Download as pdf or txt
Download as pdf or txt
You are on page 1of 16

Poznań Studies in Contemporary Linguistics 52(3), 2016, pp.

503–518
© Faculty of English, Adam Mickiewicz University, Poznań, Poland
doi: 10.1515/psicl-2016-0019

AN ACOUSTIC DESCRIPTION OF SPECTRAL


AND TEMPORAL CHARACTERISTICS
OF AZERBAIJANI VOWELS

PAYAM GHAFFARVAND MOKARI* AND STEFAN WERNER


University of Eastern Finland, Joensuu
*payam.ghaffarvand@uef.fi

ABSTRACT

This paper presents an acoustic description of the nine Azerbaijani vowels; investigating
the underlying acoustic and temporal characteristics of its vowel system. We explored
acoustic and temporal parameters including: the first three formants (F1, F2 and F3),
fundamental frequency (F0) and duration of the vowels. Participants in this study were
20 male and 23 female Azerbaijani speakers with a Tabrizi dialect. They were asked to
utter three repetitions of the nine Azerbaijani vowels in three natural word contexts, em-
bedded in carrier sentences. Results showed that the [ɯ] and [œ] vowels had a large
overlap in the F1–F2 vowel space. Further analysis suggested that F3 is an important cue
in discrimination of this vowel pair. Vowel-intrinsic duration effect seemed to be rela-
tively strong in Azerbaijani. Other universal features also were found in the production
of Azerbaijani vowels: low vowels and female speakers had lower F0 values. Surprising-
ly, in contrast with previous results for most languages, the average duration of Azerbai-
jani vowels was greater in males than females. The results of this study define the acous-
tic vowel-space of the Azerbaijani language and develop a database for further compari-
sons and investigations.

KEYWORDS: Acoustic analysis; vowels; formants; Azerbaijani.

1. Introduction

The Azerbaijani, Azerbaijanian or Azeri language (henceforth Azerbaijani) is


mainly spoken in Azerbaijan and Iran (Johanson 2010). After Farsi as the offi-
cial language, Azerbaijani, with approximately 15–20 million speakers, has the
most speakers in Iran (Crystal 2010). Most Azerbaijani speakers inhabit in the
four provinces of the northwestern part of Iran. There are various dialects of
Azerbaijani. Among these dialects is the Tabrizi dialect; which serves as the

Brought to you by | University of Liverpool Sydney Jones Library


Authenticated
Download Date | 12/17/16 10:22 PM
504 P. Ghaffarvand Mokari and S. Werner

norm for Iranian Azerbaijani speakers (Menges 1951; Johanson 1998). Azerbai-
jani has nine vowels with no length distinction and no diphthongs; including
four high vowels [i y, ɯ, u], three mid vowels [e, œ, o] and two low vowels [æ,
ɑ].
The most common way of examining the characteristics of vowel systems is
in terms of their formant frequencies. Generally, the first three formants of the
vowels are the most important indicators providing information for distinguish-
ing vowels from each other (Hagino et al. 2008). There is a close relation be-
tween first formant frequency (F1) and articulatory and/or perceptual dimension
of vowel height. The second formant frequency (F2) corresponds with the place
of maximal constriction during the production of the vowels (Wang and Van
Heuven 2006).
There are common features found cross-linguistically in many languages.
Vowel-intrinsic pitch (Whalen and Levitt 1995) is one of these features, which
shows that the F0 (in an accented syllable) tends to be higher as the vowel is
higher. For vowel-intrinsic duration, Solé (2007) observed that mean durations
of vowels vary from vowel to vowel, in correlation with vowel height, i.e., the
lower the vowel the greater its “intrinsic” duration. Gender-dependent duration
(Simpson and Ericsdotter 2003) is another feature that shows female vowel du-
rations are normally longer and also female speakers produce long and short
vowel categories with greater difference. Gender-dependent size of the vowel
space (Simpson 2001; Escudero and Boersma 2009) explains that female speak-
ers tend to have a larger vowel space than men, even along logarithmic scales,
i.e. in terms of a ratio of the F1 values of lowest versus highest vowels. The
cause of this effect has been sought in the physiology (Simpson 2001). Skewed
symmetry in F1 between front and back vowels (Peterson and Barney 1952;
Clopper et al. 2005; Escudero and Boersma 2009) is when there are front vow-
els with back vowel counterparts between which we can identify pairings, e.g.
pair of high vowels. One of the research questions is whether or not these trends
hold for production of Azerbaijani vowels.
There are many studies investigating the acoustic characteristics of vowels.
One of the best-known studies on formant measurements of vowels was for
American English conducted by Peterson and Barney (1952). To the best of our
knowledge, the only study on acoustic description of Azerbaijani vowels has
been set by Mozaffarzadeh Peivasti (2012) which is limited in various aspects.
Firstly, Mozaffarzadeh Peivasti (2012) pooled over the formant values of the
male and female participants and reported one average value of F1 and F2 for
each vowel. This does not seem a proper approach as it is well known that ana-
tomical differences between male and female vocal tracts or vocal cords, make

