AN ACOUSTIC ANALYSIS OF HINDKO ORAL VOWELS.
Hindko is a member of the Indo-Aryan language family and had several regional dialects. It is mainly spoken in Hazara division; Khyber Pukhtonkhawa province of Pakistan. This paper aims to present an acoustic analysis of Hindko oral vowels. The data is recorded involving Hindko native speakers who live in Hazara division in consonant-vowel-consonant (CVC) context. All the data is recorded using high fidelity microphone directly on PRAAT software. And the analysis is carried out using PRAAT software package (Boersma 2012). A statistical package for social sciences (SPSS) is used for statistical analysis where possible. The analysis reveals that Hindko oral vowels have a phonemic durational contrast and with the help of formant features classified as front central and back vowels. In addition to this the study of formants also determines the acoustic space as the short vowels cover less space than the long ones.
Keywords: acoustic analysis oral vowels durational contrast formants acoustic space
Hindko belongs to the Indo-Aryan family of languages which has several dialects on a regional basis (Lothers and Lothers 2010). Hazara dialect is one of its main dialects that is spoken in Hazara division; Khyber Pukhtonkhawa province of Pakistan. The majority of Hindko speakers live in this area (Haroon and Sohail 2011).
According to Haroon and Akhtar (2012) there are nine oral vowels in Hindko. Out of these nine four are front /i I e/ two are central /a 0/ and three are back /u o 0/. The length contrast is found in the pairs /i - I / Ia- e/ and /u- o/ while there is no length contrast for /e 0/ as these have no counterpart short vowels. The table given below presents word level distribution of Hindko oral vowels:
Table 1: Word Level Distribution of Hindko Oral Vowels
i###i~i###prize money on Eid###pis###grind pesi###have to
###itti###small wood to play###tJ~i1l peal###-
a###ax###say###~ap###fever alutfa plum
o###or###a vegetable###boc~ok wasp###moko to me
This table 1 illustrates that only short vowels / i 0/ do not occur word finally.
Vowel inventories differ in size from language to language. Maddieson's (1984) survey of 317 languages demonstrates that 5.7% languages have only three vowels 4.1% have 17 or more vowels. On the basis of size he classifies vowel inventories into three categories namely small (2-4) average (5-6) and large (7-14). The vowel inventory of Hindko is a large one. Haroon and Akhtar (2012) document the vowel system of Hindko.
They regard it a complex system as it has 9 oral vowels 5 nasal vowels and three diphthongs. Here the oral vowels will be discussed acoustically using temporal and spectral parameters. This study will establish the vowel inventory of oral vowels.
2. Acoustic Analysis of Vowels
The physical differences across different phonemes can be measured in terms of frequency amplitude duration and spectral distribution. Kent and Read (1996) suggest the correspondence of the perceptual correlates like pitch loudness length and quality. This section focuses on the acoustic analysis of Hindko oral vowels; examining in terms of duration of vowels and format frequencies.
2.1 Temporal Properties
Cochrane (1970) identifies that vowels phonemes can be distinguished with regard to duration for such languages and dialects that employ phonemic vowel length. The durational contrast may or may not be the only distinguishing property between two vowels. For instance in English the short vowels / i are distinct in quality from their respective counterpart long vowels. Kimiko (2002) supplements the claim that length contrast on the basis of duration may be used in some languages but not in others. He indicates that English and Hindi differ from Thai and Japanese as the latter use vowel duration as an acoustic cue for distinguishing length besides quality differences in vowels like /i:/ and / i I. Watson and Harrington (1999) believe that accuracy is enhanced when vowel duration and frequency are put together.
The main difference between long and short vowels is of total duration only. However the difference is relative rather than absolute as various contextual and prosodic factors affect the final length of vowel. Lindblom (1967) identifies that open vowels tend to be longer than close vowels.
