Amir Bedir Stefan Werner
Production of Arabic pharyngeal and pharyngealized consonants for Finns
learning Arabic as a second language
ITÄ-SUOMEN YLIOPISTO – UNIVERSITY OF EASTERN FINLAND Tiedekunta – Faculty
Philosophical Faculty
Osasto – School
Foreign Languages and Translation Studies Tekijät – Author: Amir Bedir
Työn nimi – Title: Production of Arabic pharyngeal and pharyngealized consonants for Finns learning Arabic as a second language
Pääaine – Main subject:
Linguistics and
Language Technology)
Työn laji – Level: MA Thesis Päivämäärä – Date:
29.12.2018
Sivumäärä – Number of pages:
58 Pro gradu -tutkielma X
Sivuainetutkielma Kandidaatin tutkielma Aineopintojen tutkielma Tiivistelmä – Abstract:
The main focus of our study is the pharyngeals, and the pharyngealized consonants of the Arabic language. We approach the study from a phonetic point of view. As a teacher of Arabic as a foreign language in Finland, I was privileged with the chance to record with 10 Finns learning Arabic as a second language. The recording consisted of words including the target consonants in the medial position (in between two vowels). The target consonants are the Arabic pharyngeals /ʕ/, /ħ/ and their counterparts /ʔ/, /h/, and the Arabic pharyngealized consonants /t̥/, /d̥/, /s̥/, /ð̥/, and their counterparts /t/, /d/, /d/, /ð/. Each target consonant is repeated 5 times for a reliable statistical results. The recorded files are then compared to a native speaker’s recording of the same words. Additionally, the study used the “rapid-shadowing paradigm” as a method of recording. Later on, an acoustic analysis was done answering our four research questions. In a nutshell, the analysis results show that Finnish speakers learning Arabic as a second language are better in pronouncing pharyngealized consonants than pharyngeal ones. Moreover, a deeper analysis shows that the voiced pharyngealized consonants’ results are better than the voiceless ones.
Interestingly, the results show the opposite in regards to pharyngeals, as the voiceless pharyngeal (/ħ/) records a noticeably better results than its voiced counterpart (/ʕ/.) Finally, three different second language learning hypothesis were considered to be applied to our acoustic analysis results. Majorly, the application seconds the theories in the following points: (1) It is easier to learn a new sound in TL if it is similar to a sound in NL.
(2) If a sound is an alien to the NL, then it will be hard to learn (Alwabari, 2013). (3) As a learner, it would not be possible to perceive and produce native-like new phonemes.
Perceiving and producing of new sounds will be always related to already found sounds in the NL. Then can come the creation of a new phonetic category that can be different from the native one.
Avainsanat – Keywords: pharyngeal, pharyngealized, Arabic, Finnish, phonetics, acoustic analysis, second language learning.
ITÄ-SUOMEN YLIOPISTO – UNIVERSITY OF EASTERN FINLAND Tiedekunta – Faculty
Filosofinen tiedekunta
Osasto – School
Vieraat kielet ja käännöstiede Tekijät – Author: Amir Bedir
Työn nimi – Title: Production of Arabic pharyngeal and pharyngealized consonants for Finns learning Arabic as a second language
Pääaine – Main subject:
kielitiede ja kieliteknologia
Työn laji – Level: MA Thesis Päivämäärä – Date:
29.12.2018
Sivumäärä – Number of pages:
58 Pro gradu -tutkielma X
Sivuainetutkielma Kandidaatin tutkielma Aineopintojen tutkielma Tiivistelmä – Abstract:
Tutkimukseni keskittyy faryngaaleihin ja arabian kielen faryngalisoituneisiin konsonantteihin. Tutkimusta tarkastellaan fonetiikan näkökulmasta. Arabian kielen opettajana Suomessa minulle avautui mahdollisuus äänittää kymmenen arabiaa opiskelevan suomalaisen ääntämistä. Äänitys koostuu sanoista, joissa tutkittavat konsonantit esiintyvät sanan keskellä kahden vokaalin välissä. Nämä konsonantit ovat arabian kielen faryngaalit / ʕ/, /ħ/ ja näiden vastineet /ʔ/, /h/, sekä arabian faryngalisoituneet konsonantit /t̥/, /d̥/, /s̥/, /ð̥/, ja näiden vastineet /t/, /d/, /d/, /ð/. Kukin näistä konsonanteista toistuu äänityksessä viisi kertaa, jotta tulokset olisivat tilastollisesti luotettavat. Näitä sanoja verrataan samoihin sanoihin äidinkielenään arabiaa puhuvien ääntäminä. Äänityksessä käytetään ”rapid- shadowing paradigm” -menetelmää. Tuloksia tarkastellaan akustisella analyysillä.
Analyysin tulokset osoittavat, että arabiaa toisena kielenä opiskelevat suomenkieliset puhujat ääntävät faryngalisoituneita konsonantteja paremmin kuin faryngaaleja. Tarkempi tarkastelu osoittaa vielä, että soinnillisia faryngalisoituneita konsonantteja on helpompi ääntää kuin soinnittomia faryngalisoituneita konsonantteja. Toisaalta soinnittoman faryngaalin (/ħ/) ääntäminen osoittautuu helpommaksi kuin soinnittoman vastineen (/ʕ/).
Akustisen analyysin tuloksiin harkitaan sovellettavan toisen kielen oppimisen kolmea hypoteesia. Pääasiassa soveltaminen puoltaa teorioita seuraavasti: (1) Uusi äänne on helpompaa oppia kohdekielellä, jos äänne on samankaltainen kuin lähdekielessä. (2) Jos äänne on vieras lähdekielessä, se on vaikea oppia. (Alwabari, 2013). (3) Oppijalle uusien äidinkielen kaltaisten foneemien havaitseminen ja tuottaminen ei ole mahdollista. Uusien äänteiden havaitseminen ja ääntäminen on aina sidoksissa jo lähdekielestä löytyviin äänteisiin. Vasta tämän jälkeen voidaan luoda uusi foneettinen kategoria, joka voi olla erilainen kuin lähdekielen kategoria.
Avainsanat – Keywords: faryngaaleihin, faryngalisoituneisiin, arabia, suomea, fonetiikka, akustisen analyysi, toisen kielen oppimisen.
