June 16 (Thu) W6B201

Guest Lecture 1

Learning to perceive Japanese length contrasts: What the training research informs us

Professor Yukari Hirata, Colgate University, USA


This talk will give an overview of the experimental research conducted in the past few years that trained learners of Japanese to auditorily perceive Japanese length contrasts. Research has highlighted several crucial factors for effective perceptual learning, such as usefulness of variability in presentation contexts (i.e., isolated words versus sentences), variability in speaking rate, and simultaneous presentations of auditory (speech) and visual (lips) information. These findings have not only helped us understand the nature of human perception and learning, which is a theoretical pursuit in and of itself, but also provided useful ideas for practical applications to teaching. I will also discuss factors that need to be considered from language teachers’ points of view in order for theoretical findings to be applied to useful pedagogical tools. This discussion is intended to move toward filling the gap that still exists between speech science research and practice of teaching.


Presentation 1

A comparative study on perception of foreign-accented Japanese by L1 and L2 Japanese listeners, and the implications for Japanese language education: Focusing on pitch and timing errors

Dr. Shunichi Ishihara, Australian National University


This study concerns good pronunciation from the L1 and the L2 Japanese listeners’ points of view.  More precisely, this study investigates 1) whether L2 Japanese listeners from different L1 backgrounds perceive foreign‑accented Japanese differently; 2) how the perceptions of L2 Japanese listeners changes as a function of their proficiency; and 3) how different the perceptions of L2 Japanese listeners are from that of L1 Japanese listeners. In this study, we focus on the prosodic variations of timing and pitch‑accent in Japanese.  Eight groups of informants participated in a series of perception experiments where they were asked to assess the stimuli which were extracted from the recordings of L2 Japanese learners.  Six groups among these eight groups are L2 Japanese learners having three different L1 backgrounds (English, Chinese or Indonesian) at two levels of proficiency (beginner/intermediate or advanced), and two groups are native Japanese speakers with and without formal teaching experience of L2 Japanese.  We will report that the learners’ assessment of good pronunciation is not straightforward, being different from the logical expectation that learners will behave more like native speakers as their L2 Japanese proficiency develops.  We will also discuss possible implications for L2 Japanese education. 


Presentation 2

The production of Japanese vowel and moraic nasal sounds: Comparison of native and non-native speakers

Dr. Kimiko Tsukada, Macquarie University


This study compared durational characteristics of native (NJ) and non-native Japanese (NNJ) pronunciation which could form a basis for future resources and directions in the Japanese language education. Eight NNJ speakers participated in 4-week pronunciation training and recordings were made twice, once each before and after the training. The speakers read the words minimally contrasting in vowel length embedded in a short carrier sentence. Overall, the NNJ pronunciation did not change measurably before and after the training. At both recording times, NNJ speakers hardly differed from NJ speakers in the mean vowel duration, but differed in the duration of the moraic nasal sound. NNJ speakers produced a longer nasal than did NJ speakers in both 2- and 3-mora words and they were much more variable in their duration, particularly, for the nasal in 3-mora words. Given this native vs. non-native difference in the duration of the moraic nasal, it is expected that the contrast between 2- and 3-mora word duration is more distinct in the NJ than in NNJ speech. This, in turn, may affect the intelligibility of contrastive words even though native and non-native speakers may not differ in their vowel duration.


Guest Lecture 2

Native and non-native perceptual cues of length contrasts in Japanese

Dr. Hiroaki Kato, National Institute of Information and Communications Technology (NICT)/Advanced Telecommunications Research Institute International (ATR), Japan


The temporal resolution of human hearing in general is surprisingly high. Nevertheless, length contrasts in Japanese are said to be difficult to perceive and produce for people whose native language does not use duration alone to signal lexical contrasts. This must be due to native speakers commonly using types of internal measures that non-native speakers do not have. In this talk, I will introduce some of our findings on the perception of temporal aspects of speech obtained over the past decade and attempt to reveal the nature of the native speakers’ discriminative measures. Finally, I will show an empirical study using computer-based perception training and discuss whether such intensive training can enable non-native learners to acquire at least some of the native-type perceptual measures.