Brought to you by | University of Liverpool Sydney Jones Library


Authenticated
Download Date | 12/17/16 10:22 PM
An acoustic description of Azerbaijani vowels 505

differences in acoustic vowel space dimension (Simpson 2002). Diehl et al.


(1996) suggest that there is a non-uniform scaling between male and female
vowels, stating that adult females’ vocal tracts tend to be shorter than those of
adult males, which leads to higher frequencies for female formants. Therefore,
the precise values for vowel characteristics may not be adequately represented
in the literature. Mozaffarzadeh Peivasti (2012) only provided the information
about the first two formants of the vowels and the information about the F0 and
the F3 are missing. Temporal characteristics of the vowels have also been found
to play an important role in vowel perception (Nearey 1989; Strange 1989),
which also have not been examined in the study of Mozaffarzadeh Peivasti
(2012).
The present study has tried to improve on the existing limitations in the lit-
erature by presenting a comprehensive acoustic description of Azerbaijani vow-
els. The universal trends in production of Azerbaijani vowels were also studied.
Finally, we aimed to develop a more accurate database for further comparisons
and investigations.

2. Materials and methods

2.1. Subjects

The participants were volunteers that were selected based on the following crite-
ria: born and grown up in Tabriz and using the Tabrizi dialect for daily commu-
nications, aged 20-35 years old and were university students or graduated stu-
dents with no history of speech or language disorders. In total there were 43
subjects including 20 male and 23 female. They had a mean age of 26.6 ±3.8
years.

2.2. Stimuli and procedure

The nine Azerbaijani vowels (æ, ɑ, o, e, œ, ɯ, u, i, y) were embedded in fixed


carrier phrase “burdaki, /cVc/ kælmæsidi”, that were presented to the speakers
on a printed list. A phrase including a non-target word was added to the begin-
ning and another to the end of the list to reduce the list effect. Each target vowel
was produced in a carrier phrase. The target words were in consonantal contexts
/bVd/ that “V” refers to the target vowel in the target words. In cases that there
were no existing natural Azerbaijani words with /bVd/ context we used other

Brought to you by | University of Liverpool Sydney Jones Library


Authenticated
Download Date | 12/17/16 10:22 PM
506 P. Ghaffarvand Mokari and S. Werner

voiced stop consonant contexts. Participants were asked to read the list of carri-
er phrases three times. Also participants were allowed to repeat the phrases
when making a mistake. A total of 1161 vowel tokens were recorded (9 vowels
× 3 repetitions × 43 speakers). The misproduced and creaky tokens were
dropped out from the recordings. Finally we had 1103 tokens for analysis.

2.3. Recordings

Subjects’ utterances were recorded digitally in a quiet room using a M-Audio


USB Producer microphone connected to a laptop. The microphone-mouth dis-
tance was constant at 18–20 cm. Recordings were saved at 44,100 Hz sampling
rate and at 16-bit resolution. The recordings were manually segmented and la-
beled using Praat speech analysis software, version 5.4.10 (Boersma and Ween-
ink 2015).

