Computer Assisted Language Learning 0958–8221/99/1205–0427$15.

00
1999, Vol. 12, No. 5, pp. 427–440 © Swets & Zeitlinger

Computer-Aided Pronunciation Pedagogy: Promise,

Limitations, Directions*

Martha C. Pennington
The Spires Research Centre, University of Luton

ABSTRACT

An overview is presented of the promise and limitations of working on computer to improve

pronunciation in a second language. It is maintained that the considerable promise of the com-
puter as an instructional tool for developing language learners’ pronunciation has yet to be
realized in practice, primarily because of lack of attention to pedagogical design rather than
because of inherent limitations of the technology. On the basis of this overview, suggestions are
made in the way of ten design principles.

1. THE NEED FOR CAP

The sound pattern, or phonology, of a language or language variety (dialect)

‘is its surface form in spoken mode’ (Pennington, 1999, p.13). Phonological
competence is therefore crucial for both production and reception (decoding
and comprehension) of spoken language in every aspect—lexis, syntax,
semantics and pragmatics. In addition to being the medium through which oral
language is produced by speakers and received by listeners, phonology is ‘a
main medium for presentation of self’ to others (Pennington, 1999, p.13),
functioning as an indicator of such characteristics as gender, age, national or
geographical origin, socio-cultural grouping(s) and other acquired or desired
affiliations. A speaker’s pronunciation, both (i) the articulation of individual
sounds (phonemes) or sound segments and (ii) prosody, including tone and
intonation (patterns of pitch on words or longer utterances), stress (force of
articulation) and rhythm (timing), is thus:

*This paper is revised from a presentation given under the title “Pronunciation work on com-
puter: Promise, limitations, and directions” in the TCIS Colloquium on the Uses and Limitations
of Pronunciation Technology at the 32nd Annual TESOL Convention, Seattle, Washington, 21
March 1998.
Correspondence: Martha C. Pennington, The Spires Research Centre, University of Luton, 2
Adelaide St., Luton LU1 5DU, England. Tel: 01582 743790. Fax: 01582 743701. E-mail:
[email protected].

Manuscript submitted: May, 1999

Accepted for publication: July, 1999
428 M.C. PENNINGTON

a primary indicator of a speaker’s national or cultural group as well

as any other groups to which the speaker belongs, wishes to belong,
or wishes to emulate. Those who can adapt their pronunciation to
different groups and situations therefore have an advantage over
those whose pronunciation is limited to one accent or style of
speaking. (Pennington, 1999, p.13)
During childhood, the sound pattern of one’s native language (and any other
languages learned in childhood) will become ingrained in both a physiologi-
cal (‘hard-wired’) sense and a psychological and socio-cultural (‘soft-wired’)
sense (Pennington, 1999, forthcoming b). After the ‘critical period’ for child-
hood development—comprising phonological as well as general motor,
psychological, social and cultural development—has been passed, it becomes
difficult to alter a person’s pronunciation patterns and associated behaviours
in any substantial way. Adolescent and adult language learners generally reach
a point of ‘fossilization’ or ‘diminishing returns’ at a relatively early (inter-
mediate) stage of learning a new language or variety. From this point onwards,
most adult learners will hardly be able to improve their productive and recep-
tive competence of a new sound system without explicit instruction
(Pennington, 1998).
Computer-aided pronunciation (CAP) offers a medium for increasing users’
access to their own and others’ pronunciation performance and underlying
phonological systems, for focusing their attention on phonology, and for
acquiring new pronunciation patterns. In so doing, it offers considerable
promise for language pedagogy, as a medium for improving adolescent and
adult language learners’ productive and receptive competence in pronuncia-
tion of a target language or variety (dialect).

2. TECHNOLOGY FOR CAP

Systems for performing and displaying an acoustic analysis of speech in both

segmental and prosodic aspects, using input from a microphone or recorded
speech sample, are available commercially and may be used for improving
pronunciation in a second language (L2) or variety. The majority of CAP sys-
tems run on a standard PC, though some require special hardware. Other sorts
of computer capabilities such as graphics, animation, pictures, video and audio
may be used as well to develop CAP pedagogy.
 COMPUTER-AIDED PRONUNCIATION PEDAGOGY 429

Straightforward accounts of CAP technology in relation to second language

pedagogy, including sample screen displays, can be found in Anderson-Hsieh
(1992), Chun (1989), Molholt (1988), Pennington (1989a), and Pennington
and Esling (1996). Anderson-Hsieh (1998) offers an up-to-date breakdown of
the capabilities and costs of a wide range of systems, and evaluates these in
terms of their practicality and potential for pronunciation instruction. More
detailed, technically oriented reviews of available systems are provided by
Read et al. (1990, 1992).

