VIP Call Girl Cuttack Aashi 8250192130 Independent Escort Service Cuttack
ship
1. Izabelle Grenon — The University of Tokyo, Japan (grenon@boz.c.u-tokyo.ac.jp)
Chris Sheppard — Waseda University, Japan (chris@waseda.jp)
John Archibald — The University of Victoria, Canada (johnarch@uvic.ca)
At the 18th Annual Conference of the Japan Second Language Association
Gakushuin University, Tokyo, Japan
June 16, 2018
2. 1. The target contrast: /i/-/I/ as in ‘sheep’ and ‘ship’
2. The identification task
3. Limitation of the identification task
4. The discrimination task
1. Method
2. Results
5. General discussion and conclusion
3. 1. The target contrast: /i/-/I/ as in ‘sheep’ and ‘ship’
2. The identification task
3. Limitation of the identification task
4. The discrimination task
1. Method
2. Results
5. General discussion and conclusion
7. 1. The target contrast: /i/-/I/ as in ‘sheep’ and ‘ship’
2. The identification task
3. Limitation of the identification task
4. The discrimination task
1. Method
2. Results
5. General discussion and conclusion
8. Key features:
Multiple talkers
Multiple contexts (i.e., the target vowels are surrounded by
different consonants)
Identification task
15. Grenon, Kubota, Sheppard
(in preparation)
All 23
Japanese
used vowel
duration
before
training
11 of them (46%) stopped
using vowel duration
after training
17. Identification training can change speech perception from focus on
duration to focus on spectral (formant) information.
A change in cue-weighting (focus on formants instead of vowel
duration) leads to BETTER improvement and better generalization
to new tokens and talkers.
18. 1. The target contrast: /i/-/I/ as in ‘sheep’ and ‘ship’
2. The identification task
3. Limitation of the identification task
4. The discrimination task
1. Method
2. Results
5. General discussion and conclusion
19. Learners must be literate in the L2
Young Japanese
children learning
English
Beginner Japanese
learners of Russian
20. Using pictures instead of letters
(require knowing the meaning of the words)
Visual cue association tasks (Grenon et al. in progress…)
Discrimination task (AX task)
(other possibilities are AXB task and ABX task)
21. Using pictures instead of letters
(require knowing the meaning of the words)
Visual cue association tasks (Grenon et al. in progress…)
Discrimination task (AX task)
(other possibilities are AXB task and ABX task)
22. 1. The target contrast: /i/-/I/ as in ‘sheep’ and ‘ship’
2. The identification task
3. Limitation of the identification task
4. The discrimination task
1. Method
2. Results
5. General discussion and conclusion
23. 20 native Japanese speakers (recruited University of Tokyo)
Aged between 18 and 27 (Mean: 20)
Never stayed in English-speaking country more than 8 weeks (M: 1.7week)
(1 extra participant rejected)
40 North American English speakers (recruited University of Victoria)
Aged between 17 and 28 (Mean: 21)
(14 extra participants rejected)
25. English speakers
Pre-test only
Japanese speakers
Pre-test (day 1)
Training 30 min. (day 2)
Training 30 min. (day 3)
Training rose-roads (d4)
Training rose-roads (d5)
Post-test (day 6)
26. English speakers
Pre-test only
Japanese speakers
Pre-test (day 1)
Training rose-roads (d2)
Training rose-roads (d3)
Training 30 min. (day 4)
Training 30 min. (day 5)
Post-test (day 6)
27. Pre- and post-test
Identification task
All 28 stimuli
Presented randomly
Repeated 4 times
(discarded first round as
practice)
Training
Discrimination task
31. Pre- and post-test
Identification task
All 28 stimuli
Presented randomly
Repeated 4 times
(discarded first round as
practice)
Training
Discrimination task
Subset of 16 stimuli
34. Pre- and post-test
Identification task
All 28 stimuli
Presented randomly
Repeated 4 times
(discarded first round as
practice)
Training
Discrimination task
Subset of 16 stimuli
16 ‘same’ pairs
16 ‘different’ pairs
reverse order = 64 pairs
Presented 4 times (512 words)
About 30 min. per training
session
35. Discrimination training improved significantly from first
to second training session :
First session (M: 88.3%, St.dev. 10.6)
Second session (M: 93.58%, St. dev. 8.10)
38. Repeated Measures ANOVA
Time (Pre-Test, Post-Test) and Duration (90ms, 120ms, 150ms, 180ms)
The data was not spherical
Duration