2.4. Measurements

2.4.1. Formant frequencies

Formant frequency analysis was performed using the Burg algorithm available
in Praat with its default standard settings for females (range of 5,500 Hz for five
formants) and males (range of 5,000 Hz for five formants). F1, F2 and F3 values
were then measured for each vowel by taking an average of the central 40% of
the vowel centered around the midpoint. Results were checked for probable
formant tracking errors by hand and manually modified where necessary.

2.4.2. Fundamental frequency

For each vowel token, F0 was measured automatically by the auto-correlation


method in Praat at 25%, 50% and 75% of the vowel duration. The pitch range
for the analysis was set to 50–300 Hz for men and 100–500 Hz for women. The
obtained values were checked for outliers (cases > 1.5 IQR + Q3, cases < Q1 –
1.5 IQR) for each participant. The outliers were verified manually and unrelia-
ble values were replaced by the mean F0 of the target vowel produced by the
speakers.

Brought to you by | University of Liverpool Sydney Jones Library


Authenticated
Download Date | 12/17/16 10:22 PM
An acoustic description of Azerbaijani vowels 507

2.4.3. Duration

The interval between the onset and offset of the glottal vibrations of the vocalic
portion of the /cVc/ was considered as vowel duration. Boundaries for the dura-
tion of each token were labeled manually in the digitized speech wave using the
Praat (Boersma and Weenink 2015). The beginning and end of segment bound-
aries were taken at the zero crossings. The speech rate of the speakers was
measured for each carrier phrase production time excluding the target vowels.

2.5. Statistical analysis

Descriptive statistics were performed to describe mean and standard deviation


values for output measures (F1, F2, F3, duration and F0). The linear mixed
model analysis was used to measure the effects of vowel category, gender and
their interaction on output measures.

3. Results

3.1. Descriptive statistics

Table 1 represents the gender-specific average values for the first three formants
of the central 40% of the tokens for the nine Azerbaijani vowels. The mean val-
ues of F0 and duration are shown in Table 2.
As shown in Table 1, the highest and the lowest mean F1 were for the [æ]
and [y] vowels and the highest and the lowest mean F2 were for the [i] and [o]
vowels, respectively in both genders. The highest mean F3 values were for the
[o] in males and the [i] in females. As presented in Table 2, the highest and low-
est mean F0 were for the [u] and [ɑ] vowels, respectively. The average duration
of all vowels seems to be higher for males.

3.2. Formant frequencies

Distribution of the vowels in a F1 by F2 plane is given in Figure 2 that provides


an overview of the acoustic vowel space for Azerbaijani for male and female

Brought to you by | University of Liverpool Sydney Jones Library


Authenticated
Download Date | 12/17/16 10:22 PM
508 P. Ghaffarvand Mokari and S. Werner

Table 1. F1, F2 and F3 (Hz) average (standard deviation)


for Azerbaijani vowel tokens.

F1 F2 F3
Vowel
Male Female Male Female Male Female
æ 605 (45) 788 (71) 1517 (81) 1782 (180) 2398 (163) 2769 (396)
ɑ 567 (42) 653 (75) 1088 (121) 1243 (104) 2453 (259) 2906 (205)
e 398 (33) 489 (56) 1851 (109) 2163 (128) 2517 (104) 2903 (248)
œ 399 (31) 467 (54) 1420 (88) 1527 (145) 2234 (138) 2644 (208)
o 425 (31) 485 (46) 895 (98) 975 (80) 2552 (189) 2883 (158)
i 296 (30) 367 (54) 1898 (120) 2218 (307) 2532 (156) 2940 (193)
y 300 (40) 338 (47) 1561 (140) 1648 (194) 2243 (125) 2575 (170)
ɯ 355 (39) 445 (44) 1364 (167) 1526 (158) 2458 (135) 2903 (210)
u 336 (29) 401 (52) 1004 (109) 1112 (148) 2428 (170) 2805 (158)

Table 2. F0 (Hz) and duration (ms) average (standard deviation)


for Azerbaijani vowel tokens.