Table of contents:
1. Introduction ………..………. 06
2. Theoretical background ……….……… 08
2.1. Introduction ……… 08
2.2. Arabic and Finnish phonological backgrounds ……….. 09
2.3. A closer look into pharyngeal and pharyngealized consonants ……….. 10
2.4. Observation and recording of pronunciation organs’ movements …………. 12
2.5. Acoustic analysis ………. 15
2.6. Second language learning theoretical background ……… 18
2.6.1. Lado’s (1957) contrastive analysis hypothesis (CA) ……… 18
2.6.2. Best’s (1995) perceptual assimilation model (PAM) ……… 19
2.6.3. Flege’s (1995) speech learning model (SLM) ……….. 20
3. Data and methodology ……….……….. 23
3.1. Pilot study ……… 23
3.2. Data collection methodologies background ……… 27
3.2.1. Methodologies definitions ……….. 27
3.2.1.1. Nasoendoscopy ……….. 27
3.2.1.2. Videofluouroscopy ………. 27
3.2.1.3. Electromagnetic articulography (EMA) ………. 28
3.3. Data collection and methodology ……… 30
4. Results ………. 32
4.1.How good are Finnish speakers learning Arabic as L2 in pharyngealization (turning an allophonic feature into a phonemic one)? ……….……….. 34
4.1.1. Data presentation and acoustic analysis ……….. 34
4.1.2. Data against L2 learning theories ………..………. 39
4.2. Pharyngeals, a novel two consonants for the Finnish language. Is it possible for Finnish speakers to learn them? ………. 40
4.2.1. Data presentation and acoustic analysis ……….. 40
4.2.2. Data against L2 learning theories ………..………. 43
4.3. Which is easier to learn, for Finnish speakers learning Arabic as L2: Pharyngealized consonants or pharyngeals? ……….. 45
4.3.1.Data presentation and acoustic analysis ………..……….… 45
4.4. Is it true that because voiceless consonants outweigh voiced ones in Finnish that Voiceless pharyngealized consonant will be easier for them to learn than voiceless pharyngealized ones? ……… 49
4.4.1. Data presentation and acoustic analysis ………. 49 5. Conclusion……….…….. 52
1. Introduction
This study tests the pronunciation of pharyngeal and pharyngealized Arabic consonants. It is mainly a phonetic study, but also a pedagogical one.
Arabic is our target language here. We will get to learn some basic notions about the sound system of it, along with a deeper analysis of its most common feature (i.e., pharyngealization). On the other hand, Finnish is the native language of the current study.
Learning about its phonological system is also a crucial step that we will go through in this study.
The subjects of the study are 10 Finnish speakers learning Arabic as a second language in Joensuu, Finland. All the subjects volunteered to do the sound recording in the language lab of the University of Eastern Finland. The tools were as simple as a computer, a speaker, and a microphone. All placed in a room that minimizes noise for the clarity of the recording, and the accuracy of the analysis.
The method followed for the recording is called “rapid-shadowing paradigm.” This method, we found is optimal as it does not involve reading (very difficult because of the difficulty of the Arabic language alphabets for the Finnish speakers). Instead, it is a rapid listening-and-repeating of Arabic words.
Another method of cards reading was followed in a pilot study. The card reading method showed so inaccurate results due to the interference of the reading difficulty.
Switching to rapid-shadowing paradigm solved this problem. Moreover, got us more accurate results.
A total of 600 sound files were analyzed against 60 sound files. The 60 sound files belonged to the native speaker of the Arabic language. Those 60 files included words having the target consonants in between two vowels. Each target consonant repeated five times for a reliable statistical results. The target files included: On the one hand, the four pharyngealized Arabic consonants, and their plain counterparts. On the other hand, the two Arabic language pharyngeals, and their glottal counterparts.
Acoustic analysis is our tool for results’ digging. We used Praat along the way in all of its procedures (i.e., voice recording, annotation, and formant extraction).
Previous studies are our references, that supports our findings. We made use of many previous studies’ findings to link with ours. This linking helped a lot in understanding our
Finally, we tested our results to three second language learning hypothesis. Those three hypothesis have gone popular in the recent research period. Exclusively, they are the contrastive analysis by Lado (1957,) Best’s Perceptual assimilation model (PAM), and Flege’s (1995) speech learning model (SLM).
2. Theoretical background
2.1. Introduction
Arabic pharyngeals /ʕ/, /ħ/ and their counterparts /ʔ/, /h/, and pharyngealized consonant /t̥/, /d̥/, /s̥/, /ð̥/, and their counterparts /t/, /d/, /d/, /ð/ are the subjects of this study. We will start by a little introduction about the Arabic (target language,) and Finnish (native-language).
Then we will move towards reviewing some of the previous studies that were concerned about these phonemes (Arabic gutturals and pharyngealized coronal) understanding. Later on we will review some second language learning hypothesis on which we are going to base our hypothesis. So that, the results would have a backbone to support it.
2.2. Arabic and Finnish phonological backgrounds
Semitic languages are marked by a limited vocalic system and a rich consonantal system. There are typically three basic vowels a, i, u, which are attested in both their short and long forms. Semitic languages are also marked by a rich inventory of guttural consonants, which include both the laryngeals /ʔ/, /h/, the pharyngeals /ʕ/, /ħ/, and the uvular fricatives /χ/ and /ʁ/. The consonantal phonemes of Semitic languages usually constitute triads of voiceless, voiced and ‘emphatic’ in certain sub-sets of the coronal set.
The major lexical contrasts in Arabic are indicated through the consonants. This is reflected in the Arabic script which is based on (mainly triconsonantal) roots of consonants and glides, and which inserts short vowels when necessary as diacritics above or below the consonant. Thus, Arabic has a very rich consonantal system and a relatively impoverished vocalic system.
On the other hand, The classification of [Finnish] consonants according to manner of articulation includes three major classes (i.e., obstruents, glottals, and resonants.) The class of glottals is very small. Using the nearest IPA cardinal vowel symbols, the eight vowel phonemes could be given as /i/, /e/, /y/, /ø/, /æ/, /a/, /o/ and /u/. They occur e.g. in the series of word forms mikin – mekin – mykin – mökin – mäkin – makin – mokin – mukin.
Six of the Arabic consonants are produced at the pharynx area. Being not so familiar to other languages- the example we have here is Finnish- makes them difficult to learn. The pharyngeal consonants are primarily articulated in the pharynx and they have quite a tight constriction in the pharynx. In Arabic there are two pharyngeals; /ʔ/, and /ħ/. On the other hand, the pharyngealized consonants (commonly known as emphatics) have a lower degree of constriction at the pharynx.
2.3. A closer look into pharyngeal and pharyngealized consonants
Laufer & Baer made a study on emphatic and pharyngealized sounds in Hebrew and Arabic. Their results clearly show that all the emphatic sounds, when pronounced as such, share pharyngealization as a secondary articulation. A constriction is formed between the pharyngeal walls and the tip of the epiglottis, which tilts backwards. To a lesser degree, the lower part of the root of the tongue is also retracted. The data show that all the emphatic and pharyngeal sounds we studied are made with qualitatively the same pharyngeal constriction.
However, the pharyngeal constriction is more extreme and less variable for the pharyngeal sounds, where it is the primary articulation, than for the emphatic sounds, where it is a secondary articulation. (Laufer, A., & Baer, T., 1988).
Furtherly explained in one of Laufer's earlier studies: A survey of the literature dealing with emphatics indicates that most investigators, although some did not state so explicitly, believe that emphaticness in Hebrew and Arabic involves a secondary articulation.
Secondary articulation is defined as an articulation performed separately from and in addition to the primary articulation associated with a sound. "The secondary articulation, according to definition, is less constricted than the primary articulation; if the primary articulation is constricted to the degree of 'stop', for instance, the secondary articulation can be constricted to the degree of 'fricative' or 'frictionless continuant' If the primary articulation is to the degree of 'fricative', the secondary articulation must be wider." (Laufer, 1985, p. 83, translated from Hebrew).
A more detailed analysis of how the pharyngeal and pharyngealized sounds are produced is given by Panconcelli-Calzia (1924, pp. 4 8 - 4 9 ). [Where he suggests] six processes for the emphatic consonants: (1) Contraction of the muscles of the hyoid bone; (2) lowering of the epiglottis towards the glottis; (3) raising of the larynx; (4) constriction of the pharynx due to the actions of the constrictor muscles; (5) shorter bursts for the emphatics; (6) earlier voice onset after emphatics than after non-emphatics.
