reported on
the influence of tone on the ability to signal affect
through prosody in three tone languages (Taiwanese, Mandarin

and Thai) as compared to English.

The results...clearly indicate that the presence
of tone in a language adversely impacts the free
use of F0 for affective-prosodic signalling. Of
the five acoustical parameters characterizing the
F0 contour, three...show statistically that
English singularly surpassed all the tone
languages in the study by at least a factor of 2
...Those acoustical parameters in which English
singularly surpassed the others, directly or
indirectly, indicate that English has greater
freedom than tone languages in locally varying the
intonational contour for affective signalling.

The results involving timing and intensity
measurements showed no significant influence of
language type (tone vs. non-tone) on these
parameters as opposed to certain F0 parameters.
Thus, intensity and timing parameters may be used
freely in a tone language for affective-prosodic
purposes. One might even be tempted to assume that
intensity and timing parameters could be used in
tone languages to compensate for tone imposed
constraints on affective intonation. However, our
data...does not indicate that tone languages use
such compensation, since English statistically
performed as well as the best of the tone
languages in all intensity and timing parameters.
(Ross, et al. 1986:298-301)

Another difference between Mandarin and English is that
the pitch range used in the two languages is different.

G-T. Chen (1974) investigated the pitch range of English and

Chinese speakers. His study included two types of speech

42
utterances: words and sentences. He reached the following

conclusions (p.170):

1) When the test subjects spoke their native
languages, the average pitch range of the four
Chinese subjects was 258% wider than that of the
four English-speaking subjects established by
UTTERANCE WORD and the average pitch range of the
four Chinese subjects was 154% wider than that of
the four English-speaking subjects established by
UTTERANCE SENTENCE.

2) When the four English-speaking subjects
switched from their native language to Chinese,
their pitch range increased by 123% when UTTERANCE
WORDS were used and by 62% when UTTERANCE
SENTENCES were used.

G-T. Chen (1974:169-170) also found that the pitch ranges of
Chinese speakers increased only 4.4% from utterance words to
utterance sentences. On the other hand, the pitch ranges of
English speakers increased 47% from utterance words to
utterance sentences when they speak English.

A contrastive analysis of Mandarin tone and English
intonation by White (1980, 1981) compared the pitch patterns
occurring in Mandarin tones and English intonation:

With what is known about the relative pitch

levels and contours of Mandarin tone, and through

extrapolation from Chen's (1974) data, we can

compare the pitch levels and contours of Mandarin

and English tones....Mandarin has a high level, a

rising, a low falling-rising, and a falling tone;

English has at least level, falling, rising,

falling-rising, and rising-falling tones. (White

1980:23)

Only the English pitch patterns which are similar to
the Mandarin tones described in section 2.1, will be

described below:

43

The h1gn_1eye1_tgng occupies a high or medium
voice pitch and is not normally pitched lower than
any preceding unstressed syllables. High level
tones are used to give secondary prominence to
words in the head of the tone-group, where they
usually form a gradually descending sequence.
(Lindstrom 1978:78)

The rise is frequent as a nuclear tone. It is
usually independent but may, in a compound tonic,
be subordinate to a preceding nuclear tone. In
prenuclear position it may serve to build up a
rising head.

The rise begins on a low or medium pitch and
reaches a medium or high pitch. Its range varies a
good deal. Wide rises normally start from a low
pitch. A "broken" variant which occurs only as a
nuclear tone, may begin on a higher pitch and is
realised by a falling-rising pitch movement.

(pp.79-80)

The {a11;11§g is always independent and
nuclear. It falls from a high or medium pitch to a
lower pitch and then rises to a medium or high
pitch. In non-final position in the utterance it
tends to be high and narrow. In final position its
range varies. The rising part of the tone does not
normally end higher than the onset of the
preceding falling element. (p.80)

The fall may begin in any voice pitch region

and ends up on a lower pitch. When it forms the

nuclear tone of the tone-group, its range varies

with the pitch of the nucleus. In the head it is

usually high and narrow. Postnuclear falls are

always low and narrow. (p. 79)

White showed that many tonal errors produced by
learners of Mandarin can be attributed to the transfer of
English intonation in certain sentence types or in
particular sentences. For example, the intonation on simple
declarative sentences typically begins on a low, unstressed

sYllable, builds to the nuclear (or stressed) syllable, and

then glides down. This pattern may account for the low-tone

 

44
initial syllable and gliding final syllable heard when
English speakers produce Mandarin sentences. Mandarin
V-not-V questions are frequently pronounced with a rising
intonation by English speakers in approximation of the
intonation in the English equivalent (yes/no questions).
Many Mandarin descriptive adjectives or stative verbs which
are not pronounced with a falling tone (e.g., hen ”very")
are often mispronounced by English speakers because emphatic
stress is associated with a high-falling tone in English,
and such words are often stressed in English.

Because the function of pitch in a tone language like
Mandarin and an intonation language like English is
different, pitch ranges used in the two languages are
different, and some specific similar pitch patterns are used
in different ways for different functions in the two
languages, it is not surprising that English speakers

experience difficulties in their learning of Mandarin tones.

3. PERCEPTION AND PRODUCTION OF SONE ASPECTS OP L2 PBONOLOGY

In this chapter, I will first review literature on the
perception and production of some aspects of L2 phonology,
with emphasis on the adult learners: then I will review the
literature on the perception and production of Mandarin

tones by L2 learners.

3.1 Perception and production of L2 phonology by adults

Most studies on L2 phonology have primarily been
concerned with production rather than perception. The major
reason for this is probably that production is easier to
observe. In fact, even a slight accent produced by L2
learners is easily detected by native speakers (Scovel
1988:105ff). However, since the learning of L2 phonology
should include both production and perception, literature on
both aspects will be reviewed.

Studies that examine the variables affecting L2
phonological learning (pronunciation) have always found age
at arrival in the L2 environment the most important variable
(Asher and Garcia 1969, Oyama 1976, Seliger, et al. 1975,
Tahta, et al. 1981a, Piper and Cansin 1988). The younger an
L2 learner starts learning in the L2 environment, the more
likely he/she will achieve native-like pronunciation. The
critical period hypothesis, which "is the notion that

language is best learned during the early years of

45

46
childhood, and that after about the first dozen years of
life, everyone faces certain constrains in the ability to
pick up a new language" (Scovel 1988:2), has been
controversial in many areas of L2 learning, but it is
strongly supported in terms of learning native-like
pronunciation (Scovel 1988).

However, despite the overwhelming evidence of
relationship between age-at-arrival and L2 pronunciation,
research has found that adults are better than children in
learning L2 pronunciation in the short term (Olson and
Samuels 1973). It is also unclear whether the younger-better
effect exists at the very beginning of L2 learning. To
determine whether the younger-better effect of L2
pronunciation attainment exists because younger learners are
better at imitating new sounds, some studies investigated
the abilities of subjects of various ages to imitate foreign
sounds they were not familiar with. Some studies found an
older-better effect (Loewenthal and Bull 1984: Snow and
Hoefnagel-Héhle 1977, laboratory study), while others found
a younger-better effect (Yamada et al. 1980; Tahta et al.
1981b). Snow and Hoefnagel-Hohle (1977, naturalistic study)
reported a positive (older-better) correlation between age
and L2 pronunciation among L2 learners at the beginning of
their L2 learning in the L2 environment: however, this
older-better effect changed over time. Thus, "despite the

general initial advantage of the older subjects, both for

47
total scores and for many individual sounds, the younger
subjects had become better at pronouncing some of the sounds
after a period of 10-11 months' learning" (p.362).

Whether there is a younger-better relation in imitating
foreign sounds or not, and whatever the age-related causes
involved in the acquisition of L2 phonology, it is clear
that there are differences between children and adults in L2
phonological learning, and it is virtually impossible for
adult L2 learners to achieve native-like accent. However,
the pronunciations among adult learners still vary greatly.
The adult learners' learning of L2 phonology, including both

perception and production, will be our main concern below.

3.1.1 Variables affecting L2 phonology learning

On the production aspect of L2 phonological learning,
Purcell and Suter (1980) found that the most effective
predicting variable of L2 pronunciation for adult learners
from four different L1 countries is the first language (L1),
followed by aptitude for oral mimicry, residency in L2
environment, and concern for pronunciation accuracy.

The stepwise results indicate that only four
of the predictors are useful in accounting for the
variance of subjects' pronunciation accuracy
scores. First Language was by far the most
important predictor accounting for more variance
than any other variable. Aptitude for Oral
Mimicry was the second most important predictor.
The third most important predictor was a composite
variable, CV1, the only composite variable to
contribute significantly to the regression. CV1
combined Years in an English Speaking Country and
Months Residing with a Native Speaker of English.

48

We interpret CV1 as reflecting length of residency

in an English speaking environment. The last

predictor to contribute significantly to the

explanation of the criterion's variance was

Strength of Concern for Pronunciation Accuracy.

The remainder of the predictors were deemed

useless, after the above four were taken into

account. (Purcell and Suter 1980: 282-283)

Thompson (1984) investigated Russian immigrants who
came to the United States at various ages and found that the
best four-variable predictor model of pronunciation accuracy
includes (1) age at arrival (0.66), (2) sex (0.77),

(3) aptitude for oral mimicry (0.82), and (4) speaking
proficiency in English (0.85) (p.94).

On the perception aspect of L2 phonological learning,
Thompson (1984) found that the best four-variable predictor
model of L2 sound discrimination includes (1) ear preference
(0.20), (2) percent of English speaking friends (0.26), (3)
importance of English for the job (0.31), and (4) American
orientation (0.34). She noted that the battery of
predictors was far more effective in accounting for
differences in production than it was in explaining the
differences in perception.

In the following sections, however, the variables that
will be discussed are (1) phonological context, (2) task,

(3) sex, and (4) L2 experience. These variables may be more

important when L2 learners are adults of relatively narrow

range of age and are from the same L1 background. These are

49
the variables that will be examined in my analyses of

American learners' learning of Mandarin tones.

3.1.1.1 Phonological context

(A) Production
Phonological context has been found to have an effect

on the pronunciation of L2 sounds (Flege 1988:315):

Several L2 studies have demonstrated that

authenticity of pronunciation varies as a function
of phonetic context. Gatbcnton (1975, discussed in
Segalowitz and Gatbcnton, 1977) observed that the
difficult /d/ sound of English was produced
authentically by French Canadians more often when
it precedes a vowel or fricative than a voiceless
stop. Dickerson (1974, 1975) found that the
perceived authenticity with which Japanese
learners produced English sounds varied as a
function of phonetic context. In both studies, the
effect of phonetic context on authenticity

persisted over time even though pronunciation
improved overall.

L. Dickerson (1975) found that L2 learners produced L2
sounds better in word-list reading than in dialogue reading.
Borden et al. (1983) investigated Korean learners'
production (repeating after native speakers) of /r/ and /l/
in nonsense syllables, words, and sentences and found that
"[n]onsense syllables were judged correct a larger
percentage of the time than were words, and the [r] and [l]

sounds in sentences were judged to be least correct"

(p.509).

50
(B) Perception
The effect of phonological context on the perception of
L2 sounds has been found in studies investigating the
perception of English /r/ and /l/ by Japanese learners:

In examining identification by Japanese learners,
Shimizu and Dansuji (1983) found that English /r/
and /l/ were identified better in the context of
front than back vowels (about 90% versus 80%
correct), and better in isolated words than words
embedded in a carrier phrase. The authors
suggested this was due to a ”more complete"
articulatory specification of sounds in isolated
words, but it is also possible that the demands of
processing a longer stretch of speech makes it
more difficult for listeners to identify new L2
sounds. (Flege 1988:278)

Mochizuki (1981) also presented native
English-produced minimal pairs with /r/ and /l/ to
native Japanese-speaking listeners. Correct
identification averaged only 82% overall, compared
to the near-perfect identification by English-
speaking listeners. However, percent correct
identification was higher for sounds in word-final
(98%) than word-initial position (88%) which, in
turn, was higher than intervocalic (73%) and word-
initial clusters (69%). Sheldon and Strange (1982)
obtained very similar results. (Flege 1988:330)

Clearly, both production and perception of L2 phonology
are likely to be affected by phonological context. The
findings of L. Dickerson (1975) and Borden et al. (1983)
that L2 pronunciation in isolated words was better than that
in longer utterances, and Shimizu and Dansuji's (1983)
finding that /r/ and /l/ were identified better in isolated
words than words embedded in a carrier phrase, are of

special interest since the phonological contexts examined in

this study will be isolated syllables, 2-syllable words, and

51

short sentences. Findings in these studies suggest that
sound perception and production in longer utterances are
more difficult for L2 learners than in isolated words.

However, other factors may sometimes change this. Beebe
(1987) found that Thai speakers produced more errors of
English initial /r/ in word-list reading than in
conversation. A major reason for the low accuracy rates of
/r/ in word-list reading is interference from Thai. Thai
has a sociolinguistic rule of using trilled /r/ in formal
settings. Since word-list reading is more formal than
conversation, there is more negative transfer in the word-
list reading data than in the conversation. In contrast to
initial /r/, final /r/ was pronounced correctly by the Thais
significantly more often in word-list reading than in the

conversation.

3.1.1.2 Task
(A) Production

Speaking task has also been found to influence the
pronunciation of L2 sounds (Flege 1988:259):

One other study showed that L2 pronunciation
varied according to speaking task. L. Dickerson
(1974, 1975) examined the ability of Japanese L2
learners to approximate the phonetic norms of
English for a number of sounds. Pronunciation was
more often--and more nearly--correct in word-list
reading than in dialogue reading, and in dialogue
reading than in conversational speech. Although
the subjects' pronunciation of the sounds examined
improved over the course of one year, the effect
of speaking task remained. W. Dickerson (1976,
1977a, 1977b; L. Dickerson & W. Dickerson, 1977)

52

hypothesized that the variability observed

resulted from the same kind of tacit

sociolinguistic knowledge native speakers use in

L1 to vary systematically pronunciation. He

asserted that the L2 learner's developing

competence is best described by variable

phonological rules specifying the probablistic

effect of speech styles and task, as well as

phonetic context, on the selection of sound

variants.

Because the two production tasks performed by subjects
in this study are imitation and reading aloud, the task
effect of imitation and reading in L2 pronunciation is
examined. However, very little has been done in this area.

E. V. Pike (1959) found that the ability to imitate
foreign sounds correlates with students' final grade (based
on their ability to take dictation, to pronounce sound
sequences, and to understand phonetic theory) in a forty-
five hour course in phonetics. The ability to imitate
foreign sounds in general has also been found to be a
significant predictor of specific L2 pronunciation accuracy
(Purcell and Suter 1980, Thompson 1984). These findings
indicate that the ability to imitate foreign sounds probably
represents some phonological learning aptitude that cannot
be modified easily. Nevertheless, there is evidence that
with practice some improvement in imitation is immediately
possible. Snow and Hoefnagel-Héhle (1977, laboratory study)
found that the imitation of Dutch words by English-speaking
subjects improved between the first ten trials (block 1) and

the second ten trials (block 2). However, not all sounds

53
showed improvement. Some sounds showed significant
improvement (/k/, initial /x/, /o/, /l/ and yei/), while
others (final /f/, /x/ in a cluster, /Ay/, and initial /r/)

showed no improvement.

A. Imitation B. Spontaneous

 

a
'0
0 5' t
a ..- § c‘.".--.~\ i -'.
9 7 K __ ‘3 ; _,e—---O" K
/ ' .N t " . ,/"\I
40 N t. .-— \Y
’,.e
I
g 31’ r'
o (I
u ./
u 20
- e-ettme‘l
e——-stime2
10 “ r-Qtimca
a I
U
I
8 o

 

 

315 6:7 331012-115 «for: 3:: 3:7 films-1s ...-La:

AGE IN YEARS

Figure 3. Median scores of two pronunciation tasks
(Snow and Hoefnagel-Hohel 1977:363)

Snow and Hoefnageléfiohle (1977, naturalistic study)
used both imitation and spontaneous speech (in response to a
picture) in analyzing the pronunciation of L2 learners. They
did not compare the results of the two tasks. Their data
are shown in two graphs in Figure 3 (Snow and Hoefnagel-
Hohle 1977:363), and some observations can be made regarding

the two tasks. The spontaneous scores show significant

54
positive correlation with age (older-better) at time 1. The
imitation scores were better than the spontaneous scores for
all age groups at time 1, but the differences were greater
for younger groups than for adults. At time 2 the
differences between the two tasks became very small as a
result of greater improvement in spontaneous production than
in imitation. At time 3 the scores were almost the same for
the two tasks. Overall, the improvement in imitation was
not as great as in spontaneous speech for all age groups.
In fact, for adult learners, the imitation scores were

almost unchanged from time 1 to time 3.

3.1.1.3 Sex
(A) Production

Sex is a variable which has frequently been examined in
studies of L2 pronunciation. Sex was found to be a
significant variable in L2 pronunciation in some studies
(Asher and Garcia 1969, Thompson 1984) in which females were
found to have better pronunciation than males. However, the
effect seemed to diminish with length of residence (Asher
and Garcia 1969). Purcell and Suter (1980) found sex not a
significant variable in L2 pronunciation. Tahta, et al.
(1981a) found sex contributed significantly to variance in
accent, though its contribution is relatively small (2.2%).

We might note the observation that women were less

likely to show accent transfer than men. This

difference did not emerge in the second multiple
regression analysis: the effect (in the first

55

regression analysis) might have been due to the
subjects in the sample who had begun to learn
Epglish after the age of 13. However, a separate
1 ‘test of this diffe ence did not yield a
significant result (1 :- 2.45 p > 0.05). (Tahta et
al. 1981a:271)

Studies that investigated the abilities to imitate
foreign sounds has found no sex differences, either. Snow
and Hoefnagel-Hohle (1977, laboratory study) found no sex
differences in the abilities to imitate new sounds in
subjects aged five to 31. Tahta et al. (1981b), who looked
at the abilities of 5-15 year old monolingual English
schoolchildren to replicate foreign pronunciation and
intonation, also found no sex differences. They concluded
(Tahta et al. 1981b:370):

We found no reliable sex differences for either
languages at any of the ages studied. Asher and
Garcia (1969) found female superiority among
fluent L2 speakers of one or more years residence,
but sex differences disappeared with length of
stay. Other studies (e.g. Oyama, 1976: Tahta et
al., 1980, 1981) restricted observation to L2
speakers with greater length of stay and failed to
report reliable sex differences. We must conclude
that if there are sex differences among L2
speakers, these disappear with practice and
exposure, and our findings suggest they are not
attributable to reliable intrinsic differences
between the sexes in abilities to replicate novel
foreign language sounds.

(B) Perception
Few studies investigated the effect of sex on L2

phonological perception. Thompson (1984) found sex not

significant in the perception of L2 sounds by L2 learners.

56

3.1.1.4 L2 experience
(A) Production

It is reasonable to assume that previous L2 learning
experience will have some influence on subsequent language
learning. Some people believe that learning a second
language makes the learning of subsequent languages much
easier (see Larsen-Freeman and Long 1991:206). Although the
statement is usually made regarding the learning of a
language as a whole rather than the learning of L2 phonology
specifically, if prior L2 learning experience helps, it
could also help in phonology learning. A few studies on L2
pronunciation learning considered prior other L2 learning
experience. Suter (1976) found ”number of languages the
speaker is able to converse in" not a significant predictor
of pronunciation accuracy. Thompson (1984) investigated
variables predicting L2 pronunciation and also found

knowledge of other languages insignificant.

(B) Perception

Few studies investigated the effect of prior L2
experience on the perception of L2 phonology. Thompson
(1984) found knowledge of other L2s insignificant in the

perception of L2 sounds.

57
3.1.2 Relationship between the perception and production of
L2 sounds in isolation and in sequence

Sounds and suprasegmentals are produced and perceived
as part of words and utterances in speech. Although one word
utterances do occur in natural speech, words are usually
spoken in sequence rather than in isolation. Learning a
language means learning to produce and perceive natural,
fluent utterances.

Since utterances are sequences of words and adult L2
learners have this knowledge from their L1, it is assumed
that adult L2 learners normally learn individual words
first, before they can put words together in sentences or be
able to recognize and understand them in utterances.
(Unanalyzed chunks or routines can be viewed as long words.)
However, as we will see below, the relationship between the
perception of sounds in isolated words and connected speech,
and the relationship between the production of sounds in
isolated words and connected speech may not be a direct and
simple one because they involve some different factors.
Similar relationship may exist in both perception and
production between tone in isolated syllables and in

sequences which will be analyzed in this study.

3.1.2.1 Perception
Acoustic studies have found that the acoustic signals

of connected speech are continuous, with no clear boundaries

58
between words, and that sounds produced in different
phonetic contexts exhibit great variation. The lack of
acoustic-phonetic invariance is one of the most important
problems in speech perception:

For more than 30 years, it has been extremely
difficult to identify acoustic segments and
features that uniquely match the perceived
phonemes independently of the surrounding context.
As a result of coarticulation in speech
production, there is typically a great deal of
contextual variability in the acoustic signal
correlated with any single phoneme. Often a single
acoustic segment contains information about
several neighboring linguistic segments (i.e.,
parallel transmission), and, conversely, the same
linguistic segment is often represented
acoustically in quite different ways depending on
the surrounding phonetic context, the rate of
speaking and the talker (i.e., context-conditioned
variation). In addition, the acoustic
characteristics of individual speech sounds and
words exhibit even greater variabilities in fluent
speech because of the influence of the surrounding
context than when speech sounds are produced in
isolation. (Pisoni and Luce 1986:3-4)

The acoustic-phonetic invariance is a problem for the theory
of speech perception because listeners perceive the sound
variants as the same. In addition to the problems arising
from the lack of acoustic-phonetic invariance, there are
also related problems having to do with the normalization of
the speech signal. That is, despite great variations in
acoustic signals produced by different speakers or different
rates of speaking, listeners are able to normalize the

differences and perceive sounds produced by different

speakers or at different rates to be the same. The

59
normalization problem of speaking rate is relevant to the
perception of isolated words and fluent speech:

A second normalization problem concerns time and
rate normalization in speech perception. It is
well known that the duration of individual
segments is influenced substantially by speaking
rate. However, the acoustic duration of segments
is also affected by the location of various
syntactic boundaries in connected speech, by
syllabic stress, and by the component features of
adjacent segments in words....In addition, there
are substantial differences in the duration of
segments of words when produced in sentence
contexts compared to the same words spoken in
isolation. The duration of vowels produced in
sentences is roughly half the duration of the same
vowels spoken in isolation. Speaking rate also
influences the duration and acoustic correlates of
various phonetic features and segments. Numerous
low-level phonetic and phonological effects such
as vowel reduction, deletion, and various types of
assimilation phenomena have been well documented
in the literature. These effects seem to be
influenced a great deal by speaking tempo,
dialect, and surrounding phonetic context. (Pisoni
and Luce 1986:11)

Perception of connected speech differs from perception
of isolated words because it involves segmentation or
locating word boundaries. Segmentation and recognition of
words in connected speech is a complicated process that
involves the interaction of sound perception and much other
knowledge. Some evidence for the contribution of other
knowledge was found in research on the perception of speech
in noise (e.g. Miller et al. 1953), which found that words

presented in noise are recognized more accurately in a

sentence context than in isolation (even though words spoken

60
in isolation are acoustically better defined than words in
fluent speech).

However, acoustic information, especially when it is
clearly defined, is still an important basis of speech
perception. One piece of evidence for this view is that
mispronunciations can be detected by listeners. Cole's
(1981) study on children's detection of mispronunciation
showed that sounds in fluent speech and isolated words are
not perceived equally well.

The results showed that children detect

mispronunciations in isolated words much more

accurately than in fluent speech. Across all

groups, mispronunciations of words spoken in

isolation were detected 95% of the time, compared

to 59% detection of mispronunciations in nursery

rhymes. (Cole 1981:101)

Thus the sounds children have learned to perceive in
isolated words have not been learned to the same extent in
fluent speech.

Cole (1981) has also found that error detection
performance of children was affected by the type of phonetic
substitutions. The most detectable mispronunciation
involved the substitution of one stop for another or the
substitution of a nasal for a stop. By contrast,
substitution among nasals and fricatives were detected 20%
or 30% less often than were substitutions among stops. On
the other hand, adults (college students) detected all word-

initial mispronunciations with at least 90% accuracy. Thus,

all of the phonetic substitutions are highly detectable by

61
adult standards. Some researchers have found similar
processes in adult L2 acquisition and L1 acquisition by
children (e.g. Krashen 1982). If the difference in
detecting mispronunciation between adult and children
indicates that children are still in the process of learning
the language, we may expect adult L2 learners to behave more
like children in their learning of L2 phonology.

Goldstein (1983) studied word recognition in foreign
speech on the basis of acoustic signals alone. English-
speaking subjects not familiar with French were trained to
remember the pronunciation of four French words. They were
then asked to detect the four French words in a prose
passage read in French by a native speaker. Nouns were
deleted from the original French passage about every four to
five sentences. Each test word was in turn substituted into
the passage. Mean detection rates for the four words ranged
from 64% to 94%. There were significantly more false alarms
in detecting the 1-syllable word taux than scan: a result
may be due to a greater probability of words or parts of
words that sounded like taux. Thus in contrast to native
listeners, foreign language listeners appear to be greatly
affected by the degree to which target words resemble other
words and syllables phonetically. This is more likely to
happen in listening to connected speech since there are more
"other words" and the available processing time is likely to

be shorter. A fluent French speaker listening to the same

62
French passage made errorless and much more rapid detection

with only a few false alarms.

3.1.2.2 Production

It was mentioned that acoustic signals produced in
connected speech and in isolated words have been found to be
different in many respects. For instance, the average
duration of segments is shorter in sentences than in
isolated words, and the effects of vowel reduction and
assimilation are greater (Pisoni and Luce 1986:11, see also
Lyovin 1978 for similar average duration of segments in
Chines).

The difference between production in connected speech
and in isolated words results mainly from mechanical
constraints on the movement of articulatory organs during
speech production. In speech production, the shape of the
vocal tract for any given sound often accommodates to the
shape needed for surrounding sounds, a phenomenon called
coarticulation. The result of coarticulation is referred to
as the undershooting of targets. In producing isolated
words, since a pause precedes and follows the words, no
coarticulation occurs at word boundaries. On the other
hand, connected speech is produced at a faster rate, which
is also achieved by undershooting articulatory targets, and

coarticulation at word boundaries is normal.

63

One of the differences found in the production of
isolated words and connected speech is VOT (Voice Onset
Time) in step consonants. Languages have been shown to use
VOT in perception and production of voiced and voiceless
stops (Lisker and Abramson 1964: Williams 1979, 1980). VOT
refers to the relative timing of voicing onset with
reference to the release of articulatory closure. By
convention, negative VOT values stand for voicing lead
(onset of glottal vibration prior to articulatory release),
and positive values, voicing lag (onset of glottal vibration
following release) (Williams 1980:187). Locke (1983:101)
cites Baran, Laufer, and Daniloff (1977) who:

...measured the VOTs of voiced and voiceless stops

in three adults while speaking conversationally,

reading a passage, and reading single-word

citation forms. They found VOT values to be most

discrete for citation forms, where there was no

overlap between voicing categories (essentially
confirming Lisker 8 Abramson, 1967). In speaking

and reading, there was less difference between

voiced and voiceless stops: in conversational

speaking there was a "noted trend for increased

rate of utterance and less precise articulation

[p.348]."

Many of the differences found in isolated words and
fluent speech, such as segment duration and VOT, are
probably universal. We would expect to find similar
variations in isolated words and connected speech in the
production of L2 learners.

L2 learners have been found to produce L2 sounds better

in word-list reading than in dialogue reading and

64
conversational speech, and the discrepancy remained over the
course of one year, despite overall improvement in the
sounds examined (L. Dickerson 1975). This indicates a
systematic relation between L2 sounds produced in isolated
words and in connected speech. It seems reasonable to
assume that adult L2 learners when they learn to produce L2
sounds in isolated words, whether their pronunciation is
accurate or not, will be able to produce them in connected
speech in a similar way. If no other factors (e.g.
sociolinguistic rules of /r/ in Thai) are involved, we may
expect that the pronunciation errors made in isolated words
are also likely to be made in connected speech, although the
pronunciation in connected speech may be expected to be less
accurate than in isolated words since more information needs
to be encoded in producing connected speech. When some
other phonological processes that apply only in connected
speech (phonological rules across word boundaries, or higher
level suprasegmentals) are different in L1 and L2, L2
learners are expected to experience even more difficulties

in producing connected speech.

3.1.3 Relationship between perception and production of L2
phonology

The relationship between speech perception and speech
production is complicated and has not been well understood.

Evidence from studies of child language acquisition (e.g.,

65
the fis phenomenon, Berke and Brown (1960)) shows that
perception and production are distinct aspects of language
competence. The difference between speech perception and
speech production can also be found easily in second
language acquisition. Studies on L2 learners' perception
and production of L2 sounds have found that L2 learners are
able to discriminate L2 sounds without the ability to
pronounce them (Thompson 1984), as well as to pronounce L2
sounds acceptably without the ability to distinguish them
correctly (Soto 1971).

However, generally, it has been found that L2
phonological perception and production abilities are related
in L2 learners. Snow and Hoefnagel-Hohle's (1979) study on
the acquisition of Dutch by English-speaking learners in
naturalistic environment found that perception and
production of L2 sounds were highly correlated after one
year of learning, and these skills were separate from the
grammatical abilities.

