Discrimination training for learning sound contrasts

Izabelle Grenon, Chris Sheppard, John Archibald


Recent studies have suggested that for training on second language (L2) vowels, the use of an identification task results in greater improvement, and generalization to new words than does the use of a discrimination task. The current study investigated why this may be the case. Twenty native Japanese speakers received two thirty-minute sessions of discrimination training with the English high front vowels, which were modified to vary along two dimensions to go from ‘ship’ to ‘sheep’: Vowel quality (i.e., spectral cues), and vowel duration (i.e., temporal cues). Using a cue-weighting task, we found that while the L2 learners significantly improved their use of vowel quality to classify the English vowel contrast after training, some of them (25%) were unable to associate the vowels with their proper orthographic representations. We conclude that unlike what was suggested by previous studies, the discrimination task appears successful in helping L2 learners create new vowel categories along the spectral dimension. However, since the discrimination task does not provide information about phoneme-grapheme associations, additional instructions about how each vowel is represented by the orthography should be provided to the learners at some point.


 DOI: 10.21437/ISAPh.2018-9

Cite as: Grenon, I., Sheppard, C., Archibald, J. (2018) Discrimination training for learning sound contrasts. Proc. ISAPh 2018 International Symposium on Applied Phonetics, 51-56, DOI: 10.21437/ISAPh.2018-9.


@inproceedings{Grenon2018,
  author={Izabelle Grenon and Chris Sheppard and John Archibald},
  title={{Discrimination training for learning sound contrasts}},
  year=2018,
  booktitle={Proc. ISAPh 2018 International Symposium on Applied Phonetics},
  pages={51--56},
  doi={10.21437/ISAPh.2018-9},
  url={http://dx.doi.org/10.21437/ISAPh.2018-9}
}