Giannini and Pettorino (1982) examined one Arabic speaker of a Baghdadi dialect, and presented acoustic and radiographic data. They concluded that the emphatic/non- emphatic distinction is one of pharyngealization. Their acoustic results are that for the
They interpreted their radiographic results as showing a constriction in the pharynx for the emphatic sounds.
The rising of F1 can be due to the more space we get horizontally in the mouth due to the curving of the tongue. Hence, it can be also an evidence for pharyngealization. Observed constriction of the pharynx was always accompanied by lowering of F2 and raising of Fl (Laufer, A., & Baer, T., 1988)
2.4. Observation and recording of pronunciation organs’ movements
Different mechanisms (i.e. cineradiography, cinefluourography, fibreoptic nasoendoscopic, X-ray, and electromyography) were used to record the movement of the tongue (especially the back and root parts), the movement of the epiglottis (and whether it moves independently or along with the tongue root), the pharynx (how far it is constricted,) and the larynx.
Ghazeli (1977: 37) reports the main constriction for /ħ/ and /ʕ/ as being formed by the approximation of the epiglottis and the rear wall of the pharynx, about the level of the fourth cervical vertebra (CV4). His cinefluourographic data revealed a narrower constriction in the case of /ħ/ (Al-Tamimi, 2011)
Laufer & Condax (1979, 1981) present fibreoptic nasoendoscopic evidence which leads them to conclude that the epiglottis is an active articulator responsible for significant acoustic effects, that it is able to move independently of the tongue root and that it does so in the production of Hebrew and Arabic pharyngeal consonants and the open /a/ vowel. However, in a later paper, Laufer & Baer (1988) modified this position to one that recognises the importance of the epiglottis as an articulator without attributing to it the ability to move independently of the root of the tongue. In this they agree with Ghazeli (1977: 36–37), Giannini & Pettorino (1982: 25) and El-Halees (1983: 466, 1985: 288). (depicted from: Al- Tamimi, 2011).
“in all cases, the constriction for the pharyngeals is made between the epiglottis and the posterior pharyngeal wall”. X-ray photographs in Ghali (1983) show the epiglottis “almost in contact with the back wall of the throat” (Ghali 1983: 442) for both pharyngeals.
Examination of xeroradiographic images by Bukshaisha (1985: 283–312) led her to the same conclusion for Qatari Arabic. Both her informants formed a narrow constriction between the epiglottis and the rear pharyngeal wall at around the level of the third cervical vertebra. For both pharyngeals, Elgendy (2001: 59) noted that the epiglottis descended onto the apices of the arytenoid cartilages “closing off most of the distance between the root of the tongue and the posterior wall of the lower pharynx”. Contrary to Giannini & Pettorino (1982) and El-
Halees (1983), he is of the opinion that the epiglottis does move independently of the tongue (depicted from: Al-Tamimi, 2011).
It has been commonly observed that contraction of the pharynx accompanies the rearward movement of the epiglottis in a kind of sphincteric action (Bukshaisha 1985: 304–
306; Elgendy 2001: 66) consistent with the engagement of what Esling (1996, 1999, 2005) calls the laryngeal articulator (see also Hassan & Esling, 2011).
That the larynx typically rises in the production of Arabic pharyngeals is something that has been noted by several researchers (e.g. Ghazeli 1977: 36; El-Halees 1983: 466; Laradi 1983: 105; Bukshaisha 1985: 299; Butcher & Ahmad 1987: 167; Elgendy 2001: 83;
Heselwood 2007: 20).
Whether the epiglottis moves under its own steam or through being pushed by the tongue, or by the tilt of the thyroid cartilage (El-Halees 1983: 466), the picture from all previous studies is that it plays a significant role in the production of the pharyngeal consonants through its influence on the shape of the resonating cavity (Esling 1999: 364) but less of a role in the production of the laryngeals because it does not move so far from its position of rest (Laufer & Baer 1988: 191) (depicted from: Al-Tamimi, 2011).
During the production of the pharyngeals /ħ, ʕ/ the articulation is characterised by retraction of the root of the tongue and slight forward displacement of the posterior wall of the lower pharynx, resulting in a place of articulation at the level of the epiglottis. For both consonants the larynx is raised by at least 0.7cm with respect to its position during non- guttural sounds. The articulatory constriction is, however, narrower for the voiceless /ħ/ than for the voiced /ʕ/ (Ghazeli 1977). The nature of the active articulator of pharyngeals is subject to disagreement among authors. A strong claim is maintained by Laufer & Condax (1979) who think that the epiglottis retracts independently from the tongue root. Such a claim has been challenged by Boff-Dkhissi (1983) and toned down by Laufer & Baer (1988), where it is shown that both the tongue root and the epiglottis covary with each other. The reason behind this disagreement is probably due to the different observation techniques used.
Fiberscopy was the main technique used by Laufer and his colleagues and by Esling, while Ghazeli and Boff-Dkhissi used X-ray films. Fiberscopy, compared to cineradiography, has its
the tongue and the laryngeal /pharyngeal articulators. Esling (1999) argues that there is not much ground for distinguishing two distinct places of articulation for sounds produced in the lower pharynx, i.e. epiglottal and pharyngeal. Rather these sounds should be distinguished as a function of manner of articulation or larynx height. Esling suggests that the active articulator is not the epiglottis but rather the aryepiglottic folds (cf. Zeroual el al. 2004;
Edmondson et al. 2007) and hence the sounds are considered to be epiglottal rather than pharyngeal. Heselwood & Al-Tamimi (this volume), using fibreoptic nasoendoscopy, have also shown the importance of the retraction and the lowering of the epiglottis in distinguishing pharyngeal consonants from laryngeal ones. (depicted from: Yeou, 2011).
A further point of uncertainty is whether the primary articulation is different in plain and emphatic coronal consonants, with some researchers claiming that the point of contact is more retracted in the emphatics (Cantineau 1960: 15; Trubetzkoy 1969: 131; Al-Ani 1970:
45; Odisho 1973: 41; Laradi 1983: 325–326; Bukshaisha 1985: 132). If the back of the tongue is raised up and backwards, or the root of the tongue is retracted, then one might expect the tip of the tongue to be more constrained in the forward direction when making a dental or alveolar gesture, even moreso if there is sulcalisation of the tongue body (Ali &
Daniloff 1972: 89; see also Zeroual et al. 2011). (depicted from: Al-Tamimi, 2011).
Using electromyography, Kuriyagawa et al. (1988) found the posterior part of the genioglossus muscle was more active for plain / t, s / than for emphatic / t̥, s̥ /, and increased more for the vowel following an emphatic consonant in JA. Activation of the genioglossus is thought to contribute to pulling the tongue forward out of the pharynx, an action needing a greater effort after an emphatic consonant. They also observed greater activity of the geniohyoid muscle in the vowel after an emphatic, speculating that this muscle contributes to forward movement of the tongue by displacing the hyoid bone in a forward direction. These results provide some insight into the physiology responsible for the lowering of F2 and raising of F1 evident in their acoustic data. (depicted from: Al-Tamimi, 2011).
