A coustic Realization and Perception of English Lexical Stress by - - PowerPoint PPT Presentation

1 1

A Acoustic Realization and Perception of English

coustic Realization and Perception of English Lexical Stress by Mandarin Learners Lexical Stress by Mandarin Learners

Yuwen Lai Yuwen Lai

University of British Columbia University of British Columbia University of Kansas University of Kansas

Joan Joan Sereno Sereno

University of Kansas University of Kansas

Allard Allard Jongman Jongman

University of Kansas University of Kansas

2 2

Background Background

Segmental learning Segmental learning

Flege Flege 1995, James 1988, Leather & James 1991 among others 1995, James 1988, Leather & James 1991 among others

Suprasegmental learning Suprasegmental learning

• Effect on foreign accent

Effect on foreign accent

( (Anderson

Anderson-

• Hsieh et al. 1992, Munro & Derwing 1998,

Hsieh et al. 1992, Munro & Derwing 1998, Trofimovich Trofimovich and and Baker, 2006) Baker, 2006)

• Stress placement

Stress placement

( (Altmann 2006, Altmann 2006, Archibald 1997, 1998) Archibald 1997, 1998)

• Acoustic realization and perception of cues

Acoustic realization and perception of cues

(Beckman 1986, (Beckman 1986, McGory McGory 1997, Mennen 2004, 1997, Mennen 2004, Ueyama Ueyama 2001) 2001)

3 3

Prosodic features of target languages Prosodic features of target languages

L1: L1: Taiwan Mandarin Taiwan Mandarin

• Tone language

Tone language

Phonemic role of F0 Phonemic role of F0

• Syllable

Syllable-

• timed rhythm

timed rhythm L2: L2: American English American English

• Stress language

Stress language

F0, intensity, duration, and spectral composition F0, intensity, duration, and spectral composition

• Stress

Stress-

• timed rhythm

timed rhythm

4 4

The present study The present study

Research questions Research questions

How do learners utilize F0 and duration differently in How do learners utilize F0 and duration differently in production of English lexical stress? production of English lexical stress? How do learners weigh these two cues in perception? How do learners weigh these two cues in perception? Is there any correlation between production and Is there any correlation between production and perception? perception?

5 5

Acoustic study Acoustic study

Stimuli Stimuli

• 14 word pairs:

14 word pairs: ob

• bject

ject-

• ob
• bject

ject (noun (noun-

• verb)

verb) Participants

• 9 beginning and 9 advanced L2 learners
• 10 native English speakers

Task

• Read target word in a carrier sentence, repeat the

Read target word in a carrier sentence, repeat the target word in isolation after each sentence. target word in isolation after each sentence.

• nly the word produced in isolation was analyzed.
• nly the word produced in isolation was analyzed.

Measurements Measurements

• Max and mean F0, duration, intensity, and F2

Max and mean F0, duration, intensity, and F2

6 6

Examples Examples

• bject

Native English speakers

• bject

Beginning learners Advanced learners

7 7

Results Results

first first-

• to

to-

• second vowel ratios

second vowel ratios

1.0

• All speaker groups utilize all correlates in nouns
• Native speakers use max F0 and intensity to a greater extent than others

nouns nouns

8 8

Results Results

first first-

• to

to-

• second vowel ratios (verbs)

second vowel ratios (verbs)

1.0

• Learners utilize all correlates in verbs
• Native speakers rely mainly on duration

9 9

Results Results

stressed-to-unstressed vowel ratio, noun: v1/v2, verb: v2/v1

Mean F0

noun verb

Max F0

noun verb

Duration

noun verb

Intensity

noun verb

10 10

Results Results

Striking asymmetry in max F0 and duration realization for nouns and verbs for native speakers

noun verb 1.0

Max F0

noun verb

Duration

11 11

Acoustic study Acoustic study-

• summary

summary

Native speakers

marked asymmetry of realization in nouns and verbs

reduce unstressed vowels in all conditions

Beginning learners

use consistent magnitudes of max F0 and duration for

nouns and verbs

do not always reduce unstressed vowels

Advanced learners

pattern between beginners and native speakers

12 12

Perception Perception-

• methodology

methodology

Stimulus construction Stimulus construction

• Spectral composition (dada and

Spectral composition (dada and d dəda da) )

