Perceptual Learning of Syllable-final Contrast ...

Perceptual Learning of Syllable-final Contrast: Perception and Training of Various “can” and “can’t” by Japanese, Chinese and French Listeners 大井川朋彦 Tomohiko Ooigawa 2017年9月4日 September 4th, 2017 上智大学言語科学研究科言語学専攻公聴会（口述試験）博士（言語学）

Thesis defense (viva), Ph.D. in Linguistics, Graduate School of Languages and Linguistics, Sophia University 1

My fundamental research interests • Phonological contrasts of non-native languages – Why difficult/easy ? – How to improve – Perception & Production

2

Table of contents 1. 2. 3. 4. 5. 6. 7.

Introduction Experiment I: Japanese, Chinese, French listeners Experiment II: Japanese only Experiment III: Japanese only Discussion Conclusion Implication 3



Table of contents 1. Introduction • • •

Objectve “Why?”s Literature review • •

Production of “can” and “can’t” Perception of “can” and “can’t”

5

Introduction

6

Objective • Final goal: to produce a listening training program to be able to identify “can” & “can’t” for all learners of English.

7

Objective • If can’t, – “can’t” should be avoided as English as Lingua Franca. – Or, we should say always “canNOT.”

8

Why “can” & “can’t” ?

9

Why “can” & “can’t” ? • “can” and “can’t”: ✓ Basic ✓ Used very often ✓ Should not be confused ✓ Comments from learners and workers (Takahashi & Ooigawa, 2012, 2016)

✓ Various pronunciation ✓ Vowels, consonants, cluster, stress, rhythm 10

Why “can” & “can’t” ? • However, few studies done: ✓ Nakano (1972) ✓ Sasaki (1993, 1995a, 1995b) ✓ Matsui (1998) ✓ Nakayama (2000) ✓ Takahashi and Ooigawa (2012, 2016) ✓ Ooigawa (2013, 2014) ✓ Ernestus, Kouwenhoven, and van Mulken (2017) ✓ But, so many studies on /l/ and /r/ 11

Why these listeners ? • Japanese

• Chinese • French

12

Why these listeners ? • [n] is very important to identify “can” & “can’t.”

13

Why these listeners ? • Different coda nasal phonological system – Japanese – Chinese – French

14

Why these listeners ? • Different coda nasal phonological system ✓ Japanese: キャン /kjaN/ can be … [m], [n], [ŋ], [ɴ], [Ṽ]…

No contrast

(e.g., Okada, 2009; Vance, 2008) 15

Why these listeners ? ✓ Chinese: /n/ vs /ŋ/: • 门 mén /mən/35 • 蒙 méng /məŋ/35

‘door’ ‘to cover’

✓ French: /Ṽ/, /n/, /m/ • son /sɔ/̃ • sonne /sɔn/ • somme /sɔm/

‘sound’ ‘(it) rings’ ‘sum’

(e.g., Fougeron & Smith, 1993, Léon, 2005, Lin, 2007, Vaissière, 2006, Wioland, 2005)

16

Literature review

17

Production of “can” and “can’t”

18

Production of “can” and “can’t” • can /kæn/ Weak form

Strong form

19

Production of “can” and “can’t” • can /kæn/ Weak

Strong

[kh(ə)̃ n]…

[ˈkhæ̃n]…

20

Production of “can” and “can’t” • can /kæn/ Weak

Strong

American

[kh(ə)̃ n]…

[ˈkhæ̃n]…

British

[kh(ə)̃ n]…

[ˈkhæ̃n]…

American: GA: General American → North American-type British: RP: Received pronunciation → British-type

21

Production of “can” and “can’t” • can’t

22

Production of “can” and “can’t” • can’t Weak

Strong

N/A

23

Production of “can” and “can’t” • can’t American (GA)

British (RP)

24

Production of “can” and “can’t” • can’t American

British

/kænt/ /kɑnt/ [ˈkhæ̃(n)t(˺)] [ˈkhɑ̃(n)t(˺)]

American: GA: General American → North American-type British: RP: Received pronunciation → British-type

25

Summary

(Jones et al., 2011, Wells, 2008, Collins & Mees, 2008, Cruttenden, 2008, Ladefoged, 2006, Cox, 2012, Malécot, 1960, Shockey, 2003, Stevens & Keyser, 2010, Wells, 1982, Ooigawa, 2014, Takahashi and Ooigawa, 2012, 2016, Sasaki, 1993)

26

Amer.