It is interesting to note how these two components

separate themselves out through time. At the first

test session, pronunciation loaded on the same

factor as several of the grammatical abilities

tests, whereas auditory discrimination loaded on a

different factor. At time 2, the various

grammatical abilities tests showed a relationship

both to vocabulary skills and to receptive skills.

Pronunciation had been separated out as a factor,

but auditory discrimination still clustered with

vocabulary and with some of the grammatical

abilities tests. At this stage of proficiency, the

correlation between pronunciation and auditory

discrimination was still very low, suggesting that

differences in pronunciation ability might have
been the result of motor rather than perceptual

66

deficits. By time 3, pronunciation and auditory

discrimination correlated highly and formed one

factor, suggesting that perceptual ability had

become the limiting factor in pronunciation skill.

(Snow and Hoefnagel-Héhle 1979:158-159)

Schneiderman, et al. (1988) found that discrimination
and production abilities of French phones were significantly
correlated in learners of French (beginning and low-
intermediate levels) before they underwent a discrimination
training program. However, the discrimination and
production abilities no longer correlated significantly
after the training. Goto (1971) and Sheldon and Strange
(1982) both examined the perception and production of
English /r/ and /l/ by Japanese learners and found that good
speakers perceive the contrast better than poor speakers.
Borden et al. (1983) had the same finding with Korean
learners.

Williams (1979) found Spanish-speaking children
learning English as an L2 showed significant changes in both
perception and production of VOT towards the English
monolingual pattern. Flege (1988) reviewed studies on VOT
and concluded:

Perhaps the most interesting result of the

VOT studies reviewed here is that the same L2

learners who approximated L2 phonetic norms for

VOT in production also showed evidence of changing

perception: their phoneme boundaries were

intermediate to those of monolingual native

speakers of L1 and L2 (Caramazza et al., 1973:

Williams, 1977b, 1979, 1980: cf. Albert & Obler,

1978: Obler, 1983). This suggests that production

and perception evolve together. Many researchers
believe that the changes in perception which occur

67

in speech acquisition or learning ”lead”

corresponding changes in production. (Flege

1988:362-363)

Whether changes in perception "lead" corresponding changes
in production in L2 learners is a controversial issue.
Hypotheses concerning phonological perception and production
in children acquiring their native language often presuppose
that perception precedes production (see Sheldon and Strange
1982:243), but the evidence from L2 acquisition is quite
different from that in child language acquisition.

Studies by Goto (1971) and Sheldon and Strange (1982)
found that adult Japanese learners of English were better in
producing identifiable /r/ and /l/ sounds than in perceiving
the contrast. In Sheldon and Strange's (1982) study, almost
all of the /r/ and /l/ sounds produced by five good Japanese
learners were correctly identified by English speaking
listeners. However, the five Japanese learners made an
average of 11% errors on identifying test words produced by
Americans.

Evidence that production precedes perception in L2
acquisition has also been reported in learning foreign stop
and fricative contrasts (Briere 1966: cited in Sheldon and
Strange 1982:254). When this relationship occurs, it has
often been found that the L2 learners had been taught to
pronounce the sounds in question by explicit reference to

articulatory parameters.

68

It is possible that an L2 learner produces L2 sounds
identifiable to native speakers but with non-native
features, and his perception of L2 sounds is also based on
some of the features that are not used by native speakers;
in another words, he may not have the same ”prototype" or
"perceptual target" as native speakers. For example, Goto
(1971:321) noted "It was found by checking the scores that
some Japanese subjects got better scores on the voices of
the Americans with whom they had had close contact.”
Mochizuki (1981:290) found that the Japanese speaker in his
study perceived the /r/-/l/ contrast she produced herself
better than the contrast produced by the American speaker:

The Japanese speaker of Identification Test II

could identify correctly 74% of the time in

Identification Test I of an American speaker, and

her tokens in Identification Test II were

correctly identified 92% of the time by the

American listeners. This result conforms to Goto's

claim that the production of intelligible /r/ and

/l/ is easier than their perception. Yet it is

interesting that this Japanese speaker could

identify her own tokens in Identification Test II

98% of the time, which is higher than the average

identification among the American listeners.
Sheldon and Strange (1982) also found that Japanese speakers
generally can identify the /r/-l/ contrast they produced
themselves better than they can identify the contrast
produced by American speakers.

Mochizuki (1981) also showed that the production of /r/
and /l/ by a Japanese speaker in different word positions is

not equally well identified by American listeners: likewise,

69
Japanese learners' identification of /r/ and /1/ produced by
an American speaker in different word positions is not
equally good. The order of difficulty according to word
position is different for production and perception. In
addition, there was an asymmetry in errors made on
identifying /r/ and /l/ in initial and consonant cluster
positions (Sheldon and Strange 1982). Whereas /l/ was
misidentified more than /r/ in initial position, /r/ was
misidentified more than /1/ in consonant clusters. Thus the
particular test stimuli chosen in a test could have an
effect on the scores of perception and production. For
example, Goto (1971) used only words with initial /r/ and
/l/ or initial cluster with /r/ and /1/: these positions
happened to be more difficult for perception but easier for
production. This positional difference for perceptual and
production difficulty may account for the greater
discrepancy between perception and production found in his
study.

If perception and production acquisition are related,
then an important inference for L2 teaching is that training
on one aspect may lead to improvement on the other. Some
studies have tried to explore this possibility.

Pimsleur (1963) conducted experiments with high school
students on the effect of discrimination training on
production. He found that experimental groups which

received perception training on the discrimination of French

70
nasalized vowels /a/ /6/ /§/ scored significantly higher
than the control group in their production of these
phonemes. However, the experimental group which received
discrimination training on the distinction between French
[0] and American English [0"] did not differ from the
control group in their production of French [0].

Mueller and Niedzielski (1968) conducted another
experiment on the effect of discrimination training on
production with university students taking French. In the
first nine weeks of the course the experimental students
spent a total of 2 hours on discrimination training in
addition to 22 hours of laboratory exercises used by both
the experimental group and the control group. The results
showed that among the seven variables studied, only two
showed significant differences between the experimental
group and the control group. The experimental group scored
significantly higher in vowels which have a correlate sound
in the English system (variable 3), while the control group
scored significantly higher in phonemes that do not have a
correlate in the English sound system (variable 1).
However, the girls in the control group showed significantly
higher aptitude, and general impressionistic evaluation
(errors involving "mismanagement of allophones" or
"deviations from phonemes", fluency, liaisons and the

general intonation features, stress and rhythm features)

71
found the experimental group superior to the control group.
Mueller and Niedzielski (1968:416) concluded:

In view of the fact that the test was heavily

weighted towards variable 1, and because the

aptitude of the experimental group was so much

inferior to the aptitude of the control group

(girls), the impressionistic superiority of the

experimental group seems to be a valid conclusion

and does not seem to be contradicted by the

quantitative evaluation. Discrimination training

seems to be an effective training device in the

learning of pronunciation.

Schneiderman et al. (1988) found that learners of
French improved significantly in both discrimination and
production after undergoing a discrimination training
program, while the control group did not show similar
improvement. However, the discrimination and production
abilities no longer correlated significantly after the
training.

Borden et al. (1983) investigated Korean speakers'
perception and production of English /r/ and /1/ before and
after two or three weekly 45-minute training sessions. They
found that there was some change in the accuracy of /r/ and
/1/ production during each training session (from an average
of 80% correct on the probe test given at the beginning of
each training session to an average of 89.4% correct on the
post-training probe test), but "there was not a significant
difference between the more extensive speech production test

given before training and the final test, indicating that

there was little carryover of training".

72

However, despite the lack of carryover effect of
perceptual training on production, there was an indication
that perception (with criteria used by native L2 speakers)
precedes production (Borden et al. 1983:516):

When subjects judged the correctness of their

own [r] and [l] productions as compared with the

taped model, the extent to which their judgments

were in agreement with the judgments of the

experimenter correlated significantly and

positively with their production scores.

Where there is a marked difference between the two

final scores among speakers undergoing more

extensive change in their speech patterns, it can

be seen that self-perception scores were always

higher than production scores. This indicates

that the ability to perceive errors in one's own

speech may precede correct production in the

timetable for learning new phonemes, or that self-

perception is simply an easier task than

production of new speech sounds.

On the other hand, Catford and Pisoni (1970) found that
subjects receiving articulatory training not only scored
higher in production but also in discrimination than
subjects receiving only aural training. They stated: "This
obviously implies some kind of carry-over from productive
competence to auditory discriminatory competence, and may,
indeed be taken to be some support for a "motor theory of
speech perception" (p.481). Because of lack of understanding
of how perception and production interact, and since most of
the training programs involved both perception and
production, the results of these studies did not show
conclusively what effect training in one aspect has on the

other.

73
3.1.4 Acquisition of L2 phonology over time
L2 learning takes time. In this section, we will
examine the time effect of L2 phonological learning.

Perception and production will be discussed separately.

3.1.4.1 Perception

Adult L2 learners can achieve high perceptual ability
similar to that of native speakers after years of experience
in the L2 environment. Neufeld (1980) reported that
advanced English speakers of French who accessed themselves
as high in oral-aural skills in French were able to detect
traces of foreign accent in French speech samples just as
French native speakers can. However, less proficient
English speaking learners of French were not able to do the
same: they performed only at chance level.

Thompson (1984) found fluent L2 learners of English
(who had lived in the USA from two to 42 years with an
average of 22 years) achieved high perception ability, which
was not much different from that of native speakers, as
measured by discrimination and identification tests of
‘English words.

Flege and Hillenbrand (1984) found advanced adult L2
learners who had studied French for at least 4 years in
school can discriminate French /tu/ and /ty/ syllables

produced by native French speakers about 90% of the time.

74

The effect of L2 experience on the perception of new L2
sound has been found in the learning of /r/ and /l/ by
Japanese speakers. Sheldon and Strange (1982) found their
Japanese subjects did better in identifying /r/ and /1/
produced by American speakers than the subjects in Goto's
(1971) study. Part of the reason for this difference is
that their subjects were in the United States and had more
experience speaking with native English speakers.

This effect of L2 experience with native English
speakers on the perception of /r/ and /1/ by Japanese
speakers has also been found in the categorical perception
of synthetic /r/ and /1/ stimuli.

MacKain et al. (1981) found that Japanese learners

with extensive experience speaking and hearing

English, but not relatively inexperienced Japanese

learners, identified and discriminated synthetic

/r/ and /l/ stimuli much like native English

speakers. The experienced Japanese learners showed

a sharp crossover from /r/ to /l/ judgments as F3

onset frequency changed, and a peak in percent

correct discrimination near the /r-l/ phoneme

boundary. (Flege 1988:331)

Studies on L2 learners' perception of voicing in terms
of VOT in prevocalic stop consonants also provide evidence
for the effect of L2 experience on perception. Different
languages use different VOT in perception and production of
voiced and voiceless stops (Lisker and Abramson 1964;
Williams 1979, 1980). Williams (1980) found that Spanish-
speaking children with varying degrees of exposure to

English in a second language environment (0-6 months, 1.5-2

75
years, and 3-3.5 years) labelled VOT series toward the
English-like pattern gradually with increasing exposure.

Gass (1984) also found that the perception of VOT by
adult L2 learners of English changes as a function of time.
A group of L2 learners enrolled in an English-language
training program were tested on their perception of VOT in a
labelling task in an English set three times at one-month
intervals. These subjects had little prior exposure to
spoken English by native speakers, and their native language
differed from English in VOT boundaries between stops. In
contrast to native English speakers, the L2 learners'
perception did not show a narrow categorical boundary:
instead, there was a larger area of uncertainty, with
gradual rise and slope reversals. Although the abruptness
of the slope and the number of slope reversals did not
change significantly over two months for the whole group,
individual data from most subjects exhibited fewer slope
reversals or more abrupt curves.

Short term L2 perceptual training has sometimes been
found to have an effect on L2 perception. Gillette (1980)
found that the identification of /r/ and /l/ by Japanese and
Korean learners can be improved by training:

Gillette (1980) studied the effect of a five-
week classroom training program on the perception

of English /r/ and /l/ by three native speakers

each of Japanese and Korean. The training

consisted of articulatory descriptions of /r/ and

/l/, individual and group practice in producing

/r/ and /l/, and auditory evaluation by the
students of the extent to which their productions

76

of /r,l/ differed from that of fellow students and
the native English-speaking instructor. Their
students' correct identification of /r/ and /l/ in
minimal-pair words increased from 70% to 92%.
(Flege 1988:330-331)

There are also indications that perception of VOT can be
modified after training:

In one study (Lisker, 1970) discrimination
training enabled native speakers of Russian, a
language with lead and short-lag stop categories,
to label synthetic short- and long-lag stimuli as
/ba/ and /pa/. (Flege 1988:354)

[A] later study (Pisoni et al., 1982) showed that
native English-speaking subjects could be trained
to identify three categories of stops through the
alternating presentation of "good” examplars of
the three categories (i.e., stimuli with -70, 0,
and 70 ms VOT values: see also McClasky et al.,
1983). (Flege 1988:355)

However, the effect of short-term perceptual training seems
to be limited in other studies. For example, Pimsleur
(1963:203) stated:

It was found in the pilot studies that students
could not be trained to discriminate reliably
beyond a certain point. Some students still
failed, even after considerable training, in the
task of telling whether vont, vent, or vain
appeared in the sentence they heard: Ils vont
venir tout l'heure.

Strange and Jenkins (1978) found that the perceptual
training of VOT had only limited effect (cited in Flege
1988:354-355):

Strange (reviewed in Strange 8 Jenkins, 1978)
attempted to train English speakers to identify
members of a lead category in a continuum of CV
syllables synthesized with VOT values ranging from
-150 to +150 ms. Native English speakers typically

77

hear only two categories in this stimulus range.

Six hours of training with either immediate or

delayed feedback resulted in only limited

improvement in subjects' ability to discriminate

lead and short-lag stops, and training did not

generalize to untrained places of articulation.

Strange (Strange & Jenkins, 1978, p.152)
concluded that ”changing the perception of the VOT
dimension...is not easily accomplished by

techniques that involve several hours of

practice."

Tees and Werker (1984, cited in Mackain 1988:57) found
that after one year of experience with Hindi as a second
language, the number of native English speaking subjects who
reached the discrimination criterion of Hindi voice and
place (retroflex vs. dental) contrasts (two of 28 prior to
second-language learning) did not increase significantly.

It is clear from the literature that L2 phonological
perception improves with exposure. However, the change can
be very gradual, and the effect is likely to be limited with

short-term exposure.

3.1.4.2 Production

Studies on the acquisition of L2 phonology in an L2
environment generally found that accent is inevitable within
the first two years of living in the L2 environment, even
for young children who are likely to acquire native-like
pronunciation (Tahta, et al. 1981a, Snow and Hoefnagel-Héhle
1977). Studies that found a significant year of residence
effect usually include subjects with less than two years of

residence (Asher and Garcia 1969), while studies that found

78
no effect of length of residence investigated subjects with
more than two years of residency (Tahta, et al. 1981a, Oyama
1976).

Snow and Hoefnagel-Hohle's (1977) study on L2 learners'
pronunciation over a 10-11 months period in a naturalistic
L2 environment showed that adult learners made smaller
improvement in pronunciation than children, and that the
improvement made during the second 4-5 months was smaller
than that made during the first 4-5 months. Although from
the literature we cannot be sure if and when adult L2
learners stop making improvement (and there may be some
variation among learners), it seems that after two years of
active learning in the L2 environment, adult L2 learners
either stop making any improvement in their L2
pronunciation, or the improvement is too small to be
perceived by native speakers as significant.

Compared to a second language environment, a foreign
language environment is clearly not ideal for L2
pronunciation learning. Learners have few or no native-like
models on which to base their hypothesis about target
pronunciation. Even if native-like models are available,
the number is likely to be small and may not provide
adequate variation for learners to develop native-like
phonological perception and production targets. Since
production in adult L2 learners seldom closely resembles

that of their models even in a second language environment

79
where native speaker input is readily available, their
deviation from authentic pronunciation can be expected to be
even greater in a foreign language environment.

Flege and Hillenbrand (1984, also Flege 1987) found
that experienced American learners of French (professors who
had used French on a fairly regular basis for an average of
10 years and had spent an average of 1.3 years in France)
produced French /t/ with shorter (more French-like) VOT
values than relatively inexperienced learners (undergraduate
students who had spent the previous academic year in Paris),
but the difference did not reach significance, and none of
them produced /t/ with an average VOT value produced by
monolingual native speakers of French. The more experienced
learners also produced a more effective contrast between
French /u/ and /y/ than less experienced learners. However,
the two groups did not produce French /y/ differently: /y/
produced by experienced and inexperienced groups was
correctly identified by French speakers an average of 71.5%
and 70.2%, respectively, compared to 93.7% correct
identification of /y/ produced by native French speakers.
Only two out of seven speakers in each American group
produced /y/ correctly over 93%. The confusion between /u/
and /y/ by less experienced learners was a result of their
production of French /u/ (in tons) more like French /y/
rather than French (or even English) /u/, suggesting that

they may have confused the identity of the vowel in tons.

80
The more experienced learners produced French /u/ very much
like their English /u/ (as measured by F2 frequency). "Thus
even after many years of experience speaking French they
seem to have done little to modify their production of /u/
in the direction of French phonetic norms" (Flege and
Hillenbrand 1984:196).

Since both groups of American learners were living in
their native country and using English as their primary
language when the study was conducted, it appears that L2
pronunciation does not improve much with experience in a
foreign language environment after learners have studied the
language for many years and gained considerable fluency, as
might be expected from the studies on L2 pronunciation in L2
environment.

It should be noted that, in the studies cited above
(Flege and Hillenbrand 1984, Flege 1987), native speakers of
French who had lived in an English-speaking environment for
an average of 12.2 years produced French /t/ with longer
(more English-like) VOT values than those produced by
monolingual French speakers. Likewise, another group of
native English speakers, who had lived in France for an
average of 11.7 years and used French as their principal
language, produced English /t/ with shorter (more French-
like) VOT values than those produced by monolingual English
speakers. Thus the language environment a speaker is in has

an effect on the pronunciation (at least VOT in stops) of

81
not only L2 but also L1. This is consistent with what was
described by Obler (1982) as "the parsimonious bilingual".
Obler (1982) reported that balanced bilinguals who spoke
both English and Hebrew without an accent produced stop
consonants in the two languages with different VOT values.
However, they did not produce the consonants exactly like
the monolinguals. Rather, they produced English voiced
consonants closer to the Hebrew monolingual norms, and the
Hebrew voiceless consonants closer to the English
monolingual norms. Table 2 from Obler (1982:342) is
reproduced below to show the VOT means produced by

monolingual and bilingual groups.

Table 2. Group Voice Onset Time Means for Production

 

 

 

 

 

Unilinguals Bilingusls
Hebrew group English group Hebrew condition English condition '
Ipl +25.6 +77.6 +58.0 +683
lb! - I I0.8 -8.5 -96.8 -64.8
It! + 33.9 + 77.0 +73.8 - +89.6
Id! -95.8 -6.8 -85.5 -55.8
Ikl +617 +89.4 +9.12 +99.4
lg! - 101.0 +15.1 -92.2 -58.3

 

Thus, bilinguals tend to maximize the contrast and produce

equivalent sounds in two languages closer to each other.

82

Gass (1984) found that adult L2 learners of English
produced stop consonants /b/ and /p/ with VOT values mostly
within the range of possible English norms. Moreover, "when
the VOT value does not fall within the possible English
range, it is often not within the native language range,
either. Rather, it is in a direction away from the native
language" (p.69). The difference of VOT values in /b/ and
/p/ produced by native English speakers and L2 learners
showed mainly in the amount of variation. It was found that
the production of VOT values by L2 learners, "while on the
surface for the most part appearing nativelike, showed a
trend from lesser variability to greater variability and
back to lesser variability" (p.71).

A difficulty frequently encountered in the study of L2
pronunciation is that variations is found across speakers as
well as within a speaker. For example, Borden et al.
(1983:509) found Korean speakers' pronunciation of [r] and
[l] as judged by the experimenter was not linear over three
or four weekly testing sessions. Two subjects improved each
time, two subjects improved at first and then regressed, and
four subjects got worse before they improved. The causes
for the non-linear development may be quite complicated,
involving not only the similarities of phonetic and
phonological properties of L1 and L2 sounds but also the
organization of the phonological system or the distance

between neighboring phonemes (Major 1987).

83

It is clear from the studies on /r,l/ learning by
Japanese speakers that articulatory training making specific
reference to articulatory parameters can help adult L2
learners to produce foreign sounds acceptably even though
they may still have difficulty perceiving the contrast.
Catford and Pisoni's (1970) study suggests that articulatory
training is more effective than aural training in helping
learners produce "exotic sounds" immediately. In their
study subjects who received about two hours of articulatory
training produced the exotic sounds better than subjects who
received only aural training (without seeing the mouth of
the teacher).

The acquisition of English /r,l/ by L2 learners also
suggests that experience in a second language environment,
even after many years of L2 learning in a foreign language
environment, has an effect on the accuracy of production.
The Japanese subjects in Sheldon and Strange's (1982) study
(who had resided in the USA for 14-39 months) produced fewer
errors than subjects in Goto's (1971) study (of whom only
two had spent years in America). In a study by Borden et
al. (1983) on Korean learners, "the half with better scores
averaged about 18 months in the United States and the half
with lower scores averaged almost 6 months" (p.509). This
is consistent with Purcell and Suter's (1980) finding that

length of residency in an L2 speaking environment is

84
significant in predicting L2 pronunciation, while the amount
of formal L2 classroom training is not significant.

The study of Pimsleur (1963) found a difference in the
effect of discrimination training on the production of
phonemic distinction and the phonetic approximation in the
learning of L2 sounds. Production of French nasalized
vowels /a/, /6/, /é/ improved significantly after training
while production of the French vowel /o/ did not. Beebe
(1987:168) believed that most pronunciation errors do not
involve confusion of two phonemes. She stated that
"phonemic contrasts are a very minor problem for
intermediate and advanced learners. Phonetic inaccuracy,
compared with a TL norm, is extensive. The phonetic
deviance is heard by the teacher/listener as a phonemic
error." That the advanced learners' pronunciation problem
is one of phonetic inaccuracy rather than of phonemic
contrast is also found in the acquisition of tone. In a
study by Ioup and Tansomboon (1987), the production of Thai
tone by advanced Thai learners (who had extensive formal
instruction in Thai and had lived many years in Thailand as
missionaries) was judged to be phonemically correct 95% of
the time, compared to only 58% for the beginning learners
(who were taking second quarter of elementary Thai at a
university). However, only 8% of the tone produced by the

advanced learners, and 3% by the beginning learners, was

85
judged to be phonetically indistinguishable from native Thai

speakers.

3.2 Perception and production of Mandarin tones by adults L2
learners

There are only a few studies investigating the
perception or production of Mandarin tones by English-
speaking adults. Tone perception and production will be

reviewed separately in this section.

3.2.1 Perception

Winitz (1981) conducted an experiment on the perception
of Mandarin tones and sounds by monolingual English-speaking
children (third-graders) and adults (college students). In
the three contexts (isolation, disyllable, and sentence)
tested, the adults achieved higher scores than did the
children. Context was found to be significant, indicating
that the tasks increased in difficulty from isolation to
disyllable to sentence. Only in the perception of sounds,
the adults showed no decrease between the disyllable and
sentence contexts.

Broselow et al. (1987) tested English speakers'
perception of Mandarin tones. A tape consisted of training
and identification tasks was made by a Mandarin speaker and

‘~was presented to English speakers who had not studied any

tone languages. The brief training introduced the four

86

lexical tones of Mandarin with the syllable ti, and the
syllable was produced three times using each tone. The
identification tasks consisted of single syllables and
series of two and three syllables. A single example was
presented before each set of identification trials, and
subjects were given feedback after the presentation of
singlets and doublets.

Tone 4 was found to be the most easily perceived tone
in single tones (one syllables). Its index of detectability
was significantly different from that of all the other
tones. The percent of correct responses declined for all
tones in longer strings. However, Tone l and Tone 2 showed
no dramatic positional difference in the percent correct
responses in 2- and 3-syllable strings, while Tone 3 and
Tone 4 showed a significant difference in detectability
between different positions. For both Tone 3 and Tone 4,
detectability in the final position was significantly
greater than detectability in non-final positions. The
difference of detectability in Tone 3 can be attributed to
allophonic variations in different positions. The
positional difference of detectability in Tone 4, however,
the authors argued, was a result of interference from
English intonation. Since the contour of Tone 4 corresponds
to the fall at the end of declaratives in English, ”[tJhe
relative ease of detectability of the fourth tone in string-

final position is a result of positive transfer: a familiar

87
item occurring in a familiar position is easy to hear. The
difficulty of hearing the fourth tone in non-final position,
on the other hand, is a result of negative transfer: the
high falling pitch contour normally does not occur in non-
final position in English strings” (p.357).

Kiriloff (1969) tested Australian students' perception
of Mandarin tones in isolated syllables after two months of
Mandarin learning and again after four and a half months of
learning. The students were asked to listen to a recording
of 20 syllables pronounced by a native speaker and to
transcribe the sounds with the Pinyin symbols. The average
results showed no significant difference in tone perception
between the two tests. However, in a third test shortly
after the second test, the students were asked to identify
the tones only. The results showed that identification of
tone was significantly better in the third test than in the
first two tests. The xg'test on the errors showed that
"[e]rrors involving the 2nd tone are considerably higher and
those with the lst and 4th tones much lower than expected."
Thus, Kiriloff maintained that ”[t]he considerable number of
incorrect identifications of the second tone and a marked
tendency to confuse with it the third tone indicate that the
second tone presents the greatest difficulty to students of
Mandarin" (p.65).

W.C.J. Lin (1985) tested tone perception in beginners

(150 contact hours), intermediate (600 contact hours), and

88
advanced (1050 contact hours) students of Chinese. The four
tones over the same segmental sequence m3 'mother', ma
'hemp', m5"horse' and ma 'to scold' were put in the
sentence frame we shuo de shi __ (”What I said is __"), and
Tade __ hen youming (”His/her __ is quite famous"). The
correct identification for tones in the sentence final
position was 70.83% for beginners, 83.33% (but I get 91.67%
based on the figures given in his Table 2) for intermediate
students, and 94.04% for advanced students. The correct
identification of tone in medial position was 75% for
beginners, 86.46% for intermediate students, and 92.85% for
advanced students. For all groups of students and in both
sentence frames, there were more errors in identifying Tone
3 than the other tones.

Lin proposed a new treatment of Tone 3: introducing the
norm of Tone 3 as low-dipping (the allophonic contour occurs
in non-final position) instead of the citation form low-
dipping-rising. The reason was that "the tonal contour of
low-dipping-rising is realized only when it occurs in
isolation, in contrast with another syllable with the same
sounds but different tone, or in the case of emphasis. The
frequency of occurrence of these cases is very low. In the
overwhelming majority of cases, including its occurrence at
the end of an utterance or before a pause, the 3rd tone
simply dips without rising" (p.33-34). (M-Y. Chen 1983 also

found that low-dipping occurs 83% of the time.) After the

89
new treatment of Tone 3 was explained and some drills were
given, the test in the sentence frame Tade _; hen youming
was given. The results were 83.33% (90.97% according to my
calculation) correct identification for beginners, 95.83%
for intermediate students, and 98.81% for advanced students.
There was a clear improvement in the identification of tones
in general and improvement in the identification of Tone 3
in particular. For students who had been taught the new
treatment of Tone 3 since the very beginning, the results
were even better: 97.22% for beginners, 98.96% for

intermediate, and 100% for advanced students.

3.2.2 Production

Neufeld (1978) reported that many English speaking
adults achieved native-like pronunciation imitating
Mandarin utterances after 18 video-taped one-hour training
sessions which focused exclusively on phonetic material.
However, on a five-point scale, only one out of twenty
subjects received a five in Mandarin, and eight received a
four (appears native with occasional English-like sounds).
Thus his claim that "nine were judged to be native speakers
of Chinese" (Neufeld 1979:234) is questionable.

Shen (1989) investigated eight American students'
production of Mandarin tones in reading a text they were
familiar with. Shen argued that the acquisition of tonal

shapes (contours) of Mandarin does not constitute a

90

difficulty for L2 learners, but the tonal registers make the
mastering of tones a difficult task. She agreed (Shen
1989:27) with Zhao's (1987) statement about errors in tonal
register:

the register in tonal production functions like

the key in singing: if, when singing in the key of

C, the singer changes suddenly to the key of F,

the audience hears a deviate tune: by the same

token, when the L2 learner utters a high register

tone with a low register, in spite of the accuracy

of the tonal shapes, L1 listeners perceive an

inaccurate tonal realization.
In her study, the production data were tape recorded and
judged by four native speakers regarding the correctness of
the tone. When a tone error occurred (as judged by at least
three judges), the type of error (register or contour) was
also auditorily analyzed by the judges. The eight subjects
can be divided into two groups: the group of four better L2
speakers made significantly fewer errors than the group of
four worse L2 speakers. The differences in errors among the
five tones (including neutral tone) were not significant for
the group of better L2 speakers, but the differences were
significant for the group of worse L2 speakers. "However,
as for the overall errors of the two groups, the differences
of the errors among the five tones are significant" (p.32).
It was found that the rates of errors in the production of
Tone l and Tone 4 were significantly higher than those of
Tones 2, 3, and 0 (neutral tone). The rate of errors of

Tone 4 was significantly higher than that of Tone 1, but the

rates of errors of Tones 2, 3, and 0 were not significantly

91
different. The auditory analysis found that most errors
were register errors rather than contour errors. Acoustic
analysis was also carried out on tones produced by one of
the subjects (the one that made the most errors), and the
results were found to be in agreement with those from the
auditory analysis.