2.2 Vowel Formant Features
Generally in articulatory phonetics vowel segments are described in shape of height of the tongue; backness of the tongue and lip rounding. Acoustic studies approach differently to description of vowels. It involves different formant configurations characterizing each vowel. By comparing the formant values the relationship among them can be examined (Olive Greenwood and Coleman 1993). Acoustically vowels are commonly described with the help of their formant structure. This indicates the vocal tract resonance and therefore articulatory shape (Fant 1960). The high-low and front-back distinctions are represented by the first formant (F1) and the second formant (F2) on the spectrogram (Fry 1996; Olive Greenwood and Coleman 1993). F1exhibits the high-low distinction while F2 front-back distinction. It means the lowerF1 is the higher the vowel. And the highF2 implies that the vowel is close to front position.
The relationship between F1 and F2 guarantees when the first two formants of a set of vowels target are plotted on axes with proper scaling features the results closely match the traditional auditory vowel map. Such vowel spaces with axes F1 and F2 rely on the notion of the target vowel. The target is the vowel which is least affected by its surrounding phonetic context.
2.3 Vowel Space
Catford (1988) argues that the idea of the cardinal vowels by Daniel Jones banks on the notion that the vowels are limited by space. Davenport and Hannahs (2005) noted that the positions of vowels on the chart indicate the vowelness rather than the vowels of any particular language. Thus these vowels are `reference points' that can be used to describe the vowels of any language as these are `language neutral' (Ball and Rahilly 1999).Theoretically a vowel in any language must have tongue position either on the vowel limit itself or within the vowel space.
Maddieson (1984) claims that the vast majority of world languages include three vowels that mark the extremes of the general vowel space namely /iGu/ known as `point vowels'. Consequently these have a special status in theories of vowel systems. Dispersion theory (DII) views that the arrangement of vowels in a given language in the acoustic space in order to minimize the perceptual confusion between distinct vowel kinds. This Computer program driven approach to generate the optimal configuration for vowel systems has been proved quite effective (Liljencrant and Lindblom 1972; Lindblom 1975 1986; Disner 1984).
However such investigations of DT placing emphasis on inter category distance as the determiner of vowel system configuration in a universally defined acoustic vowel space have failed to account for some systems like Swedish where nine vowels crowed in a small corner of the entire vowel space rather than dispersing them throughout the available space (Disner 1983). Recent developments express DT as a principle of sufficient rather than maximal contrast (Lindblom 1989 1990).
Additionally this theory has been able to explain within speakers' variation. A speaker's vowel space will be expanded relative to his/her casual speech vowel space.
Stevens (1998) in the Quantal Theory of Speech (QTS) proposes a different approach to vowel system. This theory states that there are definite regions of stability in the phonetic space. Specifically it claims the correspondence of stable regions to the point vowels /iau/. Consequently QTS suggests that the point vowels are in more or less the same location in all languages irrespective of the size of the vowel inventories. Moreover it also predicts that since the point vowels are in phonetically stable regions they should show less variability than non-point vowels.
In total ten Hindko speakers (five male and five female) participated in the study. The participants were born in Hazara Hindko dialectal area. All these are either in their twenties or thirties. Eight of them are graduate students while two are teachers at the university of AJ and K. All the participants are multilingual. At home they speak Hindko while elsewhere Urdu. However classroom communication is in English to a great extent. None of them reported any history of language impairment. They had not received any phonetic training and knowledge of this kind of experiment.
The stimuli contain Hindko words and non-words (due to non- availability indicated by `-`) with a CVC context for oral vowels where V is the target vowel.
Table 2: Stimuli for Oral Vowels
Target vowel###CVC context###Target vowel###CVC context
e###ket -###k~t -
a###kut -###kot cut
u###kut -###o###kot beat
Each participant repeated the 9 stimuli 3 times for a total of 27 stimuli per participant. Totally 270 tokens (9 vowels x 3 repetitions x 10 participants) were analyzed to determine acoustic cues. Before recording a brief session of instruction regarding how and what to do was held. For familiarization each participant rehearsed the target words. The list of target words was given and for each repetition the order was changed. The data was recorded using high fidelity microphone directly on PRAAT software. A statistical package for social sciences (SPSS) was used for statistical analysis where possible.
