You are on page 1of 11

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/275262490

Speechant: A vowel notation system to teach English


pronunciation

Article  in  ELT Journal · March 2011


DOI: 10.1093/elt/ccr019

CITATIONS READS

2 369

2 authors, including:

Valerie Hazan
University College London
152 PUBLICATIONS   2,784 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

The effect of ageing on speech communication View project

ENRICH: Enriched Communication Across the Lifespan View project

All content following this page was uploaded by Valerie Hazan on 28 January 2016.

The user has requested enhancement of the downloaded file.


Speechant: a vowel notation system
to teach English pronunciation
Jorge dos Reis and Valerie Hazan

Downloaded from http://eltj.oxfordjournals.org/ at University College London on March 6, 2015


This paper introduces a new vowel notation system aimed at aiding the teaching of
English pronunciation. This notation system, designed as an enhancement to
orthographic text, was designed to use concepts borrowed from the representation
of musical notes and is also linked to the acoustic characteristics of vowel sounds.
Vowel timbre is represented in terms of the height of the symbol and vowel
duration in terms of the length of the symbol. The Speechant system was
evaluated in E F L adult education classes in Portugal. A formal assessment that
measured the impact of a term’s tuition by looking at changes in accent ratings of
the learners over that period showed that the group taught using the Speechant
system showed greater improvements in pronunciation than the control group.
Speechant may be an especially useful aid to pronunciation teaching in situations
in which foreign languages are taught without the benefit of technological support.

Introduction The importance of developing good oral competence when learning


a foreign language has been recognized in a framework document
(Common European Framework of Reference for Language) published by the
Council of Europe (2001), although the means of achieving good
competence in a new language, especially if acquired in adulthood, is not
spelt out. Pronunciation is an important element of oral competence, as
a non-native pronunciation of the sounds or intonation patterns of an L2 can
affect speech comprehension, even when L2 speakers are fluent and have
a good level of language proficiency (Derwing and Munro 2005). The
difficulties that learners have in acquiring speech sounds in an L2 that do
not occur or which have a different phonological status in the L1 is well
attested in countless research studies published over the last three decades
(see Flege 2003 for a review). Age of L2 acquisition is a key factor that
impacts on the development of good pronunciation and accurate
discrimination of sounds in an L2. In the last two decades, with advances in
speech technology and in multimedia computing, new approaches to
teaching pronunciation have been developed within the framework of
computer-assisted learning. There are now commercial products on the
market that diagnose a learner’s mispronunciation of sounds in the L2 and
suggest ways of correcting these errors, using animated articulatory displays
(for example Neri, Cucchiarini, Strik, and Boves 2002). Other systems
immerse language learners within a virtual world in which they interact
with avatars in the L2 to complete simple tasks (for example Johnson and

156 E LT Journal Volume 66/2 April 2012; doi:10.1093/elt/ccr019


ª The Author 2011. Published by Oxford University Press; all rights reserved.
Advance Access publication June 17, 2011
Valente 2008). Although such technological approaches are no doubt
extremely valuable, it is still the case that much foreign language teaching,
especially within the framework of adult education classes, is done without
the benefit of computer-assisting learning or even without the benefit of less
sophisticated tools such as audio or video recordings. In E F L classes at adult
education centres in Portugal, for example, which served as a test bed for the
work described here, English pronunciation is typically taught by a native
Portuguese teacher, using words and sentences printed on flashcards. The
challenge in Portugal and in many other countries where foreign languages
are taught without the benefit of technological support is to provide an ‘aid
to pronunciation teaching’ that may enhance the teaching of pronunciation
based on orthographic text.
Learning to pronounce English accurately, using tuition methods based on

Downloaded from http://eltj.oxfordjournals.org/ at University College London on March 6, 2015