Miracle (1989) also studied American students'
production of Mandarin tones. His speech samples were 1- or
2-syllable words taken from sentences read by subjects.

Only the words in the topic (sentence initial) position were
analyzed. The tones were analyzed acoustically. The F0
range of each subject was calculated using the highest and
lowest F0 found in the voiced portion of the words analyzed.
Tonal errors were classified by him as either tone register
errors (based on each individual's tonal space) or tone
contour errors based on comparison with the F0 contours
obtained from native speaker models. He found that the
errors were fairly evenly divided among the five tones
(including the neutral tone) and were evenly divided between
tone register errors and tone contour errors. The overall
error rate was 42.9%. In the tone combination (2-syllable)
words, the error rate in the initial syllables (49.6%) was
significantly higher than that in the final syllables (32%).

Shen (1989) and Miracle (1989) appeared to have quite
different findings about American learners' learning of

Mandarin tones. Miracle noted that Shen's subjects had only

92

had one semester of Chinese while his subjects had been
studying Chinese for almost twice as long. ”This fact may
suggest that while the first and fourth tones are initially
troublesome, with further study, the difficulties even out
across all the tones.” (p.56) However, other differences may
also have contributed to their different results. The
evaluation procedures adopted in the two studies were quite
different. Miracle based his analysis entirely on acoustic
information, while Shen's study was based primarily on the
auditory judgement of four Chinese judges. Although
acoustic analysis was also carried out by Shen on the
production of one subject and the results were found to be
in agreement with those from the auditory analysis, it has
generally been found that physical differences in acoustic
signals do not always correspond to differences in human
perception. For instance, a slightly falling F0 contour has
frequently been produced as Tone 1 by native speakers in
some environments (Miracle, 1989: Shen 1985: Dreher and Lee,
1968) and would be perceived as Tone 1 (level) by native
Mandarin speakers (Tseng, 1981). Furthermore, the F0 range
of each subject in Miracle's study was based on the Mandarin
words analyzed (words taken out from sentence initial
position), which may not be the same range if the subject
were a native Mandarin speaker.

Another important difference between the two studies

was the material read by subjects. Miracle restricted his

93
data to words in the topic position ”to reduce the possible
interference of English sentence intonation" (p.51). On the
other hand, the sentences read by Shen's subjects were
longer and connected. For example:

You shihou ta gén pengyoumen qfi kin dianying,
Sometimes she and friends go see movie,

you shihou 951 babe, mama he péngyou xie xin.
sometimes to Dad, mom and friend write letter

These sentences were very likely to have been influenced by
English sentence intonation.

In long stretches of connected speech, some tonal
contours (and other phonological features) may not be
preserved well in native speakers' production because of the
shorter duration of syllables in connected speech
(Tseng,1981: Lyovin, 1978). Thus in the perception of tone
in long stretches of connected speech when contour features
are not clear, tonal register may be more important than
tonal contours (Massaro et al. 1985). Although we do not
know how fast the American subjects' in Shen's study read
the text, it is quite possible that at least some syllables
were short since the text they read was familiar to them.

If that was the case, then it is not surprising that more
register errors were found by the Chinese judges.

Basically, studies on Mandarin tone learning by English
speakers were concerned with the errors the learners made,
the source of errors, and the ways to correct the errors.

Only Kiriloff's (1969) study compared the results of

94
identification tests conducted at two different times and

found no improvement.

4. ISSUES AND NYPOTEESES

It has been mentioned in the Introduction that the
purpose of this study is to see how tones are perceived and
produced by English speaking adults at the beginning of
their learning of Mandarin. More specifically, this study
will try to answer the following questions:

(1) Is tone perception and production affected by
variables of tonal context, task, sex, and L2 learning
experience?

(2) Is there a relationship between the perception of
tone in isolated syllables and in sequences, and a
relationship between the production of tone in isolated
syllables and in sequences?

(3) Is there a relationship between tone perception and
tone production?

(4) Do tone perception and tone production improve over
one term of learning?

(5) Among the four Mandarin tones, is there a
difficulty order for English speaking learners? If there is,
what can account for the order of difficulty?

(6) Are there general patterns in errors of tone made
by English speaking adults learning Mandarin? If such

patterns can be found, what accounts for them?

95

96

These question are important questions for researchers
in second language acquisition and for teachers of Mandarin
as a second language.

The first question concerns two external variables
(phonological context and task) which have been found to
affect L2 performance. But no studies have been done on the
effect of the particular phonological context (isolated
single syllables, 2-syllable words and short sentences) and
task (imitation and reading aloud), or on two learner
variables (sex, L2 experience) in the perception and
production of Mandarin tone by L2 learners. The second
question concerns an assumption often made by L2 researchers
and L2 teachers that the perception or production of a sound
or tone in its citation form is closely related to its
perception or production in utterances. But few studies
have been done on this. The third question is about the
relation between perception and production, which has often
been asked in L2 acquisition research as well as in other
related areas (e.g., L1 acquisition and psycholinguistics)
because of its theoretical importance and its implication
for L2 learning and teaching. The fourth question concerns
the development of L2 phonological perception and production
in L2 learners, which may be especially important in
developing an explanation for the wide-spread phenomenon
among adult L2 learners of foreign accent. The fifth

question is about the difficulty order of the four Mandarin

97
tones for L2 learners, and the sixth question is about the
type of errors L2 learners make. Both questions have been
investigated by L2 researchers in many aspects of L2
acquisition (phonology, morphology, syntax, etc.) in many
different languages, and theories of L2 acquisition have
been proposed and modified based on the findings of these
studies. Findings on the difficulty order of Mandarin tones
and the type of errors L2 learners make in learning tone
also need to be accounted for by theories of L2 acquisition,
as they may provide valuable evidence for or against
particular theories.

In this chapter, based on the literature reviewed in
the previous two chapters, I will propose some hypotheses
concerning the above questions. Except for question (1),
which includes four variables each of which corresponds to
one hypothesis (Hypotheses 1 to 4), questions (2) to (4)
each corresponds to one hypothesis. Thus, Hypothesis 5
correspond to question (2), Hypothesis 6 to question (3),
Hypothesis 7 to question (4). These hypotheses will then be
examined with the data obtained. No specific hypotheses
will be proposed for questions (5) and (6), concerning the
difficulty order of Mandarin tones and error patterns, which

will be obtained by analysis of the data.

98
4.1 Effects of some variables on L2 tone perception and
production

In research on variables that predict accuracy of L2
pronunciation, while age of arrival in L2 environment has
been found to be the best predictor for accuracy of
pronunciation (Asher and Garcia 1969, Oyama 1976, Seliger et
al. 1975, Thompson 1984), native language (L1) was found to
be the best predictor for foreign accent among adult L2
learners (Purcell and Suter, 1980). There are few studies
that investigate variables predicting L2 phonological
perception. One study found ear preference to be the best
predictor for perception ability but it was not a strong
predictor (Thompson 1984).

In this study, all subjects are native speakers of
American English, and all subjects have passed puberty (the
presumed critical or sensitive period for achieving native-
like accent) and the age range is relatively narrow.
Individual differences are kept constant by testing the same
subjects twice on tone perception and production, although
some subjects were absent or dropped out for the second
test. The variables to be analyzed in this study are tonal
context (tone in isolated syllables, 2-syllable words and
short sentences), taSk (imitation and reading), sex, and L2

learning experience.

99
4.1.1 Tonal context

It is noted by many experienced teachers of Mandarin as
an L2 that many students can pronounce Mandarin tones in
citation forms quite well, yet cannot do as well when
speaking a sentence. The same may also be true in tone
perception. Winitz (1981) and Broselow et al. (1987) found
that English speakers with no knowledge of Mandarin
perceived Mandarin tones in citation form better than in two
or more syllables.

In this study, tone in three tonal contexts (isolated
syllables, 2-syllable words and short sentences of three or
four syllables) will be studied. The difference in
utterance length may affect L2 perception and production
(see 3.1.1.1), and other factors, such as transfer of L1
suprasegmental patterns (Odlin 1989), are also expected to
come into play in different contexts. For instance, English
sentence intonation is more likely to be transferred to a
Chinese sentence than to a syllable or a word. The English
stress pattern of 2-syllable words is more likely to be
transferred to 2-syllable Chinese words rather than to a 1-
syllable words. Thus the perception and production of
Mandarin tones in isolated single syllables are expected to
be easier for English learners of Mandarin than perception
and production in 2-syllable words or short sentences.

Hypothesis (1): Subjects perceive and produce tones

with significantly higher scores (see Chapter 5 below)

100
in isolated single syllables than in 2-syllable words

and sentences.

4.1.2 Task

L2 learners' performance can vary for different tasks
(L. Dickerson 1975, Beebe 1987, Sato 1985). The production
data in this study will be collected on two different tasks:
imitation and reading from a list. Both imitation and
reading are common tasks used in soliciting subjects' speech
samples because the control of other variables (e.g.
grammar, vocabulary, phonological context) is easier in
imitation and reading than in free speech. However, if the
two tasks result in different performance, then results
obtained from studies using different tasks may not be
directly comparable.

It is commonly observed that beginning L2 learners can
pronounce L2 sounds much better when repeating immediately
after an auditory model than when pronouncing the same
sounds without such a model. Data from a study using both
methods showed pronunciation in imitation better than that
in spontaneous production (Snow and Hoefnagel-Héhle 1977,
see 3.1.1.2). In L2 tone production, just as in other
aspects of L2 pronunciation, imitation is expected to
produce better results than reading, at least in the

beginning.

101
Hypothesis (2): Task, imitation or reading, affects
subjects' performance in tone productioné imitation
results in significantly higher scores in tone

production than does reading.

4.1.3 Sex

Sex has often been included as a variable in studies of
L2 acquisition probably because a sex difference is usually
present and many people believe that females are better than
males in learning languages (e.g., Bolinger 1975:333). It
will be interesting to see if tone learning is affected by
sex.

As reviewed in 3.1.1.3, sex was found to be a
significant variable in L2 pronunciation in some studies
(Ash and Garcia, 1969, Thompson 1984), in which females were
found to have better pronunciation than males. However, the
effect seemed to diminish with length of residence (Asher
and Garcia 1969). In some other studies (e.g. Purcell and
Suter 1980) sex was found not significant in L2
pronunciation. There is no evidence that L2 phonological
perception is affected by sex.

Hypothesis (3): Subjects' performance in Mandarin tone

perception or production is not affected by their sex.

102
4.1.4 L2 experience

Knowledge of a previously learned L2 has been reported
to facilitate the learning of another closely-related L2
(Rivers 1979, Singleton and Little 1984). Some people
believe that learning a second language makes the learning
of subsequent languages much easier (Larsen-Freeman and Long
1991:206). However, there is no evidence in the literature
on L2 acquisition that this is true in the learning of
phonology (see 3.1.1.4).

Originally it was planned to test this hypothesis by
grouping subjects into two groups: those with previous L2
learning experience and those without such experience,
assuming that students taking Chinese 101 have no prior
Chinese learning experience and no experience in learning
another tone language. However, it turned out that many of
the students had taken Chinese classes before, either in
high school or in college, though for no more than one year.
It was decided to include them in the study as well, but
they are classified into a third group: those with Mandarin
learning experience.

It is expected that students with prior Mandarin
learning experience will do better since they had already
studied the tones before. Students with other L2 learning
experience and students with no L2 learning experience are

not expected to learn tones differently.

103
Hypothesis (4): (A) Subjects with prior Mandarin
learning experience score significantly higher in
Mandarin tone perception and production than those
without such experience. (8) For subjects without
Mandarin learning experience, other L2 learning
experience does not affect performance in tone

perception or production.

4.2 Relationship between the perception and production of
tone in isolated syllables and in sequences

In most beginning Chinese classes, only tone in
isolated syllables and the tone sandhi rule for Tone 3 (Tone
3 becomes Tone 2 before another Tone 3, see 2.1.2) are
specifically taught. In addition, some proponents of
teaching tones in sequences emphasize the third tone and the
neutral tone, which are realized phonetically differently
depending on the preceding or following tones (see 2.1.2).

The focus of teaching tone in isolated syllables is not
surprising since multisyllabic words or phrases and
sentences are basically a succession of syllables with their
lexical tones. If there is a strong positive correlation
between the acquisition of tone in isolated syllables and in
sequences, then the teaching of tone in isolated syllables
may be quite adequate. If, however, there is no significant

correlation or the correlation is weak, then the teaching of

104
tones in connected speech may need to be done more than is
usual.

Acoustic studies on Mandarin tones found that the basic
tonal contours of the four tones generally remain the same
in isolated 2-syllable words and sentences as in their
citation forms (Ho 1976a, M-Y. Chen 1983). However, in
spontaneous running speech or in fast speech, the tonal
characteristics may not be well preserved in all syllables
(Tseng 1981, Wu 1982). As a result, tone in syllables taken
out of such context becomes difficult to identify even for
native speakers (Tseng 1981, Lyovin 1978). Therefore
sequences in this study are normal speed utterances of
limited length, 2-syllable words and short sentences of
three or four syllables: this eliminates the tonal variation
that results in longer sentences (see 2.1.3).

In this limited definition of tone in sequences, the
tonal characteristics of tone are expected to be clearly
produced, and the identification of tone based on acoustic
signals alone should be possible, just as is the
identification of tone in isolated syllables. Two to four
syllables should be within the short term memory span of an
average adult, especially when all his attention is directed
to the tone. Native language may have some different effect
on the perception or production of tone in isolated
syllables and in sequences (see 4.1.1), but the effect is

expected be similar for subjects speaking the same L1. It

105

is assumed that the ability to identify a tone in isolated
syllables is a prerequisite to the ability to identify tone
in sequences. Thus it is expected that those who do better
in identifying tone in isolated syllables will also do
better in identifying tone in 2-syllable words and short
sentences.

On the production side, the ability to pronounce tones
in isolated syllables correctly also seems to be a
prerequisite to pronouncing tones in a sequence of two to
four syllables correctly. Thus it is expected that L2
learners who pronounce tones in isolated syllables well will
tend to do better in pronouncing tones in 2-syllable words
and short sentences.

Hypothesis (5): There is positive correlation between

perception of tone in isolated single syllables and

syllable sequences, and positive correlation between

production of tone in isolated syllables and syllable

sequences .

4.3 Relationship between L2 tone perception and tone
production

Phonological perception and production are distinct
aspects of language competence. L2 learners have been found
to be able to discriminate L2 sounds without the ability to

pronounce them (e.g. Thompson 1984) as well as able to

106
pronounce sounds acceptably without the ability to
distinguish them fully (e.g. Goto 1971).

However, generally, it has been found that L2
phonological perception and production ability are related
in L2 learners. L2 learners who have better pronunciation
also perform better in sound perception and vice versa (see
3.1.3). Tone belongs to suprasegmental phonology yet it is
similar to segmental phonology in that it is distinctive at
the lexical level. Based on previous studies of
phonological perception and production in L2 learners, it is
expected that L2 tone perception ability and tone production
ability are related. Learners doing better in one test
(e.g. tone in isolated syllables) of tone perception are
expected to do better in the counterpart tone production
test.

It will also be interesting to compare general
performance in tone production and tone perception since in
L2 phonological learning the question of whether perception
is ahead of production is a controversial issue. In this
study, only the distinctive (phonemic) aspect of tone will
be analyzed.

Hypothesis (6): There is positive correlation between

subjects' Mandarin tone perception scores and tone

production scores.

107
4.4 Improvement of tone perception and production

It is usually expected that an L2 learner's L2
competence will improve with experience over a long period
of time. However, concerning the acquisition of L2
phonology by adults, it is not clear how long L2 phonology
will continue to improve. The commonly observed phenomenon
that adult learners typically speak an L2 with a foreign
accent seems to indicate a cessation of improvement in their
L2 phonological learning.

Studies investigating the L2 pronunciation accuracy of
subjects who have lived in the L2 environment for many years
have usually found little effect of length of residency
(Oyama 1976, Tahta et al. 1981a). However, for L2 learners
who have lived in the L2 environment for less time there
seems to be improvement with increased experience (Borden et
al. 1983). There is also evidence that adults learn faster
and have better control of pronunciation than children at
the beginning of L2 learning (Olson and Samuels 1973, Snow
and Hoefnagel-Hohle 1977). However, adults' improvement in
pronunciation slows down after the initial stage and
children's pronunciation becomes better than adults' in less
than a year (Snow and Hoefnagel-Héhle 1977). Even for
younger learners who are likely to improve more than adults
and eventually achieve a native-like accent, the learning
process can take years. It has been found that native-like

pronunciation generally cannot be achieved by L2 learners

108

regardless of age within two years in the L2 environment
(Tahta, et al. 1981a, Snow and Hoefnagel-Héhle 1977).
Studies on L2 learners' pronunciation have found a gradual
approximation toward the native speakers' norm (Flege 1987).
All these studies indicate that the learning of
pronunciation is a gradual process. It can be expected that
the progress of adults after the initial stage will be even
slower if they indeed continue to improve as children do.

On the perception aspect of L2 phonology, there is
evidence that even bilinguals who can speak two languages
with no accent have different perceptive patterns from
monolingual native speakers (e.g. categorical boundary of
VOT) (Obler 1982). Nevertheless, as far as the
identification of real L2 sounds is concerned, advanced L2
adult learners have been reported to have good perceptive
abilities that are not much different from those of native
speakers (Flege and Hillenbrand, 1984). However, there is
also evidence that no significant change in perception
occurs after one year of foreign language learning (Tees and
Werker 1984). W.C.J. Lin's (1985) test results of Mandarin
tone perception by beginning (150 hours), intermediate (450
hours) and advanced (600 hours) learners suggested a
continual improvement in learners as a group.

Studies on the variables contributing to L2
pronunciation in a second language environment have

typically found little effect of formal L2 teaching (e.g.,

109
Purcell and Suter 1980), and studies on the perception of L2
learners who have little experience in an L2 environment
also suggest that foreign language classroom experience may
not have much effect on the perception of an L2 contrast not
found in the native language (e.g., Goto 1971). However,
since many adults will begin their learning of L2 in a
foreign language environment, such as taking a foreign
language class in the university, it is important for us to
understand how perception and production of L2 phonology
develop in adults in a foreign language classroom.

There is evidence that suprasegmentals may be more
difficult than segmentals for adults to learn, since many
adults who speak L2 with only a slight accent are said to
have suprasegmental deviations rather than segmental errors.
The ability to imitate L2 intonation has been found to
decline rapidly between ages 8 and 11 (Tahta, Wood and
Loewenthal 1981b). The learning of tone may be even more
difficult since tone is utilized in a totally different way
from intonation. Research comparing the learning of tone
and other aspects of phonology and grammar (Ioup and
Tansomboon 1987) has confirmed the difficulty of tone for
adult learners.

Since learning of L2 phonological perception and
pronunciation is gradual, and may be even slower for
suprasegmentals, it is difficult to predict if the

improvement in tone perception and tone production will be

110
significant within a short time such as one term. However,
based on studies on L2 phonological learning which show that
improvement occurs more rapidly for adults at the beginning
of L2 learning, we will hypothesize that the learning of
Mandarin tone will improve over one term of learning.
Hypothesis (7): Subjects' scores in perception and
production of Mandarin tone are significantly higher at

the end than at the beginning of the term.

4.5 Order of difficulty among the Mandarin tones

Early studies on the acquisition of some English
grammatical morphemes by L2 learners have shown that there
is a similar developmental sequence or order of accuracy
among L2 learners (e.g., Dulay and Burt 1973), and this
order has been found by some to correspond to the
acquisition order of children acquiring their first language
(Krashen 1977). Thus the order found by comparing
success/error rates in L2 performance is not random and the
order can reveal important differences among the elements
under study. Theories of L2 acquisition as well as relevant
linguistic theories need to explain this fact.

A few studies on the learning of Mandarin tones by L2
learners have discussed the relative difficulty of the four
tones (see section 3.2). However, the available data are
limited and not all are in agreement. On the perception

aspect, one study found Tone 2 to be the most difficult

111
(Kiriloff 1969) while another found Tone 3 the most
difficult (W.C.J. Lin 1985). On the production aspect, one
study found the four tones were equally difficult (Miracle
1989), while another found Tone 1 and Tone 4 to be more
difficult than Tone 2 and Tone 3 (Shen 1989).

This study will provide us with more data and hopefully
can shed some light on the issue. The results will be
compared with the developmental order of tone in children
acquiring Mandarin as their L1, and explanation will be

sought to account for the results.

4.6 Error patterns

Previous studies have found some patterns of tone
errors in English learners of Mandarin (see 3.2). For .
example, in the perception aspect, confusion between Tone 2
and Tone 3 has frequently been reported (W.C.J. Lin 1985,
Kiriloff 1969). On the production aspect, some found tone
register errors predominant (Shen 1989) while some found
that both register errors and contour errors are frequent
(Miracle 1989).

The classification of tone errors in this study will be
based on the Mandarin tonal system (i.e., a tone is
misperceived or mispronounced as another Mandarin tone or is
unidentifiable), and the categorization could give us some
clues as to whether a register or contour error is involved.

An examination of the tone errors in this study will give us

112

data in different tonal contexts at two different times, and
the error patterns involved may be seen more clearly.

Theories of L2 acquisition will be employed to account
for the errors. Both transfer and developmental factors
have been found to be important in the acquisition of L2
phonology (Hecht and Mulford 1982), and both are expected to
play a role in errors of tone perception and production

discovered in this study.

5. METHODS AND PROCEDURE

The data used in this study were collected from
students taking Elementary Chinese 101 at Michigan State
University in the Fall term, 1989. Both the tone perception
test and the tone production test were conducted twice: the
first time two or three weeks after the course had begun
when the sound system of Mandarin had been mostly covered
and some basic words and sentences had been taught, and the

second time near the end of the ten-week term.

5.1 Tone Perception Test

There were three parts to the tone perception test:
(A) Identification of tone in isolated syllables,
(B) Identification of tones in isolated 2-syllable words,
(C) Identification of sentences with the correct tones in a

multiple choice format. Each part will be discussed below.

(A) Identification of tone in isolated syllables

PerceptiOn of tone in isolated syllables is usually
explicitly taught in a Chinese-as-a-Second-Language class.
In this part of the perception test, students were asked to
identify the tones of isolated syllables spoken twice by a
native speaker. For example, the students would hear the
syllable "df". They would then have to identify the tone of

this syllable and write down the corresponding tone mark (-,

113

114
/, v4 \), or tone number (1, 2, 3, 4). In this case, the
correct answer is 1 or -. More examples can be found in a

complete perception test in Appendix A.

(B) Identification of tones in isolated 2-syllable words.

The difference between parts (A) and (B) is that a tone
in 2-syllable words is either preceded or followed by
another tone. This tonal context may affect the perception
of tone by an English speaking learner.

In this part of the test, students were asked to
identify both tones in a 2-syllable word spoken by a native
speaker. For example, the students would hear the word
"faxian" spoken three times. They would have to identify
the tone of the first syllable "f3" and the second syllable
"Xian", and write down the corresponding tone numbers or
tone marks of both syllables. In this case, the correct

answer would be 14 or -\. See Appendix A for more examples.

(C) Identification of short sentences with correct tones in
a multiple choice format.

In contrast with the identification of tone in isolated
syllables or two syllable words, in this case subjects were
not asked to identify all the tones in a sentence. Although
native Mandarin speakers can easily identify the tones of
all the words in a sentence, they don't necessarily listen

for these tones before making an identification. It is more

115
likely that they process the sentence by taking in all the
information available, and may identify the tones based on
other levels of information, especially when a sentence is
spoken at a fast tempo (see section 2.1.3). If we give
learners of Mandarin a sentence that they do not understand
and ask them to identify the tones, the exercise seems to be
unnatural and therefore unjustifiable. We can predict that
the identification of tones will be less successful simply
because there are more tones to pay attention to in a
sentence. In Broselow, et al. 1987, tone identification of
three syllables was less successful than that of two
syllables. However, if we give learners a sentence that
they have learned and ask them to identify the tones, we
cannot be sure that they are perceiving the tone correctly.
They could be processing the sentence by taking in other
(segmental phonological, syntactic, etc.) information, and
be making the identification based on their memorization of
the tones for each word. A native Mandarin speaker can
understand sentences which differ only in tones. For
example, T5 mai shfi, "He sells books" and T5 m5} shfi, "He
buys books" differ only in the tone of "mai". The
perception of tone is crucial in understanding the sentences
correctly. A native Mandarin speaker can also detect a tone
mispronounced in a sentence and may ask for clarification if
uncertainty or misunderstanding arises. A good L2 learner

of Mandarin should, to some extent, be able to do the same.

116

Consequently, in this third part of the perception
test, students were asked to choose from four versions of a
Mandarin sentence spoken by a native Mandarin speaker. The
four choices differed only by one tone, and only the
sentence with the correct tones was the correct answer. The
oral English translation of the sentence preceded the four
choices. For example, the students would hear the English
sentence, "Are you busy?", followed by "A. Ni mang ma? B. Ni
mang ma? C. Ni mang ma? D. Ni man ma?". All the sentences
in the task were short (three words) and, in contrast to
words in parts (A) and (B), had been taught in class. The
English translation of each correct sentence as well as its
spelling in Pinyin were written on the answer sheet, so that
the students would know the meaning and the correct tones of
the sentence they were going to hear even if they did not
remember what they had learned in the class. For the
example given above, the answer sheet would read:

Are you busy? (Ni mang ma?) a b c d
All they had to do was to pick the correctly spoken sentence
when they heard it. In this case, the correct answer should
be 2, and the letter 9 should be circled. See more examples

in Appendix A.

5.2 Tone Production Test
In the early weeks of the term, it would be impossible

to elicit spontaneous speech from the students. Even simple

117

conversation probably would be too difficult for some.
Thus, for this study, the use of imitation and reading tasks
seemed to be the best choice for obtaining speech samples.
This would have three advantages. First, exactly the same
speech samples, elicited from the recorded and written
material provided, would be obtained for all subjects.
Second, the length and the grammatical or phonological
difficulties of the tasks could be controlled. Third, the
speech samples obtained in the first and second test would
be similar and therefore comparable.

There were also three parts to the production test:
(A) isolated syllables, (B) 2-syllable words, and (C) short
sentences of three or four syllables. The students were
asked to perform two tasks for each part: (1) to imitate a
recorded model produced by a native speaker, and (2) to read
from a word/sentence list without an immediate auditory
model. For each part, they would first repeat the word or
sentence spoken by a native speaker as they heard it on the
tape, while silently scanning the word/sentence list
provided to them. After they had finished repeating all the
words or sentences in a particular part, they would then
read the same items aloud from the list. All the words and
sentences were written in Pinyin with tone marks on them.
In contrast to the words in parts (A) and (B), all the short
sentences had been taught in class and the sentence

translations were also given, so that the students would

118
know what sentence they were reading even if they had
forgotten what they had learned in class. It was hoped that
by first repeating the words while having a chance to look
at them spelled out in Pinyin, the students would be better
prepared to read them aloud on their own within a limited
time. A production word/sentence list can be found in

Appendix A.

5.3 Test Material

The test material used in both perception and
production tests were similar. Both consisted of three
types of test stimuli: (A) isolated syllables, (B) 2- A
syllable words, and (C) short sentences of three or four
syllables.

For isolated syllables and 2-syllable words, the test
items were prepared based on four considerations. First, in
order to reduce any possible distraction of unfamiliar
consonants or vowels in the perception or production of
tones, sounds that are similar in both English and Mandarin
were used as much as possible, especially for the tests
conducted early in the term and in the production test.
Second, certain Pinyin spellings that were likely to cause
difficulties or errors were not used in the production test,
which required students to read words spelled in Pinyin.
These Pinyin spellings include consonant initials x (which

represents a palatal fricative [C]) and g (which represents

119

an aspirated palatal affricate [t$fi), and 1 after the
consonant initial z, c, and s (where i represents an apical
vowel, or a syllabic [z]). They tend to cause problems
because they represent very different sounds in the English
spelling, which students at the level of those who took the
first test had not yet mastered. A few words whose spelling
may also be likely to cause errors (1 in shi, where i
represents a retroflex vowel) were nontheless used because
they had been taught in class. Third, the four Mandarin
tones have equal occurrences in parts (A) and (B), so that
the number of errors would not be affected by the unequal
distribution of certain tones. Fourth, in the perception
test, the 2-syllable words chosen were not familiar to the
students. This was done to eliminate the possibility that
the tone of the words would be known without perception.
For the production test, there were simply too few 2-
syllable words that had been learned by the students. While
most of the 2-syllable words were new to the students, a few
2-syllable words that had been taught in the class were used
as test items in the production test to reduce the
difficulty of reading all unfamiliar words. The same
familiar items were used in the first and the second
production tests.