• Max F0

Max F0

five first five first-

• to

to-

• second vowel ratios

second vowel ratios

• Duration

Duration

five first five first-

• to

to-

• second vowel ratios

second vowel ratios

13 13

Stimulus construction Stimulus construction-

• spectral composition

spectral composition

da0 də0 də1 da1 də2

duration pitch tier intensity tier

də3

144 ms 197 ms 197 ms

‘DAdu da’DU

naturally produced

14 14

Stimulus construction Stimulus construction

Max F0 Duration

15 15

Perception Perception-

• Methodology

Methodology

Participants Participants

• 25 beginning, 25 advanced learners

25 beginning, 25 advanced learners

• 25 native listeners

25 native listeners Procedure Procedure-

• stress localization task

stress localization task

DAda daDA

16 16

Statistical analysis Statistical analysis

Overall analysis (four Overall analysis (four-

• way ANOVA)

way ANOVA)

• Between

Between-

• subjects factor

subjects factor-

• proficiency

proficiency (3) (3)

• Within

Within-

• subjects factors

subjects factors spectral composition spectral composition (2) (2) duration duration (5) (5) max F0 max F0 (5) (5)

• Dependent variable: localization responses

Dependent variable: localization responses DAda DAda – – first syllable stressed = 1 first syllable stressed = 1 daDA daDA – – second syllable stressed = 2 second syllable stressed = 2 Spectral composition has a similar effect on all proficiencies Spectral composition has a similar effect on all proficiencies

• d

də əda da tokens receive significantly more verb responses tokens receive significantly more verb responses

17 17

Native speakers gradient effect Beginning learners no effect Advanced learners strong effect

Results Results-

• Max F0

Max F0 × Proficiency (*) Proficiency (*)

DAda daDA

18 18

Beginning learners strong effect Advanced learners gradient effect

Results Results-

• Duration

Duration × Proficiency (*) Proficiency (*)

Native speakers gradient effect

19 19

Perception Perception -

• Discussion

Discussion

Native speakers Native speakers

• sensitive to

sensitive to max F0 and duration max F0 and duration Beginning learners Beginning learners

• sensitive to small

sensitive to small duration duration differences differences

• insensitive to small max F0 differences

insensitive to small max F0 differences Advanced learners Advanced learners

• sensitive to max F0 and duration, with a stronger

sensitive to max F0 and duration, with a stronger effect from effect from max F0 max F0 cue cue

20 20

Cue weighting in perception Cue weighting in perception

Duration outweighs max F0 cue when in conflict noun verb noun noun

21 21

Perception

Same degree of max F0 difference between stressed and unstressed syllables does not trigger same degree of responses in verb contexts (chance level) as it does in noun contexts

Max F0 Max F0 in native production and perception in native production and perception

Production use Max F0 in nouns but not in verbs noun verb n.s.

22 22

Perception verb responses require greater duration differences

noun verb

Production duration used to a greater extent in verbs

Duration Duration in native production and perception in native production and perception

23 23

General discussion General discussion

Native production and perception Native production and perception

• If a cue is

If a cue is not used not used in production in a given context, it in production in a given context, it does not affect perception in that context does not affect perception in that context

• If a cue is

If a cue is used used in production in a given context, in production in a given context, perceptual sensitivity correlates with its magnitude perceptual sensitivity correlates with its magnitude used in production used in production

24 24

General discussion General discussion

The F0 and rhythmic features in Mandarin affect the The F0 and rhythmic features in Mandarin affect the production and perception of English lexical stress production and perception of English lexical stress

• Phonemic F0

Phonemic F0

adoption of similar F0 ratios between stressed and adoption of similar F0 ratios between stressed and unstressed vowels in production (in nouns and verbs) unstressed vowels in production (in nouns and verbs) low sensitivity to small F0 differences in perception low sensitivity to small F0 differences in perception

• Syllable

Syllable-

• based rhythm

based rhythm

causes difficulties in shortening unstressed syllable but causes difficulties in shortening unstressed syllable but triggers high sensitivity in perception triggers high sensitivity in perception

25 25

Conclusion Conclusion

Prosodic features of Taiwan Mandarin affect the Prosodic features of Taiwan Mandarin affect the acquisition of English lexical stress in both acoustic acquisition of English lexical stress in both acoustic realization and perception realization and perception When comparing stress realization between non When comparing stress realization between non-