3. PROMISE OF CAP

CAP, in conjunction with the general capabilities of computer-aided instruc-

tion (CAI), has a range of advantages that give it special promise for language
instruction (Anderson-Hsieh, 1992; Chun, 1989; Pennington, 1989a, 1996a;
Pennington & Esling, 1996). It is, first of all, quick, performing an analysis
and giving feedback to the user far faster than a human being can. The analy-
sis of a user’s speech is also infinitely repeatable, precise and reliable in the
sense of being the same every time. In all these senses, CAP is superior to the
human pronunciation coach or phonetician. CAP, which does not suffer from
limitations of hearing, judgement or patience, is in many ways more authori-
tative than ‘HAP’; that is, human-aided pronunciation instruction.
CAP also provides a type of feedback which, because it comes from the
machine, is not only authoritative but also highly salient. And this salience can
easily be enhanced by utilizing computer capabilities for presenting informa-
tion in visual and auditory modalities, including multi-modal presentation. The
computer can also individualize pronunciation instruction in ways the pro-
nunciation teacher cannot, based on a mechanical analysis of individual stu-
dent problems and past trials and performance. The computer can moreover
make available a much wider range of presentations, on demand and on the
spot, than a human trainer. In this sense, the computer has the capacity to pre-
sent both highly individual and highly variable training.
Although most CAP systems are stand-alone individual machines, speech
analysis capabilities have recently been incorporated in a language lab set-up
where students work independently at their terminals and the teacher has
access to every terminal from a master-control station, as in the SONY system
described by Lambacher (1997, 1998). This access allows the teacher to set
specific work for individuals, groups or the class as a whole; to review the
430 M.C. PENNINGTON

computer analysis of students’ speech and evaluate their performance; and to

transfer analyses of speech from one terminal to another (e.g., for purposes of
comparison across students). The potential guidance and feedback of the
teacher, in the context of individualized CAP which also allows for compari-
son with the work of classmates, means that this type of system can be used
in a range of modes combining whole-class, small-group or pair, teacher-to-
student and individual work.
The capabilities of CAP just reviewed are shown in the middle column of
Table 1. Because of these capabilities, which in some basic ways and in com-
bination are unique to the electronic medium, CAP has a number of positive
potentials for instruction, as shown in the first column of Table 1.
CAP has the potential to increase language learners’ motivation and effort to
work on their pronunciation. Because it increases the accessibility and quality
of different kinds of pronunciation input, CAP can increase learners’ awareness
and understanding of key features of the phonology of different languages or
varieties, and of their own pronunciation. It thereby increases the learnability
of phonology, an important point for learners past the critical period.
By offering a medium within which to practice, CAP can help learners
increase the precision of articulation in a language or variety, the automaticity
of pronunciation mechanics, the prosodic aspects of speech and the overall flu-
ency of utterance. In providing learners with a private and individual work-
space and various tools, CAP can help build their confidence while developing
skills in the pronunciation and discrimination of sounds and sound patterns of
the target language/variety.

Table 1. Properties, Potentials and Limitations of Computer-Aided Pronunciation

(CAP) Pedagogy.

Pros CAP is Cons

Motivating quick Restricted to some features

Stimulates effort repeatable Limited for whole-class use
Raises awareness precise Analysis must be adjusted for
Increases understanding reliable different voices
Enhances learnability authoritative No baseline for acceptable
Increases automaticity highly salient performance
Fosters precision multi-modal Weak curriculum
Builds confidence individual Focus on decontextualized
Develops skills variable articulatory mechanics
 COMPUTER-AIDED PRONUNCIATION PEDAGOGY 431