For the short sentences of both tests, it was decided
that unfamiliar sentences would not be used since either

perception or production of totally incomprehensible

120
sentences would be unnatural and might not represent the
students' performance in a real situation. However, given
the limited vocabulary and syntax that the students had
learned before the first test, it was difficult to control
the segmental environment or the tone in each position.
Thus the criteria for choosing the sentences were:
(1) familiarity to the students, and (2) sentence length of
just three or four syllables. Although different sentence
types (statements, wh-questions and yes-no questions) were
included, and the tone to be tested in the perception test
occurred in different sentence positions, the occurrences of
the four tones in different sentences types or in different
positions were not equal, because of the limitations already
mentioned. The same sentences were used in the first and
second tests, but both the order of the items and the order
of the choices was different for the two tests.

In the perception test, there were 20 isolated
syllables in part (A), five of each of the four tones, and
16 2-syllable words in part (B), including all the tone
combinations possible in 2-syllable words. (However, because
tone 3 is realized as tone 2 when followed by another tone
3, there are actually only 15 patterns). There were eight
short sentences in part (C). In the production test, there
were ten isolated syllables in part (A): excluding the first
two as practice, there were two of each of the four tones.

There were also 16 2-syllable words in part (B), including

121

all the tone combinations, and eight short sentences in part
(C) .

5.4 Subjects

All the students taking Chinese 101 in the Fall, 1989
term took the tests during their regular class hours as part
of their course requirement. Most of them were
undergraduate students and a few were graduate students. No
information about their age was obtained. However, they
were probably between 18 and 30 years of age. Based on the
information regarding the students' language background (as
given by the students in a questionnaire requested by the
instructor), those students whose native language is not
English and those students who can speak a tone language
fluently were later excluded. This resulted in a total of
31 native English speaking subjects, 15 male and 16 female.
Many of them had previous second language learning
experience, including seven who had taken Chinese before,
although their Chinese learning experience had not exceeded
one year. Some students dropped the course after the first
tests. There were also students who came late to class and
missed part of a test, or who were absent and did not take a
test. As a result, two subjects took only part of the first
perception test, four did not take the second perception

test, six did not take the second production test and one

122
took only part of it. Their scores were still included in

the analyses when available.

5.5 Judges

The subjects' production was evaluated by three Chinese
judges. These judges are native Mandarin speakers from
Taiwan with backgrounds in linguistics and experience in
teaching Chinese. One judge has a Ph.D. degree in
linguistics and over seven years of Chinese teaching
experience. Another judge is a Ph.D. student in TESOL, who
has taken courses in linguistics (including phonetics) and
who has had experience teaching Chinese to American-born
Chinese children. The author is the third judge, who has
five years of experience in teaching Chinese to Americans as
well as to American-born Chinese children. (None of the
judges was teaching the Chinese 101 course when the tests
were conducted and thus the judges could not be influenced

by the students' performance in the class.

5.6 Procedures

The first tone perception test was conducted as part of
a quiz about two weeks after the course began. The first
tone production test was administered a week after that.
Originally, the first production test was conducted at the
same time as the first perception test. However, some of

the students did not understand the instructions very well

123
and went on to read the words from the part (B) 2-syllable
list aloud immediately after they had finished repeating the
part (A) isolated syllables, thus skipping the unprompted
production task of part (A). Other students who started
reading the isolated syllables aloud also switched to 2-
syllable words after a few words because they heard others
reading different words. As a result, the recordings could
not be used and the production test had to be conducted
again the following week.

The second production test was conducted in the last
week of the class, and the second perception test was
conducted during the final examination after the ten week
course ended. There were six weeks between the first and
the second production test, and about seven weeks between
the two perception tests.

The tones and most segments of Mandarin had been
introduced and practiced before the first perception test
was conducted. A tone identification test of isolated
syllables and 2-syllable words was given in the class by the
instructor the week before the first perception test, so
that the students were familiar with the format of the first

two parts of the perception test.

5.6.1 Testing
The first perception test and both production tests

were conducted in the language laboratory, where Sony ER-840

124
Booth recorders and headphone sets were used. When a test
was administered in the language laboratory, all the
instructions and test materials were pre-recorded on the
tape. The instructions for the production tests were also
written on the students' word/sentence list. Students were
first instructed to put on their headphones and put their
tape in place, and to adjust the position of the microphone.
A brief oral description of the test was given through the
headphones before the tape was played. After the
instructions were played, the tape was stopped to allow for
questions before proceeding to the test.

For the perception test, the 20 isolated syllables were
spoken twice each and the 16 2-syllable words were spoken
three times each. The students wrote down the tone mark or
tone number on their answer sheets as they heard the word.
In the short sentence part, they chose the sentence with the
correct tones by circling on their answer sheet the letter
preceding the correct choice. Each of the four choices was
spoken once, and there were eight sentences in total. The
perception test took a total of about ten minutes.

The second perception test was conducted in a classroom
instead of the language lab because the final examination
was scheduled to be held in the classroom. The test items
were read by the instructor rather than pre-recorded.
However, the content and the format were otherwise similar

to the first perception test.

125

For the production test, after the pre-recorded
instructions were played, the students listened to the
isolated syllables and immediately repeated each syllable.
After they had finished repeating all of them, the students
then read the same syllables aloud from the word list one by
one after hearing the corresponding number on the tape.

When they had finished this reading, they would then listen
to the 2-syllable words and repeat them, and subsequently
read these words aloud from the list. They then went on to
the short sentences. The students' repetitions and readings
were recorded automatically from the control panel at the
front of the language lab.

Admittedly, the testing and recording environment of
the language lab was not ideal for the production test. The
students could definitely hear others during the production
test and sometimes this background noise was even recorded
on the tape. However, since the imitation task provided
them with a good model to imitate, the students presumably
would not pay much attention to the pronunciation by other
classmates, whose production could well be less than
adequate. Even for the reading task, there would be little
motivation for a student to listen to others before reading
a word unless he knew the other student was pronouncing the
word correctly. As a matter of fact, most students read the
words or sentences at about the same time. Some students

did try to make a modification right after the first

126
attempt. However, they were not always successful, and very
often the second attempt sounded just like the first one.
It is not clear whether they were affected by other
students' productions or simply were not satisfied with

their own.

5.6.2 Evaluation

In order to check the validity of the perception test,
three Chinese listeners, one female from Beijing and two
males from Taiwan, took both the first and the second
perception tests after they had been administered to the
students. The recording of the second perception test was
obtained by using a minirecorder in the classroom when the
test was conducted. To check the validity of the production
test, two native speakers, one male speaker from Beijing and
one female speaker from Taiwan, also made the same
recordings as the subjects did. The perception test
questions were objective questions, but the production test
questions needed to be evaluated. The recordings of the
students' production were good enough for a listener to hear
the tones clearly, even when more than one voice was
recorded on the tape. The recordings were copied and edited
according to task and subtest. For example, all the
isolated syllables produced in the imitation task were put
together. Each subject was given a number and their

recordings were preceded by their number on the edited

127
tapes. The two native speakers' recordings were also given
a number and mixed with those of the subjects. Recordings of
two of the subjects were also chosen randomly to be copied
twice and to be given a second number. The edited tapes
were then judged by three native speakers.

The judges were instructed to judge only the tones, and
were told that the correctness of the segments should not be
a concern in their judgement. None of them seemed to have
any problem with the isolation of the tone from the
segments. When a word or a sentence was repeated or
modified by the subject, the judges were asked to judge only
the version with the better tone.

For each tone recorded, the judges were asked to
decide: (1) whether it could be identified as one of the
four Mandarin tones (phonemic judgement), and (2) whether it
was well pronounced (phonetic judgement). If a tone was
judged to be one of the Mandarin tones, the judges would
write down the tone number or tone mark on the evaluation
sheet. If it could not be identified as a Mandarin tone,
they would write a question mark. If a tone was
identifiable, but deviated from the norm, they would circle
the tone number or tone mark. For isolated syllables, the
judges were asked to write down each tone regardless of the
correctness of the tone. For 2-syllable words and short
sentences, the judges were asked to write down only the

incorrect or deviant tones (for example, if a tone 1 was

128
pronounced as a tone 4, or when a tone was not well
pronounced, or not identifiable), since they probably would
recognize the target word or sentence when they heard the
recording, even though it might not have been pronounced
completely correct. By not writing down all the correct
tones, they could therefore concentrate on the incorrect
ones.

The recordings total about seven hours in length, so
the evaluation was done at each judge's own convenience. It
was often necessary for them to listen to an item more than
once in order to make a judgement, so that it took much
longer than seven hours to complete the work.

Ioup and Tansomboon (1987) used a similar evaluation
procedure to evaluate tone production of learners of Thai.
Three native Thai judges rated tones in a taped conversation
and a humming test based on a three-point scale:

Level 1: phonemically and phonetically incorrect

Level 2: phonemically correct but with phonetic

deviation

Level 3: indistinguishable from native speakers

The three judges concurred completely on their
evaluation of the phonemic accuracy of tones, and they
agreed on 93 percent of their judgements in evaluating the
phonetic quality of the tones (p.340). This argues that

tone evaluation by native judges is a reliable procedure.

129
5.7 Computing the scores
The responses of each subject to each test item, and
the evaluations given by the three judges on each test item
for the production test, were entered in a computer file.
The scores were calculated by computer by comparing the

correct answers with the responses given by the subjects.

5.7.1 Perception test

There was only one correct answer to every item in the
tone perception test. The percentage of correct responses
in the two perception tests was computed separately for each
of the three parts--(A) isolated syllables, (B) 2-syllable
words, and (C) short sentences--and the percentage of
overall correct responses in the perception tests was also
computed. For 2-syllable words, the percentages of correct
responses for the first syllable and for the second syllable

were also computed separately.

5.7.2 Production test

For the production tests, each tone produced by the
subjects received one judgement from each of the three
judges. Two different methods were used to compute the
scores. First, the percentage of correct tone production in
each of the six parts (2 tasks [imitation and reading] x 3
contexts [(A) isolated syllables, (B) 2-syllable words, and

(C) short sentences]), was computed separately based on the

130
average score of the three judges. The percentage of
correct production of tones in the first and second
syllables of 2-syllable words and the percentage of correct
production of tones in the first (l-syllable) word, second
(1-syllable) word, and the third (1- or 2-syllable) word of
a sentence were also computed separately. Each subject also
had an overall correct tone production score for imitation
and reading. In computing the scores, only the phonemic
judgement was considered, and any phonetic deviation was
ignored. Thus a tone judged to be correct but not very good
would be the same as a tone judged to be correct and good.
However, it was often the case that a tone judged to be not
very good by one or two judges would be judged to be
incorrect by another judge. Thus the average of the three
judges could reflect how good the production of a subject
was.

Another way of calculating the scores was to take the
majority opinion (two out of three judges) in deciding the
correctness of each tone production. Thus a tone would be
considered Tone 2 if at least two judges judged it to be
Tone 2 even though one judge judged it to be Tone 3. If all
three judges disagreed with one another, then the tone would
be considered unidentifiable. This way of deciding the
answer is necessary to compare the error/success rates of
the four tones and to find the tone error patterns for each

tone. The percentage of correct production in each of the

131
six parts and the overall correct percentage for the two
tasks (imitation and reading) in the two tests were all also

computed this way.

5.8 Statistical Analyses

The statistical analyses were done with the SAS
statistical computer package on the IBM 3090 mainframe
computer at the Computer Center of Central Michigan
University, Mount Pleasant, Michigan. For all the
statistical analyses, the significance level was set at 0.95
(i.e., p<0.05). The production scores obtained by the two
different methods (the average score of the three judges and
majority opinion) were both analyzed when applicable.

Some experimental data were uncollectable: two subjects
took only part of the first perception test, four did not
take the second perception test, six did not take the second
production test and one took only part of it. These data
were considered as missing values. When analyses of
variance were computed separately for each subset of the
tests, the missing values in one subset did not affect the
analyses of the other subsets of the tests. However, in the
analyses of Correlations and Repeated Measures, because the
calculation was based on paired values in two subsets, all
the pairs with at least one missing value were treated as

missing values.

132
5.8.1 Analyses of the effects of some variables on tons
perception and tone production

The analysis of variance (ANOVA) was computed by using
the SAS generalized linear model (GLM) procedure to analyze
the effects of tonal context, task, sex and L2 experience on
perception and production scores. Three variables were
analyzed for the perception test: tonal context (isolated
syllables, 2-syllable words, short sentences), sex (male,
female), L2 experience (no L2 learning experience, previous
Mandarin learning experience, other L2 learning experience).
For the production test, the above three variables (tonal
context, sex and L2 experience) plus the variable of task
(imitation, reading) were analyzed. The significance level
was set at 0.95.

Tukey's studentized range (HSD) test at the
significance level of 0.95 was used with the ANOVA to
compare the means of the multiple levels of each variable.
Thus, if a variable was found significant, the way in which
the means of the dependent variable was affected by the
multiple levels of the variable could be determined.

The ANOVA was first performed separately on each
subset. When the repeated measures were performed to
compare the corresponding subtests in the first and the
second tests, the ANOVA was again performed on each of the
two corresponding subsets, but the missing values in one

subset would cause the corresponding values in the other

133
subset also to be treated as missing values. No missing

values were used in the repeated measures.

5.8.2 Analysis of correlations between the perception and
production of tone in isolated syllables and in sequences

The SAS CORR procedure was used to compute the
correlation coefficients (Pearson's r) between all the
subsets of both the perception and production tests (e.g.,
reading of isolated syllables in the first test would be one
subset). The significance level was set at 0.95. The
relationship between the perception or production of tone in
isolated syllables and tone in sequences (2-syllable words
or short sentences) would be determined by examining the

correlation coefficient obtained between them.

5.8.3 Analysis of correlations between tone perception and
tone production

The relationship between the perception and production
of tone in the same tonal context (e.g. 2-syllable words)
would be determined by the correlation coefficient obtained
between them using the SAS CORR procedure. The significance

level was set at 0.95.

134

5.8.4 Analysis of improvement in tone perception and tone
production

To determine whether improvement occurred between the
first test (Time 1) and the second test (Time 2), the SAS
GLM procedure with a Repeated option (Time as the repeated
measure) was performed on the perception and production data
sets as well as on each subset of the perception and
production tests. Missing values at one time would cause
the corresponding values at the other time to be treated as
missing also. No missing values were used in the repeated
measure analysis. The significance level for this analysis

was set at 0.95.

5.8.5 Analysis of order of difficulty among the Mandarin
tones

To determine whether the four Mandarin tones were
equally difficult for subjects to perceive or produce, the
SAS FREQ procedure with a chi-square test was used to
compute the two-way cross-tabulation tables of the number of
correct/incorrect responses to each of the four tones in
each test subset. The Pearson chi-square values obtained by
the FREQ procedure were used to determine if the correct/
incorrect responses were affected by the variation of
Mandarin tones. The significance level was set at 0.95 for
the chi-square test. To determine whether the pattern of

correct/incorrect responses to each of the four tones in

135
each part remained the same in both the first and second
tests, the SAS FREQ with the Cochran-Mantel-Haenszel (CMH)
option was used to compute the two-way cross-tabulation
tables in corresponding subsets of the two time-varied
tests. The general association values obtained by the FREQ
procedure were used to determine if the pattern of
correct/incorrect responses to the four tones in each subset
remained the same in the first and the second tests. The
significance level was set at 0.95 for the general

association values.

6. RESULTS AND DISCUSSION

6.1 Test validity and reliability

As controls for the perception test, three native
speakers of Chinese were asked to take both the first and
the second perception tests. One control's answers were
completely correct. Two made one mistake each on their
tests. However, when they were asked to do part of the test
again (which included the item on which they had made the
mistake), they changed their answers. Even with the
mistakes, the percentage of correct answers of native
speaker controls was over 99%.

For the production test, the judges agreed almost
completely on the two native speakers' production. Their
production was judged to be all correct and good except for
one isolated syllable, which was judged to be not very good
by one judge. The two native speakers were recognized
easily by the other two judges besides the author.

The production of two subjects was randomly chosen to
be judged twice. A comparison was made between these two
judgments by each of the three judges. The average
consistency rate for these judgments of the three judges was
95:1.66%, when only the phonemic judgment was considered
(96.1Sil.66% for Judge 1, 93.85i3.72% for Judge 2, and
97.1112.03% for Judge 3). If phonetic judgment was also

taken into consideration, then the three judges had an

136

137
average consistent rate of 93.46 12.75% (93.66il.7% for
Judge 1, 92.50i3.84% for Judge 2, and 94.23i2184% for
Judge 3).

For all subjects, the three judges agreed on 89% of
their judgments on the phonemic identification: Judge 1 and
Judge 2 agreed on 90.15% of their judgment, Judge 1 and
Judge 3 agreed on 92.16%, and Judge 2 and Judge 3 agreed on
91.39%. Because the phonetic judgments of the production
was not used in computing the scores, we did not calculate
the percentage of agreement among judges on the phonetic
judgments. Obviously it would be lower than the percentage
of agreement on phonemic judgments only.

Subjects' scores, based on the average of the three

judges, can be found in Appendix B.

6.2 Variables affecting tone perception and production

The effects of tonal context, task, sex, and L2
experience on tone perception and production were determined
by the results of ANOVA, with Tukey's test revealing the
difference between the various levels of each variable.

However, the results of Tukey's test are not always
consistent with those of ANOVA. A variable considered
significant in ANOVA (usually with p close to 0.05)

sometimes turns out to be insignificant in Tukey's test.

138

6.2.1 Tonal context
Perception

The ANOVA procedure revealed that subjects performed
significantly differently (p=0.0001) in the three tonal
contexts ((A) isolated syllables, (B) 2-syllable words,
(C) short sentences) in the first perception test. Tukey's
test revealed that the differences were significant between
all three contexts (A and B, B and C, as well as A and C).
However, the differences between the three contexts in the
second perception test narrowed and the results of ANOVA did
not reach significance. The mean scores of tone perception

in the three contexts in the two perception tests are shown

below:

A B C
lst perception 85.52 64.69 50.40
2nd perception 92.78 87.96 85.19

This change was also reflected in a repeated measures
analysis, which showed significant Time*Context interaction

effect (p=0.0001).

Production

There was a significant context effect in both the
first and the second production tests. In the first test,
the difference between part A (isolated syllables) and part
B (2—syllable words) as well as part A (isolated syllables)
and part C (short sentences) was significant. The

difference between part B and part C was not significant.

139
In the second test, only the difference between part A and
part B remained significant. The difference between part A
and part C was no longer significant.

There was also a significant Context*Task interaction
effect found in both the first and the second test. When
the two tasks were analyzed separately, the Context factor
was significant only in the reading task in both the first
test (p=0.0001) and the second test (p=0.0013), but not in
the imitating task. The mean scores of reading and

imitation are shown below:

A B C
lst reading 93.15 72.31 76.04
2nd reading 88.67 74.18 78.61
lst imitation 98.52 96.88 95.70
2nd imitation 97.17 97.17 95.06

Hypothesis 1, "Subjects perceive and produce tones with
significantly higher scores in isolated single syllables
than in two-syllable words and sentences," is confirmed for
perception in the first test and for reading production, but
not for imitation. The results indicate that L2 tone
perception and reading production, just as L2 sound
perception and production, are affected by different
phonological contexts (Flege 1988:278, 315, see 3.1.1.1).

As expected, tone in isolated syllables was perceived and
produced with significantly higher scores than tone in 2-
syllable words and sentences, just as L2 sounds in isolated

words are perceived and produced better than in longer

140
utterances (Shimizu and Dansuji 1983, L. Dickerson 1975,
Borden et al. 1983). However, this context effect soon
disappeared for perception. At the end of the term,
although the perception scores in isolated syllables are
still higher than those in 2-syllable words and short
sentences, the difference did not reach significance. Thus
the effect of different tonal contexts (as defined in this
study) on tone perception decreased in less than two months
as learning progressed.

It should also be noted that while the perception of
tone in 2-syllable words was much better than that in short
sentences at the beginning, the reading of tone in short
sentences was a little better than that in 2-syllable words.
The fact that all short sentences had been taught in the
class, but most words had not, may have some effect on the
scores. Thus, for the production aspect, difference in
familiarity with the test items seems to have affected
subjects' performance. Consequently, the context effect
that may normally be expected (tone in 2-syllable words is
produced better than tone in sentences) did not show up.
For the perception aspect, familiarity with the test items
did not seem to have the same effect as in production,
probably because it is necessary to produce the tone in
producing a sentence but it is not necessary to perceive the
tone when listening to a sentence. Another possible

confounding factor is that perceiving tone in 2-syllable

141
words and in short sentences are different tasks (see 5.1)
and the task of perceiving tone in short sentences might be
more difficult.

6.2.2 Task

The variable task (imitation and reading) was analyzed
for the production test. Results of ANOVA showed that the
differences between imitation and reading were significant
(p<0.02) in all three contexts (isolated syllables, 2-
syllable words, and short sentences) in both the first and
the second production tests. The mean scores on the two

tasks are listed below:

lst test A B C Mean
imitation 98.52 96.88 95.70 97.03
reading 93.15 72.31 76.04 80.50
2nd test

imitation 97.17 97.17 95.06 96.48
reading 88.67 74.18 78.61 80.60

Hypothesis 2, "Task, imitation or reading, affects
subjects' performance in tone production: imitation results
in significantly higher scores in tone production than
reading," is clearly confirmed. The discrepancy between
reading and imitation was greater in tone sequences (2-
syllable words and short sentences) than in isolated-
syllable tones.

This result is similar to that obtained by Snow and
Hoefnagel-Hohle (1977, naturalistic study) from L2 learning

in an L2 environment, which also showed higher scores of

142
pronunciation in imitated than in non-imitated speech (in
response to a picture) (see 3.1.1.2). However, their data
also showed a tendency for non-imitation scores to improve
faster and gradually catch up with the imitation scores as
learning progressed. The lack of improvement in reading
scores in the present study (see 6.5.2 below) does not show
the same tendency for tone learning.

Nevertheless, there were other signs that reading and
imitation were getting closer. Examining the correlation
between reading and imitating tone in the same context in
the first production test, we found no significant
correlation between the two tasks. However, in the second
production test, a significant correlation was found between
reading and imitating of isolated syllables (r-0.66208,
p=0.0003), and a lower correlation was found between reading
and imitating of short sentences (r=0.46459, p=0.0222).

Also found significant was the correlationbetween
reading of tone in isolated syllables in the first test and
the imitation of tone in isolated syllables in the second
test (r=0.61164, p=0.0012), a correlation better than that
between imitation in the two tests. A significant
correlation was also found between the reading of short
sentences in the first test and the imitation of short

sentences in the second test (r80.60441, p=0.0018).

143
Thus, the imitation of tone in isolated syllables and
short sentences seemed to become more related to the reading

of tone in corresponding contexts as learning progressed.

6.2.3 Sex
Perception

Results of ANOVA on the perception tests showed that
the difference between male and female was significant
(p=0.03l9) in the first perception test. The T-test
procedure on each of the three subtests, however, revealed
that only part C (short sentences) showed a significant
difference between male and female (p=0.0413). The sex
difference in part A (isolated syllables) and part B (2-
syllable words) did not reach significance. In the second
perception test, sex difference did not reach significance

in any of the three parts. The scores are listed below:

lst perception A B C Mean
male 89.00 68.33 57.50 71.62
female 81.79 61.04 43.75 61.35

2nd perception
male
female

95.42
90.67

90.63
85.83

88.54
82.50

91.52
86.33

However, when a repeated measure analysis was done, and

no observations with missing values were used, then sex was

not a significant variable in either of the two perception

tests.

This suggests that the significant difference

between the sexes was probably due to the small number of

subjects which does not represent the general population of

144
students. Perhaps the few individuals who dropped the
course later in the term or who came late and missed part of

a test affected these results significantly.

Production

Based on the result of ANOVA, sex should be a
significant (p=0.0319) variable in the first production
test. However, Tukey's test indicated that the difference
between males (mean=89.41) and females (mean=88.15) did not
reach significance at the 5% level. The second production
test did not show a significant sex difference.

When each part was analyzed separately, only part (A)
(isolated syllables) of the first production test showed a
significant sex difference (p=0.0416) according to ANOVA.
Again, the difference was significant only in ANOVA, Tukey's
test showing the difference between males (mean=96.39) and
females (mean=95.31) to be insignificant at the 5% level.

The scores are shown below:

lst production A B C Mean
Male 96.39 85.90 85.96 89.41
Female 95.31 83.37 85.79 88.15
2nd Production

Male 94.27 86.55 87.18 89.20
Female 91.67 85.34 86.42 88.06

When each part was analyzed according to Task
(imitation or reading), we found that the imitation task of
part A (as well as part B and part C) did not show a

significant sex difference. Only the reading task, in part

145
A of the first production test, showed a significant
(p=0.0490) sex difference. Again, the difference between
males (mean-94.44) and females (mean-91.93) did not reach
significance in Tukey's test. The scores on the reading

task are shown below:

lst reading A B C
Male 94.44 74.72 77.07
Female 91.93 70.05 75.08
2nd reading

Male 89.24 74.43 77.46
Female 88.14 79.59

73.96

When male and female production data were analyzed
separately, for the first test, the same significant
differences were found in Task (imitation or reading) and
Context (isolated syllables, 2-syllable words, or short
sentences), as well as the Task*Context interaction (see
Task and Context above). However, there were also
significant differences found only in the female data in L2
Experience (p=0.0034) and L2*Task interaction (p=0.0176),
but not in the male data. For the second test, only Task
remained a significant factor for both male and female
production. A Time*Context interaction was also found for
both male and female data.

Hypothesis 3, "Subjects' performance in Mandarin tone
perception or production is not affected by their sex,"
cannot be clearly confirmed or rejected. The difference
between sexes was greater in the first test than in the

second test. While the difference in the second test was

146
insignificant, the significance of sex difference in the
first test cannot be definitely determined because the
results from two statistical procedures were inconsistent.
The seemingly better performance in males than females in
tone learning is opposite to the findings of some other
studies on L2 pronunciation, which showed that females had
better pronunciation than males (Asher and Garcia 1969,
Thompson 1984). However, in the second test, females seemed
to perform a little better than males in reading short
sentences, which had been taught in class and were most
familiar to subjects (see 5.3), although the difference was
not significant. This might suggest a greater improvement
for females than males in reading test items they were most
familiar with. More research is needed to determine whether
sex has an effect on tone perception and production by L2

learners.

6.2.4 L2 experience
Perception

ANOVA on the perception test indicated that L2
experience did not make a significant difference in either
the first or the second perception tests. Subjects with
Mandarin learning experience or with other L2 learning
experience performed better than subjects with no second
language learning experience: however, the difference was

not significant.

147
Production

L2 is not a significant factor in the subjects' tone
production performance in either of the two production
tests. When ANOVA was carried out for different contexts
and tasks, the same insignificant results were found, with
only one exception. A significant result (p=0.0013) was
found in the imitating of short sentences in the second
production test. Tukey's test showed that the significant
difference occurred between scores of subjects with Mandarin
learning experience (mean=89.00) and scores of subjects with
no L2 learning experience (mean=96.67) as well as scores of
subjects with other L2 experience (mean-96.17). Subjects
with previous Mandarin learning experience scored
significantly lower than the other two groups.

L2 experience No L2 Other L2 Mandarin
96.67 (n-4) 96.17 (N-l6) 89.00 (N=4)

L2 experience and L2*Task interaction were also found
to be significant in the female data in the first test, but
not in the male data. However, the number of subjects is
probably too small to make a valid generalization.

Hypothesis (4A), "Subjects with prior Mandarin learning
experience score significantly higher in Mandarin tone
perception and production than those without such
experience," was not confirmed. There were no significant
differences between subjects with and without prior Mandarin

learning experience in any of the perception and reading

148
production tests. The only significant difference found
between the two groups was in imitating sentences, and the
scores were contrary to expectation: subjects with prior
Mandarin learning experience scored significantly lower than
subjects without such experience. Remember that imitation
scores were not much related to perception and reading
scores at the beginning, but the scores on imitating single
syllables and short sentences became closer to counter
perception and reading scores in the second test (see
6.2.2).

It appears that two possible factors may account for
these lower scores of subjects with prior Mandarin learning
experience. First, when subjects were very familiar with
the test items, they may already have had a "prototype" in
their mind and they paid less attention to the auditory
model they were imitating. Second, they may already have
formed a habit of producing tones in their own way and it
became harder to modify this. In fact, it has been observed
that L2 learners with prior learning experience with non-
native speakers in a foreign language environment often have
worse pronunciation than thoSe who do not have such
experience (Seliger et al. 1975:22). The lower scores of
subjects with prior Mandarin learning experience may be
considered consistent with the trend of closer relationship
between imitation and non-imitation in utterances they had

been taught. It has been found that greater variation in

149
production is characteristic of an active learning process,
and the variation tends to be reduced resulting in an
increased amount of within-subject stability (Gass 1984).

The insignificance of Mandarin learning experience on
perception seemed to contradict the result concerning Time
effect found in the perception tests. Among the seven
subjects who had had Mandarin learning experience, their
experience ranged from less than one quarter to one year of
high school, and may have included classes like "Chinese
culture and language", so it was not clear what they had
learned before coming to the Chinese 101 class in Fall 1989.
The great variation found in tone perception and the small
number of subjects in this group may also account for the
insignificant result.