orthographic text, is notoriously difficult due to two factors. First, at the
phonetic level, as mentioned above, the E F L learner may have problems in
discriminating and pronouncing certain speech sounds in English because
they do not occur in their native language. For example, many Portuguese
learners of English struggle to discriminate and correctly pronounce the
extensive English vowel inventory as Portuguese only has five vowels and no
long/short vowel distinction. A second complicating factor is the poor letter-
to-sound correspondence in English, where a same speech sound can be
represented by different letters (for example the sound / / can be spelt as:
‘ea’ as in ‘pea’, ‘ee’ as in ‘tree’, ‘ie’ as in ‘piece’, ‘e’ as in ‘me’) and a same letter
can be pronounced as different sounds. This is especially challenging for
students who are native speakers of languages such as Portuguese, Italian,
or Greek, which have a more direct and transparent letter-to-sound
correspondence as these students will have a natural expectation that a given
letter will always be pronounced in the same way.
One early approach to providing aids to pronunciation by ‘enhancing’
orthographic text was the Colour Story Reading scheme (Jones 1967). In
this system, different letters with the same sound had the same colour. Four
different letter colours were used and also nine coloured backgrounds in the
shape of a circle, square, or triangle on which black letters may be printed.
This system showed the potential of graphic design as an applied linguistic
device. However, in Jones’s scheme, there was no connection between the
shapes and the characteristics of sound: the colours were arbitrary and did
not provide any phonetic information.
Another approach appeared in a language-teaching textbook by Gilbert
(2005). Gilbert took music as a reference to stress the importance of
acquiring the appropriate rhythm and melody of the language. Gilbert also
employed some typographical manipulations as aids to pronunciation of
the text. It is quite common to use a curved line over a sentence to show the
suprasegmental elements of pronunciation, which Gilbert also uses.
However, Gilbert also stressed the representation of vowel quality. Four
kinds of typographical manipulation were used in this approach:
1 distortions or expansion of typeface
2 insertion of international phonetic alphabet symbols inside Latin
alphabet letters

Speechant: a vowel notation system to teach English pronunciation 157


3 repetition of letters and diminishing of size
4 over imposition and repetition of letters.

Description of The main objective of our research was the construction of a new vowel
Speechant notation notation system as a pronunciation aid in the teaching of E F L, mainly for
system adults. The Speechant Vowel Notation System is a new device that works
with print: the source is written text, as this is the medium used for teaching
pronunciation, and the notation provides a method to translate from written
text to sound. The design of this notation system was strongly influenced by
the first author’s background as an artist, typographical designer, and
singer.
As the learning of English vowels is particularly challenging for L2 learners
whose native languages have restricted vowel inventories and no long/short
vowel distinction, the focus of the notation system was to mark vowel timbre

Downloaded from http://eltj.oxfordjournals.org/ at University College London on March 6, 2015


and duration. The first step in the design of this notation system was the
realization that English vowels can be described as being scaled from ‘high’
to ‘low’ in terms of their timbre (related to the frequency spectrum of the
sound), just as musical notes can be scaled in terms of their pitch. Vowels
can therefore be organized on a kind of musical scale, ranging from high
timbre vowels, like the sound / / in ‘she’ to low timbre vowels like the sound
of / / in ‘port’.
A second connection between music and vowel production is that English
vowels are of different lengths just as a musical note can be short (for
example semiquaver) or long (for example a breve in musical notation).
Duration markers can therefore provide information about the length of the
vowel (relevant to the distinction between ‘bit’ and ‘beat’ for example). Some
English vowel sounds, diphthongs, involve a change in timbre from one
vowel to another. This dynamic aspect was addressed through marking the
sound route of the diphthongs graphically (for example the diphthong / /,
as in ‘boy’, starts with a low timbre sound and changes into a high timbre
sound). Some of the Speechant notation for different vowel types is shown
in Figure 1.

Assessing the An essential component of the development of any aid to language learning
Speechant system in is its assessment in a classroom situation with the student population for
a classroom setting which it is intended. The Speechant Vowel Notation System was evaluated
in an adult education setting in Lisbon, Portugal, over a period of a term. The
method of language education used in these adult education classes is quite
traditional and text based although there is also strong emphasis on the oral
aspects of the language, with students repeating phrases and words
presented on the board or on flashcards.

Participants The learners were all adults aged between 22- and 58-years old who attended
four different ‘beginner’ classes of English as a Foreign Language. Their
teachers were Portuguese speakers, who had satisfactory oral English
proficiency, but with a foreign accent in their English pronunciation.
Teachers also used tapes produced by native speakers in their teaching. The
Speechant system was used with four groups of students in adult education
classes (total of around 100 students) but was formally evaluated in one
beginner class of 20 adult learners. These were typically adults in

158 Jorge dos Reis and Valerie Hazan


Downloaded from http://eltj.oxfordjournals.org/ at University College London on March 6, 2015
figure 1
Speechant notations for
a range of lax and tense
vowels and diphthongs

employment (for example nurses, taxi drivers) who attended English classes
in the evening for two sessions of two hours a week. A further group of 27
learners attending another class in the school were taught with the same
curriculum but with normal orthography rather than with the addition of
the Speechant vowel notation; 20 of those students formed the control
group.