4. LIMITATIONS OF CAP

In spite of these substantial positive attributes, CAP remains more a set of

exciting potentials for instruction than an exciting reality. Although there have
been some interesting developments in instructional applications of CAP,
especially in the last decade, this medium—to a greater extent than other com-
puter applications for language pedagogy—has been slow to attract the atten-
tion of top-notch instructional developers. When compared with the innovative
software developed in the last twenty years for teaching science and mathe-
matics or the creative advances in computer arcade software during the same
period, CAP is clearly lagging behind mainstream instructional and entertain-
ment applications of computer technology. Thus, it could be said that CAP has
yet to achieve a state-of-the-art status in language instruction.
As a main limitation of the technology (see the third column of Table 1),
certain aspects of pronunciation do not show up well in the visual representa-
tions of the speech analysis such as (simplified or modified) waveforms and
so cannot generally be trained by such representations. Therefore, developers
need to be aware of the limitations of pronunciation analysis and visualization
and select areas for training with these restrictions in mind. Another main fea-
ture (which is both a strength and a weakness of CAP) is that, because they
are designed for individual use, most pronunciation training systems have lim-
ited utility for whole-class instruction and may therefore be impractical under
many conditions of instruction. In addition, much speech analysis software
must be adjusted or ‘trained’ every time it is applied to a new voice and is in
this sense also restricted for use by more than one student at a time.
The most serious limitations of CAP, however, are pedagogical. First, most
software is not based on any particular theory or model of pronunciation which
differentiates variation from (true) error. Most software therefore has no base-
line or standard for pronunciation targets nor for allowable deviations from
these, other than whatever voice(s) may have been recorded as a model to imi-
tate. In most cases, the learner must simply estimate by eye (e.g., from a sim-
plified waveform) whether an acceptable match with the pre-recorded voice(s)
has been achieved, though in some software pre-set targets have been built in
to specific tasks. In the best software (e.g., some pedagogical software for
modifying the user’s pitch which is available with Kay Elemetrics speech
analysis systems), the degree of match with, or the achievement of, the target
is shown in a motivating graphic (e.g., a giraffe whose neck grows as the pitch
of the input utterance increases).
432 M.C. PENNINGTON

The lack of a model which differentiates (acceptable) variation from (unac-

ceptable) deviation from norms or targets can lead to the problem of feedback
in the way of ‘false negatives’ and ‘false positives’. False negatives are indi-
cations of failure to match a particular speaker’s performance or other pre-set
target, where the learner has in fact achieved a target that would be the same
as, or within the range of variation of, that for a (native) speaker of the lan-
guage. This happens, for example, because the target is based on only one
speaker or variety. False positives in CAP are indications that a learner has
achieved a target when in fact he or she is not actually within the range of
acceptable performance. This happens when the criterion for acceptable per-
formance is set within parameters which are too broad, or which do not dis-
criminate the right features of correct and incorrect, native and non-native, or
central (unmarked) and peripheral (marked) performance.
An equally important problem is that the overwhelming emphasis in com-
puter-based work on pronunciation has been towards the decontextualized
mechanics of articulation. Most of the available pronunciation software con-
tains no curriculum or a limited curriculum, and few applications of the tech-
nology link mechanical and meaningful dimensions of phonology
(Pennington, 1989a). This may be an important reason why pronunciation
technology has enjoyed only modest success in the teaching of English as a
second language, where a focus on meaning and the context of communica-
tion is primary. While some developers cite computer memory limitations as
the main reason for any limitations or restrictions in scope of their software,
clearly the amount of memory available is to some extent a design decision
and is at any rate only one of the significant factors impacting decisions about
pedagogical scope and content.

5. DIRECTIONS FOR CAP

Given these limitations, the most significant of which are conceptual, result-
ing in a rather limited scope for CAP pedagogy, I propose ten principles for
improving this area of CAI, as listed in Table 2.
(1) The first of these principles is that the CAP developer should start from
a well-articulated theoretical position. Most CAP appears to have been devel-
oped outside of any theory of pronunciation or second language phonology.
The implicit theory seems in most cases to be one of pronunciation as a seg-
mental or low-level performance phenomenon.1

1. G. Molholt (personal communication) claims that new software which he has developed,
marketed by Kay Elemetrics Corp, is based on a prosodic rather than a segmental orienta-
tion to L2 phonology.
 COMPUTER-AIDED PRONUNCIATION PEDAGOGY 433

Table 2. Suggestions for Improving CAP Pedagogy.