Hypothesis (48): "For subjects without Mandarin
learning experience, other L2 learning experience does not
affect performance in tone perception or production," was
supported by the results obtained. This finding is
consistent with the findings of other studies on L2
pronunciation, which found no significant effect of other L2
experience (Suter 1976, Thompson 1984). However, since L2
learners have been reported to rely on prior-learned L2 in
learning additional related languages (Rivers 1979,
Singleton and Little 1984), it is possible that prior L2
phonological learning will have some effect on learning

additional similar L2 phonological structure. Thus the

150
learning of non-tone languages may not have much effect on
the learning of a tone language, the learning of another
tone language may have greater effect on the learning of
additional tone languages. However, as in many aspects of
L2 learning, similarity may result in transfer, both
negative and positive, and because of the interaction of
tone and intonation, it may not be easy to sort out the

effect of prior L2 learning experience.

6.3 Relationship between the perception and production of

tone in isolated syllables and in sequences.

6.3.1 Perception

In the first perception test, the perception of tone in
isolated syllables correlated with the perception of tone in
2-syllable words (r30.77826, p=0.0001), especially with
perception of tone in the second syllable (r=0.84243,
p=0.0001). The correlation with the first syllable was also
highly significant but not as high (r30.52571, p=0.0001).
The correlation between the perception of tone in isolated
syllables and tone in short sentences was also significant,
but even lower (r=0.47642, p=0.0090). There was also a
significant correlation between perception of tone in 2-
syllable words and short sentences (r=0.56774, p=0.001l):
this correlation was higher than that between tone in

isolated syllables and tone in short sentences.

151

In the second perception test, the perception of tone
in isolated syllables correlated with the perception of tone
in 2-syllable words (r=0.60031, p=0.0009). However, unlike
in the first test, the correlations with the first syllable
(r=0.57955, p=0.0015) and with the second syllable
(r=0.55999, p=0.0024) were not much different. The
correlation between the perception of tone in isolated
syllables and the perception of tone in short sentences was
no longer significant. But the correlation between the
perception of tone in 2-syllable words and short sentences
continued to be significant (r=0.52464, p=0.0050).

In addition, the perception of tone in isolated
syllables in the first test correlated with the perception
of tone in sequences in the second test. The correlation
was highly significant with the perception of tone in 2-
syllable words (r=0.79424, p=0.0001) and significant with
the perception of tone in short sentences (r=0.47437,
p=0.0166) in the second test.

The perception of tone in 2-syllable words in the first
test also correlates with the perception of tone in short
sentences in the second test (r=0.58891, p=0.0016). The
perception of tone in short sentences in the first test
correlated with the perception of tone in 2-syllable words
in the second test (r=0.42983, p=0.0252).

The perception of tone in isolated syllables in the

first and the second tests correlated significantly

152
(r=0.60779, p=0.0013): so did the perception of tone in 2-
syllable words (r=0.68293, p=0.0001). But the perception of
tone in short sentences did not correlate significantly
between the two tests.

The results indicate that the perception of tone in
isolated syllables at the beginning of the acquisition
process correlates with the perception of tone in sequences
at the beginning as well as at a later time. The
correlation did not change over the seven weeks of learning,
despite improvement in scores in tone perception of 2-
syllable words and short sentences. On the other hand,
improvement in the perception of tone in isolated syllables
results in weakening (2-syllable words) or insignificance
(short sentences) of the correlation of this with the
perception of tone in sequences in the second test.

The first part of Hypothesis 5, "There is positive
correlation between perception of tone in isolated single
syllables and syllable sequences", was generally confirmed.
The perception of tone in isolated syllables was definitely
related to the perception of tone in 2-syllable words. It
was also correlated with the perception of tone in
sentences, but only at the very beginning of the acquisition
process, and the correlation was not high. On the other
hand, the correlation between tone perception in 2-syllable
words and short sentences was stronger than that between

tone perception in isolated syllables and in sentences,

153
suggesting that the sequencing of tone, whether it was in 2-
syllable words or in sentences, was the common factor. The
correlation between the perception of tone in 2-syllable
words at the beginning of learning and the perception of
tone in short sentences did not seem to change after seven
weeks of learning.

The assumption that perception of tone in isolated
syllables is a prerequisite to the perception of tone in
sequences seems to be supported by the significant
correlation found between them. However, the results also
show that the correlation became lower when the tone
sequences became increasingly longer and structurally more
complicated. It should be remembered that only short
sequences of tone, which supposedly resembles tone in
citation form, was tested in this study. The insignificance
of correlation between tone perception in isolated syllables
and short sentences in the second test may be partially due
to the high scores on tone perception in isolated syllables,
but it is also possible that the teaching of these short
sentences in the class has somehow affected the correlation.
Thus after L2 learners had been taught and had been actively
learning certain constructions of L2 for some time, their
performance on these constructions may have been strongly
affected by variables other than their perception of the

citation form.

154
6.3.2 Production
Since the results of the two production tasks, reading
and imitating, are quite different, they will be discussed

separately.

6.3.2.1 Reading task

The reading of tone in isolated syllables did not
correlate significantly with the reading of tone in 2-
syllable words or short sentences in the first test or in
the second test. On the other hand, the reading of tone in
2-syllable words and in short sentences were significantly
correlated in the first test (r-O.76239, p-0.0001) and still
significant though lower in the second test (r=0.47544,
p=0.0189).

The reading of tone in 2-syllable words in the first
production test also correlated with the reading of tone in
short sentences in the second test (r=0.68713, p=0.0002) and
with the reading of tone in isolated syllables in the second
test (r=0.40045, p=0.0473). Note that scores for the reading
of tone in isolated syllables decreased in the second test.

The reading of tone in short sentences in the first
test correlated with the reading of tone in 2-syllable words
in the second test (r=0.78581, p=0.0001).

Between the two tests, the reading of tone in parallel
contexts were always correlated. The reading of isolated

syllables was correlated (r=0.62552, p=0.0008) as was that

155
of the 2-syllable words (r-O.83865, p=0.0001) and short
sentences (r-O.77655, p=0.0001).

The second part of Hypothesis 5, ”There is positive
correlation between production of tone in isolated syllables
and syllable sequences", was not confirmed for reading.
Production of tone in isolated syllables did not correlate
significantly with production of tone in sequences. Thus,
the ability to produce Mandarin tone in isolated syllables
was not related to the ability to produce Mandarin tone in
sequences. As mentioned in 4.1.1, L1 transfer may have a
different effect on the production of tone in isolated
syllables, 2-syllable words, and sentences. The results
suggest that L1 transfer, or some other factors that may
affect tone production, did not affect all L2 learners in
the three tonal contexts in the same way. However, the
production of tone in 2-syllable words and short sentences
were very much related at the beginning. The relation
became less close at the end of the term. Subjects'
increased familiarity with the test items in short sentences
through extended practice during the term (see 5.3) may have
affected the relation between producing tone in 2-syllable

words and short sentences.

6.3.2.2 Imitation task
The imitation task showed no significant correlation

between any contexts in either the first or the second test.

156

Between the two tests, the imitation of tone in
parallel contexts was not significantly correlated except
for isolated syllables (r-0.57763, p=0.0025).

Hypothesis 5 was not confirmed for imitation, either.
Imitation seemed to have reached a near peak of performance
in all three parts of the test from the beginning. (Scores
do not improve, presumably because they were so high on the
first test that there was little if any room for
improvement. Such absence of improvement is a so-called
ceiling effect). This may explain the lack of significant
correlation of imitation with reading and perception, and
the lack of correlation between the two tests or different

parts of the same test.

6.4 Relationship between tone perception and tone production
Relationship between tone perception and tone
production will be discussed according to the three tonal

contexts.

6.4.1 Tone in isolated syllables

For tone in isolated syllables, a significant
correlation was found between perception and the reading
task in the first test (r=0.63099, p=0.0002). The
correlation between perception and the imitation task was
not significant. Perception of tone in isolated syllables

in the second test correlated with both reading (r=0.61137,

157
p=0.0012) and imitation (r-0.72245, p=0.0001) of tone in
isolated syllables.

There were also significant correlations found in the
perception and production of isolated syllables across the
two tests. A significant correlation was found between the
perception of isolated syllables in the first test and the
production of isolated syllables in the second test for
reading (r=0.72925, p=0.0001), and for imitation (r-0.78515,
p=0.0001). The perception of isolated syllables in the
second test correlated significantly with the production of
isolated syllables in the first test for reading (r=0.56153,
p=0.0023) and imitation (r=0.48150, p=0.0110).

The perception and production of tone in isolated
syllables were positively correlated (except for imitation
in the beginning). This result is similar to previous
findings on the correlation between perception and
production of L2 sounds (Goto 1971, Snow and Hoefnagel-Hohle
1979, Borden et al. 1983, Schneiderman et al. 1988).

It is interesting to note that the correlation between
the first perception test and the second production test
seems higher than that between the first perception test and
the first production test. There may be a delayed effect
between the development of perception and production of
tone, and a similar effect has been proposed between input

frequency and accuracy order relationship on other aspects

158
(morpheme, syntactic structure) of L2 learning (e.g., Skehan

1982, cited in Larsen-Freeman and Long 1991:134).

6.4.2 Tone in 2-syllable words

The correlation between perception and reading of 2-
syllable words was significant in the first test (r=0.58289,
p=0.0007) and in the second test (r=0.69208, p=0.0002).
Imitation did not correlate significantly with perception in
either test.

There were also significant correlations found between
perception and production across the two tests. The first
perception test correlated with the second reading test
(r=0.64227, p=0.0010), and the second perception test also
correlated with the first reading test (r=0.67733,
p=0.0001). Again, the correlation between the first
perception test and the second production test seems higher
than that between the first perception test and the first
production test, suggesting a delayed effect, though the

difference was not as great as in isolated syllables.

6.4.3 Tone in short sentences

The correlation between perception and reading of tone
in short sentences was significant (r=0.44455,p=0.0122) in
the first test but not significant in the second test. The
imitation task did not show significant correlation with the

perception in either test.

159

Tone perception in short sentences in the second
perception test, however, showed significant correlation
with tone perception in short sentences in the first reading
test (r=0.47454, p=0.0124).

The percentage of overall correct tone perception and
tone production in the reading task and the imitation task
was also computed. The correlation between the perception
and reading of tone was significant for the first test
(r=0.62424, p=0.0003) and the second test (r=0.69484,
p=0.0002). No significant correlation was found between
perception and imitation in either of the two tests. The
first perception test also correlated with reading in the
second test (r=0.72090, p=0.0002) and imitation in the
second test (r-0.52237, p-0.0126). The second perception
test correlated with reading in the first test (r=0.63795,
p=0.0003).

There was a significant correlation between the first
and the second perception tests (r-O.77090, p=0.0001) and
between reading in the first and the second tests
(r=0.95260, p=0.0001). But imitation in the two tests was
not significantly correlated. Reading and imitation tasks
were not significantly correlated in the first test and
significantly but not highly correlated in the second test
(r=0.43977, p=0.0315).

Hypothesis 6, "There is positive correlation between

subjects' Mandarin tone perception scores and production

160
scores," was generally confirmed between tone perception and
tone reading but not for imitation. The significant
correlation between perception and reading scores indicates
that the two aspects of tone learning are very much related,
a finding consistent with the findings on the perception and
production of L2 phonology in general (see 3.1.1). The
significant correlations often found between perception and
reading across the two tests, sometimes higher than those
found in tests conducted at about the same time, also
suggest that there were interactions between them which
occurred over time.

The only exception was found between perception and
reading of short sentences in the second test. The
correlation concerning short sentences was weak at the
beginning and became insignificant after one term of
learning. Remember that all short sentences had been taught
in the class. The lack of correlation in short sentences at
the end of the term reminds us of the results reported by
Schneiderman et al. (1988), who found significant
correlation between discrimination and production of French
phones in learners of French before they underwent
discrimination training, but no significant correlation
after the training. Since we have very little understanding
of how L2 phonological perception and production interact in
L2 learners, and how they are affected by classroom

instruction, we cannot explain how classroom instruction may

161
have affected the relation between perception and production
of L2 phonology. However, the results obtained here suggest
that instruction has some effect on the relationship between

L2 phonological perception and production.

6.5 Improvement over one term of learning
6.5.1 Perception

A repeated measure analysis for perception showed a
significant time effect between the first perception test
and the second perception test (p=0.0001). There was also a
significant Time*Context interaction (p=0.0001) as mentioned
in 6.2.1. When a repeated measure was done for each subtest
separately, the same significant time effect was found in
all three parts (p<0.05). The scores of the two tests are

shown below:

A B C Mean
lst test 86.40 65.38 49.07 66.48
2nd test 92.60 87.74 85.19 88.41

The perception aspect of Hypothesis 7, "Subjects'
scores in the perception of Mandarin tone are significantly
higher at the end than at the beginning of the term", is
confirmed. This indicates that English-speaking learners'
perception of Mandarin tones improved in about seven weeks
(five 50-minute classes per week) of Mandarin learning. The
significant time effect found in all three parts of the

perception test means that improvement of tone perception

162

occurred generally in isolated syllables, 2-syllable words,
and short sentences.

L2 learners' L2 phonological perception has been found
to improve with their L2 experience (MacKain et al. 1981)
and some of them can achieve native-like perceptual ability
(Neufeld 1980, see 3.1.4.1). However, the effect of short-
term perceptual training seems limited (e.g., Pimsleur
1963). For studies on the perception of Mandarin tones by
English-speaking learners, Kiriloff (1969) found no
significant improvement for tone in isolated syllables
between two tests, the first conducted after two and a half
months of Mandarin learning and the second after four months
of learning. In the present study, the second perception
test was conducted ten weeks (about two and a half months)
after (most) subjects started learning Mandarin. It is
possible that significant improvement can only occur before
two and a half months of learning, and that after two and a
half months (at least till four months) no significant
improvement may be made.

W.J.C. Lin (1985) found that for a syllable embedded in
a sentence, beginners (150 contact hours) made more errors
in tone identification than intermediate (600 contact hours)
students, who in turn made more errors than advanced (1050
contact hours) students. Most of the subjects in the
present study had had only about 50 hours of instruction

when the second perception test was conducted. They seemed

163
to have made great progress in their perception of tone in
short sentences in a relatively short time. If the test
scores reflect the perceptual ability for tone of L2
learners, English-speaking learners seemed to make greater
improvement in short sentences than in 2-syllable words, and
they made greater improvement in 2-syllable words than in
isolated single syllables. Of course, there was much more
room for improvement in short sentences than in 2-syllable
words, and in 2-syllable words than in isolated single
syllables.

However, we should remember that the two perception
tests were conducted under different situations: the first
in a language lab using an audio-tape pre-recorded by a
speaker whose voice the subjects were not familiar with, and
the second in a classroom read by an instructor whose voice
the subjects were familiar with. It is possible that some
other variables might also be responsible for the
significantly higher scores in the second test. If the
perception aspect of hypothesis 7 is to be tested again in
future studies, the testing situations should be made as

similar as possible.

6.5.2 Production
A repeated measure analysis showed that the difference
between the first and the second production tests was not

significant. When the scores of the two tasks, reading and

164
imitation, were analyzed separately, there was still no
significant difference in either the imitation task or the
reading task. The repeated measure analysis was also
carried out on each of the subtests [(A) isolated syllables,
(B) 2-syllable words, (C) short sentences] separately, and
none of the parts showed significant time effect. The

scores are listed below:

Reading Task A B C Mean
lst test 93.15 72.31 76.04 80.50
2nd test 88.67 74.18 78.61 80.60
Imitation task

lst test 98.52 96.88 95.70 97.03
2nd test 97.17 97.17 95.06 96.48

The production aspect of Hypothesis 7, ”Subjects'
scores in production of Mandarin tone are significantly
higher at the end than at the beginning of the term," is not
confirmed for either reading or imitation. This means no
improvement in tone production was made between the
beginning and the end of the term. This result in imitation
can be attributed to the ceiling effect reached from the
beginning of the term.

It is widely observed that adult L2 learners often
cease to improve well before achieving native-like
pronunciation, and a lack of improvement in L2 pronunciation
has been reported in many aspects of L2 phonology (e.g., VOT
in stops and vowel quality, see Flege and Hillenbrand 1984).
Studies have also found that the amount of formal classroom

training is not a significant variable predicting L2

165
pronunciation (e.g., Purcell and Suter 1980). However, the
lack of significant improvement in the production of tone at
the beginning (the third to the tenth week) of Mandarin
learning may still be a little surprising. After all, adult
L2 learners have been reported to be able to produce
phonemic distinctions better with increased L2 experience
(Beebe 1987), including tone (Ioup and Tansomboon 1987), and
there is some evidence that adult L2 learners' pronunciation
improves the most at the beginning of their L2 learning
(e.g., Snow and Hoefnagel-Hohle 1979).

If the results in this study are valid, then one
possible explanation is that most of the learning had taken
place before the first test (the third week of the class).
It is also possible that there was improvement which could
not be measured easily by impressionistic judgment of the
correctness of a tone. For instance, there may be a trend
from more variation to less variation (Gass 1984:70), or
from more phonetic deviation to less phonetic deviation.

The results suggest that improvement in tone learning
occurs more rapidly in the perception aspect. It seems that
at the beginning of Mandarin learning, tone production is
ahead of perception (even if we disregard imitation scores
here, which did not appear to be related much to learning).
Students could pretty much read an isolated syllable with
correct identifiable tone but made more errors in

identifying it. This finding is similar to that concerning

166
the production and perception of the /r/-/l/ contrast by
Japanese speakers (Goto, 1971: Sheldon and Strange, 1982).
Just as pronunciation training may have helped adult L2
learners producing sound contrasts which they could not
perceive reliably, it is possible that the teaching of tone
in isolated syllables, including specific description of the
tonal contour of the four tones, may have helped many of the
learners to produce tones better before they were able to
perceive the distinction reliably.

While producing acceptable Mandarin tones in isolated
single syllables did not seem to be too difficult to achieve
for most adult English speaking learners, tone production in
2-syllable words and short sentences was much more
difficult: however, reading was still better than perception
at the beginning of the term. At the end of the term,
perception of tone in 2-syllable words and short sentences
improved significantly while the reading scores had not
changed much. As a result, the perception of tone was

better than reading at the end of the term.

6.6 Order of difficulty of the four tones
Recall the nature of the four tones in Mandarin:
Tone 1 high-level ( 1)
Tone 2 high-rising ( 1)
Tone 3 low-dipping ((A) in final position

low-falling (.J) in non-final position

167
and not followed by another Tone 3
high-rising ('4) when followed by Tone 3
Tone 4 high-falling (\J)
Comparison of the error/success rates of the perception or
production of the four Mandarin tones in a given test often
reveals an order of difficulty of the four tones. The
success rates of the perception of the four tones in the
perception tests were apparent, since there was only one
correct answer to each test item. Production, however, was
subjectively evaluated by the three judges and disagreement
was not uncommon. Therefore, the success rate of tone
production was based on majority opinion (two out of three
judges). A tone produced by a subject was identified by
judges as one of the four tones, or a neutral tone, or
unidentifiable. When at least two judges agreed on the tone
category (phonetic deviation was ignored), that would be the
final decision. Thus the correctness of a tone produced by
a subject or how it was mispronounced (e.g. Tone l
mispronounced as Tone 2) would be decided this way.

The percentage of correct identification and production
for the four tones in different contexts will be presented
below. The number of successes and errors of the four tones
for each task in each context was analyzed with a chi-square
test to see if there was any significant difference among
the four tones. The results were significant (p<0.05) for

most parts of the tests, indicating that the four tones had

168
significantly different success rates for perception or
production. When the success rates of the perception or
production of the four tones were not significantly
different, this will be made clear.

6.6.1 Perception
6.6.1.1 Tone in isolated syllables
The percentage of correct answers in the perception of

tone in isolated syllables is shown below:

Tone 1 Tone 2 Tone 3 Tone 4 Average
lst test 90% 72% 96% 83% 86%
2nd test 95% 90% 92% 95% 93%

In the first perception test, in isolated syllables
Tone 2 was the least successfully identified tone, while
Tone 3 was the most successfully identified. In the second
perception test, all four tones in isolated syllables were
correctly identified over 90% of the time. A chi-square
test showed that the difference in success rates of the
perception of the four tones was not significant.

Kiriloff (1969) also studied students' tone perception
in isolated syllables (see 2.3). He found that "errors
involving the 2nd tone are considerably higher and those
with the 1st and 4th tones much lower than expected" and
thus "[tJhe considerable number of incorrect identifications
of the second tone and a marked tendency to confuse with it
the third tone indicate that the second tone presents the

greatest difficulty to students of Mandarin" (p.65).

169

In this study, in agreement with Kiriloff's finding,
Tone 2 was also the least successfully identified tone in
isolated syllables in the first test, and continued to be in
the second test, though the difference between errors on
Tone 2 and on other tones was insignificant. Error rates of
the perception of Tone l and Tone 4 were not always much
lower. Tone 3 was correctly identified at a higher rate
than Tone 1 and Tone 4 in the first test. However, judging
from the lower identification rate for Tone 3 in the second
test and the error patterns of Tone 2 and Tone 3 in the two
tests (see 6.7.1), it seems that the high identification
rate of Tone 3 in the first test was due to the tendency of
subjects to identify both Tone 2 and Tone 3 as Tone 3. As
they learned to adjust the boundary between Tone 2 and Tone
3, Tone 2 was more correctly identified but Tone 3 was often
misidentified as Tone 2. Nonetheless, error rates for the
perception of the four tones in the second test were not

significantly different.

6.6.1.2 Tone in 2-syllable words

The percentages of correct answers in the perception of
tone in 2-syllable words, including the first syllable and
the second syllable as well as the overall scores, are

listed below:

170

lst test

Tone 1 Tone 2 Tone 3 Tone 4 Average
1st syll 68% 54% 46% 73% 61%
2nd syll 67% 51% 86% 72% 69%
Overall 67% 53% 69% 73% 65%
2nd test

Tone 1 Tone 2 Tone 3 Tone 4 Average
lst syll 93% 88% 86% 82% 88%
2nd syll 93% 84% 94% 82% 88%
Overall 93% 86% 91% 82% 88%

Chi-squire shows that in the first perception test, the
four tones were identified at significantly different rates
in both the first and the second syllable of 2-syllable
words. Tone 3 was more difficult to identify than the other
tones in the first syllable but much easier to identify than
the other tones in the second syllable. Tones l, 2, and 4
seemed to be correctly identified at a similar rate in the
first syllable and the second syllable. There are at least
two possible reasons for the different success rates in
identifying Tone 3 in different positions. The first is the
allophonic variation of Tone 3 in (emphasized) final and
non-final positions (see 2.1.2). Since the variation that
occurred in non-final position was different from that in
isolation, while the variation which occurred in final
position was the same as that in isolation, it is natural
that the tone which occurred in final position was
identified better than in non-final position. Another
reason for the better identification rate of Tone 3 in the
second syllable (final position) was probably due to the

same reason as in single syllables: both Tone 2 and Tone 3

171
tended to be identified as Tone 3. Overall, as when tone
was identified in isolated syllables, Tone 2 had the lowest
identification rate, and Tone 4 had the highest
identification rate.

In the second perception test, the chi-square test
shows that the successful identification rate for the four
tones was not significantly different in the first syllable
of 2-syllable words but was significantly different in the
second syllable, in which Tone 1 and Tone 3 were identified
more correctly than Tone 2 and Tone 4. However, Tone 3 still
had the biggest discrepancy of successful identification
rate between the first and the second syllable, though the
discrepancy was much smaller than that in the first test.
Tones 1, 2, and 4 in the first syllable and the second
syllable still seemed to be correctly identified at similar
rates. The overall identification rates of all four tones
in the second test were much better than those in the first
test. But now Tone I seemed to be the most successfully
identified tone and Tone 4 was the least successfully
identified, the result of much more improvement in the other

tones than in Tone 4.

6.6.1.3 Tone in short sentences
The percentages of correct perception of tone in

sentences are shown below:

172

Tone 1 Tone 2 Tone 3 Tone 4 Average
lst test 24% 50% 68% 52% 50%
2nd test 91% 72% 90% 85% 85%

The Chi-square test shows that the four tones had
significantly different success rates on both the first and
the second perception test. In the first perception test,
the success rates for tone perception in short sentences
were not high for all four tones, and the success rate of
Tone 1 perception was not higher than chance. In the second
test, however, success rates of identification for all four
tones greatly improved. Tone 2 was the least successfully
identified tone, and Tone I became the most successfully
identified.

W.C.J. Lin (1985) tested tone perception of his
students with two sentence frames: W0 shuo de shi __. "What
I said is __" and Tade __ hen youming "His/her __ is quite
famous". Only the syllable ma with four different tones was
tested (see 3.2.1). His test was somewhat similar to the
test of tone perception in short sentences in this study,
although the test in this study involved recognizing the
sentence with the correct tones rather than identifying the
tone of a syllable in a sentence. However, because of the
limitations mentioned in 5.3, in this study the four tones
did not always occur in the same position or with the same
syllable.

According to Lin's results Tone 3 was the most

problematic tone for his students to identify in both

173
sentence frames, and he suggested a new treatment of it.
However, even though the treatment seemed to be effective
and the students' performance improved, Tone 3 was still the
most difficult to identify. In the present study Tone 3 was
not the least successfully identified tone in short
sentences, though it was the least successfully identified
tone as the first syllable of 2-syllable words. The
difference of the test methods employed in the two studies
mentioned above may be part of the reason for this
difference.

In the study of Broselow, et al. (1987), Tone 4 was the
tone most correctly identified in isolated syllables by
American English speakers. However, Tone 4 was identified
at a significantly lower rate in non-final position than in
final position. This difference cannot be explained by
allophonic variation as can be that for Tone 3, and as was
argued as evidence for perceptual transfer of American
English intonation (see 3.2.1). Broselow et al.'s subjects
had no Mandarin Chinese learning experience and lacked
knowledge of the phonological characteristics of the four
tones (e.g., falling, level, etc.) other than the examples
they were given before the test. Their results may be
generally valid for American listeners with no Mandarin
learning experience, but presumably after learning the
Mandarin tonal system and being exposed to more examples,

the learners' perception would change. Tone 4 was not the

174
most correctly identified tone in isolated syllables in our
first perception test, less than three weeks after the class
began, and the positional difference was not found.

There may be more complicated factors involved in the
perception of tone. The position of the tone and the tonal
context can certainly affect perception. Other factors
include syllable structure and degree of familiarity with
the vocabulary. For example, examining the original data, I
found a particularly high error rate for tones in words with
the same syllable structure as words the students had
learned right from the beginning. The syllable haoz was
perceived to be Tone 3 (hao3 "good”) and the syllables n12
and n14 were perceived to be Tone 3 (n13 "you”) at a much
higher rate than other Tone 2 and Tone 4 words. The fact
that tests were conducted in the class may have lead the
students to expect that the words they have learned would be
more likely to occur in the test.

The actual variation of tone production the students
were exposed to may also have some effect. They may have
developed some criteria that were more effective in
identifying tones produced by speakers whose voices they
were familiar with. Since the students were not familiar
with the tape-recorded voice in the first perception test,
but were familiar with the voice of the instructor who read
the test items in the second perception test, their

perception could have been influenced by this fact.

 

175
6.6.2 Reading
In the production of tone, the two tasks of reading and
imitation produced quite different results and results on
the reading task were more closely related to those of

perception (see 6.2.2 and 6.4).

6.6.2.1 Tone in isolated syllables
The percentage of correct production of tone in

isolated syllables in the reading task is shown below:

Tone 1 Tone 2 Tone 3 Tone 4 Average
lst test 97% 92% 94% 97% 95%
2nd test 92% 68% 96% 96% 88%

In the first production test, the rates of correct tone
reading were not significantly different for the four tones,
and all four tones were correctly produced over 90% of the
time. However, the second test showed a dramatic change for
Tone 2: only 68% of Tone 2 produced by subjects was judged
to be correct. The other three tones did not show such

change from the first test.

6.6.2.2 Tone in 2-syllable words
The percentage of correct tone production in 2-syllable
words, both the first syllable and the second syllable as

well as the overall scores, is shown below:

lst test

Tone 1 Tone 2 Tone 3 Tone 4 Average
lst syll 73% 65% 52% 81% 69%
2nd syll 73% 83% 86% 64% 76%

Overall 73% 73% 71% 73% 73%

 

176

2nd test

Tone 1 Tone 2 Tone 3 Tone 4 Average
lst syll 91% 77% 48% 69% 72%
2nd syll 75% 75% 94% 60% 76%
Overall 83% 76% 71% 65% 74%

In the first production test, the overall success rates
for reading the four tones in 2-syllable words did not
differ significantly. However, the production of the four
tones had significantly different success rates in the first
syllable as well as in the second syllable. And except for
Tone 1, every tone had quite different success rates of
production in the two syllables. In the first syllable Tone
4 was the most successfully produced by subjects, and Tone 3
was the least successfully produced. The second syllable
showed exactly the opposite picture. Tone 3 was the most
successfully produced while Tone 4 was the least
successfully produced.

In the second production test, the reading of the four
tones had significantly different overall success rates.
Tone 1 was the most successfully produced tone, and Tone 4
was the least successfully produced. The success rates of
the production of the four tones in the first and the second
syllable were different too: Tone 1 was the most successful
in the first syllable and Tone 3 the least successful. Tone
3 was the most successful in the second syllable and Tone 4
the least successful.