2.5. Acoustic analysis
It is well known that the frequency of the second formant resonance (F2) of vowels adjacent to emphatic consonants is typically lower than when adjacent to corresponding plain consonants (Obrecht 1968; Al-Ani 1970; Ghazeli 1977; Giannini & Pettorino 1982; Card 1983; Bukshaisha 1985; Norlin 1987; Heselwood 1992; Zawaydeh 1998; Khattab et al. 2006;
Embarki et al. this volume; Hassan & Esling this volume; Zawaydeh & De Jong this volume;
Zeroual et al. this volume). (depicted from: Al-Tamimi, 2011).
All acoustic studies of these sounds have noted a lowering of the second formant resonance (F2) of vowels adjoining pharyngeal consonants (Al-Ani 1970; Ghazeli 1977;
Adamson 1981; Bukshaisha 1985; Laufer & Baer 1988; Butcher & Ahmad 1987; Heselwood 1992) and F2 lowering in vowels next to the laryngeals has also been reported (Heselwood 1992). Researchers agree however that F2 lowering is greater in vowels next to the pharyngeals. Al-Ani (1970: 59), for example, identifies this as their “distinguishing factor”.
(depicted from: Al-Tamimi, 2011).
F1 tends to rise in vowels adjoining pharyngeals, a feature that has been observed to a lesser extent in the vicinity of laryngeals as well (Heselwood 1992). The perturbation theory of vowel formant resonance predicts that F2 will be lowered if there is a narrowing of the vocal tract close to a point of velocity maximum for F2, and F1 will be raised if there is a narrowing close to a velocity minimum for F1. In a typical male vocal tract F1 will rise and F2 fall if there is a narrowing in the pharynx at least 2.83 cm above the glottis which is about the level of the laryngeal additus and the aryepiglottic folds (Heselwood 2007: 12). (depicted from: Al-Tamimi, 2011).
Larynx and hyoid raising were also observed by Elgendy (2001: 83) for Egyptian Arabic. Upward displacement of the larynx is predicted to raise the frequencies of all formants due to reduction in the length of the vocal tract. Part of the increase in F1 may be attributed to this, but clearly the lowering of F2 is contrary and shows that the effects of the articulatory constriction on the resonance pattern far outweigh those of the larynx-raising.
(depicted from: Al-Tamimi, 2011).
Two studies have conducted perceptual experiments with synthetic speech to investigate these perceptual cues (El-Halees 1985; Alwan 1989). Both agree that only F1 (onset value) is critical in discriminating between these categories as listeners categorically move from uvulars to pharyngeals with raised F1. As for coarticulation, it seems that pharyngeals and particularly uvulars have a much lower coarticulatory resistance from vocalic influence, compared to pharyngealised consonants (Ghazeli 1977; Boff-Dkhissi 1983;
Yeou 1997).
Acoustically, there is much acoustic variability in the realisation of uvular and pharyngeal sounds, depending on the language variety and the context. The voiceless pharyngeal /ħ/ is either a fricative or an approximant and is found to have aperiodic noise together with marked formant structure (Ghazeli 1977). For the voiced pharyngeal /ʕ/, allophonic variation of manner (stop and approximant) has been attested. Al-Ani (1970) reports that Iraqi speakers realise /ʕ/ as a stop (cf. El-Halees 1985). The same stopped realisation is found in Sudanese Arabic (Adamson 1981). Ghazeli (1977) finds that /ʕ/ is not realised as a stop by any of his informants, including one Iraqi. Butcher & Ahmad (1987:
170) report that with three Iraqi speakers /ʕ/ is “realised as a voiced approximant, which in final position is often followed by a stop articulation, and which is almost invariably accompanied by creaky voice”. Alwan (1986) finds stopped as well as approximant /ʕ/ with her Iraqi informants, but the most common realisation is the approximant. With Lebanese (Klatt & Stevens 1969), Egyptian (Norlin 1983) and Moroccan subjects (Yeou & Maeda 1995), /ʕ/ is mostly realised as an approximant. Heselwood (2007) identifies a variant of /ʕ/
which he calls a ‘tight approximant’; it is acoustically characterised by amplitude reduction of the first harmonics, including the fundamental. The study of Heselwood (2007), which is based on 21 speakers from different Arab countries, shows that the prevalence of /ʕ/ variants is as follows:
1. the normal approximant (49%), 2. the tight approximant (47%), 3. and the stop (2%).
The term ‘tight’ “refers to the impression that a high degree of constriction is present in the articulatory mechanism, higher than is normally associated with strictures of open approximation” (Heselwood 2007: 9). (depicted from: Yeou, 2011).
Khattab et al. (2006: 138–139) discuss the possibility that using the acoustic evidence of F1 and F3 values to determine the location of the secondary constriction may be confounded by the lowering effects on all formants of lip-protrusion and / or sulcalisation of the tongue. (depicted from: Al-Tamimi, 2011).
In her 1998 study, Zawaydeh measured tokens of her own speech and found that F1 is higher in emphatic coronals than in plain coronals, and F2 is lower. Although she does not give overall average values, these results are in general agreement with previous studies of other Arabic varieties. (depicted from: Al-Tamimi, 2011).
Al-Masri & Jongman (2004) found no durational effects of emphasis in JA but did find the expected lowering of F2 in vowels adjacent to emphatic coronals. They report that, on average across gender and different vowel contexts, F2 was 521 Hz lower when adjacent to emphatics compared to plain coronals. They present no data for F1 or F3. (depicted from: Al- Tamimi, 2011).
In addition to confirming the importance of a lowered F2 in distinguishing emphatic from plain coronals, Khattab et al. (2006) provide evidence in support of Zawaydeh (1998) for a raised F1. They found that across gender and vowel context, F2 was on average 520 Hz lower while F1 was on average 96 Hz higher, a difference that is statistically significant.
(depicted from: Embarki, 2011).
Hassan (2005) showed that pharyngealisation introduced spectral modifications, especially visible in a slight upward shift of F1 and a clear downward shift of F2. (depicted from: Embarki, 2011).
2.6. Second language learning theoretical background
2.6.1. Lado’s (1957) contrastive analysis hypothesis (CA)
Contrastive Analysis hypothesis (CA) simply says that if the target language (TL) and (NL) are similar then it would be easier to learn than if they were different. Phonemically speaking, if the target phoneme the learning from the TL is found in his / her NL, then it would be easy to learn, and vice versa. It is based on a structural comparison between the sound systems of both NL and TL.