(1) Start from a theoretical position

(2) Establish a baseline for pronunciation
(3) Set an overall goal for performance
(4) Build in specific targets for performance
(5) Build skills in stages
(6) Link pronunciation to other learning and communicative goals
(7) Design on a principled curriculum
(8) Design based on creative use of properties of computer medium
(9) Raise awareness of contrast with L1 and range of targets for L2
(10) Provide for exploration of database

Such a theory implies that the teaching of pronunciation should be orient-

ed to the level of individual sound segments, or phonemes. An alternative to
the segmental view of pronunciation is the perspective of prosodic phonology.
From this perspective, teaching the prosodic aspects of phonology—intona-
tion, rhythm, rules of linking words, etc.—will have a far greater pay-off than
the teaching of individual sounds, as these inevitably vary greatly in context.
An alternative view to the notion of phonology as a low-level performance
phenomenon, as I have discussed above and elsewhere (Pennington, 1989b,
1996a, 1997, 1998, 1999), is that pronunciation is a key element of one’s self-
image and the image one projects to others. It is therefore as much a social as
a purely mechanical phenomenon. Moreover, pronunciation is a central feature
of pragmatic competence, and every type of phonological error or transfer has
pragmatic consequences (Pennington & Zegarac, 1998). Such a social view of
pronunciation suggests the desirability of linking the mechanics of articula-
tion to communicative contexts or goals.
(2) The second principle for CAP design is to establish a baseline for pro-
nunciation in terms of one or more reference accents. There are issues here in
terms of how a learner wishes to sound, where the learner will be living, who
he or she will be communicating with, and the type of communication (its
structure or genre and degree of formality) engaged in. In my view, a range of
accents should be provided, as a way to aid the learner to develop an aware-
ness of the range of variation that exists among speakers of the target language
or variety.
(3) The third point is that the developer should set an overall goal for per-
formance. This goal should be determined by the learner’s characteristics, such
as language proficiency and needs. The pronunciation goal may be global or
434 M.C. PENNINGTON

else focused on particular lexis, grammatical structures, conversational rou-

tines or patterns; or it may be focused on developing a certain type or types of
skills, such as asking different types of questions, disambiguating information
by pronunciation, etc. A different sort of goal is for learners to be able to
accomplish certain tasks successfully, such as making various types of tele-
phone calls (e.g., different types that would be relevant for someone working
in an international firm).
Another aspect of setting a goal for CAP is whether to focus on intelligi-
bility, accuracy or fluency. Nearly all CAP is focused on accuracy, usually as
an implicit goal with no specific theoretical or contextual underpinning.
Moreover, the focus is usually on accuracy at the level of specific phonemes,
though focusing a learner’s performance on the articulation of individual
phonemes generally disrupts fluency and can even be counterproductive for
intelligibility if the learner concentrates on phonemes to the exclusion of into-
nation, rhythm and the stress of individual words.
For those who have achieved a degree of fluency but not intelligibility, accu-
racy is a defensible goal—assuming the problem with intelligibility is one
involving the articulation of individual sounds rather than one of incorrect
prosody. Intelligibility is of course a priority over accuracy per se and may
lead to a different focus in instruction, such as on key words, stress, rhythm
and intonation rather than on the articulation of individual sounds. For many
speakers who are already focused on accuracy, CAP lessons which focus their
attention away from accuracy and towards fluency may be useful.
(4) The fourth point is to build in specific targets for performance; that is,
as to what performance will count as having achieved or made progress
towards a desired target. Thus, the developer needs to think about how far one
may diverge from, say, a visual target and still be within an acceptable native
speaker range, or within the parameters of one language or variety rather than
another. The developer will also need to consider what items, structures, skills
or tasks will be good indicators of the learner’s progress or achievement.
(5) The fifth point is to build skills in stages; for example, move from eas-
ier to more challenging tasks and link pre-production with in-production and
post-production training. Very little of the CAP available today even has the
basic pedagogical design feature of building from easier to more challenging
tasks in stages, and virtually none conceives of pronunciation training as con-
sisting of pre-production, in-production and post-production phases. In fact,
much of instructional CAP is in essence a form of decontextualized articula-
tion practice, without any attempt to link from one learning stage to another.
 COMPUTER-AIDED PRONUNCIATION PEDAGOGY 435