Presentation 3

How good is good enough? Towards the measurement of intelligible speech

Dr. Chiharu Tsurutani, Griffith University


In this presentation, I would like to introduce the method to create stimuli (using speech synthesis - STRAIGHT) for the experiment which aims to identify the most crucial prosodic factor for achieving intelligible pronunciation. First, prosodic errors by English speaking learners of Japanese will be illustrated in the scope of contributing to the advancement of computer-based pronunciation assessment programs. Then, we will discuss how to determine the acceptable level of intelligibility in spoken language communication for second language (L2) learners. L2 speakers spend significant time and effort before reaching the intelligible stage without knowing the cause of their pronunciation difficulty. The development of pedagogic research infrastructure for pronunciation enhancement will help L2 learners attain their goal efficiently. 


MQ Lecture

Learning vowel length contrasts: Perspectives from L1 acquisition

Professor Katherine Demuth, Macquarie University


Vowel length can vary as a function of both intrinsic/phonemic contrasts, and extrinsic contrasts that are a function of context (voicing of the following coda, phrase-final lengthening, etc). Of the few studies that have examined the acquisition of phonemically long vowels, most report that these are acquired later than simple vowel quality contrasts. Interestingly, however, the few studies that have examined the acquisition of extrinsic vowel length contrasts find that these are acquired quite early. The implications of these studies for understanding processes of L2 learning are discussed.


June 17 (Fri) W6B225


Demonstration 1: Perception training program

Dr. Hiroaki Kato, NICT/ATR, Japan


In this demonstration, I will present the minimum set of our perception-training program. This version is sufficient to let you experience yourself every kind of phonemic length contrast in Japanese as well as understand the various promising aspects of the program. Just as with the full-length training, the demonstration uses a length-based minimal-pair identification paradigm with feedback. Four different types of length contrasts are employed, i.e., long vs. short vowel, geminate vs. singleton consonant, with vs. without a syllabic nasal, and long vs. short i-like sound (e.g., kiyaku vs. kyaku). The demonstration comprises three sessions that differ in the degree of presentation variability. The first session contains four minimal word pairs corresponding to the four different contrast types and a minimal triplet spoken as isolated words, as well as another set of four pairs and a triplet embedded in a carrier sentence. A male and a female professionally trained speaker produced them at a normal speaking rate. The number of speakers is doubled in the second session and then tripled in the third session, which in addition incorporates higher variability in the speaking rate. In addition to presentation variability, I would like to use the demonstration to show what else we consider important for a training program. Specifically, this involves excluding to the extent possible factors that do not appear in people-to-people communication; learners are exposed to nothing but naturally produced speech samples and are asked to do nothing but identify a word.

Demonstration 2: Mora Program

Professor Yukari Hirata, Colgate University, USA


The Mora Program I developed in Hirata (2004), called Good-Bye Syllable!, aims at helping beginning learners of Japanese to grasp the notion of moras and learn to perceive them correctly in Japanese words. In perceiving moras, which are distinct from syllables, learners need to pay attention to length of vowels and consonants because a long one gives an additional count to the number of syllables they hear, e.g., the words, i ‘stomach’ (a short vowel; one mora) and ii ‘good’ (a long vowel; two moras). In this training program, 600 words of one to six moras are presented, and learners are asked to choose the correct number of moras in each word they hear. Immediate feedback is given for every response. At the end of each session, learners are given a summary of their performance. A set of ten training sessions focuses on hearing isolated words, and another set of ten sessions focuses on hearing these words in sentences. The Mora Program comes with a pre-test and a post-test that measure their initial performance and their performance after finishing the training sessions. In this demonstration I will show how the program works, discuss how it can be integrated into a Japanese language curriculum, and suggest further research possibilities.


Demonstration 3: Pronunciation Check

Dr. Chiharu Tsurutani, Griffith University

Dr. Shunichi Ishihara, Australian National University


In this demonstration we will introduce the functions of the computer program “Pronunciation check” which was developed as a self assessment program for Japanese pronunciation by English-speaking learners. The program was developed using a language model built with input from a Japanese language teacher in collaboration with speech engineers. This collaboration enhanced the program’s capacity for accurate assessment and provides practical support to users by 1) linking evaluation with feedback, and 2) an editorial function to work on error patterns. The program drew positive responses from participants in a trial run and through experimental usage in university courses teaching Japanese language.





Contact ASSTA: Either email The ASSTA Secretary, or

G.P.O. Box 143, Canberra City, ACT, 2601.

Copyright © ASSTA