GemCon Glasgow August 2015 HIRATA

August 2015
GEMCON in Glasgow
Production of Japanese geminates by
native English speakers: Durational
accuracy and native speaker evaluation
Yukari Hirata
Colgate University
Izumi Takiguchi
Bunkyo Gakuin University
Dept. of East Asian
Languages and Literatures
Center for
Language and Brain
1
Background
•  Native English speakers learning Japanese have problems producing
Japanese consonant length (C) and vowel length (V) distinctions.
–  /kako/ ‘past’ versus /kakːo/ ‘parenthesis’
–  /kake/ ‘gambling’ versus /kakeː/ ‘family accounting’
•  For C production,
–  Beginning learners improve in the first year (Masuda and HayesHarb, 2005).
–  Beginning learners improve more readily in V than C production
(Toda, 1997).
–  Advanced learners did not reach the native-speaker level in C
production, even after a year or more living in Japan (Han, 1992).
2
Questions
1. Focus on intermediate learners: Improve or not?
To what extent does their ability to produce length distinctions
improve over the first several months in Japan? Do they improve
without focused production training?
2. Types of length distinctions?
Do intermediate learners improve less in C distinction than in V
distinction at the intermediate level, following the beginning
learners’ trend in Toda (1997)?
! Analysis of produced duration measures
(Compare with native speaker data; Hirata, 2004; Hirata & Whiton,
2005)
! Analysis of accuracy and degree of foreign accent evaluated by native
Japanese (NJ) listeners
3
Method
Participants
–  Seven students (ages: 19-21) who were native speakers of
American English and had taken two years of Japanese language
classes, totaling 300 hours, at Colgate University.
–  All students participated in a study abroad program in Japan for
four months.
–  None of them received special production or perception training
on length distinctions during this longitudinal experiment.
4
Method
Stimuli
–  C-pair: /kako/ ‘past’ (2 moras) vs /kakːo/ ‘parenthesis’ (3 moras)
–  V-pair: /ɾika/ ‘science’ (2 moras) vs /ɾikaː/ ‘liquor’ (3 moras)
–  Participants read these words (among others) in a carrier
sentence /soko wa ___ to jomimasu/ at three speaking rates,
each repeated three times.
–  Identical materials were recorded before and after their study in
Japan.
5
Results
Duration of contrasting vowels and consonants
6
Learners’ results
Vowel duration in rika(ː) pair
400
Vowel duration (ms)
350
300
pre fast
pre normal
250
pre slow
200
post fast
150
post normal
post slow
100
50
0
short
•  The Quantity x Time
interaction was
marginally significant
(p = 0.057).
•  A greater durational
difference between
the short and long
vowels at postJapan.
long
Long/short ratios
NJs: 2.22 (Hirata, 2004)
Learners: Pre-Japan 1.81 ! Post-Japan 2.24
! Improvement found for V-pair
7
Learners’ results
Consonant duration in kak(ː)o pair
400
Closure duration (ms)
350
pre fast
pre normal
pre slow
post fast
post normal
post slow
300
250
200
150
100
•  No significant
Quantity x Time
interaction
! No improvement
for C-pair
50
0
single
geminate
Geminate/single ratios
NJs: 3.06-3.18 (Hirata & Whiton, 2005)
Learners: Pre-Japan 1.97 ! Post-Japan 2.11
8
Vowel-to-word ratio &
consonant-to-word ratio
9
Vowel-to-Word (V/W) ratio by NJs
SLOW SPEECH
r i
r i
Short vowel
ka
k a a
Long vowel
Short vowel
FAST SPEECH
Long vowel
10
V/W ratio by NJ (Hirata, 2004)
Vowel to Word ratio
0.8
Long vowels
0.7
0.6
0.5
0.4
0.38
0.3
with 94%
accuracy
0.2
Short vowels
0.1
0
0
200
400
600
800
Word duration (ms)
1000
1200
11
Learners’ results
V/W ratio in rika(ː)
•  Classification accuracy
Pre-Japan: 66%
!
Post-Japan: 85%
•  A significant Quantity x Time interaction, indicating improvement
in V/W ratio at post-Japan
12
Consonant-to-Word (C/W) ratio by NJs
in kak(ː)o pair
(Hirata & Whiton, 2005)
k a ko
ka
k
ko
13
Learners’ results
C/W ratio in kak(ː)o
•  Classification accuracy
Pre-Japan: 73%
!
Post-Japan: 74%
•  No significant Quantity x Time interaction, indicating their
distinction did not change over time.
14
How learners’ production was perceived
by native Japanese (NJ) speakers
15
Method
Listeners
–  13 & 16 monolingual NJ speakers for Accuracy Test and Accent
Rating Test, respectively.