W = 0.24, p = .003
Time X Duration
W = 0.19, p < .001
Greenhouse-Geisser correction was used
Time
F (1, 14) = 0.216, p = 0.65, ηp
2 = 0.003
Duration
F (3, 42) = 47.6, GGe = 0.53, p < .001, ηp
2 = 0.57
Time X Duration
F (3, 42) = 22.9, GGe = 0.55, p < .001, ηp
2 = 0.28
39. Mixed Effects ANOVA
Group (English Speakers, Post-Test) and Duration (90ms, 120ms, 150ms, 180ms)
The data was not spherical
Duration
W = 0.34, p < .001
Group X Duration
W = 0.34, p < .001
Greenhouse-Geisser correction used
Group
F (1, 53) = 26.0, p < .001, ηp
2 = 0.13
Duration
F (3, 159) = 22.8., GGe = 0.57, p < .001, ηp
2 = 0.23
Group X Duration
F (3, 159) = 4.10, GGe = 0.57, p = 0.024., ηp
2 = 0.05
43. Repeated Measures ANOVA
Time (Pre-Test, Post-Test) and Formant (Formant 1 – Formant 7)
The data was not spherical
Formant
W = 0.003, p < .001
Time X Formant
W = 0.07, p = .049
Greenhouse-Geisser correction used
Time
F (1, 14) = 0.22, p = 0.65, ηp
2 = 0.003
Formant
F (6, 84) = 107, GGe = 0.30, p < .001, ηp
2 = 0.77
Time X Formant
F (6, 84) = 12.0, GGe = 0.51, p < .001, ηp
2 = 0.19
44. Mixed Effects ANOVA
Group (English Speakers, Post-Test) and Formant (Formant 1 – Formant 7)
The data was not spherical
Formant
W = 0.03, p < .001
Group X Formant
W = 0.03, p < .001
Greenhouse-Geisser correction used
Group
F (1, 52) = 26.7, p < .001, ηp
2 = 0.08
Formant
F (6, 312) = 604, GGe = 0.57, p < .001, ηp
2 = 0.91
Group X Formant
F (6, 312) = 14.3, GGe = 0.57, p < .001., ηp
2 = 0.19
47. Native Japanese
speakers’ (N=15) use
of formants became
more categorical after
training
Ship- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - Sheep
48. Japanese speakers could reduce their reliance on vowel
duration and focus more on spectral information to
perceive the vowels as in ’ship’ and ‘sheep’ after only 1h of
discrimination training.
However, 5/20 (25% of them) could not form the proper
grapheme-phoneme association, as assessed with the post-
test (using an identification task).
49. 1. The target contrast: /i/-/I/ as in ‘sheep’ and ‘ship’
2. The identification task
3. Limitation of the identification task
4. The discrimination task
1. Method
2. Results
5. General discussion and conclusion
50. Discrimination training seems to help with differentiating
sounds;
However, doesn’t necessarily help with creating the proper
grapheme-phoneme correspondence;
Maybe best used before or along with proper instructions
about the sound-letter association?
Maybe can be used before identification training?
51. Shinohara & Iverson (2018)
Looked at effect of high variability phonetic training using
identification and discrimination tasks
53. Shinohara &
Iverson 2018
But discrimination training
results in some
improvement if before
identification training
54. Shinohara &
Iverson 2018
Identification training
provided superior results to
discrimination training
However, the training
methods used different sets
of stimuli…
55. Comparing Identification training with Discrimination
training using the same set of 512 words per training
session (i.e., same 16 tokens presented repetitively).
Evaluating the effect of discrimination and identification
training with the ‘rose-roads’ contrast (which is more
difficult than the ’ship-sheep’ contrast, with discrimination
training scores around chance level, i.e., 50%).
56. What about if used with true adult beginners (like
Japanese adults learning Russian)?
What if used with children younger than 12 years old?
Would the results be different (more native-like)?
Possible advantage if used at an early stage: may avoid
fossilization of acoustic forms that are different from the
target forms (e.g., producing the word ’right’ with an ‘l’
even after being able to distinguish ‘r’ from ‘l’).
57. Results in perception impacts production when using
HVPT with identification task. (e.g., Bradlow, Akahane-
Yamada, Pisoni & Tohkura 1999).
• What about effect of discrimination task on production?
58. A discrimination task could potentially be used for
training with novel sound contrasts in an L2.
However, this kind of training does not guarantee the
proper grapheme-phoneme association.
Hence, this training program may be best used before or in
conjunction with instructions about the association
between sound and letter(s) (e.g., with true beginners).