Gemination in Swedish and Arabic with a Particular Reference to the Preceding Vowel Duration

Dept. for Speech, Music and Hearing Quarterly Progress and Status Report

Gemination in Swedish and Arabic with a particular reference to the preceding vowel duration. An instrumental and comprative approach.

Majeed, H. Z.

journal: Proceedings of Fonetik, TMH-QPSR volume: 44 number: 1 year: 2002 pages: 081-084

http://www.speech.kth.se/qpsr

TMH-QPSR Vol. 44 – Fonetik 2002

Gemination in Swedish & Arabic with a particular reference to the preceding vowel duration. An instrumental & comparative approach

Zeki M. Hassan Associate Professor. A Guest Researcher at the Dept of Linguistic, University of Gothenburg.

Abstract Speech materials recorded by native speakers of Swedish and Arabic have been transferred to a computer system ”Sound Edit 16 version 2”. Acoustic time measurements in msec. have indicated that geminate durations outweigh those of non- geminates in both languages and go beyond the JNDs or DLs. The durations of vowels preceding non-geminates in Swedish also outweigh those preceding geminates and go beyond the JNDs whereas in Arabic this difference is found to be insignificant and hover around the JNDs. It does seem feasible to speculate that the first three differences are maintained by the speakers of both languages for phonological reasons whereas the latter is exerted by the articulatory nature of gemination and hence is phonologically redundant.

Phonological background In classical Arabic (C.A), on the other hand, the vowel system is triangular, i.e it consists of The Swedish vowel system is traditionally three long vowels /aa/, /uu/ and /ii/ with three known to consist of nine vowels with long and short counterparts /a/, /u/, /i/. Two more long short counterparts. The long short dichotomy is vowels /ee/ and /oo/ are added to the system in referred to by most Swedish phonologists as Iraqi Spoken Arabic (I.S.A) to make the number allophonic. (Elert, 1964; Eliasson & La Pelle, eight vowels. (Al-Ani, 1970; Odisho, 1973; 1973; Czigler, 1998, among others ). This Hassan, 1981, among others). Vowel length is dichotomy is accompanied by striking difference indisputably distinctive in both C.A and I.S.A. in quality in almost all nine vowels except contrary to length in Swedish vowel system perhaps e.g. / ε / as in /vε:g/ vs./ vεg:/. where it is often referred to as allophonic. However, it has been debated for quite Consonant length (gemination) is also sometime whether quality or quantity has the indisputably phonemic in Arabic no matter most dominant distinctive role. Recent studies whether it is preceded by short or long vowel have shown that length is the primary parameter e.g. /'adad/ vs. /'addad/. Consonant clusters distinguishing Swedish vowels, although quality are not as prevalent in Arabic as in Swedish and could not be excluded as another perceptual cue. the most dominant syllabic structure in Arabic is (Hadding-Koch & Abranson, 1964; Czigler & of the type CVCV /'kataba/, /ki'taaba/, Behne, 1996; Czigler, 1998). /ki'taabun/ and the most dominant post vocalic The most pertinent fact to this study is that consonant cluster is of the type /CVC:/, /CV:C:/, within the same morpheme in Swedish a phono- /CVCC/, e.g. /samm/, /saamm/, /dars/. (Al-Ani, logically long stressed vowel is followed by a 1978; Odisho, 1973; Hassan, 2000, among short consonant and a phonologically short others). vowel is followed by a long consonant There has been a myodynamic, aero- (geminate) or a consonant cluster e.g.[ ti:ta] vs. dynamic and acoustic evidence (Hassan, [ ] [ [ ] tIt:a or ø:p] vs. œpt . Swedish phonologists 1981) that geminates shorten the preceding do not seem certain whether consonant length vowel but the durational differences have (gemination) or vowel length has the more been found phonologically redundant. distinctive role.