Stimuli
–  All learners’ production were presented to each listener in a
randomized order.
Procedure
–  For Accuracy Test, listeners chose one of six response
alternatives, e.g.,
kako, kaako, kakoo, kaakoo, kakko, kakkoo
–  For Accentedness Test, listeners marked the degree of foreign
accent in each stimulus in a 1-7 scale.
16
Mean production accuracy
Consonant vs. Vowel length pairs
Percent Correct
80
70
60
rika(a)
ka(k)ko
52.4%
50
40
68.9%
42.4%
33.5%
•  A significant Contrast type x
Time interaction: The amount of
improvement was greater for Vpair than C-pair.
V-pair: 16.5%
C-pair: 8.9%
30
Pre
Post
17
Production accuracy by rate
Slow Rate
Percent Correct
90
Normal Rate
Fast Rate
90
90
80
80
70
70
60
60
60
50
50
50
40
40
40
30
30
30
20
20
20
80
rika(a)
70
ka(k)ko
Pre
Post
Pre
Post
Pre
•  A significant Contrast type x Time x Speaking rate interaction
Post
18
Production accuracy by speaker
rika(a)
ka(k)ko
•  A significant Contrast type x Time x Speaker interaction
19
NJ’ accent rating
on learners’ production
Normal Rate
Slow Rate
7
7
7
6
6
5
5
4
4
4
3
3
3
2
2
2
1
1
1
6
Accent Rating
Fast Rate
5
rika(a)
ka(k)ko
Pre
• 
• 
Post
Pre
Post
Pre
7 = Native level; 1 = Not at all native-like
Significant main effects of Time and Rate; No effect of Contrast type
Post
20
Summary
1. Improve or not?
! YES, participants with two years of Japanese language
study in their home country continued to improve without
intensive training.
2. Types of length distinctions?
Consonant length pair
! NO improvement in any acoustic measures
!  A small but significant improvement in perceived accuracy
and accent rating
!  Perceived accuracy improved less for C-pair than V-pair.
! Accent rating improved similarly between C-Pair and V-pair.
Vowel length pair
!  Significant improvement in all acoustic measures, perceived
accuracy, and accent rating.
! Greatest accuracy improvement at the fastest rate.
21
Summary
3. Worth noting…
•  Speaking rate of production was a significant factor.
•  Large room for further improvement on C-pair
(post-test accuracy mean: 42%; range: 3%-85%).
•  Large individual variation.
•  Very small amount of accent improvement.
•  The accuracy difference between C-pair and V-pair does not
appear to affect the overall accent rating.
22
Conclusions
•  Intermediate learners are not yet fixed in their ability to produce
vowel length distinction (Toda, 1997), but about to reach a plateau
for consonant length (Han, 1992).
•  Focused training on consonant length distinction may be necessary
for a notable amount of improvement for intermediate learners.
•  However, the amount and type of improvement depend greatly on
individual learners.
•  Methodological: Native speaker durational measures were useful in
characterizing production learning.
But some discrepancy was observed between durational measures
and NJ auditory judgments for the consonant length pair.
23
Discussion
Further research is necessary...
•  To investigate why learners improved more on vowel length than on
consonant length distinctions.
-  Generalize to other vowels and consonants?
-  Infants’ first language acquisition
-  Sonority of segments
-  Similarities and differences across languages
24
References
Han, M. (1992) The timing control of geminate and single stop consonants in Japanese:
A challenge for non-native speakers. Phonetica 49, 102-127.
Han, M. (1994) Acoustic manifestations of mora timing in Japanese. Journal of the
Acoustical Society of America 96(1), 73-82.
Hirata, Y. (2004) Effects of speaking rate on the vowel length distinction in Japanese.
Journal of Phonetics 32, 565-589.
Hirata, Y. and Whiton, J. (2005) Effects of speaking rate on the single/geminate stop
distinction in Japanese. Journal of the Acoustical Society of America 118(3),
1647-1660.
Masuda, K. and Hayes-Harb, R. (2005) Eigo bogo washa ni yoru sokuon seisei no
sutoratejī [Native English speakers’ strategies of producing moraic obstruents]. In
Minami, M. (ed.), Gengogaku to nihongo kyōiku IV: New directions in applied
linguistics of Japanese, pp. 259–270. Tokyo: Kurosio.
Toda, T. (1997) Strategies for producing mora timing by non-native speakers of
Japanese. In Toda, T. (ed.), Acquisition of Japanese as a Second Language, pp.
157-197. Tokyo: Bonjinsha.
25
Thank you!
yhirata@colgate.edu
Acknowledgments
Grants: Colgate’s Research Council
Students: Ian McNally, Connor Forbes
26