F0 Duration
Vowel
Male Female Male Female
æ 147 (24) 239 (24) 169 (50) 151 (23)
ɑ 137 (17) 224 (22) 161 (47) 150 (31)
e 142 (21) 227 (23) 157 (51) 134 (25)
œ 142 (19) 231 (27) 161 (43) 150 (28)
o 145 (20) 233 (24) 171 (49) 160 (28)
i 142 (21) 233 (22) 146 (46) 130 (33)
y 146 (22) 240 (22) 150 (55) 137 (28)
ɯ 140 (17) 226 (25) 134 (53) 127 (28)
u 150 (24) 240 (26) 166 (56) 143 (31)

formants, respectively. Spreading ellipses were drawn at ± 2 SD. The ellipse


theoretically contains 95.4 percent of the tokens belonging to the shown catego-
ry.

Brought to you by | University of Liverpool Sydney Jones Library


Authenticated
Download Date | 12/17/16 10:22 PM
An acoustic description of Azerbaijani vowels 509

Figure 1. IPA chart of Azerbaijani vowels.

Figure 2. Distribution of the vowels in F1 × F2 (Bark) space


with ellipses representing two standard deviations from the mean.

3.3. Linear Discriminant Analysis

A Linear Discriminant Analysis (LDA; Klecka 1980; Weenink 2006, Wang and
Van Heuven 2006) was used to observe the degree of correct classification in
accurate distinction of the nine Azerbaijani vowels. As mentioned by Wang and
Van Heuven (2006), the more distinct the categories are in the dataset, the fewer
the number of classification errors yielded by the algorithm. In Table 3 (a), the
discriminant functions are presented based on linear combinations of two spec-
tral parameters F1, F2. Afterwards, we ran the LDA algorithm two more times
with extended set of predictors by adding once the vowel duration and once the
F3 to the previous parameters (Table 3 b,c). Errors less than 5% are not present-
ed in the tables.

Brought to you by | University of Liverpool Sydney Jones Library


Authenticated
Download Date | 12/17/16 10:22 PM
510 P. Ghaffarvand Mokari and S. Werner

Table 3. Percentage of classification of Azerbaijani vowels based on LDA.

(a) F1 and F2 as predictors.

æ ɯ œ ɑ e i o u y
æ 86.6 6.3 7.1
ɯ 37.1 32.3 11.3 16.9
œ 43.3 47.8
ɑ 84 14.5
e 5.6 72 13.6 8.8
i 9.6 78.4 12
o 6.8 82.9 10.3
u 8.8 19.6 69.6
y 11.9 6.8 73.7

(b) F1, F2 and duration as predictors.

æ ɯ œ ɑ e i o u y
æ 89 5.5 5.5
ɯ 42.7 28.2 9.7 16.9
œ 35.8 56
ɑ 83.2 15.3
e 6.4 72 12 8.8
i 8.8 79.2 12
o 6.8 84.6 8.5
u 6.9 18.6 70.6
y 11.9 6.8 73.7

The overall percentage of correct classification when predictor parameters were


only F1 and F2 was 70.1%, while with addition of F3 and duration as predictor
parameters, the overall percentags of correct classification were 72.3% and
77.3%, respectively. With F1 and F2 as predictors, the largest error was in dis-
crimination of [ɯ] and [œ] vowels. After including duration as the predictive
parameter, the classification of [ɯ] and [œ] vowels improved by 5.6% and
8.2%, respectively. But when the predictors were F1, F2 and F3, the classi-

Brought to you by | University of Liverpool Sydney Jones Library


Authenticated
Download Date | 12/17/16 10:22 PM
An acoustic description of Azerbaijani vowels 511

Table 3. Percentage of classification of Azerbaijani vowels based on LDA.

(c) F1, F2 and F3 as predictors.

æ ɯ œ ɑ e i o u y
æ 91.3 7.1
ɯ 60.5 12.9 11.3 13.7
œ 16.4 75.4
ɑ 87.8 9.9
e 5.6 70.4 13.6 9.6
i 8 80 12
o 88 8.5
u 7.8 19.6 70.6
y 10.2 6.8 6.8 70.3

fication of [ɯ] and [œ] vowels improved by greater degrees of 23.4% and
27.6%, respectively.