Curriculum A new set of didactic materials was developed in collaboration between the
teachers, who selected the words and sentences to be worked on, and the
first author who then prepared flashcards using the Speechant notation.
These focused on teaching English vowels in a hierarchical way, focusing on
high timbre vowel sounds before moving to low timbre sounds. The order of
sounds was organized into three month-long periods. This approach to the
teaching of vowel pronunciation is in line with the usual teaching methods
in such adult education classes. In the first period, the following vowel
sounds were the focus of the training: , as in
‘bit, bet, bat, but, beat, Bert, Bart, bait, bite, boy’. In the second period, the
focus was on the following vowel sounds: as in
‘pot, put, putt, boot, port, boat, bout, bier’. Finally, in the third period, the
following vowel sounds were added: as in ‘pear, pure, poor,
pew’. In the lessons, an emphasis was put on teaching the students about
the lack of consistent letter-to-sound correspondence in English. If, for
example, the vowel / / was the focus of a part of the teaching session, the
flashcards included a set of words where the sound / / was represented by
different letters.

Speechant: a vowel notation system to teach English pronunciation 159


The set of didactic materials developed for use with the Speechant vowel
notation system included flash cards (with individual words and small
sentences), spelling board (long sentences, specific groups of organized
vocabulary), and individual cards (words and sentences and texts). An
example of sentences produced with Speechant notation is given in
Figure 2. These three didactic reading platforms are also commonly used in
regular methods of teaching in Portugal. Two sets of didactic materials were
prepared. For the classes using Speechant, the texts on these cards carried
the Speechant notations; for the control classes, the cards contained normal
orthography only. These flashcards were used in the classroom as prompts
for oral pronunciation lessons. The students read the words aloud as a group
or individually many times throughout the lesson, so that the link between
the sounds in the words and their letter representation was established. The
aim of using the flashcards was to help visualize and retain the connection

Downloaded from http://eltj.oxfordjournals.org/ at University College London on March 6, 2015


between sound and word. Exercises involving the flashcards and other
materials (with Speechant for the test group, and plain orthography with the
control group) typically took up at least three-quarters of the two-hour
lessons.

figure 2
Example of flashcard
materials produced with
Speechant notations

Materials used in the Speechant evaluation


An evaluation was carried out to investigate whether the students who had
received the vowel pronunciation lessons using flashcards with Speechant
notation improved their pronunciation more than the students who had
done similar pronunciation lessons but without the addition of the
Speechant notation. The evaluation of the learners’ pronunciation was
based on the reading of three sentences typed in plain orthography (i.e.
without the Speechant notation). These sentences were designed so that
Sentence 1 primarily contained vowels trained in Period 1, Sentence 2
additionally contained vowels trained in Period 2, and Sentence 3
additionally contained vowels studied in Period 3. These sentences are given
below, with an indication of the vowels trained in that period that were
included in the sentence.

160 Jorge dos Reis and Valerie Hazan


Downloaded from http://eltj.oxfordjournals.org/ at University College London on March 6, 2015
The three sentences were new for the learners but the words and sounds
present in each of the three sentences had previously been studied in the
classroom. For evaluation purposes, the three sentences were read by each
learner in the Speechant and control groups at the end of each month-long
period (i.e. on three occasions), without correction by the teacher, and were
recorded on tape. The expectation is as follows: if the Speechant notation
system is helping learners to pronounce the vowels, the improvement in
vowel pronunciation should be greater for the Speechant group and should
follow the order in which sounds were taught (i.e. the readings of Sentences
2 and 3 should improve later than Sentence 1).

Evaluation procedure The complete data set (i.e. the recording of the three sentences at each of
three time periods) was collected for 20 learners in the Speechant group and
20 learners in the control group. The recordings were digitized and, for each
learner group, a set of three CDs was created. Each CD included the files for
one of the three sentences randomized across learners and across the three
recording times. The 180 speech files were presented to 45 undergraduate
students from the Linguistics Department of Lisbon University, who were
experienced in transcription and in judging the quality and accuracy of
pronunciation. Their task was to rate each sentence in terms of how ‘native’
or ‘foreign’ it sounded on a scale of 7 (native English) to 0 (foreign English).
Such rating tasks have been used in many studies of the phonetic aspects of
L2 acquisition (for example Flege, Birdsong, Bialystok, Mack, Sung, and
Tsukada 2006) and have been shown to reliably show the impact of age of
learning, for example. Although a judgement of 7 (native English) is
unlikely to be achievable for these learners, the ‘native English’ label is
a clearer benchmark against which to rate the recordings than a benchmark
such as ‘high proficiency in English pronunciation’ which is a more likely
target but less well defined. The ratings data were analysed to get a mean
rating score for each sentence by each learner (averaged over the 45
raters). This made it possible to look for evidence of improvement over
time. This same procedure was later repeated to obtain ratings for the
control group.

