Vowel Space using Acoustic Phonetics

advertisement
24. An Acoustic Study of Vowels of English
in ESL and EFL speech:
Case of Arabic speakers Learning English in Delhi (ESL) and Those Learning
English in Cairo (EFL)
Asher R. Jesudoss
(Published in ‘Issues in Learning Theories and Pedagogical Practices’, an
Orient Blackswan Publication, edited by Prof. Vaishna Narang)
Introduction
The aim of this study is to measure the acoustic space of the vowels of English as spoken by two groups
of learners—speakers of Arabic learning English as a second language in Delhi, India; and those learning
English as a foreign language in Cairo, Egypt. The assumption is that learners in a second language
environment will be able to form a distinct vowel space for a language faster than those learning the
language in a foreign language environment. Also, one would expect the distinction between the vowel
spaces to be greater in Delhi than in Cairo. The objectives of this experiment are to analyse vowel
sounds produced in a controlled environment by Arabic speakers of English (both Arabic vowel sounds
and English vowel sounds) and to map the vowel spaces of an individual speaker in each language with
that in the other language. Based on this mapping, which was a result of the recording of vowel samples
and analysing them using software like Wave surfer, we wanted to find out if there is a difference
between:

Vowel space of English in ESL (Delhi) and English in EFL (Cairo)

Vowel space of English and Arabic in an ESL situation

Vowel space of English and Arabic in an EFL situation

Vowel space of Arabic spoken in Cairo and Arabic spoken in Delhi
Acoustic Phonetics
Through the spectrographic analysis of the vowel sounds recorded, we can plot the frequencies and
therefore determine the relative position of a vowel sound and also determine the vowel space of a
particular speaker of a particular language. Vowels can be categorised mainly in terms of three features:
1) Vowel height— which is inversely proportional to the frequency of the first formant.
2) Degree of backness—which is proportional to the difference between the frequencies of the
first and second formants.
3) The degree of rounding—which has complex correlates.
Given below is a three dimensional diagram or vowel space, and the arrows indicate the flow of air as
we breathe. The shape of the tongue within the vowel space is what gives the vowel sound its quality,
besides the shape of the lips. (Source: Ladefoged, 1982).
Vowel Space using Acoustic Phonetics
It is possible to analyse sounds so that we can measure the actual frequencies of the formants. We can
then represent them graphically as well. This is because of the invention of the sound spectrograph,
which translates sound into a visual representation. When one converts vowels to a waveform, there is
a certain way in which these parameters can be calculated. The first formant often corresponds to the
height of the vowel, while the difference between the first and second formants indicates the frontness
or backness of the vowel. The third formant indicates, to a certain extent, the degree of rounding of the
lips. When the difference between the first two formants is plotted against the first formant, we get
something quite close to the position of the tongue as calculated by articulatory phonetics.
Preparation and Ground Work
The first step of the study was to conduct some research on Arabic vowel sounds in order to identify the
sounds that are closest to their English counterparts, and also in order to identify the target vowel
sounds to be analysed.
The second step was to identify the vowel phonemes in each language. Next, a list of five short listed
words was created (in Arabic), and then another list of English words with similar vowel sounds was
created. The speakers were each given 5 target vowel sounds in Arabic, and the equivalent sounds in
English. All these vowel sounds were placed in carefully chosen words, wherein the vowel sounds
occurred in a similar syllabic structure (CVC), in order to minimise variations.
The next step was to look for speakers of the language in the right age group and with similar exposure
to the languages at hand. The sample we took consisted of students at the universities who were 20–25
years of age, and had learnt English as a second/foreign language in school, and had learnt Arabic in
their childhood, either as their mother tongue, or/and as a language of recitation of the Quran.
The subjects were all male, in the age group of 20–25, the age when students graduate from college and
look for jobs. One purpose of this paper is also to gauge the level of English proficiency attained by a
speaker in either country at this age, as the greater the proficiency, the better chances of getting a job in
today’s globalised world. We preferred to study subjects of the same gender, as this reduces anomlies
that may arise due to the difference of frequency between the voices of men and women. The subjects
were students from Cairo University in Egypt, and Jawharlal Nehru University (JNU) in New Delhi.
Each speaker was made to speak each word thrice (the tone of the speaker often varies during each
production, and the second articulation is usually the one where the tone is even throughout the
utterance.) In all we had 5 vowel sounds in English and 5 vowel sounds in Arabic that were to be
analysed. 10 speakers uttering each vowel sound thrice made it a total of 10*10*3 = 300 recordings to
be analysed.
The Survey
The aims and objectives, as stated at the beginning of the study, were to conduct an acoustic analysis of
the vowel sounds produced by Arabic speaking learners of English in an ESL environment (Delhi) and in
an EFL environment (Cairo) in order to find out if there is a difference between:

Vowel space of English in ESL (Delhi) and English in EFL (Cairo)

Vowel space of English and Arabic in an ESL situation

Vowel space of English and Arabic in an EFL situation

Vowel space of Arabic spoken in Cairo and Arabic spoken in Delhi
The acoustic space was calculated for the following:

Arabic vowel sounds produced by learners in Cairo

Arabic vowel sounds produced by learners in Delhi

English vowel sounds produced by learners in Cairo

English vowel sounds produced by learners in Delhi
Following this, the mapping was done at four levels

Mapping of Arabic vowels and English vowels as produced by learners in Cairo

Mapping of Arabic vowels and English vowels as produced by learners in Delhi

Mapping of Arabic vowels produced by learners in Cairo with those produced by learners
in Delhi

Mapping of English vowels produced by learners in Cairo with those produced by learners
in Delhi.
The assumption was that the distinction between the vowel space of Arabic (the native language
of the speakers) and that of English (the second/foreign language of the speakers) would be greater in
an ESLenvironment as compared to an EFL environment. For this, the following steps were taken.
1.
The cities, vowel phonemes, word-lists, subjects were identified.
2.
The word recordings of the subjects were then fed into a computer software (Wave surfer) and
the values of F1 and (F2-F1) were calculated.
3.
The averages of these values of f1 and f2-f1 were then calculated and the graph for the vowel
space for each speaker in each language was then drawn by plotting all the vowel sounds of that
speaker in that language in one chart so that the resultant plot resembled the cardinal vowel
chart.
4.
5.
These steps were followed for each speaker in both languages.
The vowel space for each speaker in one language was then mapped with the vowel space for
the same speaker in the other language.
6.
The averages of each set of subjects was then calculated
7.
These averages were then mapped against each other in four different ways:

Mapping of Arabic vowels and English vowels as produced by learners in Cairo.

Mapping of Arabic vowels and English vowels as produced by learners in Delhi.

Mapping of Arabic vowels produced by learners in Cairo with those produced by
learners in Delhi.

