Discussion

1
The 157th Meeting of Acoustical Society of
America in Portland, Oregon, May 21, 2009

4pSW35. Confusion Direction Differences in Second
Language Production and Perception Yen-Chen Hao
Kenneth de Jong (yehao_at_indiana.edu
kdejong_at_indiana.edu) Department of Linguistics,
Indiana University

• Main Points
• Second language learners may display opposite
• confusion directions in perception and
  production.
• L2 learners may show different error patterns in
  
• different prosodic locations due to the
• allophonic variation in the target language.
• Old category bias is only observed in
• production-related tasks.
• Category bias in perception and production
• Frequency biasing
• In perceptual identification by native speakers,
  
• response bias typically correlated positively
  with
• segment frequency (Silbert de Jong, 2007).
• Second language learners should be biased
• towards L2 old categories since they occur
• more frequently (both in L1 and L2) than the
• new categories (only in L2).
• Novel Category bias

• Discussion
• L1-L2 category mismatch
• Korean EFL learners have more problems
• with the allophonic variation in post-stress
• position, which mismatches the L1 categories.
• Post-stress position induces larger variation
• in bias than pre-stress position.
• Discrepancy in perception and production
• Korean learners displayed different
• directionality of post-stress errors in
• perception and production
• In perception-related tasks (Identification and
• Mimicry), learners made more voiceless ?
• voiced errors.
• In the production task (Reading), learners
• made more voiced ? voiceless errors.
• Identification Reading

Methods Participants -20 Korean learners of
English -Undergraduates (22-28 years 15 F,
5 M) in Gyeonggi Korea (near Seoul)
-Traveled 3 months or less in English-
speaking areas Stimuli -10 English anterior
obstruents Similar to L1 (Old) /p, b, t, d,
s/ Not similar to L1 (New) /f, v, ?, ð, z/
-Two contexts /aCá/ /áCa/ -Produced in
isolation by 4 Midwestern American speakers
(2 male, 2 female)

• Results
• Bias in Identification
• Large variation in bias parameter in the post-
• stress context.
• Post-stress position exhibits more bias
• towards the voiced obstruents.
• Bias in Mimicry

/áCa/
/aCá/
Orthographic /p/ prompt
Evaluation /p/

• Tasks
• Identification (Id)
• Participants identified the obstruents from a
  set of 15 response options (plus Other ____).
• 2. Mimicry (Mim)
• Participants listened to the stimuli and
  repeated after each one.
• 3. Reading (Rd)
• Participants read same list of forms as American
  speakers who produced stimuli.
• Evaluation
• 10 English native speakers judged each of the
  productions from Tasks 2 and 3 using similar
  response options as in Task 1.
• Analysis
• Similarity Choice Model Bias parameters (Luce,
  1963), estimates of an overall tendency for a
  particular category to be used, were calculated
  for each segment

English talker /áCa/
English judge /áCa/
/áCa/
Production p
Production b/p
Korean listener /aCa/
Korean talker /aCa/
Allophonic variation in English and Korean
(hypothetical)
English Pre-stress /aCá/
Identification /b/ response
Orthographic /b/ prompt
/aCá/
- Voicing
Post-stress /áCa/
/áCa/
References De Jong, K., Hao, Y-C., Park, H.,
Silbert, N.H. (2008). The English effect in
acquisition of English as a second language
Novel-category biases in second language
perception and production. JASA, 123 3332,
2pSCc53. Luce, R. D. (1963). Detection and
recognition. In R. D. Luce, R. R. Bush, E.
Galanter (Eds.), Handbook of Mathematical
Psychology (pp. 103-189). New York John Wiley
Sons. Silbert, N.H., and K. J. de Jong (2007).
Response Bias, Type and Token Frequency, and
Prosodic Context in Segment Identification. JASA,
1223019, 3aSCa15.
Korean Intervocalic /aCa/
- Voicing
Acknowledgments We thank Mi-Hee Cho for her help
in collecting the consonant data, and Noah
Silbert, and Kyoko Nagao for their work on the
design, analysis, and processing of the data.
Work supported by the NSF (grant BCS-04406540).
/aCá/
- Voicing