Brit.

I can do it. I CAN do it. I can’t do it. Next week I can. Next week I can’t. 27

Acoustic data

28

I can take it. (American)

29

I can take it. (American)

[n]

30

I CAN take it. (American)

31

I CAN take it. (American)

[n]

32

I can’t take it. (American)

33

I can’t take it. (American)

[ʔ]?

34

Next week I can. (American)

35

Next week I can. (American)

[n]

36

Next week I can’(t). (American)

37

Next week I can’(t). (American)

[ʔ]?

38

Next week I can’T. (American)

39

Next week I can’T. (American)

[n]?

40

British

41

I can take it. (British)

42

I CAN take it. (British)

43

I can’t take it. (British)

44

Next week I can. (British)

45

Next week I can’T. (British)

46

Why is [n] short or nothing ?

47

can

canned can’t American

48

Perception of “can” and “can’t”

49

Perception by native speakers

50

Perception by native speakers • Continuum/synthesized sound studies

can → can’t

51

Perception by native speakers • Continuum/synthesized sound studies

I can teach. I can get it. I can go. I can.

→ → → →

I can’t teach. I can’t get it. I can’t go. I can’t. 52

Duration modification

h k æn

̃

C

Vowel + Nasal

Closure

Sasaki (1995a, 1995b)

Sasaki (1995a, 1995b) Matsui (1998) 53

h k æn ̃

Shorter

Longer

C

More frequently identified as “can’t”

54

Perception by non-native speakers

55

Perception of “can” and “can’t” by non-natives Listeners

Japanese Mandarin Spanish Dutch

Stimuli American English

✔ ✔ ✔ ✔

Australian English

✔

(Matsui, 1998, Nakayama, 2000, Takahashi & Ooigawa, 2012, 2016, Ooigawa, 2013, 2014) (Ernestus et al., 2017) 56

Perception performance Listeners

Japanese Mandarin Spanish Dutch

Stimuli American English “can” and “can’t”

Poor Poor Poor Good

57

Why? Listeners

Mandarin Spanish Dutch

Stimuli American English “can” and “can’t”

Poor Poor Good

No coda /nt/ No coda /nt/ Coda /nt/ ok!

(Ernestus et al., 2017)

58

Perception performance Listeners

Japanese

Stimuli American English

Australian English

Poor

Better

(Takahashi & Ooigawa, 2012, 2016)

59

Therefore,

60

These are the “Real cues” nobody had pointed out! • Weak vs. Strong contrast: No use! • Vowel quality differences: No use! to discriminate/identify “can” and “can’t”

61

These are the “Real cues” nobody had pointed out! • can: long and clear [n] • can’t: audible or non-audible [t]-release [t] reduces [n] [t] makes the preceding vowel shorter [t] makes the following closure longer

62

Universal “Real cues” like F3 for /l/-/r/ • can: long and clear [n] • can’t: audible or non-audible [t]-release [t] reduces [n] ([t] makes the preceding vowel shorter) [t] makes the following closure longer

63

These are the “Real cues” nobody had pointed out!

I can do it. I c an

do it. 64


I can do it.

Positive

I c an

Negative

do it.

65


I can I c an 66


I can

h [ˈk æn]

I c an

h [ˈk æ]

̃

̃ /

h [ˈk ɑ]

̃

67

Oh, it’s like French !

I can

h [ˈk æn]

I c an

h [ˈk æ]

̃

̃ /

h [ˈk ɑ]

̃

68

And,

69

No studies on ... 1. Comparisons among Japanese and the other language listeners 2. Training study

3. Comparisons between weak and strong “can” 4. Comparisons “can(’t)” + voiceless and “can(’t)” + voiced 5. Comparisons between sentence-medial and sentence-final 6. The other accents than American and Australian Englishes

70

Hypothesis and Prediction

71

Hypothesis and Prediction

72

Speech Learning Model (SLM) • SLM (e.g., Flege, 1995) is a well-known theoretical model to account for learning of L2, including perception of non-native speech sounds.