There are three elements, according to Lado, instrumental in this comparison;
phonemic inventory, allophonic membership within these phonemes, and positional distribution of phonemes within the language. (Alwabari 2013). “Phonemic split” (Eckman &
Iverson, 1997, pp. 192- 193), can be another hurdle for learning. This, applied in a Finnish- Arabic setting would be like a Finn trying to learn the two Arabic sounds /S/, and /Sˤ/. For the Finn they are allophones of the phoneme /S/ as in "Säädä" and "Saada". However, an Arabic speaker perfectly recognizes them as two phonemes. This phenomena "researchers argue that it induces the maximal learning difficulty.” (Alwabari 2013)
Not only the availability of phonemes in TL and NL, or the matching of phonemes and allophones of both languages makes the learning process easier / more difficult, but also the phonological rules that govern the phonological system of the language has a clear impact according to Lado (1957). Alwabari gives a clarifying example about that: "both English and French have the fricative /ʒ/ and all its allophones; nonetheless, these consonants can occur in initial position in French (for example, jamais) but not in English. This distributional constraint causes English learners of French to transfer this phonological rule and, thus, encounter a difficulty learning /ʒ/ in initial position." (Alwabari, 2013)
In the area of research, there are two main streams that made use of Lado's theory. The first one supports the idea that the similar sounds in TL and NL are easy to learn, including (Brière, 1966; Eckman et. al., 1997; Flege & Port, 1981; Gass & Selinker, 2001). On the other hand, other researchers (such as Major 1994) proved that it is easier to learn a new sound in TL that is not available in NL. (Alwabari, 2013)
However, this hypothesis began to show its inadequacy, leading to opening the door for many other hypothesis like; the perceptual assimilation model (Best, 1995); and the speech learning model (Flege, 1995) whom we are going to talk about in the following sections.
2.6.2. Best’s (1995) perceptual assimilation model (PAM)
The perceptual assimilation model, developed by Best (1995) suggests that learning a second language depends majorly on how similar or different the segments are in the NL and TL. Moreover, there are two principles for that: 1) the universal phonetic domain, and 2) native phonological space. Moreover, the model uses the phonetic gesture, the basic unit of the universal phonetic domain, [it] is the coordinated formation and release of a varying- degree constriction that is harnessed for sound production and formed along the vocal tract (Browman et. al., 1990a).
Additionally, the multidimensional, universal phonetic domain defines the “biodynamic constraints” of the vocal tract (Best, 1995). To illustrate, this domain is the summation of all spatial and temporal properties of all possible phonetic gestures that are harnessed in the world’s languages. (Alwabari, 2013). Hence, this domain can range from the lips—as most of the world’s languages have bilabial consonants (Ladefoged & Maddieson, 1990)—to the glottis. Although only a few languages make use of a posterior, glottal constriction (such as in Arabic /ʔ/ and in ejective consonants /p’, t’, k’, s’/, Ashby, 2011)
On the other hand, a native phonological space is part of the universal phonetic domain and it encompasses all the sound articulations that this language uses. more details are mentioned above in the Arabic and Finnish phonological environments section.
There are four aspects of the native phonological space that would determine the similarities or differences between the TL and NL. The spatial dimension along the tube geometry (i.e. vocal tract) that encompasses all locations of a given language’s gestures (Alwabari, 2013). The spatiotemporal invariants of language gestures (i.e. the space and time parameters of a language’s gestures). The production of Arabic /s̥/ necessitates special timing
of the formation and release of the pharyngeal constriction relative to the coronal constriction (Laufer et. al., 1988)
The degrees of constriction of the articulatory gestures represent the third aspect of native phonological space, and refer to the degree of proximity between the active and passive articulators involved in gesture formation.
There are varying degrees of constriction in segmental production—for example, closed (as in plosives); critical (as in voiced fricatives); narrow (as in approximants); mid (as in most vowels); and wide such as the opening of the glottis in the production of voiceless stops (Ashby, 2011; and Best, 1995). For example, Arabic has quite a different degree of constriction in pharyngeal and pharyngealized consonants that are not found in Finnish. As we can see the Arabic /ħ, ʕ/ are narrower than the Finnish /h/, and /a/. This narrow constriction harnessed in Arabic consonants cannot be incorporated within English phonological space; therefore, it could be a source of sound learning difficulty. (Alwabari, 2013)
The fourth aspect is “gestural constellation” (Browman et. al., 1990a) or the assembly of multiple simple gestures in the production of a single utterance. (Alwabari, 2013)
To illustrate, multiple gestures coordinate together in articulating Arabic /s̥/. These gestures include forming a median depression (groove) along the upper surface of the tongue;
raising the tongue blade to touch the alveolar ridge (anterior, close constriction); retracting the root of the tongue to approach the back wall of the pharynx; elevating the hyoid bone and raising the larynx. (Alwabari, 2013). This combination cannot be found in the Finnish /s/, that's why it does not fall in the same phonological space and so can be difficult to harness.
2.6.3. Flege’s (1995) speech learning model (SLM):
perceptual assimilation model accounts for the speech production as being mediated by its perception. On the other hand, speech assimilation model pays attention to the speech perception only. However, the term native phonological space, is fundamental to both.
Central premise of SLM is the proposal that individuals’ NL phonetic systems “remain adaptive” over time and that they are susceptible to reconfigurations and additions as a result
There are Seven different sub-hypothesis. Relative to our study are First hypothesis (H1,) and Second hypothesis (H2.) Below is a brief introduction of them, along with their relation to the present study.
The first hypothesis (H1) suggests that NL and TL sounds are perceptually related at an allophonic level rather than at a phonemic level, and ability to perceive and discriminate TL sound correlates with its position within the word. (Alwabari, 2013).
An attention was paid to place the targeted pharyngealized sounds and their plain counterparts in the middle of the word. Pharyngealization effect extends extends to the adjacent vowels (before and after the pharyngealized phoneme.) In Finnish, the pharyngealization effect is seen regarded as phonemic one in vowels. For example, there is / a/, and /æ/. They are both considered different phonemes for Finnish speakers, while they are considered as different allophones for Arabic speakers. On the other hand, in Arabic it is the other way around. /s/ and /s̥/ are regarded as different phonemes, but for a Finnish speaker they have only an allophonic difference.
Pharyngealized consonants have the effect of changing the allophonic feature of the adjacent vowel. In the same time, back and front vowel change the allophonic feature of adjacent consonant.
The perception of our subjects to the Arabic pharyngealized consonants can be an allophonic, as they relate it to their NL phonological environment. However, what would be the result of that?
The second hypothesis (H2) argues that a new phonetic category can be configured to represent TL sounds that differ from the NL sound perceptually. (Alwabari, 2013)
Arabic pharyngeals is a tailored example for this hypothesis. On the one hand, /ʕ/ finds it difficult to fit in the Finnish utterances. It is a new phoneme, perceptually it is new trying to find its way through the Finnish tongue. The subjects try to place it somewhere resulting in some mispronunciation like and open vowel (e.g., /a/,) or glottal stop /ʔ/. Nevertheless, they sometimes manage to realize and repeat it. They create a new phonetic category for it, but would this new category sound like a native’s?
/ħ/ sound learning for Finnish speakers learning Arabic as L2 finds its place in this hypothesis as well. /ħ/ perception can be quite problematic. For a native speaker, it could be related to /h/, however is it perceived in the same manner for a learner?
Flege predicts that although TL learners do create new phonetic categories for
TL sounds does not necessarily permit them to produce the TL sounds as authentically as the native speakers do. (Flege, 1995, pp. 243)
3. Data and methodology
When we look at the Arabic language, we find that it is very rich in consonants. Another fundamental characteristic in it is the pharyngealized consonants. There are four of them in Arabic, namely, /t̥/, /d̥/, /s̥/, and /ð̥/. At the same time, there is the same non-pharyngealized versions of them: /t/, /d/, /s/, and /ð/.
