SST 2010
Pharyngealization in Assiri Arabic: an acoustic analysis.
Saeed Shar, John Ingram
School of Languages and Cross Cultural Studies, University of Queensland, Brisbane, Australia
[email protected], j.ingram @uq.edu.au
capital of the southern region of Saudi Arabia. The Assiri
dialect serves as a standard dialect for speakers of other local
The study is part of a wider
dialects in the region.
investigation of the acoustic and articulatory mappings of
guttural sounds in Assiri Arabic, using MRI.
Abstract
Five native speakers of Assiri Arabic read word lists
comprising contrasting pairs of plain and emphatic
(pharyngealized) consonants in three vocalic environments
(/i/, /a/, /u/). Although considerable individual variation in
expression was apparent from auditory and acoustic analysis
of the tokens, the most consistent acoustic correlate of the
plain-emphatic contrast appeared to lie in the transition phase
of the accompanying vowel formant trajectories, involving a
raising of F1 and lowering of F2. A statistical analysis
(ANOVA) of the formant targets involving interactions of
pharyngealization with, consonant , vowel, and subject factors
is presented, with discussion of implications for articulatory
targets for the plain-emphatic contrast, as assessed by MRI
imaging in the same group of subjects.
2.
2.1.
2.2.
Stimuli
The target sounds to be examined in this study are the set of
Introduction
Arabic emphatic consonants; [sˤ], [tˤ], [ðˤ] and [dˤ], and their
plain ones; [s], [t], [ð] and [d]. A set of real Arabic words were
chosen and arranged in four groups according to minimal pair
contrast. Target sounds are pronounced in different positions;
initially and medially. Each minimal pair is followed by
vowels [a] [i] and [u], which are a subgroup of the Arabic
vowels. In this acoustic experiment, there are eight emphatics
and plain sounds in two different positions followed by three
vowels and uttered by five subjects. This yielded 240 tokens
for measurement (2 [plain-emphatic] x 4 [consonants] x 2
[initial-medial] x 3 [vowels] x 5 [subjects]).
All dialects of Arabic have minimal or near-minimal pairs of
contrasting ‘emphatic’ vs. ‘plain’ consonants. The total
number of emphatics in Standard Arabic is four; [sˤ], [tˤ], [ðˤ]
and [dˤ],. Emphatics contrast with their plain equivalent
sounds /s/, /t/, /ð/ and /d/, as shown in the following examples
of minimal pairs:
/s/ & /sˤ/ /sa:r/ ‘walked’ vs. / sˤa:r/ ‘became’
/t/ & /tˤ/ /ti:n/ ‘fog’
vs. /tˤi:n/ ‘mud’
/ð/ & /ðˤ/ /ðall/ ‘cringed’ vs. /ðˤall/ ‘still’
/d/ & /dˤ/ /da:l/ ‘guiding’ vs. /dˤa:l/ ‘mislead’
The emphatic consonants are usually characterised phonetically as `pharyngealized’, involving a secondary constriction
in the pharynx, produced more or less simultaneously with the
primary place of articulation for the consonant, but spreading
into the adjacent vowel, predominantly rightward but bidirectionally, and in some cases beyond the target syllable
containing the pharyngealized consonant. The well known
acoustic effect of emphatics on neighbouring vowels is, ‘emphasis spread’, and the most reported influence is raising F1
and lowering F2 (either at the transition only or through the
whole vowel) because of retracted tongue root and raising
larynx (Shahin,1997, Trigo, 1991, Zawaydeh,1999, Muqbil,
2006) . Al-Ani (1970) reported large F2 drops in vowels following emphatics consonants as opposed to non-emphatic
ones; the vowel [a] shows the greatest fall of F2. The same
result was found by Ghazeli (1977), Also he found that F1 of
all vowels is raised by emphatics. Younes (1982) found similar
patterns in Northern Palestinian. Zawaydah (1999) confirmed
this result; furthermore, she found that F2 is lowered more by
emphatics than by gutturals which are primarily produced in
the pharyngeal region.