One similarity between the two production tests was

that Tone 3 was the least successfully produced tone in the

177
first syllable but the most successfully produced tone in
the second syllable. Another is that Tone 4 was more
successfully produced in the first syllable than in the
second syllable, where it was the least successfully
produced among the four tones. This appears to be a
transfer from the English word stress pattern. The
similarity of unstressed syllables in English words and Tone
3 in Mandarin has been found in the perceptions of Mandarin
speakers (Cheng, 1968). Cheng found that English words with
an initial unstressed syllable would cause tone sandhi in
the preceding Tone 3 (e.g., Hao professor bu duo "There are
not many good professors"), just as a Tone 3 in Mandarin
would, while a word starting with a stressed syllable would
not. A 2-syllable English word usually has the stress on
the first syllable, and the second syllable is unstressed.
The stressed syllable usually is high in pitch, and the
unstressed syllable low in pitch. It appears that this
high-low pitch pattern was carried over to the 2-syllable
Mandarin words: Tone 3 (a low tone) was produced correctly
only about half of the time in the first syllable, but much
more frequently in the second syllable. The tone that was
most successfully produced in the first syllable always
starts high (Tone 4, high falling: Tone 1, high level). The
high falling Tone 4 was the most difficult to produce in the
second syllable. Most Tone 4 errors in the second syllable

were judged unidentifiable in the first test and as Tone 3

178
in the second test, indicating that Tone 4 was probably low
in pitch, a transfer of the English stress pattern.

Results in Miracle's (1989) study on 1- and 2-syllable
words seems similar to that in the reading of 1- and 2-
syllable words in this study, although in his study the
words were those of the topic (sentence initial) position
rather than words produced in isolation. In addition,
Miracle relies in his analysis on acoustic information while
this study relies on the auditory judgments of 3 Chinese
judges.

Miracle found that the errors made by his subjects were
evenly divided among the five tones (including the neutral
tone). The same was found in this study in isolated
syllables and 2-syllable words in the first test. However,
this does not apply to the second test or to the first word
in a sentence (see 6.6.2.3 below). He also found that in 2-
syllable words students made significantly more errors in
initial syllables (49.6% error rate) than in final syllables
(32% error rate). This was also true in this study though
the difference was much smaller: 31% vs. 24% error rate in

the first test and 28% vs. 24% in the second test.

6.6.2.3 Tone in short sentences
The percentages of correct tone production in reading

short sentences are shown below:

179

lst test

Tone 1 Tone 2 Tone 3 Tone 4. Average
lst syll 83% 45% 61% 85% 76%
2nd syll 50% 89% 75% 72%
3rd word 77% 89% 73% 61% 77%
Overall 81% 65% 77% 76% 75%
2nd test

Tone 1 Tone 2 Tone 3 Tone 4 Average
lst syll 93% 67% 67% as: 85%
2nd syll 56% 73% 84% 75%
3rd word 65% 88% 83% 92% 78%
Overall 83% 71% 76% 84% 79%

The frequency of occurrence for each tone in short
sentences was not controlled for position as in 2- syllable
words. Tone 1 did not occur in the second syllable at all.
However, exactly the same sentences were produced by the
subjects in the two tests: therefore, the results in the two ,
tests should be comparable.

The success rates of tone production varied for the
four tones in each position. In the first syllable, Tone 1
and Tone 4 seemed to be the most successfully produced in
the first test and Tone 2 the least successfully. In the
second test, Tone 1 was the most successfully produced tone
while Tone 2 and Tone 3 the less successfully. In the
second syllable (Tone 1 did not occur in this position),
Tone 2 was still the least successfully produced in both
tests, but Tone 3 was most successfully produced in the
first test, and Tone 4 the most successfully produced in the
second test. In the third or fourth syllable (one 2-
syllable word occurred in one sentence), Tone 4 was produced

the least successfully and Tone 2 the most successfully in

180

the first test, but Tone 1 the least successfully and Tone 4
the most successfully in the second test. Overall, success
rates for the production of the four tones were: Tone 2 the
least successfully produced tone in both tests: Tone 1 the
most successfully produced in the first test and Tone 4 and
Tone 1 the most successfully produced in the second test.

The task in Shen's (1989) study (see 3.2.2) is similar
to the reading of sentences in this study. However, there
are also important differences. Shen's study was based on
the reading of a lesson: thus the sentences formed a
coherent text. All the sentences were statements and their
lengths varied from 3 words (6 syllables) to over 10 words.
On the other hand, the sentences read by subjects in this
study were all short (3 or 4 syllables). Both questions
(yes-no and question word questions) and statements were
included. Some were sequenced like questions and answers,
but each sentence was preceded by a number, and there was no
necessary connection between sentences.

Shen (1989) found that error rates for Tone 1 and Tone
4 were significantly higher than those for Tones 2, 3, and 0
(neutral tone), and that the errors originated principally
from tonal register rather than from tonal contour. It is
clear from an examination of the success rate in reading
short sentences in this study that errors for Tone 1 and
Tone 4 were not more common than those for Tone 2 and Tone

3. On the contrary, Tone 2 (and Tone 3 in the second test)

181

seemed more difficult than Tone 1 and Tone 4, though Tone 1
and Tone 4 seemed to have more errors in the final (third or
fourth) syllable in the second test and Tone 4 seemed more
difficult in the third syllable in the first test.

Shen argues that Tone 1 and Tone 4 were produced with
more errors than Tone 2 and Tone 3 because Tone l and Tone 4
are less marked and are more likely to be subject to L1
transfer than the other tones. This explanation obviously
cannot be applied to the results found in this study.

In studies on children's acquisition of Mandarin tones
(Li and Thompson 1977, 1978, Clumeck 1980), Tone 2 and Tone
3 have been found to be acquired later than Tone 1 and Tone
4 in both perception and production, even though there are
individual differences in children acquiring their Ll
(rising tone has been found to be produced first, Clumeck
1980). Besides the general tendency for rising tones to be
more difficult than level and falling tones for children
acquiring tone languages, the phonetic similarities between
Tone 2 and Tone 3, and the tone sandhi rule for Tone 3
(realized as Tone 2 before another Tone 3), may account for
their difficulty for children learning Mandarin as Ll. Tone
2 and Tone 3 also appeared to be more difficult for L2 adult
learners in this study, and the same reasons may account for
these difficulties. However, it should not be surprising
that adult L2 learners exhibit even more variation. Besides,

error rate may not always be a good indicator of difficulty.

182
For instance, Tone 3 may seem to be easier to perceive than
Tone 1 and Tone 4 at the beginning since it was perceived at
the lowest error rate. In fact this may be a result of
confusion between Tone 2 and Tone 3 with a boundary closer
to Tone 2: all the errors were made in the direction of
misidentifying Tone 2 as Tone 3.

Since the difficulty order of tones may be different in
different contexts (e.g., Tone 3 in the first and second
syllables of 2-syllable words), the difficulty order of
tones can not be easily determined if we consider contexts

other than isolated syllables.

6.6.3 Imitation

It is clear that imitation was a much easier task than
reading: most tones were produced with over 95% success.
The differences between the four tones and the difference

between different contexts were therefore small and mostly

insignificant.
Isolated syllables

Tone 1 Tone 2 Tone 3 Tone 4 Average
1st test 100% 97% 100% 100% 99%
2nd test 98% 100% 100% 100% 100%

There were no significant differences in the production
of the four tones in imitating isolated syllables. The
success rates in imitating the four tones were all very
high, 100% or very close, in both the first and the second
test.

2-syllable words

183

lst test

Tone 1 Tone 2 Tone 3 Tone 4 Average
1st syll 97% 91% 99% 95% 95%
2nd syll 99% 100% 98% 100% 99%
Overall 98% 95% 98% 98% 97%
2nd test

Tone 1 Tone 2 Tone 3 Tone 4 Average
lst syll 98% 92% 100% 95% 96%
2nd syll 97% 100% 96% 100% 98%
Overall 97% 96% 98% 97% 97%

The average success rates of imitating the four tones

in 2-syllable words were not significantly different.
However, the difference between the four tones was
significant in the first syllable. Tone 2 was less
successfully imitated. The overall success rates of
imitating the four tones in 2-syllable words did not change
significantly between the two tests.

Short sentences

lst test

Tone 1 Tone 2 Tone 3 Tone 4 Average
lst syll 90% 97% 100% 100% 94%
2nd syll 90% 97% 97% 95%
3rd syll 98% 98% 98% 100% 99%
Overall 92% 95% 98% 98% 96%
2nd test ,

Tone 1 Tone 2 Tone 3 Tone 4 Average
1st syll 98% 96% 96% 92% 96%
2nd syll 81% 94% 95% 91%
3rd syll 100% 94% 100% 100% 98%
Overall 99% 89% 97% 95% 95%

The success rates in imitating the four tones in the

first syllable of short sentences were significantly

different in the first test: Tone 1 was less successfully

imitated than the other tones.

In the second test, however,

 

184

the success rates in imitating the four tones were not
significantly different, a result of improvement for Tone 1
and slightly drop for the other three tones. The second
syllable did not show significant differences among the
three tones (Tone 1 did not occur in this position) in the
first test, but the difference was significant in the second
test, in which Tone 2 was produced at a significantly lower
rate than Tone 3 and tone 4. The third and fourth syllables
did not show a significant difference for the four tones in
either the first or the second test, though Tone 2 was
somewhat more difficult than the other tones in the second
test. The overall success rates in imitating the four tones
were different in the first as well as in the second test.
Tone 1 was less successfully imitated in the first test, and
Tone 2 was less successfully imitated in the second test.

However, the success rate in imitating any tone in any
part of the imitating task is higher than its counterpart in
the reading task, and, except for Tone 2 in the second
syllable in the second test, all tones were imitated with
90% or higher success rate. Differences in success rates in
producing the four tones were much smaller in any part of
imitation than in reading, and when a difference is
significant, it does not always correspond to a difference
in the reading task, indicating quite different processes

involved in imitation and reading.

185
The following table shows the percentage of overall
correct production of the four tones in reading and

imitation in all syllables.

Reading

Tone 1 Tone 2 Tone 3 Tone 4 Average
Test 1 79% 73% 77% 77% 76%
Test 2 84% 73% 76% 76% 78%
Imitation

Tone 1 Tone 2 Tone 3 Tone 4 Average
Test 1 96% 95% 98% 98% 97%
Test 2 98% 94% 98% 97% 97%

The success rates in producing the four tones did not
differ significantly in the reading task in the first test,
but they differ significantly in the second test: Tone 1 was
more successfully produced than the other tones. The success
rates in producing the four tones also differ significantly
in the imitation task. Tone 2 was produced less

successfully than the other tones.

6.7 Error Patterns

When there is a tendency for a group of learners to
make the same errors in a certain aspect of L2, it indicates
that a general language learning process may be involved. A
clear understanding of the data will help us to learn more
about the possible L2 learning processes involved.

Is there a tendency for American English speakers to
make certain common errors in learning Mandarin tones? We
will examine the data and try to find any general patterns.

In the following sections, the number as well as the

186
percentage of correct and incorrect responses from the
students will be listed according to tones. Different tasks

in different contexts will be discussed separately.

6.7.1 Perception
6.7.1.1 Tone in isolated syllables

Test 1 (isolated syllables)
Perceived as

Tone 1 Tone 2 Tone 3 Tone 4
T1 131 (90%) 8 (6%) 2 (1%) 4 (3%)
T2 2 (1%) 105 (72%) 37 (26%) 1 (1%)
T3 0 (0%) 2 (1%) 139 (96%) 4 (3%)
T4 3 (2%) 13 (9%) 8 (6%) 121 (83%)

Test 2 (isolated syllables)
Perceived as

Tone 1 Tone 2 Tone 3 Tone 4
T1 128(95%) 5 (4%) 1 (1%) 1 (1%)
T2 2 (1%) 121(90%) 12 (9%) 0 (0%)
T3 0 (0%) 11 (8%) 124(92%) 0 (0%)
T4 5 (4%) 2 (1%) 0 (0%) 128(95%)

In the first perception test, when errors were made in
identifying a tone in isolated syllables, the most obvious
error pattern was that Tone 2 (rising) tended to be
misidentified as Tone 3 (falling-rising). There also seems
to be a tendency for Tone 1 (high level) to be misidentified
as Tone 2 (rising), and Tone 4 (high-falling) to be
misidentified as Tone 2 (rising).

In the second test, there still was a clear tendency
for Tone 2 to be misidentified as Tone 3, and in addition, a
clear tendency for Tone 3 to be misidentified as Tone 2.
Tone 1 was still most often misidentified as Tone 2, but

Tone 4 was misidentified as Tone 1 more often than Tone 2.

187

A comparison of the two tests showed that Tone l was
most often misidentified as Tone 2 on both the first and the
second test, though the number of errors was relatively
small (6% and 4%). Tone 2, the least successfully
identified tone on both tests, was overwhelmingly
misidentified as Tone 3 in the first test as well as in the
second test. Tone 3 was correctly identified 96% of the
time in the first test and the few misidentifications
involved both Tone 4 (4 times) and Tone 2 (twice). However,
the error rate of Tone 3 increased in the second test and
all the errors were misidentifications as Tone 2. The most
likely explanation for this change is that some students
developed an incorrect phonemic boundary between Tone 2 and
Tone 3 closer to the Tone 2 end before the first perception
test, so all the errors were made in one direction: Tone 2
being misidentified as Tone 3. After several weeks of
learning their perception boundary shifted toward the
correct boundary and errors were made in both directions.
Tone 4 was more often misidentified as Tone 2 in the first
test, but as Tone 1 in the second test, though the number
was small (4%).

Kiriloff's study (1969), on the tone perception of
isolated syllables, specifically mentions Tone 2's "marked
tendency" to be confused with Tone 3. From his data (as
shown in Kiriloff, 1969, table II, p.66) we can see that

besides Tone 2 being often misidentified as Tone 3 and Tone

188
3 as Tone 2, Tone 1 was most often misidentified as Tone 2,
and Tone 4 as Tone 3. Comparing the error patterns in his
data to those found in our data, we find that Tone 1 and
Tone 2 error patterns are the same in both studies, and the
Tone 3 error pattern in his study is consistent with that in
our second test. However, Tone 4 was more often
misidentified as Tone 2 than Tone 3 in this study (first
test), and it was rarely misidentified as Tone 3 in all
other contexts.

Some studies on the perception of tones used
synthesized tonal continua between Tone 1 (high level) and
Tone 2 (rising) to test both native speakers and non-native
speakers (Leather, 1987: Chan, et al. 1975). It has been
found that native Mandarin speakers display a sharp category
boundary at about the middle of the continuum, while for
non-native speakers the boundary was not as sharp and it was
shifted more toward the level end of the continuum. This
may account for some of the errors made in identifying Tone
1 as Tone 2.

According to Gandour (1978), there are five dimensions
in tonal perception: average pitch, direction, length,
extreme endpoint, and slope. These five dimensions were the
results of the analysis of tone perception data from three
language groups: Thai, Yoruba, and American English. The
dimension average pitch is the most important dimension for

all three language groups. However, the relative importance

 

189
attached to this dimension is significantly different for
tone language groups, Thai and Yoruba: and the non-tone
language group, English. The English group placed
relatively more emphasis on average pitch and extreme
endpoint in comparison with either the Thai or Yoruba
groups. On the other hand, the dimension of direction was
more important to tone language groups than to the English
group. The errors made in identifying Tone 4 (high falling)
as Tone 2 (rising) may be due to this relatively lesser

importance attached to direction than average pitch by

American English speakers.

6.7.1.2 Tone in 2-syllable words

Test 1
lst syllable

Tone 1 Tone 2 Tone 3 Tone 4
T1 81(68%) 11 (9%) 1 (1%) 27(23%)
T2 19(13%) 81(54%) 23(15%) 26(17%)
T3 7 (8%) 17(19%) 41(46%) 24(27%)
T4 15(13%) 14(12%) 3 (3%) 88(73%)
2nd syllable

Tone 1 Tone 2 Tone 3 Tone 4
T1 80(67%) 34(28%) 1 (1%) 5 (4%)
T2 8 (7%) 61(51%) 36(30%) 15(13%)
T3 5 (4%) 5 (4%) 103(86%) 6 (5%)
T4 14(12%) 20(17%) 0 (0%) 86(72%)

In the first syllable of 2-syllable words in the first
perception test, Tone 1, Tone 2 and Tone 3 were all most
often misidentified as Tone 4, and Tone 4 was the most
successfully identified tone. Thus there seemed to be a

tendency for all tones in the first syllable to be

190
identified as Tone 4. However, the misidentification of
Tone 2 as Tone 4 (17%) may not represent a meaningful
pattern since there were almost as many misidentifications
as Tone 3 (15%) and Tone 1 (13%). The misidentification of
Tone 1 as Tone 4 was probably due to the similar starting
pitch level of the two tones and the shorter duration of the
first syllable of 2-syllable words, since the level and
falling pitch contours of the two tones are likely to be
less prominent in this position. This possibility was also
seen in the misidentification of Tone 4 as Tone 1. It
should be noted that there were a substantial number of
errors made in identifying Tone 1 and Tone 4 besides
misidentifying one of them as the other, and the second most
frequent error seems to be the same as the most frequent
error found in isolated syllables: Tone 1 and Tone 4 were
often misidentified as Tone 2. The misidentification of
Tone 3 as Tone 4 was probably due to the similarity of
falling pitch contour (low-falling in Tone 3 and high-
falling in Tone 4) in this position.

In the second syllable, Tone 1 was most often
misidentified as Tone 2, Tone 2 was most often misidentified
as Tone 3, Tone 3 has the lowest error rate and was
misidentified as all other tones at similar rate, and Tone 4
was most often misidentified as Tone 2. The error pattern
for the second syllable was the same as that for isolated

syllables in the first perception test.

191

Test 2
lst syllable

Tone 1 Tone 2 Tone 3 Tone 4
T1 100(93%) 7 (6%) 0 (0%) 1 (1%)
T2 5 (4%) 119(88%) 6 (4%) 5 (4%)
T3 2 (2%) 2 (2%) 70(86%) 7 (9%)
T4 5 (5%) 13(12%) 1 (1%) 89(82%)
2nd syllable

Tone 1 Tone 2 Tone 3 Tone 4
T1 100(93%) 5 (5%) 0 (0%) 2 (2%)
T2 5 (5%) 91(84%) 6 (6%) 6 (6%)
T3 0 (0%) 5 (5%) 102(94%) 1 (1%)
T4 2 (2%) 14(13%) 2 (2%) 89(82%)

In the first syllable of 2-syllable words in the second
test, Tone l was most often misidentified as Tone 2, Tone 2
was misidentified as all other tones at similar rate, Tone 3
was often misidentified as Tone 4, and Tone 4 was often
misidentified as Tone 2. This error pattern for Tone 2 and
Tone 3 was the same as in the first test. However, the
tendency to confuse Tone 1 and Tone 4 clearly decreased.

The confusion between Tone 1 and Tone 4 mostly disappeared
since Tone 1 was misidentified as Tone 4 only 1% of the
time, and Tone 4 was misidentified as Tone 1 5% of the time.
As a result, the error pattern for Tone 1 and Tone 4 was
similar to that in isolated syllables and in the second
syllable of 2-syllable words in the first test.

As the second syllable of 2-syllable words, Tone 1 was
most often misidentified as Tone 2, Tone 2 was misidentified
as all other tones, Tone 3 was most often misidentified as
Tone 2, and Tone 4 was most often misidentified as Tone 2.

That Tone 3 was most often misidentified as Tone 2 was as in

192
isolated syllables in the second test, suggesting a shift of
perception boundary as mentioned in 6.7.1.1, and also
indicating a consistent identification criterion for tone in

isolated syllables and tone in 2-syllable words.

6.7.1.3 Tone in short sentences

Test 1

Tone 1 Tone 2 Tone 3 Tone 4
T1 15(24%) 17(27%) 14(23%) 16(26%)
T2 14(23%) 31(50%) 8(13%) 9(15%)
T3 2 (2%) 26(28%) 63(68%) 1 (1%)
T4 6(19%) 1 (3%) 8(26%) 16(52%)
Test 2

Tone 1 Tone 2 Tone 3 Tone 4
T1 49(91%) 0 (0%) 0 (0%) 5 (9%)
T2 1 (2%) 39(72%) 10(19%) 4 (7%)
T3 2 (2%) 5 (6%) 73(90%) 1 (1%)
T4 0 (0%) 1 (4%) 3(11%) 23(85%)

In short sentences, the perception task was a little
different from that in isolated syllables and 2-syllable
words. When a misidentification occurred, a sentence with
an incorrect tone was chosen, rather than a correctly
produced tone being misidentified as another tone.

In the first test, more instances of Tone 2 were
misidentified as Tone 1, but in fact all four tones were
chosen at a similar rate: more instances of Tone 1 were
misidentified as Tone 2, more instances of Tone 2 were
misidentified as Tone 3, and more instances of Tone 3 were

misidentified as Tone 4. Except for Tone 1 (which was

correctly identified at only about chance level), the

193
pattern was also the same as found in isolated syllables and
2-syllable words in the first test.

In the second test, more instances of Tone 4 were
misidentified as Tone 1, more instances of Tone 3 were
misidentified as Tone 2, more instances of Tone 2 were
misidentified as Tone 3, and more instances of Tone 3 were
misidentified as Tone 4. This is the same pattern as for
isolated syllables in the second test. The major difference
between the first test and the second test was in Tone 1 and
Tone 2. There was more confusion between Tone 1 and Tone 2
in the first test but more between Tone 2 and Tone 3 in the
second test.

In W.C.J. Lin's (1985) data for beginners (150 contact
hours, closest to our subjects), Tone 3 was the most often
misidentified tone and it was mostly confused with Tone 2 in
sentence-final position (W.C.J. Lin 1985, table 1, p.36).

In sentence-medial position ("Tade __ hen youming'),
however, though Tone 3 was still the most difficult tone, it
was most often misidentified as Tone 4 (W.C.J. Lin 1985,
table 4, p.38).

The tendency for students to confuse Tone 3 with Tone 2
in emphasized sentence final position, and to confuse Tone 3
with Tone 4 in sentence medial position, can be attributed
to the pitch contours Tone 3 exhibits in these two
positions. Since the falling-rising contour in final

position is similar to Tone 2 (rising), and the low-falling

194
contour in non-final position is similar to Tone 4 (high-
falling), it was not surprising that Tone 3 was more easily
confused with Tone 2 in final position and with Tone 4 in
non-final position.

To sum up, in the perception aspect, Tone 1 was most
likely to be misidentified as Tone 2, though
misidentification of Tone 1 as Tone 4 was also common in the
first syllable of 2-syllable words at the beginning. It was
rarely misidentified as Tone 3. Tone 2 was most likely to be
misidentified as Tone 3, though misidentification as Tone 4
and Tone 1 was also common in 2-syllable words. Tone 3 was
most likely to be misidentified as Tone 4 in the first
syllable of 2-syllable words. Otherwise misidentification
as Tone 2 was more common. The positional difference was
most likely due to the allophonic variation of Tone 3 in
these positions. Tone 4 was most likely to be misidentified
as Tone 2, though misidentification as Tone 1 was also

common in 2-syllable words at the beginning.

6.7.2 Reading
6.7.2.1 Tone in isolated syllables

Test 1
Pronunciation judged as
Tone 1 Tone 2 Tone 3 Tone 4 Uniden Tone 0
T1 60(97%) 1 (2%) 1 (2%) o (0%) o (0%) 0
T2 0 (0%) 57(92%) 4 (6%) 1 (2%) o (0%) 0
T3 0 (0%) 1 (2%) 58(94%) 2 (3%) 1 (2%) o (0%)
T4 2 (3%) o (0%) o (0%) 60(97%) 0 (0%) o

195
Test 2

Pronunciation judged as
Tone 1 Tone 2 Tone 3 Tone 4 Uniden Tone 0

T1 46(92%) 2 (4%) 1 (2%) 1 (2%) o (0%) 0
T2 2 (4%) 34(68%) 14(28%) 0 (0%) o (0%) o (0%)
T3 0 (0%) 2 (4%) 48(96%) 0 (0%) 0 (0%) 0 (0%)
T4 0 (0%) 0 (0%) 0 (0%) 48(96%) 2 (4%) 0 (0%)

In the reading of isolated syllables, only a few tokens
were judged to be phonemically incorrect by at least two
judges. When the errors made by subjects were few, they can
not indicate a general error pattern. Therefore, only the
two errors exceeding 5% will be discussed. In the first
test, Tone 2 tended to be mispronounced as Tone 3 (6%). The
same tendency was found in the second test, and the number
of errors increased dramatically (28%). In the two
perception tests, the error patterns of Tone 2 and Tone 3
seem to indicate a shift of boundary from closer to Tone 2

toward Tone 3. The increase of Tone 2 mispronounced as Tone

3 may indicate the same shift of boundary.

6.7.2.2 Tone in 2-syllable words
Test 1

1st syllable
Pronunciation judged as
Tone 1 Tone 2 Tone 3 Tone 4 Uniden Tone 0
T1 91(73:) 4 (3%) 7 (6%) 19(15%) 3 (2%) 0
T2 14 (9%) 101(65%) 25(16%) 15(10:) 0 (0%) o (0%)
T3 1 (1%) 39(42%) 48(52%) 4 (4%) 1 (1%) 0
T4 7 (6%) 7 (6%) 5 (4%) 101(81%) 4 (3%) o

196

2nd syllable
Pronunciation judged as
Tone 1 Tone 2 Tone 3 Tone 4 Uniden Tone 0
T1 90(73%) 19(15%) 3 (2%) 4 (3%) 8 (6%) 0 (0%)
T2 15(12%) 103(83%) 2 (2%) 0 (0%) 4 (3%) 0 (0%)
T3 5 (4%) 3 (2%) 107(86%) 5 (4%) 4 (3%) 0 (0%)
T4 3 (2%) 11 (9%) 9 (7%) 79(64%) 15(12%) 7 (6%)

In reading 2-syllable words in the first test, the tone
error patterns in the first syllable and the second syllable
were different. In the first syllable Tone 1 was most often
mispronounced as Tone 4, and Tone 2 was most often
mispronounced as Tone 3, though mispronunciation as Tone 1
and Tone 4 was also common. Tone 3 was most often
mispronounced as Tone 2, and Tone 4 was often mispronounced
as Tone 1 and Tone 2. In the second syllable Tone 1 was
most often mispronounced as Tone 2, Tone 2 was most often
mispronounced as Tone 1, and Tone 4 errors were mostly
unidentifiable, though mispronunciation as Tone 2 and Tone 3
was also common.

Test 2
lst syllable
Pronunciation judged as

Tone 1 Tone 2 Tone 3 Tone 4 Uniden Tone 0
T1 87(91%) 3 (3%) l (1%) 3 (3%) 2 (2%) 0 (0%)
T2 14(15%) 74(77%) 4 (4%) o (0%) 4 (4%) o (0%)
T3 3 (3%) 45(47%) 46(48%) 1 (1%) 1 (1%) o (0%)
T4 10(10%) 12(13%) 5 (5%) 66(69%) 3 (3%) O (0%)
2nd syllable

Pronunciation judged as

Tone 1 Tone 2 Tone 3 Tone 4 Uniden Tone 0
T1 72(75%) 8 (8%) 4 (4%) 3 (3%) 9 (9%) o (0%)
T2 11(11%) 72(75%) 6 (6%) 2 (2%) 4 (4%) 1 (1%)

T3 1 (1%) 1 (1%) 90(94%) 3 (3%) 1 (1%) o (0%)
T4 2 (2%) o (0%) 23(24%) 58(60%) 6 (6%) 7 (7%)

197

In the second test, in the first syllable of 2-syllable
words, Tone 2 was most often mispronounced as Tone 1, Tone 3
was most often mispronounced as Tone 2 (actually,
mispronunciation as Tone 2 and correctly pronounced Tone 3
were about the same in number), and Tone 4 was most often
mispronounced as Tone 1 and Tone 2. In the second syllable,
Tone 1 errors were mostly unidentifiable or mispronounced as
Tone 2, Tone 2 was most often mispronounced as Tone 1, and
Tone 4 was most often mispronounced as Tone 3.

Comparing the error patterns of the two tests, we can
see in the first syllable that there was a decrease of Tone
1 mispronunciation as Tone 4. In the first syllable, Tone 2
mispronunciation as Tone 3 and Tone 4 decreased but
mispronunciation as Tone 1 increased. In the second
syllable, there was a decrease of Tone 1 mispronunciation as
Tone 2 but an increase of unidentifiable errors. Tone 2
seemed to be mispronounced as Tone 3 a little more in the
second test than in the first test. A similar but greater
change also occurred in isolated syllables. Tone 4
mispronunciation as Tone 2 decreased while Tone 4
mispronunciation as Tone 3 increased.