Mapping of English vowels produced by learners in Cairo with those produced by
learners in Delhi.
Observations and important results
Based on the averages of each set of speakers in each language:
C1-5/A (The average Arabic vowel space as produced by the subjects in Cairo)
1.
The vowel space in Egyptian Arabic is represented by a triangle, with the extremes being
formed by /i/, /u/ and /æ/.
C1-5/E (The average English vowel space as produced by the subjects in Cairo)
2.
The vowel space in English remains a quadrilateral, as it is supposed to be. However, the open
vowel sounds, as we can see, are not open enough. The vowels are restricted to closed and half
open vowel sounds
D1-5/A (The average Arabic vowel space as produced by the subjects in Delhi)
3.
4.
5.
A vowel quadrilateral, formed by /i/, /u/, /a/ and /æ/.
The /æ/, however remains more of a back vowel, although it is more fronted compared to the
/a/.
The /o/ is definitely a central vowel sound in this accent, and is not a peripheral vowel
D1-5/E (The average English vowel space as produced by the subjects in Delhi)
6.
7.
A vowel quadrilateral is formed by the peripheral vowels - /i/, /u/, /a/ and /æ/.
However, the figure actually is more of a pentagon, as the /o/ in English is retracted further
behind compared to the /u/ and the /a/.
Based on the mapping of the averages:
CD 1-5/A
1.
The vowel space remains fairly similar in terms of area. However, certain key differences
observed are:
2.
The Arabic /i/ produced by the speakers in Cairo is more fronted and more closed than that
produced by the speakers in Delhi. The difference in f2-f1 values is 71 Hz and the difference in f1
values is 86 Hz.
3.
The Arabic /u/ produced by both sets of speakers is fairly identical
4.
The Arabic /o/ as spoken by subjects in Delhi tend more towards the centre of the vowel space,
while the Arabic /o/ spoken by the subjects in Cairo occurs as a back vowel. It is also more open
than its counterpart produced by the subjects in Delhi. The difference in f2-f1 values is 168 Hz
and the difference in f1 values is 30 Hz.
5.
The Arabic /æ/ as spoken by the subjects in Cairo is completely a back vowel sound, while it is
partially back as spoken by the subjects in Delhi.
6.
The Arabic /a/ as spoken by the subjects in Cairo is more closed and fronted as compared to
that produced by the subjects in Delhi.
C1-5/AE
7.
The English /i/ is marginally more fronted than the Arabic /i/
8.
The English /u/ and the Arabic /u/ almost coincide with each other
9.
The English /a/ and the English /o/ are more back and open as compared to their Arabic
counterparts.
10.
The Arabic /æ/ gains a clear distinction compared to the English /æ/. While the latter is a front
open vowel, the former is a back open vowel. The difference in f2-f1 values is 600 Hz, while the
difference in f1 values is around 240 Hz.
11.
The English vowel space is a quadrilateral formed by /i/, /u/, /a/ and /æ/, while the Arabic
vowel space is a triangle formed by /i/, /u/, and /æ/.
CD 1-5/E
12.
The values for the English /i/ for both sets of speakers are fairly identical
13.
The English /u/ is marginally more back for the subjects in Cairo
14.
The English /o/ is more open and back for the subjects in Delhi.
15.
The English /a/ is way more open in Delhi, as compared to those in Cairo. The difference in f2f1 values is 78 Hz and the difference in f1 values is 172 Hz.
16.
The English /æ/ is also way more open for the subjects in Delhi, as compared to the subjects
in Cairo. The difference in f2-f1 values is 92 Hz and the difference in f1 values is 133 Hz.
17.
The English vowel space for the subjects from Cairo is restricted when it comes to the f1 range.
In articulatory terms this corresponds to the tongue height, where the vowels are restricted to
the articulatory space between closed position and half open position.
D1-5/AE
18.
19.
The English /i/ is more closed and more fronted compared to the Arabic /i/ for the subjects
from Delhi.
The English /u/ and the Arabic /u/ pretty much coincide with each other.
20.
The Arabic /o/ sound as spoken by the subjects in Delhi occurs at a central position, and is not a
peripheral vowel, while the /o/ sound in English is further back than any other vowel sound. This