81 Speech, Music and Hearing

Experimental procedures vs. non-geminates with the preceding vowel durations. They also show that acoustic Bisyllabic words have been recorded by native durations of geminates very significantly out- speakers of both Swedish and Arabic at the weigh those of non-geminates for both Department of Linguistics, Gothenburg Uni- languages. From perceptual point of view they versity. They have been used in a carrier sen- all go beyond the JNDs or DLs values. Vowel tence where each minimal pair is repeated six durations preceding non-geminates are also times for both languages. The Swedish carrier significantly longer than those preceding sentence: Jag sa --- andra gången sen “I said --- geminates and go beyond the JNDs in Swedish. once more”. The Arabic carrier sentence: /iktibu In Arabic, however, vowels preceding non- --- sit mar`raat/ “write --- six times”. geminates are slightly longer than those The informant for the Arabic material preceding geminates and the difference hover represents educated Iraqi Arabic (Hassan, 1981) around the JNDs. whereas the informant for the Swedish speech It does seem feasible to speculate that material represents standard central Swedish consonant length as well as vowel length in (Elert, 1964). Both informants are in their forties Swedish, in the phonetic context investigated, and have no history of hearing or speech are language specific phenomena, i.e., pathology. maintained by Swedish speakers for phono- Each informant took part in a non-stop logical purposes. Hence they can be regarded as recording session in a sound attenuated room very prominant distinctive cues for listeners. where the recording settings and the instruments The same goes for consonant length in Arabic used; Power Macintosh G3, sound card: E- despite the difference in their JNDs values from magic Audiowerk 8 (analogue input). Recording those of Swedish as shown in (Figure 1) which software: BIAS Peak VST v.2.5.1. Format: 16- can be attributed perhaps to the different speech bit, 44,1 kHz AIFF. Input mixer: Spirit Folio F1 rates of the two informants. (Built-in 48v Phantom power). Input fader level: The shorter vowel duration preceding 0 dB. Master fader level: 0 dB. Mic: AKG geminates in Arabic, on the other hand, seem to C460B. Mic low cut: 70 Hz flat. Mic distance: be language universal phenomenon, i.e. exerted 15 cm. by the articulatory nature of gemination and the The recordings were transferred to the resulting myodynamic and aerodynamic condi- computer system of Sound Edit 16 version 2, tions, see Hassan (1981). Hence they are copywright 1990-1996 – div – Hac 5. S/N phonologically redundant. 30/20-0400-1754-60016. Licenced from Fraun- hofer Gesellschaft / 11S. Table 1. Average values of vocoid duration in For each utterance a file containing a duplex Iraqi Spoken Arabic for minimal pairs con- oscillogram waveform and a spectrogram and a taining geminate vs. non-geminate consonants. label file were made. In order to calculate the Each is the average value of six tokens. segmental durations, acoustic measurements in msec. were made within the target V:C, VC: Minimal Pairs msec. msec sequences. These acoustic segments were / 'adad / vs. / 'addad / 100 85 identified from the waveform and spectrograms [' adad ] vs. ['ad:ad] by positioning a cursor at time points in the /'raqam / vs. / 'raqqam / 103 88 waveform as well as the segment onset and [' raqam ] vs. [' raq:am] offset on spectrograms. For segmentation / ' alam / vs. / ' allam / 97 85 criteria, see Fant (1958, 1960) and Peterson &   : Lehiste (1960). Aspiration has been excluded ['alam] vs. ['al am] from both contoid as well as vocoid durations. / fa'laa / vs. / fa'llaa / 85 67 Durations of the target segments have been [fa'la:] vs. [ fa'l:a:] automatically measured and shown by the / 'katab / vs. / 'kattab / 76 75 computer system and kept on the label file. [' katab] vs. [' kat:ab] / ha'maam / vs. / ha'mmaam / 84 53 Results and conclusions [ ha'ma:m] vs. [ ha'm:a:m] /'qada / vs. /'qadda / 96 80 Tables 1-8 as well as (Figure 1.) show individual ['qada] vs. ['qad:a] measurements and pooled values for geminates

82 TMH-QPSR Vol. 44 – Fonetik 2002

Table 2. Average values of vocoid durations in Table 5. Average values of vocoid duration in msec. in Iraqi Spoken Arabic for minimal pairs Swedish for minimal pairs containing geminate containing geminate vs. non-geminate con- vs. non-geminate consonates. Each is the aver- sonants. Contoid as well as vocoid enviroments age of six tokens. are pooled. Minimal Pairs Msec. Msec. Pre-non-geminate Pre-geminate lås vs. loss 267 128 92 76 /' lo:sa / vs. / 'los: a/ ['lo:sa] vs. [ 'ls:a] - Average vocoid duration = 84 vira vs. virra - Mean difference =16 / 'vi:ra / vs. / 'vira / 259 147 ['vi:ra] vs. ['vr:a] rama vs. ramma / 'ra:ma / vs. / 'ram:a / 267 157 Table 3. Average values of contoid duration in ['r:ma] vs. ['ram:a] msec. in Iraqi Spoken Arabic for minimal pairs leda vs. ledda containing geminate vs. non-geminate con- /'le:da / vs. /'led+d+a / 276 144 sonants. ['le:da] vs. ['ld:a] håla vs. hålla 302 115 Minimal Pairs Msec. Msec. /'ho:la / vs. /'hol:a / / 'adad / vs. /' addad / 83 137 ['ho:la ] vs. [' hl:a] [' adad ] vs. ['ad:ad] saba vs. sabba 322 140 /'raqam / vs. /' raqqam / 111 188 /'sa:ba / vs. / 'sab:a / [ 'raqam ] vs. [ 'raq:am] ['s:ba ] vs. ['sab:a] / 'alam / vs. / 'allam / 80 176 tita vs. titta 226 89 /'ti:ta / vs. / 'tit:a / ['alam] vs. [ 'al:am] [ 'ti:ta ] vs. ['tt:a ] / fa'laa / vs. / fa'llaa/ 64 162

