Acquisition of L2 Japanese Geminates: Training with Waveform Displays

The value of waveform displays as visual feedback was explored in a training study involving perception and production of L2 Japanese by beginning-level L1 English learners. A pretest-posttest design compared auditory-visual (AV) and auditory-only (A-only) Web-based training. Stimuli were singleton and geminate /t,k,s/ followed by /a,u/ in two conditions (isolated words, carrier sentences). Fillers with long vowels were included. Participants completed a forced-choice identification task involving minimal triplets: singletons, geminates, long vowels (e.g., sasu, sassu, saasu). Results revealed a) significant improvement in geminate identification following training, especially for AV; b) significant effect of geminate (lowest scores for /s/); c) no significant effect of condition; and d) no significant improvement for the control group. Most errors were misperceptions of geminates as long vowels. Test of generalization revealed 5% decline in accuracy for AV and 14% for A-only. Geminate production improved significantly (especially for AV) based on rater judgments; improvement was greatest for /k/ and smallest for /s/. Most production errors involved substitution of a singleton for a geminate. Post-study interviews produced positive comments on Web-based training. Waveforms increased awareness of durational differences. Results support the effectiveness of auditory-visual input in L2 perception training with transfer to novel stimuli and improved production.