Pre-production or pre-communication training helps to form targets and

schemas in advance of actually producing speech or communicating in the tar-
get language. Such advance training may involve visual and listening activi-
ties as well as repetition to establish routines and patterns. In-production or
in-process training in CAP is the sort that gives immediate feedback during
production of connected speech so that the learner can continually adjust per-
formance and develop automaticity and fluency. Post-production training in
the form of cumulative analysis and records of speech helps the learner to
develop what are generally referred to as ‘meta-analytical skills’ for self-
correction and self-regulation of performance. (For further discussion and
ideas for teaching, see Pennington, 1996b, Ch. 6; 1997; Pennington & Esling,
1996.)
In-production training seems a particularly important type of CAP, as it pro-
vides a form of real-time feedback that cannot be obtained except by comput-
er means and that makes possible progressive, self-managed improvement. As
I suggested some time ago:
A simple application could involve an indication on the screen of
appropriate groupings of words in a running discourse, showing the
places where linking and pausing would be likely to occur in native
speech. As another element, text could be added to the screen with
appropriate timing of clusters between potential pause-points. The
student would try to match the timing and groupings indicated on
the screen by speaking the text as it is appearing on the screen. An
indication of rhythm and pitch could also be part of the display . . .
(Pennington, 1989a, p.115)
In addition to such automated aids to timing and ‘chunking’ of spoken dis-
course, perhaps given in parallel with indications of prosodic features such as
the rhythm and the pitch of individual ‘chunks’ or longer stretches of speech:
the program might offer running or user-accessed on-line feedback
on production in the form of . . . comments based on a componen-
tial analysis of fluency . . . such as ‘Too Slow,’ ‘No Linking,’ ‘Too
Many Pauses,’ ‘Insufficient Variation in Pitch,’ etc. This type of
feedback can be seen as analogous to that provided by text analysis
programs for written language. (Pennington, 1989a, p.119)
Training in discourse intonation could be provided in an analogous way,
436 M.C. PENNINGTON

with the computer visual display indicating the changing patterns over time—
for example, analysed in terms of a distinction between ‘proclaiming’ and
‘referring’ intonation (Brazil et al., 1980). Proclaiming (falling) intonation sig-
nals completion of discourse units, while referring (non-falling) intonation sig-
nals ongoing development of discourse units. If a discourse unit—e.g., the
statement I’m going—is produced with proclaiming intonation, either the
speaker’s overall communicative purpose or some subpart of it has been
achieved. If, however, I’m going is spoken with referring intonation, this sig-
nals the speaker’s intention to continue the discourse or to have it continued
by another speaker. This contrast is illustrated by the pair of examples below:

Proclaiming Intonation I’m going.

Referring Intonation I’m going (and I won’t return)

(Pennington, forthcoming a)

This distinction or others which are relevant to varieties of English can be

trained by a simple in-progress display using arrows which appear above a text
on the screen as a discourse unit is completed, as in the following continuous
text:

That man over there is waving because I asked him to.

Do you remember last week when I told you I had a man I wanted you to meet?

Well, that’s him: that’s Greg. Come on over and I’ll introduce you.
(Pennington, forthcoming a)

For in-production as well as pre- and post-production instruction, exercises

should be developed on databases of real speech in different contexts; e.g., as
described by Jones (1996).
(6) The sixth point is to link pronunciation to other learning and commu-
nicative goals such as vocabulary, grammar, discourse and pragmatics. Many
books have appeared on the teaching of phonology (e.g., Avery & Ehrlich,
 COMPUTER-AIDED PRONUNCIATION PEDAGOGY 437