[ fa'la:] vs. [ fa'l:a:]

/ 'katab / vs. /' kattab / 111 197

[ 'katab] vs. [' kat:ab]

/ ha'maam / vs. ha'mmaam/ 63 150 Table 6. Average values of vocoid duration in [ ha'ma:m] vs. [ ha'm:a:m] msec. in Swedish for minimal pairs of bisyllabic qada qadda 76 160 /'  / vs. /'  / words containing geminate vs. non-geminate [ ] [ ] 'qada vs. 'qad:a consonants. Contoid as well as vocoid environments are pooled.

Pre-non-geminate Pre-geminate Table 4. Average values of contoid duration in 274 131 msec. in Iraqi Spoken Arabic for minimal pairs -Average vocoid duration = 202 of bisyllatic words containing geminates. Non- -Mean difference = 143 geminate consonants. Contoid as well as vocoid enviroments are pooled.

Non-geminate Geminate

84 167 -Average contoid duration =124 -Mean difference =80

83 Speech, Music and Hearing

Table 7. Average values of contoid duration for Table 8. Average values of contoid dura- minimal pairs containing geminate vs. non- tion in msec. in Swedish for minimal pairs geminate consonants. Each is the average value of bisyllabic words containing geminate vs. of six tokens. non-geminate consonants. Contoid as well as vocoid environments are pooled. Minimal Pairs Msec. Msec. Non-geminate Geminate lås vs. loss /' lo:sa / vs. / 'los: a/ 203 368 158 299 ['lo:sa] vs. [ 'ls:a] -Average Contoid Duration (C.D) = 228 vira vs. virra -Mean difference= 141 / 'vi:ra / vs. / 'vira / 137 274 ['vi:ra] vs. ['vr:a] rama vs. ramma / 'ra:ma / vs. / 'ram:a / 151 271 References ['r:ma] vs. ['ram:a] leda vs. ledda Al – Ani S H (1970). Arabic phonology. Mouton. /'le:da / vs. /'led+d+a / 125 263 The Hague. Paris. ['le:da] vs. ['ld:a] Al – Ani S H and May, D.R (1978). The håla vs. hålla Phonological Structure of the Syllable in Arabic. /'ho:la / vs. /'hol:a / 141 285 Reprinted in Al-Ani ed, Readings in Arabic Linguistics. Indiana University, Linguistics Club, ['ho:la ] vs. [' hl:a] 113-125. Saba vs. sabba Czigler P E (1988). Timing in Swedish VC(C) /'sa:ba / vs. / 'sab:a / 123 323 Sequences PHONUM. Sweden, Umeå University. ['s:ba ] vs. ['sab:a] Czigler P E & Behne, D.M. (1996). Temporal Effects tita vs. titta on Successive Consonants in Swedish. Journal of /'ti:ta / vs. / 'tit:a / 230 308 the Acoustical Society of America. 100 (4,pt.2) [ 'ti:ta ] vs. ['tt:a ] 2689. Elert C C (1964). Phonological Studies of Quantity in Swedish, Uppsala: Almqvist & Wiksell. DSL Values (Henry 48) Eliasson S & LaPelle, N. (1973). Generativa regler The data of this study för svenskans kvantitet. Arkiv för nordisk filologi, 88, 133-148. s.v S.c Fant G (1958). Modern Instruments and Methods for 140 130 th 120 Acoustic Studies of Speech. Procedings of the 8 110 100 A.C International Congress of Linguistics, 282-358 90 80 Oslo University Press. 70 60 Fant G (1960). Acoustics Theory of Speech 50 Production The Hague, Mouton. 40 Hassan Z M (1981). An Experimental Study of Vowel 30 Duration in Iraqi Spoken Arabic Unpublished A.V 20 Ph.D. thesis U.K: Dept. of Linguistics & Phonetics, University of Leeds. 10 Hassan Z M (2000). English Phonetics & Phonology for Arab Students. Al – Hamid Publishing House, Amman. Jordan. Hadding-Koch K & Abranson A S (1964). Duration

0 50100 150 200 250 300 350 400 450 500 Versus Spectrum in Swedish vowels: Some Perceptual Experimentals. Studia Linguistica, 2, 94-107. Figure 1. Absolute DSL (difference limens) Peterson & Lehiste (1960). Duration of Syllable plotted against reference duration or average Nuclei in English. Journal of the Acoustical vocoid and contoid durations to compare the Society of America, 32 (6), 693-703. just noticeable differences (JNDs) with the durational differences of this study.