Looking at the Finnish language, we find that it uses voiceless consonants (e.g., /t/ and / s/) way more than their voiced counterparts (i.e., /d/, and /z/). This was also the reason behind picking two voiceless plain consonants and their pharyngealized / voiced versions of them /s/, /s̥/, /t/, /t̥/, /d/, /d̥/, /ð/, and /ð̥/. The study aims to answer the primary question related to pharyngealization. However, a minor question to be answered is whether pharyngealizing voiced consonants is easier or pharyngealizing voiceless ones. Another question to be answered in this study is related to Pharyngeals. The hypothesis suggests that the voiceless pharyngealized consonants would be easier for Finnish speakers to pronounce as it is in the same phonological environment of the Finnish language.
3.1. Pilot study
In order to prove this hypothesis, we carried out a pilot study in which we recorded with one native Arabic speaker, from Saudi Arabia, and one Finn who has been learning the Arabic language for over two years now. Twenty sentences were prepared as cards, from which the subjects were asked to read in a language studio.
Figure 01 - An example card (though it was given without the phonetic transcriptions)
A total of eight consonants are introduced in this study. Four are voiced along with their voiceless counterparts. Four are pharyngealized and their plain counterparts. Each sound is repeated five times. Additionally, ten distractors were included.
The sentences were then trimmed into words’ audio files. Subject words were analyzed using Praat acoustic analysis software.
Figure 02 - An example waveform and spectrum of the sound /t̥/ pronounced by a native Arabic speaker (on the left) and a Finn learning Arabic as a foreign language (on the right
hand).
F1 and F2 were collected, and the results are shown in the following table:
Table 01 - the average of F2 collected in the eight different sounds from a native Arabic speaker and a Finnish learner of Arabic as a second language.
It is quite a known fact that a low F2 at consonant release is an essential cue to pharyngealization in Arabic. The results confirm that especially in /t̥/, and /d̥/.
The practice was very beneficial for me as I got involved in working on something very much related to the thesis. The biggest research project I worked on before was a baby thesis for my BA degree. I was intimidated by working on a big project like a master thesis, but this pilot study paved the way for me. I realized that all I have to do is do the same on a bigger scale, meaning doing more analysis and getting more results. However, the methodology would be quite the same. Then I just realized that I would like to do the same study on a bigger scale so that we may get more significant results.
I also faced some difficulties and made some mistakes that were avoided in this study like:
The difficulty of cutting the files to get them ready for the Praat analysis. We used some online website for cutting the audio files called http://mp3cut.net/, but sometimes it was not accurate in cutting the files, so sometimes some sounds dropped, and we had to work on the file again. It is a good idea to give some ample space before and after the target word even if it included some other sounds as it is pretty easy to exclude them later on Praat. This problem was prominent in the native speaker’s file as he was more fluent and sounds were very close
to each other. The best solution for that though is to use a reliable acoustic analysis software, like Praat, in editing and analyzing the files acoustically from A to Z.
Another mistake that we did was the repetition of words. We repeated two words in different examples. Maybe it could have been better to bring different words to get better and more accurate results in the end.
The fact that all the sounds under investigation here are put at the beginning of the words might help in getting more accurate results, but I am not sure whether this helps in the authenticity of speaking the language or not. This idea turned out after discussing with my supervisor to be a rather big mistake. Acoustic analysis of consonants is complicated in the sense that we cannot get formant values from it to analyze reliably. Hence, and after reviewing previous similar studies, came the idea of analyzing adjacent (before and after) vowels. The consonant pronunciation affects the preceding and following phoneme that is adjacent to it. The good idea is to surround the target phoneme with vowels. Vowels have the most explicit formant values, and that makes analysis easier and more reliable.
During this pilot study, we also got learn that both waveform and spectrograms are fundamental in any acoustic analysis. Waveforms were used mainly, in this study, to decide where the phoneme starts and ends. It is more evident there than in the spectrogram.
On the other hand, We need to look at the spectrogram in order to see the difference in pharyngealization as the pharyngealized consonants tend to have a lower F2. That is why we had to make use of both the waveform and the spectrogram. The methodology I used for identifying the beginning and the end of the consonant was to mark at the peak of the waveform just before it starts to have its familiar shape of the vowel and go back till the beginning since formants started to appear. It is a brief period, and that is why analyzing consonant sounds can be a bit tricky. Then we got the mean F1, and F2 of the selected area and recorded it.
3.2. Data collection methodologies background
We noticed that these methods were majorly used: nasoendoscopy, videofluoroscopy, Electromagnetic articulography (EMA), and spectrography in analyzing pharyngealization.
The study scope is only limited to acoustic analysis (sound waves and spectrograms).
However, it was crucial to read about other methodologies for a better understanding of the meaning of formants (spectrographic analysis). We found that listing them here would also help the reader for a better understanding of the thesis in general and the formants’ analysis in particular.
3.2.1. Methodologies definitions
Here we list the essential pharyngealization analysis methodologies on which most related studies use. Then we end with the methodology I followed in my study:
3.2.1.1. Nasoendoscopy
An investigation performed using a long flexible tube with a bright light at the end which is inserted through one of the subject’s nostrils and can be positioned to look down into the pharynx. It sends the digital images for display on a monitor at a sampling rate of 30 frames per second. After capture, the images were replayed in slow motion with the synchronized audio signal from the built-in audio recorder. The frame of maximum articulatory constriction for each target sound was identified and a copy taken for analysis. (Al-Tamimi, 2011)
3.2.1.2. Videofluouroscopy
A technique for viewing and recording real-time X-ray imaging using a video camera and must be carried out in the radiology department of a hospital under medical supervision. It can deliver a lateral view of the oropharyngeal areas and render visible movements of the jaw, tongue, epiglottis, and larynx in the vertical and front-back planes. Images are displayed on a monitor at the sampling rate of 30 frames per second. (Al-Tamimi, 2011)
3.2.1.3. Electromagnetic articulography (EMA)
A modern technique to record articulatory data. A sampling rate of 200 Hz gives an outstanding temporal resolution. That allows the capturing of excellent speech movements. In the following sections, we present the EMA data, extracted from an ongoing project on the direction of pharyngealization co-articulatory effects in the Arab world, to provide an interpretation of the different articulatory observations. (Embarki, 2011)
Spectrography and sound waves are analyzed after recording audio files using a decent voice recorder usually in a language studio (where it allows to capture only the voice of the subject without any background noise). Later on, these data are opened and analyzed acoustically with the help of sound analysis software (e.g., Praat, emu).
Another exciting method was observed in (Yeou, 2011), where “Idealised models based on realistic vocal-tract area functions are proposed for Arabic uvular /χ/, /ʁ/ and pharyngeal / ħ/, /ʕ/ and compared against data from four male speakers of Moroccan Arabic (MA).” (Yeou, 2011). These idealized models were used to create “versions of [aCa] sequences… and subjected to perception tests” (Yeou, 2011).
Another method was taking the “center-frequency measurements … of F1, F2 and F3 from combined FFT and LPC spectra using a 512-point Hamming window setting at the onset of the vowel following the target consonant in the context of initial position, and at the offset of the vowel preceding the target consonant in the final position context. LPC coefficients were adjusted for sampling rate (11025 Hz) and speaker gender.” (Al-Tamimi, 2011)
Something else which sounded interesting is converting the acoustic values into the Bark scale to model perception (Al-Tamimi, 2011).