Some studies (Miracle 1989, Shen 1989) classified tonal
errors into register and contour errors. The production
data in this study were judged on the tone category alone
and not classified as register or contour errors. However,

some mispronunciation errors were apparently tonal contour

198
errors. Tone 1 (level) mispronounced as Tone 4 (falling)
was such an error, and often occurred when the second
syllable started lower in pitch (Tone 2 or Tone 3). Tone 4
mispronounced as Tone 1 was another such error. Tone 4
(falling) mispronounced as Tone 2 (rising) and Tone 2
mispronounced as Tone 4 were also tonal contour error. Some
mispronunciation errors were tonal register errors, such as
Tone 4 mispronounced as Tone 3 (half Tone 3, low-falling)
and vice versa. Tone 2 mispronounced as Tone 1 usually
involved a start too high in pitch. Though the rising
contour may be maintained to a lesser degree in some cases,
it was nonetheless perceived as Tone 1 (level). Tone 2
mispronounced as Tone 3 usually resulted from starting too
low or dipping too low, and Tone 3 mispronounced as Tone 2
was a result of not falling low enough at the beginning.
Some errors may involve both tonal register and tonal
contour. The unidentifiable tone errors usually lacked
clear tonal contours and showed ambiguity of pitch level
between two tones.

The much higher error rate in producing 2-syllable
words than isolated syllables may be due to difficulty in
the control of pitch patterns in 2-syllable words, since the
length of syllables in 2-syllable words is shorter.
However, some other factors seem to be involved. Transfer of
the English word-stress pattern stressed-unstressed (high-

low in pitch) may account for the different error rates of

199
the four tones. The error patterns in 2-syllable words,
except for the Tone 2 mispronunciation as Tone 3, which also
occurred in single syllables, seem to be consistent with
this account in that most errors in the first syllable
tended to be high in pitch and the errors in the second

syllable tended to be low in pitch.

6.7.2.3 Tone in short sentences

lst syllable
Pronunciation judged as

Tone 1 Tone 2 Tone 3 Tone 4 Uniden Tone 0
T1 103(83%) 0 (0%) 1 (1%) l7(14%) 3 (2%) 0 (0%)
T2 4(13%) 14(45%) 0 (0%) 10(32%) 3(104) o (0%)
T3 5(16%) 4(134) 19(61%) 2 (6%) 1 (3%) o (0%)
T4 2 (3%) 3 (5%) 2 (3%) 53(85%) 2 (3%) 0 (0%)
2nd syllable
Pronunciation judged as
Tone 1 Tone 2 Tone 3 Tone 4 Uniden Tone 0
T2 7(114) 31(50%) 17(27%) 6(10%) 1 (2%) o (0%)
T3 2 (3%) 3 (5%) 55(89%) 1 (2%) 1 (2%) 0 (0%)
T4 4 (3%) o (0%) 8 (6%) 93(75%) 19(15%) 0 (0%)
3rd and 4th syllable
Pronunciation judged as
Tone 1 Tone 2 Tone 3 Tone 4 Uniden Tone 0
T1 48(77%) 9(15%) 1 (2%) 4 (6%) o (0%) o (0%)
T2 6(10%) 55(89%) 0 (0%) o (0%) 1 (2%) o (0%)
T3 1 (2%) 11(18%) 45(73%) 3 (5%) 2 (3%) 0 (0%)
T4 1 (3%) 6(19%) 2 (6%) 19(61%) 3(102) o (0%)
Test 2
1st syllable
Pronunciation judged as
Tone 1 Tone 2 Tone 3 Tone 4 Uniden Tone 0
T1 89(93%) 0 (0%) o (0%) 5 (5%) 2 (2%) o (0%)
T2 7(29%) 16(67%) 0 (0%) o (0%) 1 (4%) o (0%)
T3 2 (8%) 5(21%) 16(67%) 1 (4%) o (0%) o (0%)
T4 0 (0%) 1 (2%) o (0%) 42(88%) 5(10%) 0 (0%)

200

2nd syllable
Pronunciation judged as
Tone 1 Tone 2 Tone 3 Tone 4 Uniden Tone 0
T2 8(17%) 27(56%) 7(15%) 4 (at) 2 (4%) o (0%)
T3 2 (4%) 6(13%) 35(73t) 5(104) o (0%) o (0%)
T4 3 (3%) 1 (1%) o (0%) 81(84%) 11(11%) 0 (0%)

3rd and 4th syllable
Pronunciation judged as
Tone 1 Tone 2 Tone 3 Tone 4 Uniden Tone 0

T1 31(65%) 10(21%) 1 (2%) 2 (4%) 3 (6%) 1 (2%)
T2 4 (8%) 42(88%) 0 (0%) o (0%) 2 (4%) o (0%)
T3 0 (0%) 4 (8%) 40(33%) 2 (4%) 2 (4%) o (0%)
T4 1 (4%) 3(13%) 1 (4%) 18(75%) 1 (4%) o (0%)

In reading short sentences of three or four syllables,
in the first syllable, Tone 1 was most often
mispronounced as Tone 4, Tone 2 was also most often
mispronounced as Tone 4, and Tone 3 was often mispronounced
as Tone 1 and Tone 2. In the second syllable (no Tone 1
occurred in this position), Tone 2 was most often
mispronounced as Tone 3, though there were also quite a few
mispronounced as Tone 1 and Tone 4. Tone 3 was most often
mispronounced as Tone 2, and Tone 4 errors were most often
unidentifiable. When a mispronounced Tone 4 was
identifiable, it was most often identified as Tone 3. In
the third or fourth syllable, Tone 1 was most often
mispronounced as Tone 2, Tone 2 was most often mispronounced
as Tone 1, Tone 3 was most often mispronounced as Tone 2,
and Tone 4 was most often mispronounced as Tone 2.

In the second test, Tone 1 in the first syllable was
most often mispronounced as Tone 4 (as in the first test),

Tone 2 was most often mispronounced as Tone 1, Tone 3 was

201

most often mispronounced as Tone 2, and Tone 4 errors were
most often judged unidentifiable. In the second syllable,
Tone 2 was most often mispronounced as Tone 1 and Tone 3,
Tone 3 was most often mispronounced as Tone 2, and Tone 4
mispronunciation was most often judged unidentifiable. In
the third or fourth syllable, Tone 1 was most often
mispronounced as Tone 2, Tone 2 was most often mispronounced
as Tone 1, Tone 3 was most often mispronounced as Tone 2,
and Tone 4 was most often mispronounced as Tone 2.

Comparing the error patterns in the two tests, we can
see some changes. The mispronunciation of Tone 1 as Tone 4
in the first syllable was the most frequent error in both
tests but the error rate became lower in the second test.
The mispronunciation of Tone 2 as Tone 4 in the first
syllable of a sentence did not appear in the second test.
The mispronunciation of Tone 3 as Tone l in the first
syllable also decreased in the second test. In the second
syllable, the mispronunciation of Tone 2 as Tone 3 decreased
in the second test but the mispronunciation as Tone 1 did
not. The mispronunciation of Tone 3 as Tone 2 and Tone 4 in
the second syllable increased in the second test. However,
most Tone 4 errors in the second syllable remained
unidentifiable in both tests. In the third and fourth
syllable, the error patterns remained similar.

It is clear that the error rates of the four tones and

the error patterns in each position of the sentence were

202
different, and that there were changes between the two tests
although exactly the same sentences were produced in the two
tests. Some errors may be the result of transfer from
English intonation. For example, the word shei ('who", Tone
2) as the first word of a sentence was frequently
mispronounced as Tone 1 or Tone 4 in the first test and as
Tone 1 in the second test, but never as Tone 3. It is
possible that this is because the equivalent English
sentence would start with a high pitch, so that subjects
tended to mispronounce this word with a high tone (both Tone
1 and Tone 4 start high). Another example concerns the
question particle ma (neutral tone, not listed above), which
was most often misproduced as Tone 2 (rising). This error
also seems to be a transfer of English rising intonation at
the end of yes/no questions. These transfer errors can be
explained by comparing the intonation pattern of the
equivalent English sentence.

The study of Shen (1989) reported that American
learners of Mandarin made more errors in producing Tone 1
and Tone 4 (actually this claim could only be applied to the
poorer L2 speakers in her study), and Tone 1 and Tone 4
errors were all register errors rather than contour errors.
Thus she maintains that "the errors originated principally
from tonal register" (p.33). However, other studies (e.g.,
Miracle 1989 and this study) do not always find Tone 1 and

Tone 4 the most difficult, and Tone 1 and Tone 4 errors are

203
not all register errors. Since in English pitch is used to
express affect or grammatical functions and is often
affected by stress, it is likely that an American learner of
Mandarin would tend to produce tones differently depending
on the position of the tone in a sentence or the type of
sentence it is in, and possibly the position of a multi-
syllable word it is in (as we have found in reading isolated
2-syllable words). Thus more data obtained on all possible
conditions are needed before we can confidently make a
generalization about the error patterns made by American
learners of Mandarin.

To sum up findings in the reading aspect in this study:

1. Tone 1 was most likely to be mispronounced as Tone 4
or Tone 2.

2. Tone 2 was most likely to be mispronounced as Tone 3
in isolation, but also as Tone 1 and Tone 4 (in the first
syllable of a short sentence).

3. Tone 3 was most likely to be mispronounced as Tone
2. However, Tone 1 and Tone 4 were also common.

4. Tone 4 was most likely to be mispronounced as
Tone 2, though Tone 1 and Tone 3 were also common
mispronunciations. And Tone 4 was more often than the other
tones judged to be unidentifiable.

The errors made by subjects in this study included both
register and contour errors. Some of these errors can be

found in children learning Mandarin as their first language,

204
such as the confusion between Tone 2 and Tone 3 and the
substitution of level for contour pitch patterns in non-
final position (Clumeck 1980:261), and may be developmental.
Some errors such as a high tone in the first syllable and a
low tone in the second syllable of 2-syllable words, and the
rising contour at the end of questions, probably represent

transfer from English word-stress and intonation.

6.7.3 Imitation
isolated syllables

Test 1
Pronunciation judged as
Tone 1 Tone 2 Tone 3 Tone 4 Uniden Tone 0
T1 62(100%) o (0%) o (0%) o (0%) o (0%) o (0%)
T2 0 (0%) 60(97%) 2 (3%) 0 (0%) 0 (0%) 0 (0%)
T3 0 (0%) o (0%) 62(100%) o (0%) o (0%) o (0%)
T4 0 (0%) o (0%) o (0%) 62(100%) o (0%) o (0%)

Test 2
Pronunciation judged as
Tone 1 Tone 2 Tone 3 Tone 4 Uniden Tone 0
T1 49(98%) 1 (2%) 0 (0%) 0 (0%) 0 (0%) 0 (0%)
T2 0 (0%) 50(1oo%) o (0%) o (0%) o (0%) o (0%)
T3 0 (0%) o (0%) 50(1oo%) o (0%) o (0%) o (0%)
T4 0 (0%) o (0%) o (0%) 50(100%) o (0%) o (0%)

2-syllable words
Test 1
lst syllable
Pronunciation judged as

Tone 1 Tone 2 Tone 3 Tone 4 Uniden Tone 0
T1 120(974) 2 (2%) o (0%) 2 (2%) o (0%) o (0%)
T2 1 (1%) 141(913) 12 (8%) o (0%) 1 (1%) o (0%)
T3 0 (0%) 1 (1%) 92(99%) 0 (0%) 0 (0%) 0 (0%)
T4 6 (5%) o (0%) o (0%) 118(95%) o (0%) o (0%)

205

2nd syllable
Pronunciation judged as
Tone 1 Tone 2 Tone 3
T1 123(994) 1 (1%) o (0%)
T2 0 (0%) 124(1004) 0 (0%)
T3 0 (0%) 1 (1%) 121(98%)
T4 0 (0%) o (0:) o (0%)

Test 2
lst syllable
Pronunciation judged as
Tone 1 Tone 2 Tone 3
T1 96(98%) 0 (0%) o (0%)
T2 1 (1%) 91(92%) 6 (6%)

T3 0 (0%) o (0%) 98(100%)

T4 5 (5%) 0 (0%) 0 (0%)

2nd syllable
Pronunciation judged as
Tone 1 Tone 2 Tone 3
T1 96(97%) 1 (1%) 0 (0%)
T2 0 (0%) 99(1004) o (0%)
T3 0 (0%) 1 (1%) 94(96%)
T4 0 (0%) 0 (0%) 0 (0%)

Short sentences

Test 1
lst syllable
Pronunciation judged as
Tone 1 Tone 2 Tone 3
T1 111(9o%) 2 (2%) 1 (1%)
T2 1 (3%) 30(97%) 0 (0%)

T3 0 (0%) o (0%) 31(1oo%)

T4 0 (0%) o (0%) o (0%)

2nd syllable
Pronunciation judged as
Tone 1 Tone 2 Tone 3
T2 5 (8%) 56(90%) 1 (2%)
T3 1 (2%) o (0%) 60(97%)
T4 1 (1%) 1 (1%) 2 (2%)

3rd and 4th syllable
Pronunciation judged as
Tone 1 Tone 2 Tone 3
T1 61(98%) 1 (2%) o (0%)
T2 1 (2%) 61(98%) 0 (0%)
T3 0 (0%) 1 (2%) 61(98%)
T4 0 (0%) o (0%) o (0%)

Tone 4 Uniden
o (0%) o (0%)
o (0%) o (0%)
2 (2:) o (0%)
124(1004) 0 (0%)

Tone 4 Uniden
2 (2%) 0 (0%)
0 (0%) l (1%)
o (0%) o (0%)

94(95%) 0 (0%)

Tone 4 Uniden
o (0%) 2 (2%)
0 (0%) 0 (0%)
3 (3%) 0 (0%)

98(100%) o (0%)

Tone 4 Uniden
1o (3:) o (0%)
o (0%) o (0%)
o (0%) o (0%)
62(100%) o (0%)

Tone 4 Uniden
0 (0%) 0 (0%)
0 (0%) 1 (2%)
120(97%) o (0%)

Tone 4 Uniden
o (0%) o (0%)
o (0%) o (0%)
0 (0%) 0 (0%)

31(1oo%) o (0%)

Tone 0

0
0
0
0

(0%)
(0%)
(0*)
(0%)

Tone 0

0
0
0

(0%)
(0%)
(0*)

Tone 0

000°

(0%)
(0%)
(0%)
(0%)

206

Test 2
1st syllable
Pronunciation judged as
Tone 1 Tone 2 Tone 3 Tone 4 Uniden Tone 0

T1 94(98%) 1 (1%) o (0%) 1 (1%) o (0%) o (0%)
T2 0 (0%) 23(96%) 0 (0%) 1 (4%) o (0%) o (0%)
T3 0 (0%) 1 (4%) 23(96%) 0 (0%) o (0%) o (0:)
T4 0 (0%) 2 (4%) o (0%) 44(92%) 2 (4%) o (0%)

2nd syllable
Pronunciation judged as
Tone 1 Tone 2 Tone 3 Tone 4 Uniden Tone 0
T2 8(17%) 39(81%) 1 (2%) o (0%) o (0%) o (0%)
T3 0 (0%) 3 (6%) 45(94%) 0 (0%) o (0:) o (0%)
T4 1 (1%) 3 (3%) 1 (1%) 91(95%) 0 (0%) o (0%)
3rd and 4th syllable

Pronunciation judged as

Tone 1 Tone 2 Tone 3 Tone 4 Uniden Tone 0
T1 48(100%) o (0%) o (0%) o (0%) o (0%) o (0%)
T2 2 (4%) 45(94%) 0 (0%) o (0%) 1 (2%) o (0%)
T3 0 (0%) 0 (0%) 48(100%) 0 (0%) 0 (0%) 0 (0%)
T4 0 (0%) o (0%) o (0%) 24(1004) o (0%) o (0%)

The tone error rates in imitation tasks were generally
low, usually below 5% for all four tones in any part, and
the errors in any category rarely exceeded those in the
counterpart reading task. An error rate exceeding 5% in the
imitation task occurred mostly with Tone 2: it was
mispronounced as Tone 3 in the first syllable of 2-syllable
words in the first test (8%) and the second test (6%), and
as Tone 1 in the second syllable of a short sentence in the
first test (8%) and the second test (17%). Tone 1 was
imitated as Tone 4 in the first syllable of a short sentence
8% of the time in the first test. These errors were not
different from those that occurred in the reading task and

will not be discussed further.

7. CONCLUSIONS

7.1 Summary of findings

The results of this study can be summarized as follows:

1. Tonal context (isolated syllables, 2-syllable words,
and short sentences) has clearly influenced the perception
and production of tone by adult L2 learners. However, the
effect on perception was greater at the beginning and
decreased as learning progressed.

Tonal context has a similar effect on tone perception
and reading, but not on imitation: tone perception and
production in isolated syllables was better than in
sequences (2-syllable words and short sentences). Tone
perception in 2-syllable words was better than in short
sentences. However, tone production in short sentences was
better than in 2-syllable words. Subjects' familiarity with
the sentences seemed to have changed the expected context
effect in reading.

2. Different tasks (imitation and reading) have
influenced L2 learners' tone production. Imitation was
better than reading in all contexts from the beginning to
the end of the term, indicating imitation was a much easier
task than reading. The discrepancy between reading and
imitation was greater in tone sequences (2-syllable words

and short sentences) than in isolated syllables.

207

208

There was no significant correlation between the two
tasks at the beginning of Mandarin learning, suggesting that
they were not related. However, correlations between them
in isolated syllables and short sentences became significant
at the end of the term, thus they seemed to get more related
as learning progressed.

3. Whether L2 tone perception and production is
affected by sex cannot be determined by this study.

Contrary to findings in some other studies on L2
pronunciation, male subjects scored higher than female
subjects in tone perception and production. However,
statistical results were inconclusive. The difference
between males and females also decreased as learning
progressed. Females seem to have a greater improvement than
males in short sentences, which had been taught in class and
were familiar to the subjects, though the difference was not
significant.

4. Prior experience of learning other L2s has no
significant effect on the perception or production of tone.
This is consistent with findings in some studies that other
L2 experience has no effect on L2 pronunciation or sound
perception.

Prior Mandarin learning experience (no more than a
year) has no significant effect on the perception or
production of tone, with only one exception. This exception

was in imitating short sentences at the end of the term:

209
subjects who had learned Mandarin before did worse than
others in imitating short sentences they had’learned. It is
possible that they paid less attention to the auditory model
when they imitated items they were familiar with, or they
may have formed a habit of producing tones and it became
harder to modify it. Because of the small number of subjects
with prior Mandarin learning experience in this study,
future studies on more L2 learners with different lengths of
experience will be needed to provide further information on
the effect of L2 learning on the pronunciation of imitating
L2 utterances.

5. The correlation between the perception of tone in
isolated syllables and in sequences (2-syllable words and
short sentences) was generally significant. The correlation
was higher between tone in isolated syllables and 2-syllable
words, and between 2-syllable words and short sentences.

The correlation between tone in isolated syllables and in
short sentences was low at the beginning and became
insignificant at the end of the term. The ceiling effect of
tone perception in isolated syllables, and the teaching of
short sentences in the class, may have affected the
correlation.

On the other hand, the correlation between the
production of tone in isolated syllables and in sequences
was not significant, possibly because of the ceiling effect

of the production of tone in isolated syllables. The

210
correlation between reading 2-syllable words and short
sentences was significant, but the correlation was higher at
the beginning of the term than at the end of the term.

6. The correlation between tone perception and tone
reading was significant in isolated syllables and 2-syllable
words. Significant (and higher) correlation was also found
between tone perception and reading across the two tests,
suggesting a delayed effect between the development of
perception and production of tone. The correlation between
tone perception and tone reading in short sentences was
significant at first but became insignificant at the end of
the term. Since short sentences had been taught in the
class, this suggests that instruction may have some effect
on the relation between perception and reading. Tone
production in the imitating task generally did not correlate
significantly with tone perception.

7. There was significant improvement in tone perception
between the two tests conducted at the beginning and at the
end of the term. The improvement occurred in all isolated
syllables, 2-syllable words, and short sentences. However,
there was no significant improvement in tone production,
either in reading or imitation, between the beginning and
the end of the term. The ceiling effect appear to have been
reached for imitation from the beginning. Even if we

disregard imitation, production was better than perception

211
at the beginning of the term, but perception became better
at the end of the term.

8. The four Mandarin tones caused different degrees of
difficulty for L2 learners as measured by error rates.
However, the order of difficulty seemed different in
different contexts at different times. Tone 2 was the most
difficult tone to learn in isolation. Tone 3 was the most
difficult in the first syllable of 2-syllable words but the
easiest in the second syllable of 2-syllable words.

9. The tone error patterns also differ depending on the
context. Variation found in the same context shows that
there is probably more then one factor explaining the
errors. Both register and contour errors were found.
Errors may be developmental or transfer from word-stress
(see 6.6.2.2 and 6.7.2.2) and intonation (see 6.7.2.3) of
English.

Basically the learning of Mandarin tones by English-
speaking learners is not different from the learning of
other aspects of L2 phonology by L2 learners. Of course,
specific tonal errors and tonal contexts are unique to the

learning of Mandarin.

7.2 Suggestions for teaching Mandarin tones
Some suggestions of teaching Mandarin tones to adult L2

learners can be made based on these conclusions:

212

1. Tones in 2-syllable words should be taught more.
The teaching of tones in citation forms is the basis of
teaching tones in longer utterances. However, the
correlations found between tones in citation form and tones
in short sentences were usually low or insignificant. On
the other hand, the correlations found between 2-syllable
words and short sentences were always significant. Since
tones in 2-syllable words either precede or are followed by
another tone, they are more like tones in longer sequences.
Tone sandhi rules that apply in utterances basically apply
in sequences of two tones. Thus the learning of tones in
2-syllable words may have some positive effect on learning
tones in longer utterances. Besides, 2-syllable words are
almost as common as l-syllable words in Mandarin and both
are the basic building blocks of longer utterances. Lin
(1985) proposed to teach the norm of Tone 3 as low-falling
(see 3.2.1) and showed the effectiveness of this new
treatment of Tone 3. In fact, since the low-falling-rising
contour of Tone 3 will inevitably occur in the citation form
and in an emphasized position, the best way to teach the
low-falling variant is to teach it in the right context, and
2-syllable words provide such a context. The suggestion of
teaching tones in 2-syllable words has been proposed in
other studies on the teaching of Mandarin tones (M-Y. Chen

1983).

213

2. More time should be spent on the perception rather
than the production of Mandarin tones. The production of
acceptable isolated Mandarin tones seemed to be easily
learned at the beginning of the term: however, the
production of tones did not seem to be easily improved
within one term. On the other hand, the perception of tones
seemed more difficult at the beginning of learning, but it
can be improved more easily than production over one term of
learning, especially for tones in sequence. It thus seems
more effective to spend more time on perception than
production. Since the perception and the production
(reading) of tone correlated significantly, even though
evidence of a transfer of improvement in perception into
improvement in production is inconclusive, it is quite
possible that time spent on tone perception may also have
some positive effect on tone production. This suggestion
should not be taken to mean that production training is not
needed. In fact, because production does not improve
easily, it may be even more important to help students begin
to learn correct pronunciations from the very beginning.

3. Students should be made aware of transfer errors
from English word-stress and sentence intonation. Tonal
errors made in producing isolated syllables were relatively
few compared to errors made in tone sequences. Transfer of
English word-stress and sentence intonation appears to be an

important cause of tonal errors which students made in

214
reading 2-syllable words (see 6.6.2.2 and 6.7.2.2) and short
sentences (see 6.7.2.3).

There is evidence that form-focused instruction has an
effect on the accuracy of L2 learners' use of L2 structures
(Lightbown and Spada 1990), supporting the notion that
learners can benefit from "consciousness raising"
(Rutherford 1987, Rutherford and Sherwood-Smith 1988).

Since English (or other non-tone language) speakers may be
less conscious about the use of pitch in word-stress and
sentence intonation in their native language, they may
benefit from explicit knowledge of the differences in pitch
range and pitch contour between English and Mandarin. While
errors may be inevitable in learning an L2, awareness of the
source of errors may be helpful for adult learners reducing
or correcting their errors. Because English sentence
intonation is superimposed on the whole sentence and the
pitch patterns vary depending on the meaning of a particular
sentence, a particular intonation pattern is likely to be
transferred to Mandarin sentences with similar meanings.
Thus it may be necessary and more effective to call
students' attention to intonation transfer with particular

Mandarin sentences as students learn them.

BIBLIOGRAPHY

BIBLIOGRAPHY

Abramson, A. S. and Svastikula, K. 1983. Intersections of
tone and intonation in Thai. Haskins Laboratories
Status Report on Speech Research, Apr.-Sep., 143-154.

Albert, M. and Obler, L. 1978. The Bilingual Brain. New
York: Academic.

Asher, J. J. and Garcia, R. 1969. The optimal age to learn a
foreign language. Modern Language Journal 53, 334-342.

Baran,J. A., Laufer, M. Z. and Daniloff, R. 1977.
Phonological contrastivity in conversation: A
comparative study of voice onset time. Journal of
Phonetics 5, 339-350.

Beebe, L. M. 1987. Myths about interlanguage phonology.
Interlanguage Phonology. G. Ioup and S. Weinburger
(eds.), 165-175. Cambridge: Newbury House.

Beebe, L. M. 1987. Sociolinguistic variation and style
shifting in second language acquisition. G. Ioup and S.
Weinburger (eds.), 378-388. Cambridge: Newbury House.

Berke, J. and Brown, R. 1960. Psycholinguistic research
methods. Handbook of Research Methods in Child
Development. P. Mussen (ed.). New York: Wiley.

Boliger, D. L. 1958. A Theory of Pitch Accent in English.
Word 14, 109-149.

Bolinger, D. L. (ed.) 1972. Intonation. Baltimore: Penguin
Books Inc.

Boliger, D. L. 1975. Aspects of Language. Second edition.
New York: Harcourt Brace Jovanovich, Inc.

Boliger, D. L. 1978. Intonation across languages. Universals
of Human Language, vol. 2. J. Greenberg (ed.), 471-524.
Stanford, CA: Stanford University Press.

Bolinger, D. L. 1986. Intonation and Its Parts. Stanford,
CA: Stanford University Press.

215

216

Borden, G., Gerber, A. and Milsark, G. 1983. Production and
perception of the /r/-/l/ contrast in Korean adults
learning English. Language Learning 33, 499-526.

Briere, E. 1966. An investigation of phonological
interference. Language 44, 768-796.

Broselow, E., Hurtig, R. and Ringen, C. 1987. The perception
of second language prosody. Interlanguage Phonology.
G. Ioup and S. Weinburger (eds.), 350-361. Cambridge:
Newbury House.

Caramazza, A., Yeni-Komshian, G., Zurif, E., and Carbone, E.
1973. The acquisition of a new phonological contrast:
The case of stop consonants in French-English
bilinguals. Journal of the Acoustical Society of
America 54, 421-428.

Carroll, J. B. 1981. Twenty-five years of research on
foreign language aptitude. Individual Differences and
Universals in Language Learning Aptitude. K. C. Diller
(ed.), 83-118. Rowley: Newbury House Publishers.

Catford, J. C. and Pisoni, D. B. 1970. Auditory vs.
Articulatory Training in Exotic Sounds. Modern Language
Journal 54, 477-481.

Chan, S. W., Chuang, C.-K and Wang, W. S.-Y. 1975.
Cross-language study of categorical perception for
lexical tone. Manuscript of paper read at the 90th
Meeting of the Acoustical Society of America, 1975.
(Abstracted in Journal of Acoustical Society of
America, 58, 8119.)

Chang, Nien-Chuang T. 1958. Tones in Intonation in the
Chengtu Dialect (Szechuan, China). Phonetica, 2, 59-84.

Chao, Yuen-Ren. 1930. A system of tone letters. Le maitre
phonétique 45, 24-27.

Chao, Yuen-Ren. 1933. Tone and intonation in Chinese.
Bulletin of the Institute of History and Philology 4,
121-134.

Chao, Yuen-Ren. 1948. Mandarin Primer. Cambridge: Harvard
University Press.

Chao, Yuen-Ren. 1956. Tone, intonation, singsong, chanting,
recitative, tonal composition, and atonal composition
in Chinese. For Roman Jakobson. M. Halle, H. G. Lunt,
H. Mclean and C. H. Van Schooneveld (eds.), 52-59. The
Hague: Mouton.

217

Chao, Yuen-Ren. 1968. A Grammar of Spoken Chinese. Berkeley:
University of California Press.

Chen, Gwang-tsai. 1974. The pitch range of English and
Chinese speakers. Journal of Chinese Linguistics 2,
159-71.

Chen, Ming-yuan. 1981. Shuli tongji zai Hanyu yanjiu zhong
de yingyong (Application of statistics in the study of
Mandarin). Zhongguo yuwen 6, 466-474.

Chen, Ming-yuan. 1983. Hanyu zuowei di er yuyan jiaoxue
zhong de yuyin wenti (Phonetic problems in the teaching
of Chinese as a second language). Language Learning and
Communication 2:3, 249-265.

Cheng, Chin-chuan. 1968. English stresses and Chinese tones
in Chinese sentences. Phonetica 18, 77-88.

Chiang, T. 1979. Some interferences of English intonations
with Chinese tones.International Review of Applied
Linguistics 17, 245-250.

Chomsky, N. and Halle, M. 1968. The Sound Pattern of
English. New York: Harper and Row.

Clumeck, H. 1980. The acquisition of tone. Child Phonology,
Vol.1, Production. G. Yeni-Komshian, J. Kavanagh, and
C. Ferguson (eds.), 257-275. New York: Academic Press.