1992; Bowen & Marks, 1992; Pennington, 1996b), providing examples of how
such linkage can be accomplished in a non-electronic context. It should not take
exceptional imagination to develop the wealth of published ideas for teaching
second language phonology to make suitable use of computer technology.
(7) The design of CAP pedagogy should be based on a curriculum linked
to creative use of the properties of the computer medium in concert with, rather
than in place of, the other considerations of this list. Unfortunately, designers
of CAP have tended to opt for either (a) a curriculum-as-technology or ‘min-
imal technology’ approach, in which the technological application is matched
to pre-existing curriculum and teaching ideas (thereby underutilizing the tech-
nology), or (b) a technology-as-curriculum or ‘minimal curriculum’ approach,
in which the technology is seen as providing its own curriculum or teaching
approach.
(8) An essential point is that CAP should be based on a principled language
learning curriculum such as a communicative or task-based syllabus. A frame-
work for pronunciation design that is described in detail in Pennington (1996b,
Ch. 6) is given in Table 3. In reproducing this scheme here, I do not mean to
suggest that it is the only or best one to be used for CAP. The point is rather
that there should be a curriculum behind the software and, whatever curricu-
lum there is, should be a defensible one. Hiller et al. (1993) offer some advice
and direction to those trying to develop a progressive curriculum for CAP.
(9) CAP should raise learners’ awareness of the contrast of the L2 or target
variety with the native language or variety and also of the range of acceptable
or related targets and their social significance. Awareness of contrasts can be
built by showing computer analyses side by side for parallel features in the
Table 3. Curriculum Scheme for Pronunciation (Pennington 1996b, p.226).

Unit Activity Practice Cognitive Modality Participation Information

structure type level load

Presentation Focus Mechanical Low –Production –Interaction –Communicative

inference

Contextualization Contextualized

Practice Controlled Meaningful

Structured Realistic

Free Real High inference Production Interaction Communicative

438 M.C. PENNINGTON

two languages or in different dialects of the target language—for example, two

similar but not identical vowels, as in Rochet’s (1990) software for teaching
French vowels to English speakers (see discussion below), or two types of
intonation for declarative sentences, as in the essentially ‘mirror-image’ con-
trast of the typical English and Japanese declarative intonation contours
sketched below:

Typical English declarative intonation

Typical Japanese declarative intonation

Software could easily be developed that displays on one screen, for com-
parative purposes, the range of contours found in a sample of Japanese vs.
English declarative sentences, with statistics as to the frequency of each type
in the samples of speech used as the database. Such comparative information
could be provided as pre-production input for exercises in which learners
would try to match their contours to those on the screen, receiving in-
production (‘formative’) feedback as well as post-production (‘summative’)
feedback on the contour they produced. This feedback could be in the form of
an indication of the percentage of match to a given contour, e.g., as illustrated
graphically by colour-highlighting of the overlapping section of the two con-
tours laid out one on top of the other. Or it could be in terms of a numerical
display of the percentage of overlap, or of the statistical frequency of the con-
tour in one or both languages according to the database from which the con-
tours were derived.
(10) The final point is to provide opportunities for exploratory CAP, partic-
ularly for exploration of video or audio databases such as those of Esling
(1994) and Jones (1996) that include software tools for browsing and compil-
ing features of interest. As one of the most significant potentials of computer
access for individualizing instruction and promoting learner control and inde-
pendence, exploratory CALL should be a feature of CAP.

6. A PROTOTYPE FOR CAP

Rochet’s (1990) software for teaching French vowels to native speakers of

English is the only software I am aware of that begins to implement these ten
 COMPUTER-AIDED PRONUNCIATION PEDAGOGY 439

principles for CAP pedagogy. It might therefore be considered a prototype for

future developments in CAP pedagogy. Rochet’s software has an explicit cur-
riculum that is based on a detailed cross-linguistic comparison of the aspect of
pronunciation to be taught in comparison to the learner’s native language. It is
also based on a theoretical position, in this case, about how vowels are learned
in terms of central and peripheral exemplars. Rochet’s CAP program includes
early, mid-stage and late-stage targets, and it makes use of special features of
the computer. Even this carefully designed software is, however, far from the
ideal, as it is focused narrowly only on individual vowel contrasts between
French and English and does not exploit computer capabilities—especially,
those capabilities other than for speech analysis—to a very great extent.

7. CONCLUSION

I believe that a major effort is needed in pronunciation software development,

based on the many kinds of computer technology available. Given the very
considerable potential and natural advantages of the computer as an instruc-
tional delivery system, there can be no excuse for instructional designers’ con-
tinued production of CAP in which the technology is based on indefensible
pedagogy or a weak or non-existent curriculum. Although learners may be
intrigued by the technology and may in fact improve their pronunciation by
use of it, much more could be done to make optimal use of the considerable
expertise of computer technicians, by marrying that expertise to the consider-
able expertise of language curriculum specialists and making a major
commitment of time and effort to the development of CAP pedagogy.