• native

native and native speakers, care should be taken to distinguish and native speakers, care should be taken to distinguish stress location stress location Correlations between native production and perception Correlations between native production and perception warrant further investigation warrant further investigation

26 26

Acknowledgements Acknowledgements

Experiment participants Experiment participants Praat Praat scripts by scripts by Mietta Mietta Lennes Lennes , Yi , Yi Xu Xu, and , and Holger Holger Mitterer Mitterer Discussion with David Montero, Travis Wade and Discussion with David Montero, Travis Wade and University of Kansas Phonetics and Psycholinguistics University of Kansas Phonetics and Psycholinguistics Laboratory members Laboratory members

27 27

Results Results -

• F2

F2

Vowel reduction/centralization only in native production

syllable 1: syllable 1: stressed in stressed in nouns nouns, unstressed in , unstressed in verbs verbs

Vowel reduction Vowel reduction – – lowering in front vowels, rising in back vowels lowering in front vowels, rising in back vowels

28 28

Results Results-

• spectral composition

spectral composition

syllable 2 syllable 2

Vowel reduction/centralization in all groups

29 29

30 30

Advanced learners Advanced learners

Max F0 outweighs duration when in conflict

31 31

Beginning learners Beginning learners

Only sensitive to duration in all cue combinations

32 32

Advanced learners Advanced learners

Max F0 outweighs duration when in conflict

33 33

Acoustic study 2 Acoustic study 2

stimuli

• mon

monkey key-

• key

key (4 pairs) (4 pairs)

participants

18 NNS (9 beg. & 9 adv.) and 10 NS

measurements

F0 contours

34 34

Results Results

Learners’ F0 contours have greater F0 drops and the rise/plateau at the onset resembles the profile of the Mandarin high falling tone

35 35

Research questions Research questions

How do learners utilize F0 and duration differently from How do learners utilize F0 and duration differently from native speakers in production of English lexical stress? native speakers in production of English lexical stress?

• Learners use consistent magnitude of F0 and duration across

Learners use consistent magnitude of F0 and duration across different stress locations different stress locations

• Native speakers utilize max, mean F0, intensity, and duration

Native speakers utilize max, mean F0, intensity, and duration in noun readings but rely mainly on duration for verb readings in noun readings but rely mainly on duration for verb readings

How do learners weigh these two cues in perception? How do learners weigh these two cues in perception?

• Beginning learners focus on duration

Beginning learners focus on duration

• Advanced learners focus on max F0

Advanced learners focus on max F0

• Native speakers focus on both

Native speakers focus on both

Is there any correlation between production and Is there any correlation between production and perception? perception?

• Not found in learners

Not found in learners

• Native speakers

Native speakers

36 36

Conclusion and future research Conclusion and future research

F0 F0

• Contrastive to contrastive

Contrastive to contrastive

• Contrastive to non

Contrastive to non-

• contrastive

contrastive

• Non

Non-

• contrastive to contrastive

contrastive to contrastive

Duration Duration

• Contrastive to contrastive

Contrastive to contrastive

• Contrastive to non

Contrastive to non-

• contrastive

contrastive

• Non

Non-

• contrastive to contrastive

contrastive to contrastive

37 37

Previous research on American English Previous research on American English

Fry (1955) Fry (1955)

• production

production

stimuli: object, subject, digest, contract, permit stimuli: object, subject, digest, contract, permit ↑ ↑ vowel duration, vowel duration, ↑

↑ intensity (F0 not measured)

intensity (F0 not measured)

• perception

perception

stimuli: intensity (5 levels), duration (5 levels) stimuli: intensity (5 levels), duration (5 levels)

equal +5 db +10 db

• 10 db
• 5 db

1 2 3 4 5

38 38

Methodology Methodology

Participants Participants

• 25 native English speakers

25 native English speakers

• Mandarin L2 learners of English

Mandarin L2 learners of English

25 beginning 25 beginning 25 advanced 25 advanced

Procedure Procedure

DAda

250 ms ISI

daDA

39 39

Results Results

(adopted from Fry 1955)

0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% 1 2 3 4 5 increasing ratio v1/v2 Percentage of "noun" judgement duration intensity

29% 70 %

40 40

Methods Methods-

• procedure

procedure

It is a It is a re record cord of your grades.