For measuring vowel duration two extreme cursors are placed from the beginning of the segment (left) to the end (right) showing the duration in seconds as shown in the following figure:
To measure the formants of Hindko vowels the cursor is placed at the middle of the dark band after selection of the vowel area. Then frequencies are taken in Hertz (Hz) as the following spectrogram indicates for the vowel I i I:
4. Results and Discussion
The results are stated and discussed under the relevant subsections 4.1-3 below:
4.1 Temporal Properties
The following table presents the durational contrast of Hindko oral vowels:
Table 3: Mean Duration Along With Standard Deviation of Hindko Oral Vowels in Milliseconds (ms)Vowels Mean duration Standard deviationI 266 3.910I 106 2.55e 241 4.442ae 265 2.458a 255 1.587e 125 2.131o 233 3.198u 250 3.374u 103 2.173
The table 3 shows the duration of each vowel along with standard deviation. From this table it is obvious that only three vowels / i have relatively short duration than the rest of the vowels. The durational difference between long and short vowels is illustrated in the figure 3:
Besides indicating the durational contrast among vowels the bar chart also shows the averages for long and short vowels. Three vowels are paired in long-short contrast while the other three namely /e o/ do not have their respective counterpart short vowels. The average duration of long vowels is 251 while of short vowel is 111 milliseconds as indicated by the end bars on the chart above. The length contrast in pairs li-il /u- / and /u-o/ is evident as the short vowels are approximately 60% less than the long ones in each pair. The vowels /e o/ are long one as their duration is like other long vowels almost 60% more than the average duration of short vowels. However this study of vowel duration is not supporting the claim of Lindblom (1967) that open vowels tend to be longer than close vowels as the duration of close vowel lu (266) is more than the open vowel Ia! (255).
A single factor ANOVA proves that there is a significant difference between long and short vowels as the P-value is 0.00 which is smaller than the level of significance 0.05.
The results given above exhibit that durational contrast is phonemic in Hindko. Though it is clearly distinctive from a listener's perspective and useful in placing vowels in large categories like long and short but it not good enough to recognize individual vowels. For instance /i/ and /o/ can hardly be distinguished on durational grounds as both are short and have average duration 106 and 103 ms respectively.
Additionally to the data analysis above there is also morphological evidence that Hindko has only three contrastive pairs of long and short vowels while the other vowels /e o/ are without their respective counterpart short vowels. For example long vowels map into short vowels on concatenating the causative suffix `a' to imperative verb roots (base) if it contains a long vowel (Haroon and Sohail 2012) as illustrated below in the table:
Table 4: Vowel Mapping (long-short) in Hindko
Long to short vowel mapping###Imperative verb form
tfir###tS i rru###`cut'
The data given above illustrates that vowel mapping restricted to the long vowels /i a u/ as the gemination of the last consonant segments takes place while the long vowels /e o/ do not have their counterpart short vowels as a result there is no gemination of the last consonant segment on concatenating the causative suffix `a' with the imperative verb forms. Hindko geminates a consonant while suffixing only when it occurs after a short vowel (Haroon and Sohail 2011). Vowel mapping is also found in shape of vowel lengthening i-e while concatenating the causative suffix `a' to the verbs that contain a short vowel as shown in the table under:
Table 5: Vowel Mapping (short-long) in Hindko
Short to long vowel mapping###Imperative verb form
The vowel shortening and lengthening on affixing the causative suffix `a' to the base forms of the verbs as illustrated in the tables 4-5 reveal that Hindko has three vowel pairs with long-short contrast namely /i- i I Ia- / and /u-u/ while the other three /e o/ are without their respective short vowel counterparts.