This study investigates the acoustic features that distinguish
emphatic consonants from their plain counterparts in Assiri
Arabic. This dialect is spoken in the city of Abha which is the
ISBN 978-0-9581946-3-1 © 2010 ASSTA
Accepted after peer review of full paper
Subjects
The subjects were five male native speakers of the Assiri
dialect of Arabic, one of them is the first author of this paper.
The age range of the subjects is 30 - 35 years. All have normal
neurological history and no apparent speech or hearing
disorders. All recorded data were taken in Australia using the
same computer, microphone and other settings. During
recording, subjects were asked to read words casually as
normal speech.
Index Terms: Arabic, emphatics, pharyngealization, formants,
MRI.
1.
Method
2.2.
Procedures
Subjects were recorded using a microphone attached directly
to the computer, and acoustic data were stored as .wav files
digitized at 44.1 KHz on a personal computer using the
“Praat” speech manipulation software to measure formant
frequencies. Measurements were stored in an Excel
spreadsheet and statistically analysed using the S-plus
statistical graphics package.
Praat formant tracking was used with standard settings to
estimate the formant trajectories of F1 and F2. The whole
trajectory of the vowel following the consonant (both
transition and steady state portions) was used in calculating
the mean value of a formant trajectory.
5
14 -16 December 2010, Melbourne, Australia
statistically significant (p < .002) was relatively small in
comparison with the effect of vowel type or pharyngealization.
A posteriori paired comparisons (t-tests) showed that there
were significant differences between pharyngealized and plain
consonants for the voiced but not for the voiceless consonants
([ðˤ] - [ð]: t=3.7323, p = 0.0009 and [dˤ] - [d]: t= 2.3181, p =
0.028).
The two-way interaction plot, (see Fig.3) between
consonant and vowel type was difficult to interpret.
/sˤ a b
i r/
/s a b i r/
Figure 1: Spectrogram of / sˤabɪr/ and /sabɪr/, with
pharyngealized [sˤ] and plain [s] in initial position.
3.
Results
A graphical summary of the formant changes to the following
vowel associated with pharyngealization of the preceding
consonant is shown in Figure 2 (end of report), where formant
trajectories of individual subjects are color coded and the
formant change from plain to emphatic form is indicated by
the direction of the arrowhead. A series of 3-way ANOVAS
was conducted on the formant measurements, to assess the
statistical significance of factors influencing F1 and F2
measurements.
3.1.
Fig. 3 Interaction of consonant and vowel on F2 transition
The interaction plot (Fig. 3) suggested that the phonetic target
for the /i/ vowel is quite centralized following the voiced interdental fricative /ð/ (i.e.: / ði/ is realized as [ðɨ]). This lowering
of F2 for the interdental fricative might be expected to
influence F2 lowering for plain – emphatic contrast for this
consonant, thereby yielding a 3-way interaction of Consonant
by Vowel by Pharyngealization. However, no such significant
3-way interaction was observed.
The 2-way interaction of Pharyngealization by Vowel type
(see Fig. 4) and post-hoc t-test comparisons showed a greater
lowering of F2 under pharyngealization for the low vowel [a]
than the high vowels [i] and [u]. ([aˤ] – [a]: t=16.7761, p =
0.0000, [iˤ] - [i]: t= 2.3374, p = 0.0248 and [uˤ] - [u]: t=
2.3026, p = 0.0269).
ANOVA (1) Effects of Pharyngealization
The first series of ANOVAs was undertaken to assess the main
and interaction effects of Pharyngealization, Consonant, and
Vowel separately on the dependent variables F1 and F2. The
summary findings are presented in Tables 1 and 2.