3.4. F1

The linear mixed model analysis revealed a significant effect of gender and
vowel category on F1 (F[1,41] = 82.53, p < 0.001; F[8,1052] = 1081, p <
0.001). The [æ] vowel had the highest F1 value for both genders. However in
contrast to males, females produced [e] higher than [o] and [oe], and [i] higher
than [y]. The overall average of F1 for the females was about 88 Hz higher than
the males. The range of F1 for females was broader by 141 Hz.

3.4.1. F1 categories

Linear mixed model analysis was used to compare 20 adjacent contrasts. Bon-
ferroni correction was applied and significance level was set at α = 0.0025 (α =

Brought to you by | University of Liverpool Sydney Jones Library


Authenticated
Download Date | 12/17/16 10:22 PM
512 P. Ghaffarvand Mokari and S. Werner

0.05 / 20 contrast sets). There was no significant difference between pairs [i–y],
[e–œ] and [o–œ].

3.5. F2

As expected, the analysis also revealed a significant effect of gender and vowel
category on F2 (F[1,41] = 127.6, p < 0.001; F[8,1052] = 1002, p < 0.001 vow-
el). The highest and lowest F2 were for [i] and [o] vowels, respectively for both
genders. However, in contrast to males, females produced [æ] with higher F2
than [y]. Females had higher F2 in all vowels (Fig. 2).

3.5.2. F2 categories

Linear mixed model analysis was used to compare 20 adjacent contrasts. Bon-
ferroni correction was applied and significance level was set at α = 0.0025 (α =
0.05 / 20 contrast sets). There was no significant difference between the pairs [i–
e] and [œ–ɯ].

3.6. F0

The analysis revealed a significant effect of gender (F[1,41] = 128.9, p < 0.001)
and also a significant but small effect of vowel category on F0 (F[8,1052] =
25.50, p < 0.001). The mean F0 was 232.4 Hz and 142.2 Hz for females and
males, respectively. The highest F0 was for [u] vowel and the lowest F0 was for
[ɯ] and [a] vowels.

3.7. Duration

Results showed a significant effect of gender and vowel category on duration


(F[1,1085] = 38.01, p < 0.05; F[8,1085] = 9.70, p < 0.05). The mean duration
was 142.5 ms and 157.2 ms for females and males, respectively. The interaction
of the gender × vowel category was insignificant (F[8,1085] = 0.56, p = 0.80).
The [o] vowel had the highest mean duration and the [ɯ] vowel had the lowest
mean duration for both genders (Fig 3).

Brought to you by | University of Liverpool Sydney Jones Library


Authenticated
Download Date | 12/17/16 10:22 PM
An acoustic description of Azerbaijani vowels 513

180
170
160
150
140
130
120
y u ɯ œ e o ɑ æ

Males Females

Figure 3. Mean duration of Azerbaijani vowels for males and females.

There were no significant differences between speech rate of males and females
(p > 0.05).

4. Discussion

The purpose of this study was to explore the acoustic characteristics of Azerbai-
jani vowels and to provide an overview of the vowel characteristics in three dif-
ferent contexts.

4.1. Formant frequencies

Results showed that women tend to have higher F1 values than men. This is in
line with previous studies (e.g., Peterson and Barney 1952; Koopmans-van
Beinum 1980; Escudero and Boersma 2009). It has been shown that the main
reason for the aforementioned inter-gender difference is the differences in vocal
tract length between women and men.
It is hard to compare our results to those of Mozaffarzadeh Peivasti (2012)
because she has not reported the measurements separately for each gender. Con-
trary to the results of the mentioned study, our results suggested a very small F2
difference between [ɯ] and [œ] vowels. Also our results suggested a smaller dif-