The discussion above shows that the durational contrast is phonemic in Hindko. However it alone is not enough to identify individual vowels. For this quality cues are also vital which will be discussed in the section
4.2 Vowel Formant Features
The meanF1 and F2 of Hindko oral vowels are stated along with standard deviation in the following table:
Table 6: Mean Fi and F2 in Hz Along With Standard Deviation (sd) of Hindko Oral Vowels
Vowels###Mean Fi (sd)###Mean F2 (sd)
I###335 (4.492)###2337 (5.788)
I###450 (3.566)###2020 (6.058)
e###492 (6.362)###2015 (5.316)
ae###685 (5.342)###1935 (4.357)
a###750 (3.604)###1260 (10.583)
e###734 (4.602)###1432 (5363)
o###510 (5.593)###945 (3.651)
u###405 (5.963)###864 (9.3 12)
u###465 (7.407)###1000 (6.118)
As clarified in the section 4.1 Hindko has only three contrastive pairs of long-short distinction the following figure displays Fi of these pairs:
The bar diagram given above in the figure 4 shows that the difference of Fi is a slight one between paired long and short vowels. Statistically the difference between /u- / is insignificant. The comparison of F2 for each of these pairs is displayed in the next figure which shows a clear difference among these. Statistically the difference is also found significant as the P-value is smaller than 0.05 the level of significance.
The contrastive analysis of F and F2 for each oral vowel is displayed in the figures 6 and 7.
The contrast betweenF1 and F2 for each oral vowel shows that the close long vowel /i/ has the maximal gap and the minimal gap is found for the mid-back vowel /0/. A comprehensive picture is visible in the next figure below:
Figures 6-7 demonstrate that for front vowels Fi goes lower when the constriction in the oral cavity increases. As the most constricted front vowel lu has the lowest Fi. It implies that Fi increases as the tongue position gets lower for front vowels such as the highest vowel /il has F1 335 Hz while the mid-front vowel / / 685 Hz. Fi decrease for back vowels is linked with the height of the tongue. As can be noticed the mid-back long vowel /o/ has F1 510 whereas the high-back long vowel /u/ has 405 Hz. This contrast suggests that the high vowels have lowerF1 than the low vowels. The contrast also indicates that the front vowels have higher F2 than the back vowels. The gap between F1 and F2 is maximal for front-high vowels while minimal for back vowels as F2 is much closer and lower to F1.
4.3 Vowel Space
This section presents the vowel plots of Hindko that were produced with the help of formant values. F2 is plotted against horizontal axis while F1 on vertical axis as can be seen below in the graph:
As can be seen above in the figure 8 the short vowels cover less space than the long vowels. The long vowels are distributed peripherally in vowel space forming the outer layer which implies that these cover more acoustic space than the short vowels. The quadrilateral of Hindko oral vowels is developed with the help of this acoustic space of vowels as presented in the figure below:
Figure 9 indicates that the vowels pairs pairing in the long-short contrast are close to each other. This suggests that durational contrast is the primary and distinctive cue in Hindko for differentiating long vowels from short vowels. The central vowels pair /a-o/ have higher F1 than any other vowel. Another visible and noticeable trend is that the short vowels are closer to the center and lower than their counterpart long ones.
5. Description of Hindko Vowels
Hindko has front central and back vowels that are described in this section.
5.1 Front Vowels
Hindko has four front vowels namely /i i e ei. The vowel hi [sil] `ploughing grown crops' as in and /i/ as in [sil] `slab' form a pair by virtue of length. The pair /i ii not only differ in length but also in quality. As the data exhibit F1 and F2 of /i/ are 335 and 2337 Hz while F1 and F2 of hi are 450 and 2021 Hz respectively. LowerF1 and higher F2 values suggest in the entire system that these are high front vowels.
Comparatively /i/ is higher than / 1/. Being with the maximum gap of F1 and F2 /i/ is the highest vowel. These can also be called close-front vowels. The durational contrast is a clear one as the vowel /i/ has 266 ms while /i/ has only 106 ms. The lips are spread while articulating both of these.