Table 1 ANOVA Dependent variable: F1
Pr(F)
0.0000006
0.2412642
0.0000000
0.8633677
0.9734836
0.1081491
0.7771630
Vowel
2000
Df S of Sq
F Value
1 106327
28.6949
3
15801
1.4215
2 604222 216.4699
3
2744
0.2469
2
199
0.0269
6
39908
1.7950
6
11989
0.5392
96 355720
i
u
a
Df S of Sq
F Value
1 2381774
69.5563
3
573508
5.5828
2 14806062 216.1949
3
83155
0.8095
2
513505
7.4981
6
1300816
6.3314
6
117576
0.5723
96
3287270
1400
Pr(F)
0.0000000
0.0014242
0.0000000
0.4916597
0.0009424
0.0000122
0.7515039
1200
Source
phary.
cons
vow
phary:cons
phary:vow
cons:vow
phary:cons:vow
Residuals
mean of F2
Apart from the expected main effects of vowel ([i], [a], [u])
and pharyngealization, there were no other significant main or
interaction effects upon F1. Pharyngealization raised F1 by a
mean value of 59 Hz, a highly significant effect (p<.00001).
Table 2 ANOVA Dependent variable: F2
1600
1800
Source
phary.
cons
vow
phary:cons
phary:vow
cons:vow
phary:cons:vow
Residuals
nph
ph
pharyng
Figure 4. Interaction of Pharyngealization by Vowel on F2.
This additional lowering of F2 in the low vowel [a] for
pharyngealized consonants was not unexpected (see
discussion 3.3).
In addition to the highly significant main effects of vowel
and pharyngealization which were observed upon F1, F2
showed a main effect of consonant and two significant 2-way
interaction effects: 1) an interaction effect of consonant and
vowel type, and 2) an interaction of pharyngealization and
vowel. The main effect of consonant type upon F2, though
3.2.
ANOVA (2) Individual Differences
Inspection of the formant plots (Fig. 1) and MRI imaging data
(discussed below) indicated that individual differences in
articulation targets may be present in the data. Consequently, a
6
second series of 3-way ANOVAS was conducted with F1 and
F2 as dependent variables and subjects (df = 4), vowel and
pharyngealization as factors. The results are summarized in
Tables 3 and 4.
The effect of pharyngealization on F2 was stronger for
the low vowel [a] than it was for [i] or [u]. The change in
F2 for /a/ is perceptually more salient than for the high
vowels, producing a categorically distinct vowel quality
[ɑ].
No significant interaction effects involving Subjects by
Pharyngealization were found, suggesting a consistent
acoustic strategy across speakers for implementing the
plain – emphatic contrast. This finding is of interest
because it contrasts with preliminary articulatory
observations based on MRI recordings of the subjects and
impressionistically based auditory observations of the
authors which suggest substantial inter-speaker
variability in the production of emphatic consonants.
Table 3 ANOVA Dependent variable: F1
Source
su
phary
vow
su:phary
su:vow
phary:vow
su:phary:vow
Residuals
Df S of Sq
4
248467
1
106327
2
1604222
4
1429
8
47568
2
199
8
2830
90 125869
F Value
44.4154
76.0269
573.5352
0.2555
4.2516
0.0712
0.2530
Pr(F)
0.00000
0.00000
0.00000
0.90559
0.00023
0.93130
0.97888
4.
The three highly significant main effects of subject,
pharyngealization, and vowel were to be expected, given the
results reported previously and the fact that formant
measurements had not been normalized for individual
differences in vocal tract size. Apart from a statistically
significant subject by vowel interaction, whose magnitude was
small in relation to the main effects, none of the other
interactions involving the subject factor were statistically
significant. In short, no significant interactions involving
subject and pharyngealization were found for F1.
This study confirms and extends findings of previous studies
(Shahin, 1997, Zawaydah. 1999) that raising of F1 and lowering
of F2 in the accompanying vowel constitute a robust and
relatively context invariant acoustic cue to the contrast
between emphatic and plain contrasts, which are ubiquitous in
regional varieties of modern spoken Arabic. A major challenge
for future work is to reconcile this finding with the manifest
phonetic variability which is apparent from impressionistic
and instrumental articulatory observations of emphatic – plain
contrasts in Assiri Arabic consonants.