Cole, R. A. 1981. Perception of fluent speech by children
and adults. Native Language and Foreign Language
Acquisition. H. Winitz (ed.), 92-109. New York: The New
York Academy of Sciences.

Connell, B., Hogan, J. and Rozsypal, A. 1983. Experimental
evidence of interaction between tone and intonation in
Mandarin Chinese. Journal of Phonetics 11, 337-351.

Couper-Kuhlen, E. 1986. An Introduction to English Prosody.
Baltimore: Edward Arnold.

Cruttenden, A. 1986. Intonation. New York: Cambridge
University Press.

Crystal, D. 1972. The intonation system of English.
Intonation. Bolinger, D. L. (ed.), 110-136. Baltimore:
Penguin Books Inc.

Cutler, A. 1977. The context-dependence of "intonation
meanings". Chicago Linguistic Society 13, 104-115.

218

Dickerson, L. 1974. Internal and external patterning of
phonological variability in the speech of Japanese
learners of English: Toward a theory of second language
acquisition. Ph.D. dissertation, University of
Illinois.

Dickerson, L. 1975. The learner's interlanguage as a system
of variable rules. TESOL Quarterly 9, 401-407.

Dickerson, W. 1976. The psycholinguistic unity of language
learning and language change. Language Learning 26,
215-231.

Dickerson, W. 1977a. Interlanguage phonology: Current
research and future directions. The Notions of
Simplification, Interlanguages and Pidgins. S. Corder
and E. Roulet (eds.). Neuchatel: Universite de
Neuchatel.

Dickerson, W. 1977b. Language variation in applied
linguistics. International Review of Applied
Linguistics 35, 43-66.

Dickerson, L. and Dickerson, W. 1977. Interlanguage
phonology: Current research and future directions. The
Notions of Simplification, Interlanguages and Pidgins
and Their Relation to Second Language Pedagogy. S.
Corder and E. Roulet (eds.), 18-29. Geneva: Droz.

Dreher, J. J. and Lee, Pao-Chen. 1968. Instrumental
investigation of single and paired Mandarin tonemes.
Monumenta Serica 20, 343-373.

Dulay, H. and Burt, M. 1973. Should we teach children
syntax? Language Learning 23, 245-258.

Elman, J., Diehl, R., and Buchwald, S. 1977. Perceptual
switching in bilinguals. Journal of the Acoustical
Society of America 62, 971-974.

Flege, J. 1987. The production of "new" and "similar" phones
in a foreign language: evidence for the effect of
equivalence classification. Journal of Phonetics 15,
47-65.

Flege, J. 1988. The production and perception of foreign
language speech sounds. Human Communication and Its
Disorders. H. Winitz (ed.), 224-401. Norwood, NJ: Ablex
Publishing.

Flege, J. and Davidian, R. 1985. Transfer and developmental
processes in adult foreign language speech production.

219

Journal of Applied Psycholinguistic Research 5,
323-347.

Flege, J. and Hillenbrand, J. 1984. Limits on phonetic
accuracy in foreign language speech production. Journal
of the Acoustical Society of America 76, 708-721.

Fry, D. B. 1955. Duration and intensity as physical
correlates of linguistic stress. Journal of the
Acoustical Society of America 27, 765-768.

Fry, D. B. 1958. Experiments in the perception of stress.
Language and Speech 1, 126-152.

Gandour, J. 1978. The Perception of Tone. Tone: A
Linguistic Survey. Fromkin, V. (ed.), 41-76. New York:
Academic Press.

Gao, Yu-zhen. 1980. Beijing hua de qingsheng wenti (Issues
of the neutral tone in Beijing Mandarin). Yuyan Jiaoxue
yu Yanjiu 2, 82-98.

Garding, E., Kratochvil, P., Svantesson, J. and Zhang, J.
1986. Tone 4 and Tone 3 discrimination in Modern
Standard Chinese. Language and Speech 29, 281-293.

Gass, S. 1984. Development of speech perception and speech
production abilities in adult second language learners.
Applied psycholinguistics 5, 51-74.

Gass, S. and Madden, C. (eds.) 1985. Input in Second
Language Acquisition. Rowley, Mass.: Newbury House.

Gatbcnton, E. 1975. Systematic variation in second language
speech: A sociolinguistic study. Ph.D. Dissertation,
McGill University.

Gillette, S. 1980. Contextual variation in the perception of
L and R by Japanese and Korean speakers. Minnesota
Papers in Linguistics and the Philosophy of Language
(No.6, 59-72). University of Minnesota, Minneapolis,
MN.

Goldstein, H. 1983. Word recognition in a foreign language:
A study of speech perception. Journal of
Psycholinguistic Research 12, 417-427.

Goto, H. 1971. Auditory perception by normal Japanese adults
of the sounds "l" and "r". Neuropsychlogia, 9, 317-323.

Harris, J. 1969. Spanish Phonology. Cambridge: MIT Press.

220

Hecht, B. F. and Mulford, R. 1982. The acquisition of a
second language phonology: interaction of transfer and
developmental factors. Applied Psycholinguistics 3,
313-328.

Hieke, A. E. 1990. Toward listener strategies for decoding
fluent speech. International Review of Applied
Linguistics 28, 221-233.

Ho, Aichen T. 1976a. The acoustic variation of Mandarin
tones. Phonetica 33, 353-367.

Ho, Aichen T. 1976b. Mandarin tones in relation to sentence
intonation and grammatical structure. Journal of
Chinese Linguistics 4, 1-13.

Ho, Aichen T. 1977. Intonation variations in a Mandarin
sentence for three expressions: interrogative,
exclamatory, and declarative. Phonetica 34, 446-456.

Howie, J. M. 1976. Acoustical studies of Mandarin vowels and
tones. New York: Cambridge University Press.

Ioup, G. and Tansomboon, A. 1987. The acquisition of tone: A
maturation perspective. Interlanguage Phonology. G.
Ioup and S. Weinberger (eds.), 333-349. Cambridge:
Newbury House.

Kiriloff, C. 1969. On the auditory perception of tones in
Mandarin. Phonetica 20, 63-67.

Klatt, D. H. and Stevens, K. N. 1973. On the automatic
recognition of continuous speech: implications from a
spectrogram-reading experiment. IEEE Transactions on
Audio and Electroacoustics, AU-21, 210-217.

Krashen, S. D. 1977. Some issues relating to the Monitor
Model. On TESOL '77. H. Brown, C. Yorio, and R.Crymes
(eds.), 144-158. Washington D.C.: TESOL.

Krashen, S.D. 1981. Aptitude and attitude in relation to
second language acquisition and learning. Individual
Differences and Universals in Language Learning
Aptitude. K. C. Diller (ed.), 155-175. Rowley: Newbury
House Publishers.

Krashen, S.D. 1982. Principles and Practice in Second
Language Acquisition. New York: Pergamon.

Krashen, S. D. and Terrell, T. D. 1983. The Natural
Approach. New York: The Alemany Press.

221

Larsen-Freeman D. and Long, M. H. 1991. An Introduction to
Second Language Acquisition Research. New York:
Longman.

Leather, J. 1983. Speaker normalization in perception of
lexical tone. Journal of Phonetics 11, 373-382.

Leather, J. 1987. F0 pattern inference in the perceptual
acquisition of second language tone. Sound Patterns in
Second Language Acquisition, A. James and J. Leather
(eds.), 41-80. Netherlands: Foris Publications.

Lehiste, I. 1970. Suprasegmentals. Cambridge: the M.I.T.
Press.

Li, C. N. and Thompson, S. A. 1977. The acquisition of tone
in Mandarin-speaking children. Journal of Child
Language 4, 185-199.

Li, C. N. and Thompson, S. A. 1978. The acquisition of tone.
Tone: A linguistic Survey. V. A. Fromkin (ed.),
271-284. New York: Academic Press.

Liberman, M. and Sag, I. 1974. Prosodic form and discourse
function. Papers from the Tenth Regional Meeting,
Chicago Linguistic Society, 416-427.

Lieberman, P. 1960. Some acoustic correlates of word stress
in American English. The Journal of the Acoustical
Society of America 32, 451-454.

Lieberman, P. 1965. On the acoustic basis of the perception
of intonation by linguists. Word 21, 40-54.

Lightbown, P. and Spada, N. 1990. Focus-on-form and
corrective feedback in communicative language teaching.
Studies in Second Language Acquisition 12, 429-446.

Lin, Mao-can and Yan, Jing-zhu. 1980. Beijinghua qingsheng
de shengxue xingzhi (The acoustic characteristics of
neutral tone in Beijing Mandarin). Fang Yan 3, 166-178.

Lin, William C.J. 1985. Teaching Mandarin tones to adult
English speakers: Analysis of difficulties with
suggested remedies. Regional English Language Center
Journal 16, 31-47.

Lindstrom, O. 1978. Aspects of English Intonation. ACTA
Universitatis Gothoburgensis.

222

Lisker, L. 1970. On learning a new contrast. Haskins
Laboratory Status Report on Speech Research No. 24, 1-
15. New Haven, CT.

Lisker, L. and Abramson, A. 1964. A cross-language study of
voicing in initial stops: Acoustical measurements. Word
20, 384-422.

Lisker, L. and Abramson, A. 1967. Some effects of context on
voice onset time in English stops. Language and Speech
10, 1-28.

Liu, Xiun, Guo, En-Ming, and Liu, She-Hui (eds.), 1988
Practical Chinese Reader, Book I. Beijing: the
Commercial Press.

Locke, J. L. 1970. The value of repetition in articulation
learning. International Review of Applied Linguistics
8, 147-154.

Locke, J. L. 1983. Phonological Acquisition and Change. New
York: Academic Press.

Loewenthal, K. and Bull, D. 1984. Imitation of foreign
sounds: What is the effect of age? Language and Speech
27, 95-97.

Lyovin, A. 1978. Review of Tone and Intonation in Mandarin
Chinese, by Rumjancev, M. Moscow: Izdatel'stvo
Moskovskogo Universiteta. Journal of Chinese
Linguistics 6, 120-168.

MacKain, K. 1988. Filling the gap between speech and
language. The Emergent Lexicon. M. D. Smith and J. L.
Locke (eds.), 51-74. New York: Academic Press.

MacKain, K., Best, C. and Strange, W. 1981. Categorical
perception of English /r/ and /l/ by Japanese
bilinguals. Applied Psycholinguistics 2, 369-390.

Major, R. C. 1987. Phonological similarity, markedness, and
rate of L2 acquisition. Studies in Second Language
Acquisition 9, 63-82.

Massaro, D., Cohen, M. and Tseng, Chiu-yu. 1985. The
evaluation and integration of pitch height and pitch
contour in lexical tone perception in Mandarin Chinese.
Journal of Chinese Linguistics 13, 267-290.

McClasky, C., Pisoni, D. and Carrell, T. 1983. Transfer of
training of a new linguistic contrast in voicing.
Perception and psychophysics 34, 323-330.

223

Miller, G. A., Heise, G. and Lichten, W. 1953. The
intelligibibity of speech as a function of the context
of the test materials. Journal of Experimental
Psychology 41, 329-335.

Miracle, W. C. 1989. Tone production of America students of
Chinese: a preliminary acoustic study. Journal of
Chinese Language Teachers Association 14:3, 49-65.

Mochizuki, M. 1981. The identification of /r/ and /l/ in
natural and synthesized speech. Journal of Phonetics 9,
283-303.

Morton, J. and Jassem, W. 1965. Acoustic correlates of
stress. Language and Speech 8, 159-181.

Mueller, T. H. and Niedzielski, H. 1968. The influence of
discrimination training on pronunciation. Modern
Language Journal 52, 410-416.

Neufeld, G. 1978. On the acquisition of prosodic and
articulatory features in adult language learning. The
Canadian Modern Language Review 34, 163-174.

Neufeld, G. 1979. Towards a theory of language learning
ability. Language Learning 29, 227-241.

Neufeld, G. 1980. On the adult's ability to acquire
phonology. TESOL Quarterly 14, 285-298.

Neufeld, G. and Schneiderman, E. 1980. Prosodic and
articulatory features in adult language learning.
Research in Second Language Acquisition, R. Scarcella,
and S. Krashen (eds.), 105-109. Cambridge: Newbury
House Publishers.

Obler, L. 1982. The Parsimonious Bilingual. Exceptional
Language and Linguistic Theory. L. Obler and L. Menn
(eds.), 339-346. New York: Academic Press.

Odlin, T. 1989. Language Transfer. New York: Cambridge
University Press.

Olson, L. and Samuels, S. J. 1973. The relationship between
age and accuracy of foreign language pronunciation.
Journal of Educational Research 66, 263-267.

Oyama, S. 1976. A sensitive period for the acquisition of a
nonnative phonological system. Journal of
Psycholinguistic Research 5, 261-283.

224

Pike, E. V. 1959. A test for predicting phonetic ability.
Language Learning 9, 35-43.

Pike, K. L. 1972. General characteristics of Intonation. D.
L. Bolinger (ed.), 53-82. Baltimore: Penguin Books Inc.

Pimsleur, P. 1963. Discrimination training in the teaching
of French pronunciation. Modern Language Journal 47,
199-203.

Piper, T. and Cansin, D. 1988. Factors influencing the
foreign accent. The Canadian Modern Language Review 44,
335-342.

Pisoni, D., Aslin, R., Perey, A. and Hennessy, B. 1982. Some
effects of laboratory training on identification and
discrimination of voicing contrasts. Journal of
Experimental Psychology: Human Perception and
Performance 8, 297-314.

Pisoni, D. and Luce, P. 1986. Speech perception: research,
theory, and the principle issues. Pattern Recognition
by Humans and Machines, vol. 1, Speech Perception.
Schwab and Nusbaum (eds.), 1-50. Orlando: Academic
Press.

Purcell, E. T. and Suter, R. W. 1980. Predictors of
pronunciation accuracy: a reexamination. Language
Learning 30, 271-287.

Rivers, W. 1979. Learning a sixth language: an adult
learner's diary. Canadian Modern Language Review 36,
67-82.

Ross, E. D., Edmondson, J.A., and Seibert, G.B. 1986. The
effect of affect on various acoustic measures of
prosody in tone and non-tone languages. Journal of
Phonetics 14, 283-302.

Rutherford, W. 1987. Second Language Grammar: Learning and
Teaching. London: Longman.

Rutherford, W. and Sherwood-Smith, M. (eds.) 1988. Grammar
and Second Language Teaching. New York: Newbury House.

Sato, C. 1985. Task variation in interlanguage phonology.
Input in Second Language Acquisition. S. Gass and C.
Madden (eds.), 181-196. Rowley, Mass.: Newbury House.

Schneiderman, E., Bourdages, J. and Champagne, C. 1988.
Second-language accent: the relationship between

225

discrimination and perception. Language Learning 38,
1-19.

Scott, D. 1987. Prosody. New Horizons in Linguistics 2.
J. Lyons, R. Coates, M. Deuchar and G. Gazdar (eds.),
82-102. Baltimore: Penguin Books.

Scovel, T. 1988. A Time to Speak. New York: Newbury House
Publishers.

Segalowitz, N. and Gatbcnton, E. 1977. Studies of the
non-fluent bilingual. Bilingualism: Psychological,
Sociological and Educational Implications, P. Hornby
(ed.), 77-89. New York: Academic.

Seliger, H., Krashen, S. and Ladefoged, P. 1975.
Maturational constraints in the acquisition of second
language accent. Language Sciences, Aug. 1975, No. 36,
20-22 a

Sheldon, A. and Strange, W. 1982. The acquisition of /r/ and
/l/ by Japanese learners of English: Evidence that
speech production can precede speech perception.
Applied Psycholinguistics 3, 243-261.

Shen, Xiaonan S. 1985. A contrastive study of Mandarin
Chinese and French interrogative intonologies: A
contribution to the prosody of Chinese. Ph.D.
dissertation, University of California, Berkeley.

Shen, Xiaonan S. 1989. Toward a register approach in
teaching Mandarin tones. Journal of Chinese Language
Teachers Association 14:3, 27-47.

Shen, Xiaonan S. 1990. The prosody of Mandarin Chinese.
University of California Publications in Linguistics,
vol. 118.

Shimizu, K., and Dantsuji, M. 1983. A study of the
perception of /r/ and /l/ in natural and synthetic
speech sounds. Studia Phonologica 17, 1-14.

Singleton, D. and Little, D. 1984. A first encounter with
Dutch: Perceived language distance and language
transfer as factors in comprehension. Language Across
Cultures. L. MacMathuna and D. Singleton (eds.).
Dublin: Irish Association for Applied Linguistics.

Smith, M. D. and Locke, J. L. (eds.) 1988. The Emergent
Lexicon. New York: Academic Press.

226

Snow, C. E. and Hoefnagel-Héhle, M. 1977. Age differences in
the pronunciation of foreign sounds. Language and
Speech 20, 357-365.

Snow, C. E. and Hoefnagel-Héhle, M. 1978. The critical
period for language acquisition: evidence from second
language learning. Child Development 49, 1114-1128.

Snow, C. E. and Hoefnagel-Hohle, M. 1979. Individual
differences in second-language ability: a
factor-analytic study. Language and Speech 22, 151-162.

Strange, W. and Jenkins, J. 1978. The role of linguistic
experience in the perception of speech. Perception and
Experience. R. Walk and H. Pick, Jr. (eds.), 125-169.
New York: Plenum.

Suter, R. 1976. Predictors of pronunciation accuracy in
second language learning. Language Learning 26,
233-253.

Tahta, S. 1980. An investigation into second language
learning. Unpublished M. Phil. thesis. City University,
London.

Tahta, S., Wood, M. and Loewenthal, K. 1981a. Foreign
accents: factors relating to transfer of accent from
the first language to a second language. Language and
Speech 24, 265-272.

Tahta, S., Wood, M. and Loewenthal, K. 1981b. Age changes in
the ability to replicate foreign pronunciation and
intonation. Language and Speech 24, 363-372.

Tees, P. C. and Werker, J. J. 1984. Perceptual flexibility:
Maintainance or recovery of the ability to discriminate
nonnative speech sounds. Canadian Journal of Psychology
38, 579-590.

Thompson, I. 1984. An experimental study of foreign accents.
Ph.D. Dissertation, George Washington University.

Tseng, chiu-yu 1981. An Acoustic Phonetic Study on Tones in
Mandarin Chinese. Ph.D. Dissertation, Brown University.

Van Els, T. and De Bot, K. 1987. The role of intonation in
foreign accent. The Modern Language Journal 71,
147-155.

Walz, J. 1975. A longitudinal study of the acquisition of
French pronunciation. Ph.D. dissertation, Indiana
University.

227

Wang, W. S-Y. and Li, Kung-Pu. 1967. Tone 3 in Pekinese.
Journal of Speech and Hearing Research 10, 629-636.

White, C. 1980. Mandarin Tone and English Intonation: A
Contrastive Analysis. M.A. thesis, the University of
Arizona.

White, C. 1981. Tonal pronunciation errors and interference
from English intonation. Journal of Chinese Language
Teachers Association 16:2, 27-56.

Williams, L. 1977. The perception of stop consonant voicing
by Spanish-English bilinguals. Perception and
Psychophysics 21, 289-297.

Williams, L. 1979. The modification of speech perception and
production in second-language learning. Perception and
Psychophysics 26, 95-104.

Williams, L. 1980. Phonetic variation as a function of
second-language learning. Child Phonology, Vol. 2,
Perception. G. Yeni-Komshian, J. Kavanagh, and '
C. Ferguson (eds.), 185-216. New York: Academic.

Winitz, H. 1981. Input considerations in the comprehension
of first and second language. Native Language and
Foreign Language Acquisition. H. Winitz (ed.), 296-308.
New York: the New York Academy of Sciences.

Wu, Zong-ji. 1982. Putonghua yuju zhong de shengdiao bianhua
(Tonal variations in Mandarin sentences). Zhongguo
Yuwen 6, 439-449.

Yamada, J., Takatsuka, S., Kotake, N. and Kurusu, J. 1980.
On the optimum age for teaching foreign vocabulary to
children. International Review of Applied Linguistics
18, 245-247.

Zhao, J-M. 1987. Cong Yixie Shengdiao Yuyan de Shengdiao
Shuodao Hanyu Shengdiao. Paper presented at the 2nd
International Coference on Teaching Chinese. August 10-
14, 1988, Beijing.

APPENDICES

228

APPENDIX A

Tone Perception

Part (A): l-syllable words. Please listen to eacn word, then write
down the tone number or tone mark on your answer sneet. Each word
will be spoken twice. Please l1sten carefully.

1. ku4 2. baJ 3. lu2 4. d1l
5. pe12 6. pa4 7. meiJ 8. pacl
9. n12 10. rul 11. m13 12. aa2
l3. bu4 l4. gue13 15. hao4 16. reil
17. fas 18. lal 19. da14 20. pacZ

Part (B): 2-syllab1e words. Please l1sten to each word, then write
down the tone numbers or tone marks on your answer sheet. Each word
will be spoken three times. Please listen carefully.

l. faxianl4 2. pingfangZZ 3. youbangBl 4. huamaoll

5. dianji41 6. honghaiZJ. 7. ciyu44 8. naip1n932

9. daoda42 lU. malu34 ll. x1nq1n912 12. tikuzl

13. p1da124 l4. mox1e43 15. na1zu123 16. xiangshuilB

Part (C): Short sentences. You will hear an.English sentence f1rst,
followed by four choices in Chinese. Only one Chinese sentence has
the correct tones. Please choose the correct one by circling the
letter a, b, c, or d. For example: How are you? (a) 511 hao ma?
(:9) N12 hao ma? (c) N13 hao ma? ((1) N14 hao ma? The correct answer
is (b), so you should circle the letter b.

1. How is he? Ta 532 ma? (a) 2 (b) 4 (c) 1 (d) 3
2. He is fine. Ta hen hag. (a) l (b) 4 (c) 3 (d) 2
3. Are you busy? Ni mggg ma? (a) 4 (b) 3 (c) 2 (d) 1
4. I’m not busy. Wo bu mggg. (a) 3 (b) 2 (c) l (d) 4
5. They’re all busy.Tamen 922 mang. (a) 4 (b) 1 (c) 2 (c) 3

6. We are also busy.women yg mang. (a) 2 (b) 4 (c) l (d) 3
7. What is this? Zhe shi snenme? ta) 3 (b) 4 (c) l (d) 2

8. Who is he? 22 shi shei? (a) 4 (b) 2 (c) 1 (d) 3

229

Pronunciation

There are three parts in this test: 1-syllable words, 2-syllab1e

words, and short sentences. For each part, first you will hear the
words or the sentences on the tape. Please repeat each word or
sentence right after you hear it. Then, please read the word or

sentence from the following list after you hear the number.
(A) l-syllable words
1. m5 2. f6 3. féi 4. pi 5. t3

6. dEi 7. n1 8. 951 9. a? 10. 156

(B) 2-syllable words

1. Meiguo 2. mi1fi 3. nfiang hé 4. as hgi
5. laoshi 6. gSoyao 7. ditfi 8. péoma
9. gSngping 10. HuEng hai ll. hEi mEo 12. pgobfi
13. wai yl 14. nigfi 15. fabiXo 16. lfidi

(C) short sentences

1. T5 hgo ma? (How is he?)

2. T5 hen hXb. (He is fine.)

3. T5 ye mang. (He is also busy.)

4. Ni mang ma? (Are you busy?)

5. Na shi shi. (That is a book.)

6. Zhe shi bao. (This is newspaper.)
7. T5 shi shei? (Who is he?)

8. Shéi shi 156551? (Who is a teacher?)

230

APPENDIX

Perception Scores (% correct)

Subject Sex L2 Test 1
(A) (B)

1 M Y 90 75.00
2 M Y 95 68.75
3 F Y 95 75.00
4 F Y 100 96.80
5 F Y 30 25.00
6 F Y 80 43.75
7 F Y * *

8 M Y 100 78.13
9 F N 60 50.00
10 M N 95 87.50
11 F Y 90 62.50
12 M Y 70 65.63
13 M N 60 46.88
14 F M * 62.50
15 M M 90 90.63
16 M M 85 40.63
17 M M 85 68.75
18 M Y 95 65.63
19 F Y 85 65.63
20 F Y 90 62.50
21 M Y 100 81.25
22 F Y 55 37.50
23 M Y 100 78.13
24 F Y 100 68.75
25 F Y 100 68.75
26 M Y 90 62.50
27 M Y 90 59.38
28 M N 90 56.25
29 F M 85 59.38
30 F M 80 62.50
31 F M 95 75.00

N=No L2 experience
Y=Other L2 experience
M=Mandarin experience

(C)
62.5
62.5
37.5
62.5
37.5
25.0
37.5
87.5
37.5
62.5
62.5
50.0
37.5
25.0
50.0
25.0
75.0
75.0
37.5
37.5
87.5
37.5
62.5
50.0
62.5
12.5
62.5
50.0
37.5
37.5
75.0

(A)
e

90
90
95
65
100
100
100
85
85
85
100
95
90
95
100

100
85
95

100

*
95
95
95
95

90
95
90
95

Test 2
(B)
e

96.88
96.88
100.00
40.63
100.00
93.75
100.00
78.13
100.00
100.00
78.13
81.25
84.38
100.00
75.00
*
96.88
65.63
90.63
96.88
a
87.50
100.00
93.75
90.63
*
84.38
71.88
78.13
93.75

(C)
a

75.0
100.0
100.0

62.5

75.0

87.5
100.0

87.5
100.0

75.0

87.5

87.5

87.5
100.0

87.5

75.0
75.0
75.0
87.5

100.0
100.0
100.0

87.5

75.0
75.0
50.0
87.5

231

Reading Scores (% correct)

Subject

melanoma-surel-

Test 1
(A) (B)
91.67 92.71
100.00 75.00
100.00 79.17
87.50 93.75
79.17 53.13
100.00 72.92
100.00 81.25
91.67 90.63
75.00 59.38
91.67 100.00
91.67 69.79
95.83 40.63
91.67 58.33
91.67 63.54
100.00 78.13
100.00 71.88
87.50 84.38
100.00 86.46
95.83 65.63
100.00 67.71
100.00 85.42
75.00 62.50
100.00 65.63
95.83 83.33
100.00 85.42
79.17 81.25
91.67 44.79
95.83 65.63
91.67 62.50
87.50 60.42
100.00 60.42

(C)
93.33
94.67
88.00

100.00
77.33
77.33
84.00
94.67
72.00
94.67
66.67
53.33
65.33
70.67
76.00
60.00
96.00
86.67
61.33
81.33
73.33
61.33
64.00
94.67
88.00
66.67
68.00
69.33
60.00
49.33
69.33

Test 2

(A)
e

91.67
91.67
91.67
54.17
95.83
100.00
95.83
79.17
87.50
100.00
79.17
75.00
95.83
87.50
100.00
a
95.83
70.83
100.00
91.67
a
100.00
79.17
100.00
75.00
a
91.67

87.50
a

i

(B)
e

84.38
81.25
95.83
61.46
67.71
82.29
97.92
56.25
92.71
87.50
62.50
60.42
63.54
78.13
64.58
e
88.54
61.46

67.71
s

*

63.54
86.46
87.50
73.96
*
52.08

62.50
a

*

(C)
a

80.00
77.33
90.67
76.00
85.33
80.00
89.33
76.00
94.67
56.00
48.00
62.67
77.33
89.33
58.67
*
86.67
57.33

93.33
a

*

77.33
97.33
93.33
80.00
e
85.33

74.67
a

*

232

Imitation Scores (% correct)

Subject

\DQQGUIbUNH

Test 1

(A)
95.83
95.83
100.00
100.00

91.67
100.00
100.00
100.00

95.83
100.00
100.00
100.00
100.00
100.00
100.00
100.00

91.67
100.00
100.00
100.00
100.00
100.00

91.67
100.00

95.83
100.00
100.00
100.00
100.00
100.00

95.83

(B)
93.75
94.79
94.79
97.92
97.92
97.92
97.92
98.96
98.96

100.00
97.92
95.83
94.79
94.79

100.00
98.96
95.83
96.88
89.58
96.88
95.83
89.58
95.83

100.00

100.00
98.96
96.88
98.96
95.83
97.92
98.96

(C)
96.00
100.00
97.33
100.00
97.33
96.00
100.00
96.00
97.33
100.00
96.00
98.67
96.00
88.00
96.00
97.33
93.33
93.33
92.00
100.00
94.67
93.33
90.67
96.00
100.00
81.33
96.00
93.33
93.33
96.00
97.33

Test

(A)
a

100.00
95.83
100.00
62.50
91.67
100.00
100.00
87.50
100.00
100.00
100.00
100.00
100.00
100.00
100.00
a
100.00
100.00
100.00
100.00
a
100.00
100.00
100.00
91.67
*
100.00

100.00
a

*

2
(B)
e

96.88
89.58
100.00
97.92
86.46
96.88
100.00
100.00
97.92
97.92
97.92
96.88
98.96
97.92
97.92
e
98.96
95.83
96.88
100.00
a
92.71
98.96
100.00
95.83
*
98.96

97.92
*

*

(C)
e

98.67
94.67
100.00
96.00
93.33
94.67
98.67
98.67
96.00
90.67
93.33
96.00
88.00
97.33
88.00
a
96.00
97.33

98.67
s

*

94.67
100.00
97.33
94.67
*
96.00

82.67
a

*

"I1111111711111lilies