The other two front vowels are /e/ as in [ret] `sand' and / /as in [mtra] `motherless' do not have their counterpart short vowels so offer no length contrast. These are mid-front or half-close vowels as their F1 492 and 685 Hz is higher than /i/ and /I/ while F2 2015 and 1935 Hz respectively is lower than the high-front vowels. The vowel /el is below the cardinal vowel 2 and / l is above the cardinal vowel 3. From length perspective these are categorized as long because their respective durations 241 and 265 ms are more than double of the average duration of short vowels in Hindko i-e 111 ms. The lips are in neutral position while their articulation.
In the articulation of all the front vowels the body of the tongue moves forward and rises upward in the direction of hard palate.
5.2 Central Vowels
Hindko has only two central vowels that are paired in contrast of length. These are huh as found in [ku ] `big scissors' and / / as in [k L ] `sew'. Their F1- F2 mean values are 750 -1260 and 734-1432 Hz respectively. Both are in middle range. The vowel huh has the highest F1 than the rest of the vowels so it is placed the lowest of all vowels. The contrast of length is obvious between the two as the long vowel /a/ has the mean duration 255 while the short vowel l h has 125 ms which is more less than half of the former. The lips remain neutral while the mouth is open. Both are central vowels however the long vowel /a/ is open- low- central as the mean F1 is higher than its counterpart short vowel even the highest of the rest of the vowels in the whole system. In the production of these vowels the body of the tongue raises to hard palate.
5.3 Back vowels
Hindko has a series of three back vowels namely /0 0 u/. The first one is classified as mid-back and the other two as high-back vowels. All of these have F1 in the middle range whileF2 in low range. The low range Ft identifies them as back vowels.
The mid-back vowel /o/ has F1 510 and F2 945 Hz. Due to higher F1 in the series it is categorized as mid-back vowel. As can be found in words like [sot] `put in something by force'. The location of this vowel is in between cardinal vowel 6 and 7. The duration is 233 ms which more than double of the short vowel average duration i-e. 111 ms. Hence it is recognized as a long vowel.
The other two back vowels are /u/ and /o/ as found in words such as [kul] `a water path' and [kut] `beat' respectively. These vowels are in the region of cardinal vowel 8. These form the pair in terms of length contrast as their respective mean durations are 250 and 103 ms. Their averageF1 and F2 are as 405-864 and 465-1000 Hz respectively. Comparatively the higherF1 and gets /o/ closer to central vowels and lower to /u/. The lowest F1 and F2 values in the series of back vowels /u/ becomes the high-back or close-back vowel of Hindko. Both /u/ and can be identified as close-back-long and close-back-short vowels respectively.
From articulation perspective of back vowel the lips are rounded and the mouth is half open. The body of the tongue moves upward and backward in the direction of soft palate.
The following table states the summary of the vowel description of oral vowels:
Table 7: Summary of Hindko Oral Vowels Description
The acoustic study of Hindko oral vowels finds that:
a. The durational contrast is phonemic in Hindko. The analysis reveals that only three vowel phonemes have a long-short contrast namely /i i I Ia- / and /u- o/. The other three which are Ie 0/ do not have their counterpart short vowels. However on the basis of average duration of long and short vowels these are treated as long. Statistically there is a significant difference of duration.
. The study of formant features demonstrates that the high vowels have lower F1 than the low vowels. It also shows that the front vowels have higher F2 than the back vowels. The gap between F1 and F2 is maximal for front-high vowels while minimal for back vowels as F2 is much closer and lower to F1. On the basis of formants Hindko oral vowels are classified as front i i e central ct and back o o u.
c. The acoustic space is worked out as the short vowels cover less space than the long vowels forming the inner layer. The long vowels are distributed peripherally in vowel space forming the outer layer which implies that these cover more acoustic space than the short vowels.
