Acquisition of L2 Japanese geminates: Training with waveform displays

T he value of waveform display s 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 (Aonly) Web-base d training.Stimuli were sin gleton and geminate /t,k,s/ followe d by /a, u/ in two conditions (isolated words, carrier sentences).Fillers with long vowels were include d.Participants completed a forced-choice identification task involving minimal triplets: singletons, geminates, long vowe ls (e.g., sa su, 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.T est of generalization revealed 5% decline in accuracy for AV an d 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. INTRO DUCTIO NT he duration of vowels and consonants is a contrastive feature in Japanese unlike English.In Japanese, singleton consonants contrast with their longer geminate counterparts, which are a special type of mora.Accurate perception and production of morae are important for comprehension; however, they can be problematic elements for second-language (L2) learners.T raining studies to date have not addressed this particular issue.Briefly, a mora is a unit of timing, which plays a role in the temporal organization of speech in production (e.g., Kubozono, 1999).Neighboring moraic units tend to show equal duration (Port, Dalby, & O'Dell, 1987).As a unit of rhythm, the mora also plays a role in the segmentation of speech sounds for lexical recognition (e.g., Cutler & Otake, 2002).Special morae, such as geminates, do not constitute syllables by themselves as shown in the following examples, where the moraic units are separated by a dot and syllables by a hyphen: a) moraic obstruent: /ki.t.te/, /kit-te/ " postage stamp," and b) the second half of a long vowel: /ki.i.te/, /kii-te / " listening." 2 An important acoustic cue for geminate obstruent perception by native speakers of Japanese is stop closure duration (e.g., for /p/, /t/, /k/) or period of frication (e.g., for /s/) (e.g., Fukui, 1978).However, perception is influenced by other factors such as pitch accent, which is also contrastive in Japanese (Ofuka, 2003).For example, kata produced with a H(igh)-L(ow) pattern means " shoulder," while katta (HLL) with a geminate stop means "won."In contrast, kata with a LH pattern means "form" and katta (LHH) means " bought."Using synthesized stimuli to alter the closure duration of /t/ along a kata -katta continuum for both pitch accent patterns, Ofuka found native speakers required a lon ger closure duration to perceive a geminate in the LHH pattern in contrast to the HLL.In addition, Hirata (1990) found that for stimuli embedded in carrier sentences, a singleton was perceived as a geminate if postconsonantal elements were spoken quickly, and geminates were perceived as sin gletons if these elements were spoken