Table 4 ANOVA Dependent variable: F2
source
su
phary
vow
su:phary
su:vow
phary:vow
su:phary:vow
Residuals
Df
4
1
2
4
8
2
8
90
Sum of Sq
682714
2381774
14806062
58473
981299
513505
168250
3471588
F Value
4.424
61.746
191.921
0.379
3.180
6.656
0.545
5.
Pr(F)
0.00260
0.00000
0.00000
0.82311
0.00320
0.00201
0.81939
Acknowledgements
This research by the first Author is supported by King Khalid
University (Abha city). The first author is grateful to his
supervisor Dr. John Ingram for his continuous support and
encouragement. We also thank our subjects for their role in
this experiment.
6.
There were significant interaction effects involving the
subject factor in terms of effects upon F1 and F2. The
significant vowel by pharyngeal interaction upon F2 (which
we observed previously in ANOVA 1) involved a greater
lowering of F2 under pharyngealization for [a] than for [i] or
[u].
3.3.
Implications
References
1. Al-Ani, S. 1970. Arabic phonology; an acoustical and physiological
investigation. The Hague: Mouton.
2. Butcher, A. & Ahmad, K. Some acoustic and aerodynamic
characteristics of pharyngeal consonants in Iraqi Arabic. Phonetica
1987;44:156-172
3. Ghazeli, Salem. 1977. Back consonants and back articulation in
Arabic. Ph.D. Dissertation, The University of Texas; Austin.
4. Muqbil, M. 2006. Phonetics and phonological aspects of Arabic
amphatics and gutturals. PhD Dissertation, University of WisconsinMadison.
5. Shahin, K. 1997. Postvelar harmony: an examination of its bases
and crosslinguistic variation, Ph.D. Dissertation, University of British Columbia.
6. Trigo. L. 1991. On pharynx-larynx interactions. Phonology 8: 113136.
7. Younes, M. 1982. Problems in the segmental phonology of
Palestinian Arabic, Ph.D. Dissertation, University of Texas at
Austin.
8. Zawaydah, B. A. 1999. The phonetics and phonology of gutturals in
Arabic. Ph.D. Dissertation; Indiana University, Indiana.
Discussion of ANOVA results
In summary, results from the statistical analysis of F1 and F2
changes to vowel transitions accompanying pharyngealization
of consonants in Assiri Arabic indicated:
A significant and robust effect of raising of F1 and
lowering of F2 for emphatic consonants expressed in the
following vowel transition. The strength of this main
effect of pharyngealization across the three vowels and
the four consonants attests to the robustness of this
acoustic cue for the plain – emphatic contrast in Assiri
Arabic.
7
Figure 2: F1 – F2 formant change of vowels [a], [i] and [u] from plain to emphatic (pharyngealized) consonants of each
subject; s1: red, s2: green, s3: blue, s4: brown, and s5: orange.
Appendix
sˤabir- sabir
Ɲiƞsˤar- Ɲiƞsar
maqasˤ-ras
sˤulb- sunnah
sˤiam-sinan
yansur- yuhajir
yasˤiir- yasiir
lusˤuusˤ- fluus
qamisˤ-tamiis
tˤalib-yaƝib
qitˤar-ƞitab
balatˤ- hibat
tˤub-tum
butˤuun-yaƝtuun
χutˤuutˤ- quut
tˤiin-tiin
yatˤiir- yatiim
laqiitˤ- ƞamiit
±ˤahir- ±akir
ma±ˤalim-yu±akir
mala±ˤ- mala±
±ˤulm-±ul
±ˤil- ±immah
yan±ˤur- ya±uq
yu±ˤil-yu±il
yalu±ˤ- yalu±
dˤaal- daal
nidˤal-midad
ƞadˤ- ƞad
dˤulmah- durrah
yadˤum-yadul
baƞudˤ- rudud
dˤiaƝ-dimaƝ
yadˤiƞ- Ɲadim
yabidˤ- ƞamid
8
qai±ˤ- nabi±