The acoustic analysis of oral vowels shows that Hindko has nine segments. Four are front two are central and three are back vowels. The data reveals that only three vowel phonemes have a long-short contrast namely /i- i/ Ia- / and /u- u/. The other three which are /e 0/ do not have their counterpart short vowels. The vowels pairing in long- short relationship are significantly distinctive from each other with respect to duration as the long ones have 60% longer duration than their counterpart short ones. The vowels that do not have their respective counterpart short vowels are classified as long on the basis of their duration which is more than double of the average duration of short vowels. It is found that the durational contrast is phonemic in Hindko and is primary and distinctive cue for vowel segments.
The study of formants shows that front vowels /i i e / have higher F2 open Ia o/ in the middle range and back vowels Io o u/ in the low range. Fi is higher for the open vowels. The gap between F2-Fi is the maximum for front vowels and minimum for the back vowels. The study of formant features determines in locating the acoustic space covered by the vowels. This reveals that the short vowels / i o/ form inner layer and remain close to the center while the long vowels /i e ao u/ form outer layer and are distributed peripherally over the acoustic space. This implies that the long vowels enclose more acoustic space than short vowels.
Ball J. M. and Rahilly J. (1999).Phonetics. The science of speech. NewYork: Oxford University Press.
Boersma P. (2010). Praat. Accessed from: http://. fon. hum. uva.nl. praat.
Catford J. C. (1988). A practical introduction to phonetics. New York: Oxford University Press.
Cochrane G. R. (1970). Some vowel variation in Australian English. Phonetica 22 240-250.
Davenport M. and Hannahs S. J. (2005). Introducing phonetics phonology. UK: Hodder Education.
Disner S. F. (1983). Vowel quality: the relationship between universal and language specific factors. Ph. D dissertion. University of California: Los Angeles.
Disner S. F. (1984). Insights on vowel spacing. In I. Maddieson Pattern of sounds 136-55 .Cambridge: Cambridge University Press.
Fant G. (1960). Acoustic theory ofspeech production. The Hague: Mouton Co.
Fry D. B. (1996).The physics of speech. Cambridge: Cambridge University Press.
Haroon R. and Sohail A. (2011). A brief introduction of Hindko language. Language in India 11(11) 471-482.
Haroon R. and Akhtar N. R. (2012). Hindko Vowel System. Kashmir Journal of Language Research. 15 (2) 55-76.
Kent R. D. and Charles R. (1996). Acoustic analysis of speech. San Diego: Singular Publishing Group.
Kimiko T. (2002). An acoustic comparison between American and Australian English. Paper presented at 7th International Conference on Spoken Language Processing. Colorado: USA.
Liljencrants J. and Lindblom B. (1972). Numerical simulation of vowel quality systems: the role of perceptual contrast. Language 48 83-62.
Lindblom B. (1967). Vowel duration and a model of lip-mandible coordination. Speech Transmission Labortary Progress Status Report 4 1-29.
Lindblom B. (1990). On the notion of possible speech sound. Journal of Phonetics 18 135-152.
Loothers M. and Loothers L. (2010).Pahari and Pothwari: A sociolinguistics survey Islamabad: Summer Institute of Linguistics.
Maddieson I. (1984). Patterns of sounds. Cambridge: Cambridge University Press.
Olive J. Greenwood A. and Coleman J. (1993).Acoustics of American English speech. New York: Springer-Verlag.
Stevens K. N. (1998). Acoustic Phonetics. Cambridge: MIT Press.
Watson C. Harrington J. (1999). Acoustic evidence for dynamic formant trajectories in Australian English vowels. Journal of Acoustical Society of America 106 458-468.
|Printer friendly Cite/link Email Feedback|
|Publication:||Kashmir Journal of Language Research|
|Date:||Dec 31, 2013|
|Previous Article:||GENDER CONSTRUCTION IN OBITUARIES: AN ANALYSIS OF PAKISTANI NEWSPAPERS.|
|Next Article:||LEARNER AUTONOMY: BUILDING A CURRICULUM AROUND AUTONOMOUS LEARNING WITH EXCLUSIVE REFERENCE TO THE UNIVERSITY OF SINDH PAKISTAN.|