73

Speech Learning Model (SLM) SLM: • We can learn L2 anytime (P1). • Phonetic categories in our brain (P2). • Phonetic categories are changeable (P3). • A new phonetic category can be established for an L2 sound, if we can discern the differences (H2). • We can discern them more with greater input (H3). 74

Speech Learning Model (SLM) SLM: • We can learn L2 anytime (P1). • Phonetic categories in our brain (P2). • Phonetic categories are changeable (P3). • A new phonetic category can be established for an L2 sound, if we can discern the differences (H2). • We can discern them more with greater input (H3). We can change by 75 trainings !?

However, SLM: • “The phonetic category established for L2 sounds by a bilingual may differ from a monolingual’s if: 1) the bilingual’s category is “deflected” away from an L1 category to maintain phonetic contrast between categories in a common L1-L2 phonological space;

• or 2) the bilingual’s representation is based on different features, or feature weights, than a monolingual’s” (H6).

*bilingual = L2 learner

76

However, SLM: • “The phonetic category established for L2 sounds by a bilingual may differ from a monolingual’s if: 1) the bilingual’s category is “deflected” away from an L1 category to maintain phonetic contrast between categories in a common L1-L2 phonological space;

• or 2) the bilingual’s representation is based on different features, or feature weights, than a monolingual’s” (H6). We can focus on different acoustic cues !?

77

Example

beat /bit/ • Natives: • Spanish: • Mandarin:

vs.

bit /bɪt/

Quality Quality < Duration Duration (Bohn, 1995) 78

Therefore,

79

Therefore, • Japanese listeners have learned different cue(s) from native speakers.

80

Myth?

Weak

→

Strong →

“can” “can’t”

81

Real cues

“can”

→

“can’t” →

caN ca ! n

82

Hypothesis

83

Hypothesis 1. Chinese and French listeners identify “can” and “can’t” better than Japanese listeners, because Chinese and French listeners must be sensitive to coda [n] (important cue) than Japanese listeners.

84

Hypothesis 2. Japanese learners of English will be better at identification of “can” and “can’t,” if we teach them the “Real cues,” and they train the perception.

85



Table of contents 1. Introduction

2. 3. 4. 5. 6. 7.

Experiment I: Japanese, Chinese, French listeners Experiment II: Japanese only Experiment III: Japanese only Discussion Conclusion Implication 87

Stimuli

88

Speakers North American-type (/kæn/ vs. /kænt/) • Speaker 1: American English speaker, female, 20 years old, from West Coast (Vacaville, California, US). • Speaker 2: Canadian English speaker, male, 44 years old, from South West (Calgary, Alberta, Canada). • Speaker 3: American English speaker, female, 26 years old, from East Coast (Philadelphia, Pennsylvania, US). • Speaker 4: American English speaker, male, 32 years old, from East Coast (Springfield, Pennsylvania, US). British-type (/kæn/ vs. /kɑnt/) • Speaker 5: Australian English speaker female, 27 years old, from South West (Maitland, New South Wales, Australia). • Speaker 6: South African English speaker female, 30 years old, from North East (Pretoria, Gauteng, South Africa). • Speaker 7: British English speaker, male, 26 years old, from South East England (Maidstone, Kent, UK). 89

Speakers North American-type (/kæn/ vs. /kænt/) Speaker 1: US West female Speaker 2: Canada West male Speaker 3: US East female Speaker 4: US East male

20 yrs 44 yrs 26 yrs 32 yrs

British-type (/kæn/ vs. /kɑnt/) Speaker 5: Australia East female 27 yrs Speaker 6: South Africa North East female 30 yrs Speaker 7: England South East male 26 yrs

90

Recording • In a soundproof room • Digitized at 48 kHz with 16-bit accuracy • Reading out – Do not read carefully but casually – Imagine as if you were speaking to your friend(s)

91

Stimuli I can do it. I CAN do it. I can’t do it.

I can take it. I CAN take it. I can’t take it.