Speechant: a vowel notation system to teach English pronunciation 161


figure 3
Accent ratings for
Sentence 2 for the
Speechant (black circles)
and control (white
circles) group for
recordings made at the
end of each of the three

Downloaded from http://eltj.oxfordjournals.org/ at University College London on March 6, 2015


time periods

Evaluation results The accent rating data were examined for each sentence. For each session,
repeated-measures analyses of variance (ANOVA) were used to look at the
within-group effect of recording session (T1, T2, T3) and the between-group
effect of learner group (Speechant, control). For Sentence 1, the effect of
recording session was significant (P , 0.001) and pairwise comparisons
showed that there was a significant difference between recordings T1 and
T2, and T2 and T3 showing a steady improvement. Significantly higher
ratings were obtained by the Speechant group.
For Sentence 2 (see Figure 3), which contained vowel sounds trained
between T1 and T2, the effect of recording session was significant
(P , 0.001): there was a significant difference between T1 and T2 and
T2 and T3. Higher ratings were again obtained by the Speechant group
(P , 0.001). The group by session interaction was also significant
(P , 0.01), so post hoc analyses were carried out. At Session T1, the group
difference was just significant (P , 0.05); ratings for the Speechant group
were significantly higher than for the control group at Sessions T2
(P , 0.001) and T3 (P , 0.001), i.e. after these sounds had been practised in
the classroom.
For Sentence 3 (see Figure 4), the effect of recording session was significant
(P , 0.001): there was a weakly significant difference (P ¼ .02) between
sessions T1 and T2 and strongly significant difference (P ¼ 0.001) between
T2 and T3. Higher ratings were again obtained by the Speechant group
(P , 0.001). The group by session interaction was also significant: the
difference between groups was weakly significant at T1 (P , 0.05) and T2
(P , 0.05) but the accent ratings were significantly higher for the Speechant
group at Session 3 (0.001).
It did seem to be the case that accent ratings for the Speechant group were
slightly higher overall than those for the control group at Session 1,
suggesting that some learners in the Speechant group may have had
a higher oral proficiency despite efforts to match the two groups for
proficiency. However, there is also clear evidence that the Speechant group
showed higher increases in accent ratings than the control group following
training with a particular group of vowels. For Sentence 1, which included

162 Jorge dos Reis and Valerie Hazan


figure 4
Accent ratings for
Sentence 3 for the
Speechant (black circles)
and control (white
circles) group for
recordings made at the
end of each of the three

Downloaded from http://eltj.oxfordjournals.org/ at University College London on March 6, 2015


time periods

the ‘easiest’ vowels, there was a steady improvement in accent ratings for
both groups from T1 to T3, but with consistently higher ratings for the
Speechant group. For Sentence 2, the two groups barely differed at T1,
before these sounds were trained, and this small difference is likely to be due
to the slightly higher oral proficiency of the Speechant group. However,
much greater gains in accent ratings can be seen for the Speechant group
than for the control group at T2 and T3. For Sentence 3, containing sounds
trained in the final phase, the group difference was barely significant for the
first two sessions, but a much greater increase in rating for the Speechant
group was obtained at T3. There is therefore evidence that increases in
accent rating were related to the period at which the sounds were trained
and that the Speechant group improved their accent more than the control
group.

Subjective Learners were also asked for their opinion of working with the Speechant
assessment system and in general gave positive feedback. In response to the main
question (When you first started using Speechant with the flash cards and
the spelling board what did you think?), many learners reported the
following: ‘I thought this was an easy way to learn English’, ‘I thought this
would improve phonetics’, ‘I thought this was new’, and ‘I found it fun and
enjoyable’. Each of these statements was given by at least five students.
When asked how they would convince someone in another class to try this
system, they wrote:
n it’s easy to memorise words
n it’s very easy
n it improves pronunciation
n it’s simple
n it’s faster
n it differentiates the length of the sound
n it shows when a sound goes up and down
n it’s innovative
n it improves your spelling
n it’s a lot of fun
n it really works for me.

Speechant: a vowel notation system to teach English pronunciation 163


Again, each of these statements was given by at least five students.
The two teachers who used Speechant in their adult education classes
commented that ‘the learning process was successful and useful for the
students who quickly acquired a good English pronunciation in an
enjoyable way. The students liked learning this way because it was appealing
and easy to follow’. Another teacher commented:
In November I wanted to know my students’ opinion on the experience
and I was surprised at how enthusiastically they agreed with it. I share the
opinion that good pronunciation should be a high priority for beginners
of a foreign language. Regular pronunciation work builds students’
confidence and helps them communicate more successfully.