A different way of analyzing the acoustic data was Locus equations (LEs), outlined by Lindblom (1963). They “are linear regression functions derived by plotting the onset frequencies of the F2 transitions of different vowels on the y-axis against their F2 steady states on the x-axis. The formula is given in (1):
F2mid – F2onset = k*F2mid+c (where k and c are slope and intercept, respectively).
A relatively flat slope indicates minimal vowel co-articulatory effects, in which case F2
articulatory resistance of the consonant articulation to vowel effects). On the other hand, a relatively steep slope indicates maximal coarticulation of the consonant with the vowel as F2 onset tends to have the same frequency as the vowel steady state (minimal coarticulatory resistance of the consonant articulation to the vowel). (Embarki, 2011)
One way to study the coarticulation effect of pharyngealization is to compare the dynamics of the articulation of sequences containing pharyngealized consonants with similar sequences containing their plain cognates. (Embarki, 2011)
Frequency measurements were made of vowels immediately following and preceding the pharyngeal consonants /ʕ/ and /ħ/. (Hassan, 2011)
3.3. Data collection and methodology
Many things have changed from the pilot study conducted above. The data was collected from 10 Finnish speakers learning Arabic as a second language. The methodology is called “rapid-reading paradigm,” and did not involve any reading. The reading process during the pilot study proved to have an effect of pronunciation. Not all the learners are in the same level of recognizing the Arabic alphabets and reading them correctly. That is why the “rapid- shadowing paradigm” was used here, and it proved a success in this study. It involves a beforehand recording of a native speaker. Then recording is then played to be repeated directly after listening by the subjects (learners). Meaning that the subject listens to the recorded word or phrase in Arabic and tries to repeat it, as best as s/he can.
Then, we created simple powerpoint slides. Each slide includes one Arabic word that plays automatically when the slide is turned. The subject then has his/her headphones. The subject listens and repeats. At the same time, Praat is on recording in the background mono .wav file.
*
Figure 03 - a snapshot of one of the slides used for conducting the study using the “rapid- shadowing paradigm."
After each session (lasted for 10 minutes on average). We saved the recording and then cut each word in a separate file (having 12 target sounds * 5 repetitions = 60 different files).
there was also the native recording that was used to create the slides (12 sounds* 5 repetitions= 60 files).
Later on, the files were saved as an extended file each containing 60 words in .wav format. Each of these files then was cut into 60 different ".wav" files using Praat. When cutting the files, special attention to leave a silent gap before and after the targeted sound carrying words were considered.
Each of these files was taken then to be annotated, using Praat also. The words were annotated in a manner that: Not all the phonemes annotated, but rather to annotate the target sound, preceding vowel, and following vowel. Those phonemes are the ones related to our study, so we did not waste time and effort annotating the whole words. What helped in that was reviewing previously conducted studies. For a thesis study, in my opinion, it is crucial to run a study that other researches has gone through previously. It is our first study, and if there is no reference to get back to it will be very hard and messy.
Later on, the files were saved and categorized as learners and native. Where the native category included only one speaker with 60 different words, and the learners included ten different speakers categories named in initials and categorized. Each category of them included 60 different files.
Praat was used again in opening these files and collecting the formants (F1, F2, and F3) of the target phoneme, and adjacent vowels (before and after). Then, the formants were collected and categorized in a spreadsheet in which many tables were created for investigating the goodness of the files acoustically.
Many tables were created for different purposes. The most important ones of them will be included in the next chapter of “results and discussions” for displaying what results the study uncovers. By the end of the chapter, there will be a discussion, where we can offer some findings based on this study and discuss them. Moreover, there will be suggestions about future studies that may help in pushing the research into new areas in that field.
4. Results
The results of the pilot study showed that the production of voiceless pharyngealized consonants is easier for Finnish speakers than producing voiced pharyngealized consonants.
However, is too simplistic to rely on its results. That is why a more advanced study was conducted. I would not say it is a fully fledged one, as we rely here only on acoustic analysis data. Instead, it is good enough to produce reliable results along with references from similar previous studies.
In this chapter, we are going to preview some figures of the data collected. Then, we are going to comment on them and discuss what these results may be interpreted. The main aim though has to be stressed once again, as the data is quite big and can answer many questions.
However, instead, we would focus our attention on answering the following questions:
A. How good are Finnish speakers learning Arabic as L2 in pharyngealization (turning an allophonic feature into a phonemic one)?
B. Pharyngeals, a novel two consonants for the Finnish language. Is it possible for Finnish speakers to learn them?
C. Which is easier to learn, for Finnish speakers learning Arabic as L2: Pharyngealized consonants or pharyngeals?
D. Is it true that because voiceless consonants outweigh voiced ones in Finnish that Voiceless pharyngealized consonant will be easier for them to learn than voiced pharyngealized ones?
The structure of this section will be in the following manner: We are going to take each question in a subsection and provide the needed analysis for answering it. This chapter will include data from this study and background results of previous studies to support our discussion. We find this crucial in our study for many reasons. First, we want our study to be linked with previous ones so all can form a more significant scientific body that always grows. Second, our study uses acoustic analysis methods only, and we find it necessary to make use of other physiological analysis methods as a background to it. They will give depth
to our acoustic analysis findings, and simply will make more sense. Moreover, it will help the reader to understand more what we are arguing about.
4.1. How good are Finnish speakers learning Arabic as L2 in pharyngealization (turning an allophonic feature into a phonemic one)?
In order to answer this intriguing question we need to follow some steps: First, fetch Finnish speakers learning Arabic as L2 data of pronouncing Pharyngealized consonants, and their counterparts. Second, compare those data to a native’s. Third, apply some statistical analysis methods to get a decent percentage (if there is any). Fourth, review previous studies related findings along with ours. Fourth, link second language learning theories to our findings, especially in the area of phonemes and allophones. Fifth, relate everything together and come up with a result. Finally, discuss this result and its possible meaning, and effect on language learning methodologies.
4.1.1. Data presentation and acoustic analysis
Table 02 - Pharyngealized Vs plain in Learners
Learners’ values
Phoneme F # Av adjacent Vs Phoneme F # Av adjacent
Vs
/t/̥
F1 682
/t/
F1 701
F2 1464 F2 1620
F3 2381 F3 2492
/s̥/
F1 702
/s/
F1 671
F2 1460 F2 1573
F3 2541 F3 2512
/d̥/
F1 695
/d/
F1 666
F2 1343 F2 1671
F3 2575 F3 2549
/ð̥/
F1 674
/ð/
F1 674
F2 1371 F2 1635
F3 2490 F3 2523
Average
F1 688 1.015
Average
F1 678
F2 1409 0.8671 F2 1625
Table 02 shows the learners’ data formants on the one hand. On the other hand, table 03 shows the native’s data formants. Most important part to look for in both tables would be the ratio of difference at the bottom middle of each.