stark contrast is a very interesting observation. The difference in f2-f1 values is 372 Hz and the
difference in f1 values is 115 Hz.
21.
The Arabic /æ/ produced by speakers in Delhi is further back as compared to the English /æ/,
which remains a front open vowel sound. The difference in f2-f1 values is 401 Hz and the
difference in f1 values is 117 Hz.
22.
The Arabic /a/ is more closed as compared to the English /a/. The difference in f2-f1 values is
43 Hz and the difference in f1 values is 75 Hz.
23.
The Arabic vowel space is clearly affected probably owing to contact with other local and
regional languages, since all the subjects in the study are multilingual.
Variance
The following graphs were generated to sum up the variance in (f2-f1) values and (-f1) values for each
of the vowel sounds. The key to the graphs has been mentioned below:
Vowel 1
Vowel 2
Vowel 3
Vowel 4
Vowel 5
/a/
/æ/
/i/
/u/
/o/
The following graph represents the variance in (f2-f1) and f1 values between the Arabic vowels as
spoken in Cairo and the English vowels as spoken in Cairo:
The following graph represents the variance in (f2-f1) and f1 values between the Arabic vowels as
spoken in Cairo and the Arabic vowels as spoken in Delhi:
The following graph represents the variance in (f2-f1) and f1 values between the English vowels
as spoken in Cairo and the English vowels as spoken in Delhi:
The following graph represents the variance in (f2-f1) and f1 values between the Arabic vowels
as spoken in Delhi and the English vowels as spoken in Delhi:
The source data for the graphs above has also been tabulated below. The chart clearly highlights
the vowel sounds that have a high variance in f2-f1/f1 values.
Cairo Arabic/Cairo English
Cairo Arabic/Delhi Arabic
a
-120.055
50.12446786
a
-84.8974
147.2450839
æ
630.4651
-243.328196
æ 136.8698
5.875185547
i
84.88694
12.35648402
i
-70.9781
86.25538979
u
-8.67678
4.623538668
u
34.66744
-4.396566156
o
-82.4708
40.61876422
o
168.7524
-29.97306819
Cairo English/Delhi English
Delhi Arabic/Delhi English
a
77.93661
171.9784699
a
42.77911
74.85785381
æ
-92.2689
132.703081
æ 401.3264
-116.5003005
i
30.27152
22.08870701
i
186.1365
-51.81019876
u
58.94456
3.600179957
u
15.60034
12.62028478
o
-120.751
44.57447069
o
-371.974
115.1663031
Colour
Code
Legend
Low (0-50 Hz)
Mid (50-100Hz)
High (100-200 Hz)
Extremely
High
(>200 Hz)
Conclusion
As can be confirmed from the bar graphs above, the open vowels are the ones with the most variance,
while the variance amongst the closed vowels is minimal. Some vowel sounds that stand out in terms of
contrast are the /æ/ sound in Arabic and English as spoken in Cairo, the /æ/ sound in Arabic and English,
as spoken in Delhi, and the /o/ sound in Arabic and English, as spoken in Delhi. Some others with a less
degree, yet substantial amount of variance are Arabic /a/ in Cairo, the Arabic /æ/ and /o/ between Cairo
and Delhi, the English /a/, /æ/ and /o/ between Cairo and Delhi, and the English /i/ between Delhi
English and Delhi Arabic. From the graphs and tables above, a number of interesting details have sprung
up. In articulatory terms, one can easily observe that in Delhi, the Arabic vowel /æ/ has moved to a
more fronted position as compared to the Arabic /æ/ in Cairo. This could be because of the contact with
several of the local Indian languages, wherein /æ/ is an open vowel. We can also see that the Arabic /a/
in Delhi is at a more open position as compared to the Arabic /a/ in Cairo. This could also be attributed
to the language contact situation in India. We can also observe that although the Arabic /æ/ produced
in Cairo is an open, back vowel, this does not get translated to the English /æ/, which remains a front
vowel, however, only partially open. The vowel space in the English spoken by the subjects in Cairo is
restricted in terms of openness. This is not the case in English, as spoken by the subjects in Delhi, where
the vowel chart resembles the Cardinal vowel chart more closely. We see that all three open vowels /æ/, /a/ and /o/ - are far more open than their counterparts in the vowel chart representing the sounds
produced by the subjects in Cairo. We see that the English vowel space for the speakers in Delhi is larger
