ATR Human Information Processing Research Labs
Scientific Inquiry: Improving English Listening and Pronunciation
Skills
Reiko Akahane-Yamada
To better understand how humans process spoken language, our research group is
studying how language learners develop listening and pronunciation skills in a
foreign language. For the past several years, we have been conducting extensive
research on native Japanese speakers' acquisition of English sounds, using state-of-the-art
technology. Our research findings not only advance our understanding of human
speech processing, but they also help develop better technology for human-computer
interaction and effective instructional tools for second-language education.
1. Introduction
There must be some way to resolve the complex that Japanese have with regard to
the English language. At our laboratory, we are studying the ways language learners
develop spe4ech perception and production abilities. Through a series of experiments,
we have demonstrated that even adults can learn to properly perceive and produce
speech segments in a second language. We found that this learning process, which
was once thought to be unattainable for adults, can be facilitated through the
use of state-of-the-art speech technologies, and through a consideration of human
information processing mechanisms. Many important factors related to this learning
process, such as the link between perception and production and the effects of
age, have also been clarified. We believe that research in these areas will be
helpful in resolving the "English complex of Japanese", and over the past two
years, we have published three practical books under the title "Scientific Inquiry:
Improving English Listening and Pronunciation Skills" (see figure
1).1,2,3
This paper will describe the research related to the learning methods taken up
in these three books.
2. Learning Listening Skills
We conducted perception training experiments, with adult Japanese speakers (university
students) as subjects, focusing on the identification ability of American English
(AE)/r/ and /l/, which are particularly difficult for Japanese speakers. In this
training, pairs of English words contrasting the /r/ and /l/ in various positions
within words, such as "red" - "led", "poor" - "pool", etc., were used as training
materials. We presented these training speech stimuli over headphones, had the
subjects respond with one of the items in the pairs, and gave feedback as to the
correct answers. We also used utterances produced by a number of talkers as the
training stimuli, rather than using utterances from only one speaker. Subjects'
identification ability of /r/ and /l/ improved significantly from the pre-training
test to the post-training test, and the training effects generalized to untrained
words and untrained talkers. We also found that the effects of the tr5aining were
maintained even six months after the training was completed. The results obtained
have overturned the common belief that "new sounds can only be learned during
childhood".
Furthermore, despite the fact that we conducted only perception training, production
ability (the level to which native speakers can correctly distinguish the speech
segments uttered) by the subjects also improved from pre-test to post-test, suggesting
that there is an obvious link between perception and production during the course
of second language acquisition.
3. Learning Pronunciation Skills
We also conducted production training experiments to determine whether computer-based
production training would be effective or not. Because it is difficult to monitor
the movement of the tongue and other organs related to articulation of speech
sounds, there has been a problem with methods of feeding back information to learners
on whether their pronunciation was adequate or not. We devised several methods
to resolve this problem. In the first stage of the learning process, we instructed
the position and shape of the tongue for producing /r/ and /l/ sounds by using
3-D computer animation to create a talking head with a semitransparent image of
the cheek. In this way, we displayed to the learners, in visible form, the movements
of the tongue and other organs related to articulation (Figure
2). By doing so, learners became able to pronounce /r/ or /l/ alone (that
is, a prolonged /r/ or /l/ sound) in a surprisingly short period of time.
In the next stage, we conducted production training of English word pairs contrasting
in /r/ and /l/, by giving feedback to learners on whether or not they were able
to pronounce the words adequately. We designed two types of feedback. As our second
method, we trained subjects by providing feedback using a spectrographic display.
Furthermore, as our third method, we used evaluation scores generated by a speech
recognition system so that the learners could clearly understand the quality of
their pronunciation in quantitative terms.
Before and after the above production training, a pre-test and post-test were
conducted in which learners' utterances were recorded. These utterances were later
evaluation by native AE speakers. It was found that the production ability of
/r/ and /l/ sounds improved dramatically from pre-test post-test.
We also measured listening abilities before and after production training, and
found that the production training improved perception ability in addition to
production ability.
4. Learning and Age
All of the above experiments were conducted using university students as subjects.
What would be the effects of this training, however, in the case of younger or
older learners? To answer this question, we are currently exposing subjects in
a wide range of age groups － from elementary school students to middle-aged and
elderly persons － to exactly the same learning experiences as the above university
students. The experiments are currently in progress, but at the present time,
it has become clear that high school students have obtained an effect from training
that is equivalent to that of the university students. It has also become clear
that there is a significant effect from the training even in the case of learners
in their 50s and 60s. It appears that in the context of language learning, the
excuse "I'n too old" will soon become unacceptable.
5. Moving toward Distance Learning
In April 1998, we began an "open experiment concerning listening and speaking
of English sounds using the Internet" (URL: http://bluebacks.hip.atr.co.jp/en/index.html)
with two main goals: to investigate the possibilities of distance learning, and
to gather data from a wide variety of individuals. At the present stage, we are
still operating the system experimentally with Japanese persons as targets, but
in the past 18 months, we have received over 3,000 items of thought-provoking
experimental data. This demonstrates that the use of a network linking people
in remote locations is effective as one form of language learning.
6. Relationship with the Native Language
We are often asked, "Why can't Japanese speak English well?" There are many different
factors involved, including education methods. If we were to reply from the standpoint
of speech science, however, we could say that the difference in linguistic structure
between English and Japanese, our native language, is a major factor. In fact,
we have found that Korean university students and Japanese university students
use entirely different ways of distinguishing between /r/ and /l/. It is possible
that we may find a more fundamental method of facilitating second language learning
by investigating the interference effects of the native language.
7. Phoneme Learning and English Learning
We are also asked at times, "How does a study of phonemes (vowels and consonants)
affect English communication abilities?" It goes without saying that an ability
to distinguish phonemes is not enough to enable a person to speak English.
Nevertheless, consideration of human speech processing in general suggests to
us the importance of a study of phonemes. The speech signals that enter the human
ear are encoded, through an early stage of processing, as a series of phonemes,
then as words and sentences, and later meaning is understood using background
knowledge. English learning in the past has for the most part focused on the latter
stage of this information processing; we could say that the learning of early
stages has been neglected. But if learners are not able to properly encode phonemes,
the initial process, there is a large possibility that an excessive burden is
placed on the latter process. We have been able to obtain consistent results by
focusing on the study of phonemes. We believe that in the future, it will be possible
to dramatically increase English communication abilities by systematically linking
the study of phonemes with the latter stages of information processing.
8. Conclusion
Our method of improving English listening and pronunciation skills is characterized
by four elements:
(1) the method is based on knowledge about mechanisms underlying human information
processing,
(2) the effects of training using this method have been proven through learning
experiments,
(3) the method has effectively adopted advanced technologies such as speech recognition
and multimedia, and
(4) it is highly efficient in its use of the information infrastructure.
The learning methods presented in the books mentioned in the introduction have
not attained an increase in overall communication abilities in English, but nevertheless
they have received excellent evaluation from the general public. After repeated
printings of the books, the total number of publications for the three books has
exceeded 80,000 copies. We would go so far as to say that this result might be
an expression of a new place for "information" in the context of foreign language
learning. As we approach the beginning of the 21st century, we are confident that
the results of our research will lead to increased use of information technologies
in language education, and to effective distance education as well.
Reference