Brought to you by | University of Liverpool Sydney Jones Library


Authenticated
Download Date | 12/17/16 10:22 PM
514 P. Ghaffarvand Mokari and S. Werner

ference between F2 of [i] and [e] vowels. Despite she reported almost no differ-
ence between F2 values of [u] and [o] vowels, our results revealed about 100 Hz
difference between the F2 values of these two vowels. As mentioned before,
there is no uniform scale between the formant values of males and females
(Diehl et al. 1996). Therefore, the differences between the results of the two
studies could be due to the fact that Mozaffarzadeh Peivasti (2012) pooled both
genders’ formant values together without using any normalization procedure,
which in turn could lead to a non-precise representation of the vowel space.
According to the descriptive analysis, the [ɯ] and [œ] vowels were almost
equally peripheral in their F1–F2 vowel space. This issue can raise the question
that how this pair can be discriminated by the listeners. On one hand, our results
revealed that there is about 30 ms durational difference between the [ɯ] and [œ]
vowels. Further addition of the duration to F1 and F2 as the predictors in LDA
made a small boost in the discrimination of [ɯ] and [œ] vowels. On the other
hand, there was a substantial difference between the F3 values of this pair. Fur-
ther inclusion of F3 to the predictive factors of LDA improved the correct clas-
sification of this pair pretty dramatically. Therefore, it can be assumed that F3
plays an important role in discrimination of this vowel pair. Regarding the rela-
tionship between F3 and lip configuration, it is suggested vowels’ peripherality
of some languages to be defined in terms of both tongue position and lip con-
figuration (e.g. Reetz and Jongman 2011). Norwegian [i] and [y] are equally pe-
ripheral in an F1–F2 vowel space; they differ only with respect to F3 frequency
(Best and Faber, 2009). Thus, it seems for this contrast like the Azerbaijani [ɯ]
and [œ], F1 and F2 are inadequate for describing the peripheral difference of
these vowels.
Our results showed the Azerbaijani vowels’ F1 space size is 1.20 times larg-
er for females than for males. Escudero et al. (2009) reported a ratio of 1.20 for
the Brazilian Portuguese, and a ratio of 1.09 for the European Portuguese
speakers. Peterson and Barney (1952) and Hillenbrand et al. (1995), reported
the ratio as 0.97 for the American English speakers. Adank et al. (2004) also re-
ported a ratio of 1.26 for Northern Standard Dutch speakers and the ratio of 1.03
for the Southern Standard Dutch. The Azerbaijani vowel values are closer to the
reported ratio for Brazilian Portuguese.
Azerbaijani female speakers not only have larger F1 space sizes than men,
they also have larger F2 space sizes. The average Azerbaijani female-to-male F2
space size ratio is 1.12. Escudero and Boersma (2009) reported a ratio of 1.17
for the Portuguese speakers. Peterson and Barney (1952) and Hillenbrand et al.
(1995), reported the ratio as 1.11 and 1.08 for American English, respectively.
Adank et al. (2004) also reported a ratio of 1.00 for Northern Standard Dutch

Brought to you by | University of Liverpool Sydney Jones Library


Authenticated
Download Date | 12/17/16 10:22 PM
An acoustic description of Azerbaijani vowels 515

speakers and the ratio of 1.16 for the Southern Standard Dutch. The Azerbaijani
speakers’ ratio seems to be smaller than that of Portuguese and Dutch and closer
to the ratio reported by Peterson and Barney (1952) for American English. Be-
cause studies utilize a variety of data collection, comparing female-male F1 and
F2 ratios is difficult. The average Azerbaijani female-to-male F3 space size ratio
is 1.12. One of the possibilities for different ratios reported in studies can be
traced in sociolinguistic differences between speaker groups (Van Bezooijen
1995).

4.2. F0

As expected, our results revealed that female speakers had a higher F0 than
male speakers, as it has been shown for other languages (Most et al. 2005;
Adank et al. 2004; Hillenbrand et al. 1995; Peterson and Barney 1952). Addi-
tionally, our results agree with the study by Whalen and Levitt’s (1995) study on
the universality of the vowel-intrinsic F0, with our results showing the [u] and
[ɑ] vowels to have the highest and lowest F0 values, respectively.
The F0 of the Azerbaijani vowels were higher for females than males with
the ratio of 1.63. The study conducted by Escudero et al. (2009) reported that
Portuguese-speaking women had a higher average F0 than men with a 1.73 ra-
tio. Also Peterson and Barney (1952) and Hillenbrand et al (1995), reported ra-
tios of 1.68 and 1.69 for American English, respectively. Most et al. (2005) re-
ported a ratio of 1.51 for Hebrew, and Adank et al (2004) reported ratios of 1.49
for Northern Dutch and 1.73 for Southern Dutch. Readers should keep in mind
that comparing these ratios across different studies must be done cautiously as
F0 difference can be affected by a variety of factors. For instance age of the sub-
jects can affect the F0 values (Reubold et al, 2010).