and broader as compared to their Arabic vowel space.
Also, a couple of other extremely interesting observations are that the Arabic /a/ being a back
vowel in Cairo and the Arabic /o/ being a central vowel in Delhi. In terms of the vowel spaces, it is clearly
visible that the English vowel space of the subjects in Cairo is restricted in terms of f1 values, and is
therefore more or less limited within their Arabic vowel space. The English vowel space for the subjects
from Delhi, on the other hand, is wider, wherein the open vowel sounds are completely realized, and we
can see a clear separation of the English vowel space from the Arabic vowel space. This supports our
initial assumption that the distinction between two vowel spaces would be greater in
an ESL environment (Delhi) as compared to an EFL environment (Cairo).
Future Projections
The study has certainly presented us with a lot of existing data to base future hypotheses on. A
number of conclusions can be drawn by the location of all vowels in their respective acoustic spaces.
There is a lot more potential and a lot of analyses can be done on the data available. This data gives us a
direction for further study. Also, similar studies with much larger data bases from both ESL and EFL
environments is required in order to be able to draw more conclusive results. Such studies with a larger
database will have implications in both learner language studies and in language pedagogy. Moreover,
English is taught as a second language/foreign language almost throughout the world. Similar exercises
at other locations would perhaps yield hypotheses regarding the distinction between acquisition (in a
second language environment) and learning (in a foreign language environment). The possibilities, given
sufficient time and resources, can be be further explored, streamlined, mapped and studied in a manner
appropriate to the variety of possibilities that have opened up.
References
Al-Ani, Salman.1970). ‘Arabic Phonology’, Janua Linguarum Series Practica 61.
Al-Badawi, al-Said. 1973.Mustwayat al-`Arabiyya al- mu`asira fi Misr (Levels of contemporary
Arabic in Egypt). Cairo: Dar al-Ma`arif.
Baken, R. and R. Daniloff.(eds). 1991. Readings in Clinical Spectrograph of Speech. San Diego:
Singular Publishing Group.
Cameron, D.A.1892. Arabic-English Vocabulary for the Use of English Students of Modern
Egyptian Arabic. London: Gilbert and Irvington.
Cenoz, J. and F. Genesee. 2001. Trends in Bilingual Acquisition. Amsterdam: John Benjamins.
Crystal, David. 2003. The Cambridge Encyclopedia
edn. Cambridge: Cambridge University Press.
of
the
English
Language. 2nd
Fant, G. 1960. Acoustic Theory of Speech Production. The Hague: Mouton.
Fry,
D.
B.
1976. Acoustic
Phonetics:
Readings. Cambridge: Cambridge University Press.
a
course
of
Basic
Gairdner, W. H. T. 1925. The phonetics of Arabic. A phonetic inquiry and
Practical manual for the pronunciation of Classical Arabic and of one colloquial
(Egyptian), London: Humphrey Milford, Oxford University Press.
Johnson, K. 2003. Acoustic and Auditory Phonetics. Oxford: Blackwell.
Ladefoged, Peter. 1962 Elements of Acoustic Phonetics. Chicago: University of Chicago Press.
–-1967. Three Areas of Experimental Phonetics. Oxford: Oxford University Press. ––2006
A Course in Phonetics.5th edition. Boston: Thomson Wadsworth, 189.
Ladefoged, Peter. J. DeClerk, M. Lindau and G. Papcun.1972 An Auditory Motor Theory of Speech
Production. UCLA Working Papers in Phonetics
Lehiste, Ilse. ( ed.).1967, Readings in Acoustic Phonetics. Cambridge, MA: MIT Press.
Mitleb, Faris. 1984. ‘Vowel length contrast in Arabic and English: a spectrographic test’. Journal
of Phonetics 12, 229–235.
Nasr, Raja T. 1966. Colloquial Arabic: An Oral Approach. Beruit: Librairie du Liban.
Norlin, Kjell.1987. ‘A phonetic study of emphasis and vowels in Egyptian
Arabic’. University of Lund, Department of Linguistics and Phonetics, Working Papers, 30,
Ohala, John J. 1983. ‘ The origins of sound patterns in vocal tract constraints’ in Peter F.
MacNeilage. (ed.) The Production of Speech. New York: Springer-Verlag, , 189–216.
Watson, Janet C. E. (2002) The Phonology and Morphology of Arabic, Oxford University Press,
The Phonology of the World's Languages.
Download