The data we have here goes under previous studies in pharyngealization. Here we have
“a consistent raising of F1 frequencies” (Al-Tamimi, 2011). The ratio is slightly higher in the native’s data than the learners’. Nevertheless, the fact that in the learners’ F1 data was always higher than in their plain counterparts proves that there was a successful attempt for pharyngealization. Which takes us to the fact that there was an attempt for a constriction (pharyngealization) in the rear half of the vocal tract. “A formant frequency is raised by a constriction which is closer to a velocity minimum. For F1, the velocity maximum is at the
Table 03 - Pharyngealized Vs plain in Native
Native’s values
Phoneme F # Av adjacent Vs Phoneme F # Av adjacent
Vs
/t/̥
F1 639
/t/
F1 627
F2 1240 F2 1634
F3 2698 F3 2477
/s̥/
F1 654
/s/
F1 610
F2 1202 F2 1594
F3 2714 F3 2517
/d̥/
F1 642
/d/
F1 610
F2 1099 F2 1609
F3 2794 F3 2514
/ð̥/
F1 622
/ð/
F1 591
F2 1108 F2 1580
F3 2705 F3 2546
Average
F1 639 1.0492
Average
F1 609
F2 1162 0.7244 F2 1604
F3 2728 1.0855 F3 2513
lips and the minimum at the glottis, so any constriction in the rear half of the vocal tract will tend to raise the F1 frequency."
“The most consistent and obvious acoustic difference between plain and emphatic coronals is seen in the F2 frequency in adjacent vowels at the border with the consonant.” (Al-Tamimi, 2011). Data from tables 02 and 03 follows the same pattern. There is a consistent undeniable decrease of the F2 formants in both learners’ and native’s F2 formant frequency values. The lowering of F2 is established to be the “principal acoustic correlate” (Watson 2002: 270). Hence, this is a substantial proof of a successful attempt from the learners of pharyngealizing the subject consonants. They listened to them, noticed the difference and repeated them correctly. However, not in the same values of the native values as here also we find stronger pharyngealization values in the native’s data. (Al-Tamimi, 2011) F2 lowering is a reflection of a constriction at its velocity maximum. “There is an F2 velocity maximum in the oropharynx, in the zone bounded approximately at the lower extremity by the laryngeal aditus and at the upper extremity by the uvula. The results from measurements of F2 frequencies are therefore consistent with the constriction seen in the videofluoroscopic images to occur at the level of CV2 which is at about the mid-point of the oropharynx where the velocity maximum for F2 will be located. It is also consistent with the constriction seen in the nasoendoscopic images formed by the epiglottis and the pharyngeal walls, often with the tongue root bunching up against it in a dome-like shape.” (Al-Tamimi, 2011)
“The situation with F3 is more ambiguous.” (Al-Tamimi, 2011) Likewise, it is ambiguous in our study formant values. Sometimes higher (e.g. /s̥/, and /d̥/), sometimes lower (namely, /t̥/, and /ð̥/). However, in the native’s data F3 is consistently higher in pharyngealized consonants. Nevertheless, most previous studies did not rely on it and only relied on F1 and F2. “In all vowels after an initial emphatic, the mean frequency for F1 was higher than after a plain consonant, while for F2 was lower. All tokens conformed to this pattern.” (Al-Tamimi, 2011). Hence, the F3 value is a topic that is open for discussion;
however, we are not going to rely on its values here in any results, because of its ambiguity situation.
In the instances of the ambiguity of F3, a discussion of the vowel effect was carried out in Al-Tamimi’s (2011): “The context in which there is the greatest difference in F3 values between plain and emphatic consonants is at the onset of /uː/. Differences in other vowel contexts are quite small and less consistent. After an emphatic, F3 at the onset of /uː/ is some 300–450 Hz higher than after a plain consonant. There is a velocity minimum for F3 in the upper part of the oropharynx in the region of the uvula. It would appear, then, that the secondary articulatory constriction for emphatics in the context of /uː/ in JA may be higher in the oropharynx than it is for other vowel contexts. The difference in location of the constriction may even be greater than is suggested by the F3 values because of the resonance- lowering effect of the lip-rounding inherent in /uː/.” Hence, it can be a complication to rely on its results.
F2 lowering is the primary clue for pharyngealization. “Lowering of F2 has been the most consistently reported acoustic exponent of emphatics for different Arabic varieties.” (Al-Ani 1970; Odisho 1973; Ghazeli 1977; Giannini & Pettorino 1982;
Bukshaisha 1985; Laufer & Baer 1988; Heselwood 1992; Al-Masri & Jongman 2004;
Zawaydeh & de Jong; Al-Tamimi 2011 & Heselwood 2011).
Additionally, “A raised F1 suggests a constriction in the post-uvular pharyngeal area, an unchanged F1 suggests uvular constriction, while velar constriction would cause F1 to lower (Khattab et al. 2006: 138).
How does this happen is explained in Al-Tamimi (2011): The bunching of the tongue in front of the epiglottis observed in the production of many of the emphatics has the effect of reducing the pharyngeal volume and, together with the epiglottal retraction, lends support to the suggestion that emphatic coronals are produced with pharyngealization.
In another instance in the same study, Al-Tamimi (2011) mentions: “The epiglottis retracts and lowers consistently to a greater degree during the production of an emphatic coronal than during the production of a plain coronal.” This clarification concludes that the epiglottis also plays a significant role in pharyngealization, and not only in pharyngeals.
However, his cannot be considered exclusive though as Ali and Daniloff’s (1972) study suggest that it can also be uvularization: “cinefluorographic study of Iraqi Arabic found emphatics to exhibit simultaneous slight depression of the palatine dorsum, a rearward movement of the pharyngeal dorsum towards the posterior pharyngeal wall, and a lowering
velum towards a rising tongue dorsum. Although the data suggest pharyngealization, they do not exclude uvularisation, and the height of the larynx was not evaluated.” Ghazeli’s (1977) cinefluorographic study of Tunisian reports similar results as well.
Another observation of the tongue movement is reported in “Al-Nassir’s (1993) X-ray study of Iraqi, Laradi’s (1983) videofluorographic and endoscopic study of Libyan, Laufer and Baer’s (1988) endoscopic study and Al-Tamimi & Heselwood’s (2011) nasoendoscopic and video-fluoroscopic study of Jordanian coronal emphatics, have all been able to observe a rearward movement of the tongue root towards the back of the upper pharynx and a depression in the tongue dorsum.” (Embarki, 2011).
Another explanation for F1 and F2 formant frequencies in emphatics is suggested by Hassan (2011): If F1 were consistently higher, and F2 lowered, this would constitute good evidence for laryngeal constriction as a function of aryepiglottic sphinctering, tongue retraction, and larynx raising. However, the characteristic lowering of F2 could be indicative of a pharyngeal co-articulatory gesture where the larynx is lowered.
Embarki (2011) conducted a study on emphatics and their plain counterparts in Arabic.
Our study presents similar values of F1, and F2 (F1 raising and F2 lowering in emphatics) as his. As he mentions: “The pharyngealized consonants /t̥, d̥, s̥, ð̥/ in MSA presented variable slope values, a one-way ANOVA showed significant effects of the consonant on slope values (F(3, 63) = 4.86, p < .01). For the plain consonants /t, d, s, ð/, the effects of consonant on slope values were not significant (F(3, 63) = 1.23, p = .304).
We come out of the previous studies findings that F2 is the primary cue for pharyngealization and that F1 raising is the secondary one. On the other hand, the F3 case is a complication that is not necessary for analyzing pharyngealization. Our results show a consistent rise in F1 along with a consistent considerable drop in F2 frequency values. That is why we can say that Finnish speakers learning Arabic as L2 managed successfully to turn an allophonic characteristic in Finnish language (pharyngealization) into a phonemic one. Not in a native-like manner, but reasonably good enough.