4.3. Duration

The interaction of vowel quality and vowel duration has also been noted in
many languages (Lehiste 1970). It is also found for languages without a phono-
logical length contrast, such as Portuguese (Escudero et al. 2009), Peruvian
Spanish (Morrison and Escudero 2007) and European French (Rochet and
Rochet 1991; Strange et al. 2007). The vowel-intrinsic duration effect concludes
that lower vowels are longer than higher vowels. Our results revealed that this is
not a strong effect for Azerbaijani vowels.

Brought to you by | University of Liverpool Sydney Jones Library


Authenticated
Download Date | 12/17/16 10:22 PM
516 P. Ghaffarvand Mokari and S. Werner

Escudero et al. (2007) reported that most back vowels were longer than their
front counterparts. This was also found in our results for Azerbaijani vowels.
The Azerbaijani back [u] vowel had a longer duration than the front [i] and the
back [o] vowel had a longer duration than the front [e]. Escudero et al (2009)
suggested that this might be due to commonly higher F1 values for back vowels
and F1 covariation with duration that leads to longer duration for back vowels.
In contrast with previous studies reported by Simpson and Ericsdotter
(2003), which found that in many languages vowels produced by female speak-
ers had a longer duration than vowels produced by male speakers, our study
showed that vowels produced by male speakers had a longer duration than vow-
els produced by female speakers, with a 1.15 ratio. One of the possibilities
could be the different speech rate and that the speech rate can affect the vowel
duration. Further consideration of the data revealed no significant difference be-
tween male and female speaker speech rate. Lehsite (1970) states that duration
may be considered as one of the manifestations of stress. There are languages
that regularly stressed syllable is longer than unstressed syllable, while other
factors being the same. Although, all target vowels were embedded in the same
carrier sentence, the large number of recordings made it difficult to control the
pitch pattern of the production of the vowels. Therefore, we assume this can be
the reason for the results contrary to the previous literature.

REFERENCES
Adank, P., R. Van Hout and R. Smits. 2004. “An acoustic description of the vowels of
Northern and Southern Standard Dutch”. Journal of the Acoustical Society of Amer-
ica 3. 1729–1738.
Bezooijen, R. van. 1995. “Sociocultural aspects of pitch differences between Japanese
and Dutch women.” Language and Speech 38. 253–266.
Best, C.T. and A. Faber. 2000. “Developmental increase in infants discrimination of
nonnative vowels that adults assimilate to a single native vowel”. In: Proceedings of
the International Conference on Infant Studies, Brighton, UK, 16–19 July 2000.
Boersma, P and D. Weenink. 2015. Praat: doing phonetics by computer [Computer pro-
gram]. Version 5.4.10, retrieved 27 Jun 2015 from <http://www.praat.org/>.
Clopper, C.G., D. Pisoni and K. De Jong. 2005. “Acoustic characteristics of the vowel
systems of six regional varieties of American English”. Journal of the Acoustical
Society of America. 3. 1661–1676.
Crystal, D. 2010. The Cambridge encyclopedia of language. Cambridge and New York:
Cambridge University Press.
Diehl, R.L., B. Lindblom, K.A. Hoemeke and R.P. Fahey. 1996. “On explaining certain
male-female differences in the phonetic realization of vowel categories”. Journal of
Phonetics 24. 187–208.