• f your grades.

read the sentence, repeat highlighted word underlined syllable is stressed

41 41

Acoustic measurements Acoustic measurements

vowel 1 vowel 2 Time (s) 0.8054

F2 F2 F2 F2

• bject
• bject

vowe 1 vowel 2 Time (s) 0.8923

42 42

Production Production-

• Lieberman (1960)

Lieberman (1960)

American English American English

• 6 female and 10 male speakers

6 female and 10 male speakers

Noun Noun-

• Verb pair

Verb pair

• 2 listeners listen to the production twice

2 listeners listen to the production twice

select the stressed syllable for further processing select the stressed syllable for further processing

• Results

Results

stressed syllable stressed syllable

• 90% (of all cases) higher in F0

90% (of all cases) higher in F0

• 87% higher in amplitude

87% higher in amplitude

• 66% longer in duration

66% longer in duration

• Conclusion

Conclusion

higher F0 and amplitude most relevant correlates higher F0 and amplitude most relevant correlates

43 43

Previous research on stress in L2 Previous research on stress in L2

French to English French to English-

• Per

Per (Fry1972)

(Fry1972)

• basic

basic not as sensitive to duration not as sensitive to duration as NS as NS

• advanced patterns like NS

advanced patterns like NS

Japanese to English Japanese to English-

• Per (

Per (Beckman 1986)

Beckman 1986)

• NNS

NNS more sensitive to F0 more sensitive to F0 than NS than NS

Japanese to English Japanese to English-

• Pro

Pro (

(Ueyama Ueyama 2000) 2000)

• NNS use higher F0 ratio than NS

NNS use higher F0 ratio than NS

• Basic do not use duration; advanced pattern in between

Basic do not use duration; advanced pattern in between

Chen et. al (2002) Mandarin to English Chen et. al (2002) Mandarin to English

• NNS use

NNS use ↑ ↑ F0, duration, and intensity as NS F0, duration, and intensity as NS

44 44

Production of sentential stress Production of sentential stress

Mandarin to English Mandarin to English

Ss Ss 40 (Beijing) Mandarin speakers and 40 English 40 (Beijing) Mandarin speakers and 40 English speakers speakers Stimuli Stimuli “ “I I bought a cat there. bought a cat there.” ” “ “I I BOUGHT BOUGHT a cat there a cat there” ” “ “I bought a I bought a CAT CAT there. there.” ” “ “I bought a cat I bought a cat THERE THERE. .” ” measurements measurements F F0

0, duration, intensity

, duration, intensity

• F

F0

0=F

=F0[stressed]

0[stressed] –

– (average F (average F0 [unstressed]

0 [unstressed])

) (Chen et. al. 2001)

45 45

Chen et. al. (2001) Chen et. al. (2001)-

• Results

Results

Duration stressed unstressed F0 stressed unstressed

• -- Mandarin speaker
• -- English speaker

Intensity stressed unstressed

46 46

Methodology Methodology-

• Procedure

Procedure

DAda

250 ms ISI

daDA

click circle to center mouse and activate stimulus Identify the stressed syllable by clicking text

max F0 ratios (5) × duration ratios (5) × vowel type (2) × 3 repetitions

47 47

48 48

49 49

Results Results-

• spectral composition

spectral composition

Reduction observed only in native speakers’ productions degree of reduction greater in back vowels than that of front vowels.

50 50

Results Results-

• spectral composition

spectral composition

Reduction observed in all speakers’ productions degree of reduction greater in back vowels than that of front vowels.

51 51

Native English speakers Native English speakers

Native speakers have more verb responses as max F0 and duration ratio decreases (cue verbs) Effects are gradient

52 52

Beginning learners Beginning learners

53 53

Advanced learners Advanced learners

54 54

Discussion Discussion-

• duration asymmetry

duration asymmetry

Final lengthening enhance duration cue in verbs Greater duration magnitude in verbs- especially by native speakers

55 55

Discussion Discussion-

• F0 and intensity asymmetry

F0 and intensity asymmetry

Down-stepping enhances F0 and intensity in nouns

56 56

Background Background

Speech Learning Model Speech Learning Model

• Flege

Flege and colleagues (1987, 1988, 1992, 1995) and colleagues (1987, 1988, 1992, 1995)

Native Language Magnet Native Language Magnet

• Kuhl

Kuhl and colleagues (1989, 1991, 1992, 1994, 1998) and colleagues (1989, 1991, 1992, 1994, 1998)

Perceptual Assimilation Model Perceptual Assimilation Model

• Best 1994

Best 1994