Plans for further The statistical evaluation of the students’ pronunciation suggests that

Downloaded from http://eltj.oxfordjournals.org/ at University College London on March 6, 2015


development of Speechant can be a useful tool for teaching English pronunciation. During
Speechant its development phase and evaluation in the adult education classes in
Portugal, the design of flashcards using Speechant notation was carried out
by the first author, using materials (words and sentence) suggested by the
teachers. The essential next phase of development is to put the development
of new materials under control of the teachers themselves. To this end, the
Speechant notation system is being developed by the first author into a new
font (‘Kensington Gore’) that teachers will be able to download in order to
construct their own materials. This font will include for each vowel sound
a set of glyphs/symbols which will encompass all the potential grapheme
(letter) representations of this sound in English so that teachers can
construct words or sentences in orthographic text but including the
appropriate Speechant markers.

Conclusion Many factors impact on the learning of accurate pronunciation in


a foreign language, and a strong focus on the oral aspects of language is
more likely to achieve good pronunciation than a didactic approach
strongly based on the use of text. However, given that much language
learning still occurs through the use of sentences or small dialogues
learnt through text, Speechant appears to facilitate the learning of
accurate pronunciation in this medium. The formal evaluation of
Speechant with groups of learners receiving the same curriculum with or
without Speechant notation showed that accent ratings improved more
for learners taught with Speechant. Subjective student and teacher
feedback was also positive. Broader evaluations of Speechant are
necessary to validate the system. First, Speechant needs to be evaluated
with a wider range of students taught with different didactic approaches;
it is also necessary to obtain formal ratings using a broader range of
materials than the small set of sentences used here, and to assess
learners’ pronunciation in spontaneous speech in addition to read
sentences. However, we believe that initial evaluations of this approach
are promising and that this approach is particularly relevant in situations
where more technologically sophisticated approaches are not likely to be
available.
Final revised version received November 2010

164 Jorge dos Reis and Valerie Hazan


References Applications of Artificial Intelligence Conference,
Council of Europe. 2001. Common European Chicago, Illinois.
Framework of Reference for Languages: Learning, Jones, J. K. 1967. Colour Story Reading. London:
Teaching, Assessment. Cambridge: Cambridge Nelson.
University Press. Neri, A., C. Cucchiarini, H. Strik, and L. Boves. 2002.
Derwing, T. M. and M. J. Munro. 2005. ‘Second ‘The pedagogy-technology interface in Computer-
language accent and pronunciation teaching: Assisted Pronunciation Training’. Computer-Assisted
a research-based approach’. T ES O L Quarterly 39/3: Language Learning 15/5: 441–67.
379–97.
Flege, J. 2003. ‘Assessing constraints on second- The authors
language segmental production and perception’ in Jorge dos Reis is a Lecturer at Lisbon University
A. Meyer and N. Schiller (eds.). Phonetics and (Faculty of Fine Arts) and a researcher in the field of
Phonology in Language Comprehension and typography. He has published in the area of
Production, Differences and Similarities. Berlin: letterpress, language art, and type design. He holds
Mouton de Gruyter. an MPhil from the Royal College of Art in London,

Downloaded from http://eltj.oxfordjournals.org/ at University College London on March 6, 2015


Flege, J., D. Birdsong, E. Bialystok, M. Mack, H. Sung, where he developed the Speechant vowel notation
and K. Tsukada. 2006. ‘Degree of foreign accent system. Since 1996, he has worked as
in English sentences produced by Korean a multidisciplinary typographic designer in Lisbon.
children and adults’. Journal of Phonetics 34/2: Email: the.jorge.dos.reis.studio@clix.pt
153–75.
Gilbert, J. 2005. Clear Speech from the Start: Basic
Pronunciation and Listening Comprehension in North Valerie Hazan is a Professor of Speech Sciences in the
American English. (Third edition). Cambridge: Department of Speech, Hearing, and Phonetic
Cambridge University Press. Sciences at University College London, with
Johnson, W. L. and A. Valente. 2008. ‘Tactical research interests in the phonetic aspects of L2
language and culture training systems: using acquisition, speech perception development, and
artificial intelligence to teach foreign languages and computer-assisted training.
cultures’. Paper presented at the Innovative Email: v.hazan@ucl.ac.uk

Speechant: a vowel notation system to teach English pronunciation 165

View publication stats

You might also like