Brought to you by | University of Liverpool Sydney Jones Library


Authenticated
Download Date | 12/17/16 10:22 PM
An acoustic description of Azerbaijani vowels 517

Escudero, P., P. Boersma, A.S. Rauber and R.A. Bion. 2009. “A cross-dialect acoustic
description of vowels: Brazilian and European Portuguese”. Journal of the Acousti-
cal Society of America 3. 1379–1393.
Hagino, A., K. Inohara, Y.I. Sumita and H. Taniguchi. 2008. “Investigation of the fac-
tors influencing the outcome of prostheses on speech rehabilitation of mandibulec-
tomy patients”. Nihon Hotetsu Shika Gakkai Zasshi 52(4). 543–549.
Hillenbrand, J., L. Getty, M.J. Clark and K. Wheeler. 1995. “Acoustic characteristics of
American English vowels”. Journal of the Acoustical Society of America 5. 3099–
3111.
Johanson, L. 1988. “Code-copying in Irano-Turkic”. Language Sciences 20(3). 325–
337.
Johanson, L. 2010. “Azerbaijanian”. In: Brown, K and S. Ogilvie (ed.), Concise ency-
clopedia of languages of the world. Amsterdam: Elsevier. 110–113.
Koopmans-Van Beinum, F.J. 1980. Vowel contrast reduction, an acoustic and perceptu-
al study of Dutch vowels in various speech conditions. (PhD dissertation, Universi-
ty of Amsterdam.)
Lehiste, I. 1970. Suprasegmentals. Cambridge, MA: MIT Press.
Menges, K.H. 1951. “Research in the Turkic dialects of Iran: preliminary report”. Ori-
ens 4(1). 273–279.
Mozaffarzadeh Peivasti, S. 2012. “An acoustic analysis of Azerbaijani vowels in Tabrizi
dialect”. Journal of Basic and Applied Scientific Research 2(7). 7181–7184.
Peterson, G.E. and H.L. Barney. 1952. “Control methods used in a study of the vowels”.
Journal of the Acoustical Society of America 2. 175–184.
Prieto, P, and F. Torreira. 2007. “The segmental anchoring hypothesis revisited: Syllable
structure and speech rate effects on peak timing in Spanish”. Journal of Phonetics
35(4). 473–500.
Reetz, H. and A. Jongman. 2011. Phonetics: Transcription, production, acoustics, and
perception. Oxford: John Wiley & Sons.
Reubold, U., J. Harrington and F. Kleber. 2010. “Vocal aging effects on F0 and the first
formant: a longitudinal analysis in adult speakers”. Speech Communication 52(7).
638–651.
Rochet, A. and B.L. Rochet. 1991. “The effect of vowel height on patterns of assimila-
tion nasality in French and English”. Proceedings of the 12th Int. Cong. Phonetic
Sciences, Aix (vol. 3). 54–57.
Simpson, A.P. 2001. “Dynamic consequences of differences in male and female vocal
tract dimensions”. Journal of the Acoustical Society of America 5. 2153–2164.
Simpson, A.P. “Gender-specific articulatory–acoustic relations in vowel sequences”.
Journal of Phonetics 30(3). 417–435.
Simpson, A.P. and C. Ericsdotter. 2003. “Sex-specific durational differences in English
and Swedish”. Proceedings of the 15th Int. Cong. Phonetic Sciences, 2003. 1113–
1116.
Strange, W., A. Weber, E.S. Levy, V. Shafiro, M. Hisagi and K. Nishi. 2007. “Acoustic
variability within and across German, French, and American English vowels: Pho-
netic context effects”. Journal of the Acoustical Society of America 2. 1111–1129.

Brought to you by | University of Liverpool Sydney Jones Library


Authenticated
Download Date | 12/17/16 10:22 PM
518 P. Ghaffarvand Mokari and S. Werner

Wang, H. and V. Van Heuven. 2006. “Acoustical analysis of English vowels produced
by Chinese, Dutch and American speakers”. Linguistics in the Netherlands 237–
248.
Whalen, D.H. and A.G. Levitt. 1995. “The universality of intrinsic F0 of vowels”. Jour-
nal of Phonetics 3. 349–366.

Address correspondence to:


Payam Ghaffarvand Mokari
General Linguistics and Language Technology
University of Eastern Finland
Room AG101
Joensuu, 80100
Finland
payam.ghaffarvand@uef.fi

Brought to you by | University of Liverpool Sydney Jones Library


Authenticated
Download Date | 12/17/16 10:22 PM

You might also like