REFERENCES

Anderson-Hsieh, J. (1992) ‘Using electronic visual feedback to teach suprasegmentals’, System

20: 51–62.
Anderson-Hsieh, J. (1998) ‘Considerations in selecting and using pronunciation technology’,
TCIS Colloquium on the Uses and Limitations of Pronunciation Technology. 32nd
Annual TESOL Convention, Seattle, March 1998.
Avery, P. & Ehrlich, S. (eds) (1992) Teaching American English Pronunciation. Oxford: Oxford
University Press.
Bowen, T. & Marks, J. (1992) The Pronunciation Book. London: Pilgrims/Longman.
Brazil, D., Coulthard, M. & Johns, C. (1980) Discourse Intonation and Language Teaching.
London: Longman.
440 M.C. PENNINGTON

Chun, D.M. (1989) ‘Teaching tone and intonation with microcomputers’, CALICO Journal 7:
21–46.
Esling, J. (1994) Victoria University Phonetic Database, Version 3. Victoria, BC: Speech
Technology Research, Ltd.
Hiller, S., Rooney, E. & Jack, M. (1993) ‘SPELL: An automated system for computer-aided
pronunciation teaching’, Speech Communication 13: 463–73.
Jones, C.M. (1996) ‘The oral language archive (OLA): A digital audio database for foreign lan-
guage study’, Computer Assisted Language Learning 9: 235–50.
Lambacher, S.G. (1997) ‘Developing English pronunciation skills of Japanese learners using
computerized instruction’, Journal of the Phonetic Society of Japan 1 (2): 45–52.
Lambacher, S.G. (1998) ‘The uses and limitations of pronunciation technology’, TCIS
Colloquium on the Uses and Limitations of Pronunciation Technology. 32nd Annual
TESOL Convention, Seattle, March 1998.
Molholt, G. (1988) ‘Computer-assisted instruction in pronunciation for Chinese speakers of
American English’, TESOL Quarterly 22: 91–111.
Pennington, M.C. (1989a) ‘Applications of computers in the development of speaking and lis-
tening proficiency’, in M. C. Pennington (ed.) Teaching Languages with Computers: The
State of the Art. Houston, TX: Athelstan.
Pennington, M.C. (1989b) ‘Teaching pronunciation from the top down’, RELC Journal 20 (1):
20–38.
Pennington, M.C. (1996a) ‘The power of the computer in language education’, in M.C.
Pennington (ed.) The Power of CALL. Houston, TX: Athelstan.
Pennington, M.C. (1996b) Phonology in English Language Teaching: An International
Approach. London: Longman.
Pennington, M.C. (1997) ‘Phonology in language teaching: Essentials of theory and practice’,
in K. Bardovi-Harlig and B. Hartford (eds) Beyond Methods: Companion Components
in Language Teacher Education. New York: McGraw-Hill.
Pennington, M.C. (1998) ‘The teachability of pronunciation in adulthood: A reconsideration’,
International Review of Applied Linguistics 36: 323–41.
Pennington, M.C. (1999) ‘Phonology in the context of communication and language learning’,
Research Report, Series 2. University of Luton.
Pennington, M.C. (forthcoming a) ‘Grammar and communication: New directions in theory and
practice’, in E. Hinkel and S. Fotos (eds) New Perspectives on Grammar Teaching. New
York: Cambridge University Press.
Pennington, M.C. (forthcoming b) Language Learning: An Introduction. London: Arnold.
Pennington, M.C. & Esling, J.H. (1996) ‘Computer-assisted development of spoken language
skills’, in M.C. Pennington (ed.) The Power of CALL. Houston, TX: Athelstan.
Pennington, M.C. & Zegarac, V. (1998) ‘What is pragmatic transfer?’, Conference of the
International Pragmatics Association, Reims, France, July 1998.
Read, C., Buder, E.H. & Kent, R.D. (1990) ‘Speech analysis systems: A survey’, Journal of
Speech and Hearing Research 33: 363–74.
Read, C., Buder, E.H. & Kent, R.D. (1992) ‘Speech analysis systems: An evaluation’, Journal
of Speech and Hearing Research 35: 314–32.
Rochet, B. (1990) ‘Training non-native speech contrasts on the Macintosh’, in M.-L. Craven,
R. Sinyor and D. Paramskas (eds) CALL: Papers and Reports. Houston, TX: Athelstan.