Next week I can. Next week I can’t. 92

Listeners (learners of English)

*Accent varieties, esp, Chinese listeners

93

Task

94

Demo

95

Task • In front of the display of a computer • With headphones • In a quiet room

96

Result

97

Overall

98



Table of contents 1. Introduction 2. Experiment I: Japanese, Chinese, French listeners 3. 4. 5. 6. 7.

Experiment II: Japanese only Experiment III: Japanese only Discussion Conclusion Implication 100

Experiment II

101


102

Stimuli

103

Additional stimuli

104

Task

105

Task

106

Demo

107

108

Task • In front of the display of a computer • With headphones • In a quiet room

109

Result

110

Result

111

Overall

112

Overall

113

What !?

114

Once again !

115



Table of contents 1. Introduction 2. Experiment I: Japanese, Chinese, French listeners 3. Experiment II: Japanese only

4. 5. 6. 7.

Experiment III: Japanese only Discussion Conclusion Implication 117


118

Stimuli

119

Task

120

121

Result

122

Result

123

Overall

124

Overall

125

126

127

U-shaped development McLaughlin (1990) • U-shaped (developmental) curve → Restructuring “Performance may follow a U-shaped curve, declining as more complex internal representations replace less complex ones, and increasing again as skill becomes expertise.”

128

Interpretation

129

Interpretation • Control group: just use to the sounds. • Experimental group: restructuring: changing strategy: “Real cues” were acquiring.

130



Table of contents 1. 2. 3. 4.

Introduction Experiment I: Japanese, Chinese, French listeners Experiment II: Japanese only Experiment III: Japanese only

5. Discussion 6. Conclusion 7. Implication 132

Hypothesis 1. Chinese and French listeners identify “can” and “can’t” better than Japanese listeners, because Chinese and French listeners must be sensitive to coda [n] (important cue) than Japanese listeners.

133

Hypothesis 1. Chinese and French listeners identify “can” and “can’t” better than Japanese listeners, because Chinese and French listeners must be sensitive to coda [n] (important cue) than Japanese listeners. Confirmed ! 134

Overall

135


136


Confirmed ! 137


Better performance U-shaped curve More liner development 138

Overall

139

140

141



Table of contents 1. 2. 3. 4. 5.

Introduction Experiment I: Japanese, Chinese, French listeners Experiment II: Japanese only Experiment III: Japanese only Discussion

6. Conclusion 7. Implication 143

Conclusion

144

Conclusion & future studies • “can”-“can’t” perception performance: Jpn. < Chn. = Fre. – The others should be examined.

• Even Chn. & Fre. are not perfect – Training studies with “Real cues” for them.

• Training with “Real cues” emphasis is effective – Examined with other (supra-/segmental) conditions, – Training with feedback, – “Real cues” should be re-examined by natives. 145



Table of contents 1. 2. 3. 4. 5. 6.

Introduction Experiment I: Japanese, Chinese, French listeners Experiment II: Japanese only Experiment III: Japanese only Discussion Conclusion

7. Implication 147

Implication

148

Implication

149

Pedagogical implication • Chinese and French learners of English also should be trained. • North American-type “can’t” should be prohibited as English as Lingua Franca: – British-type “can’t” is recommended.

• Or, we should say always “canNOT.”

150

No development for American “can’t”

151


152


153


154

British-type “can’t” is recommended

155


156


157


158

Theoritical implication • Another example of SLM: – Wrong cues used by the non-natives – But, changeable

159

Myths should be (re-)examined

Weak

→

Strong →


160


Weak

→

Strong →


NO ! 161


“can”

→

“can’t” →

caN ca ! n

162


“can”

→

“can’t” →

caN ca ! n

Great ! 163

Implication • Another example of …

They look complex rules, but actually a simple law is hidden behind.

164

Thank you

165

Acknowledgement

166

Acknowledgement • • • •

All the participants My family members The committee members LPP members (Université Paris 3), esp: ✓Dr. Shigeko Shinohara ✓Dr. Pierre Hallé

• Mr. Jeffrey Moore (Sophia University) • Mr. Romain Isely (University of Geneva) 167