phonological use of the larynx, Schemes and Mind Maps of Voice

Download phonological use of the larynx and more Schemes and Mind Maps Voice in PDF only on Docsity! Vaissière, J., (1997), "Phonological use of the larynx: a tutorial", Larynx 97, Marseille, 115-126. PHONOLOGICAL USE OF THE LARYNX J. Vaissière UPRESA-CNRS 1027, Institut de Phonétique, Paris, France RÉSUMÉ Cette communication concerne le rôle du larynx dans l'acte de communication. Toutes les langues du monde utilisent des configurations caractéristiques du larynx, aux niveaux segmental, lexical, et supralexical. Nous présentons d'abord l'utilisation des différents types de phonation pour distinguer entre les consonnes et les voyelles dans les langues du monde, et également du larynx comme lieu d'articulation des glottales, et la production des éjectives et des implosives. Nous abordons ensuite l'utilisation de patrons caractéristiques de la fréquence du fondamental au niveau lexical pour contraster entre les mots dans les langues à tons et à accent mélodique. La participation du larynx pour transmettre des informations linguistiques et paralinguistiques au niveau de la phrase est traitée en dernière partie. INTRODUCTION The larynx is one of the main articulators in speech production, with the tongue root, the tongue body, the tongue blade, the velum and the lips. The goal of this tutorial is to review succinctly the different uses of the larynx for communicative purposes. The first part concerns the segmental use of the larynx: (i) the phonation types for differentiating consonants and vowels, (ii) the larynx as a place of articulation for the glottal consonants, and (iii) the initiation of airstream in glottalic consonants. The second part concerns the nonsegmental (prosodic) use: the participation of the larynx for contrasting higher linguistic units such as words, phrases and sentences (tone, accent and intonation ). The third part tackles the problem of the larynx used as a carrier of paralinguistic information . THE PRIMARY FUNCTION OF THE LARYNX IS PROTECTIVE As stated by Sapir, 1923, physiologically, "speech is an overlaid function, or to be more precise, a group of overlaid functions. It gets what service it can out of organs and functions, nervous and muscular, that come into being and are maintained for very different ends than its own" [1]. The primary function of the larynx is protective. It prevents the entrance of foreign material into the lungs. It also permits the building up of intrathoracic pressure. PHONATION IS NOT PARTICULAR TO HUMAN SPEECH The secondary function of the larynx is to provide phonation. During the production of speech, the vocal folds (VFs) are mainly in adducted position. They vibrate, alternatively opening and closing the glottis (the space between the VFs) for a very short period of time. The principles underlying vibration are now rather well understood. The excessive subglottal airpressure forces the closed VFs to go apart and the elastic recoil and the Bernoulli effect suck them again together without any muscle action. The vibrations divide the continuous stream of expired air coming from the lungs into puffs of air that will excite the supraglottic cavities. The quasiperiodic modulation of the respiratory airstream by the vibrations of the VFs provides the primary source of energy for the production of voiced sounds and this process is called phonation. Most of the speech sounds (vowels and most of the consonants) are voiced. The degree of voicing however greatly varies Vaissière, J., (1997), "Phonological use of the larynx: a tutorial", Larynx 97, Marseille, 115-126. EVOLUTION HAS FAVORED FINER VOCAL ADJUSTMENTS from one language to the next. A study by Catford has shown that French is voiced 78% of the time, while Chinese (Canton) only 41% (English: 72%, Russian: 61%, and Alkhaz: 56%) [2]. Phonation is not properly linguistic. The acoustic energy in nonspeech production of amphibians, primates, and human infants is generated in the same manner. The interruption of phonation (mainly during the consonants) will be used to create contrasts between consonants (as seen later). A morphofunctional consequence of the various neuro-cranial rearrangements caused by an increase in extreme prefrontal cortex size during hominization [3] has been a descent of the larynx in the neck. The descensus of the larynx in the human neck had rendered larynx and tongue movements more independent than otherwise. The intrinsic fundamental frequency (FO) of the vowels shows however a tongue-larynx interaction, supposedly via the hyoid bone [4]. A lower position of the larynx has made possible finer adjustments at the level of the larynx, particularly of the tension of the vocal folds (VFs) and the spacing between them. In amphibians, the tension of the VFs cannot be varied and the vocal repertoire is monotonous. For what concerns the spacing, an interesting fact can be noticed. The posterior cricoarytenoid (PCA) is the only abductor muscle and it is found only in man [5]: it will be particularly active for the production of voiceless and aspirated sounds. THE POSSIBLE CONTROLS OF THE LARYNX Let us resume succinctly the controls which are used linguistically. (consult [6] on the mechanisms underlying the control fundamental frequency). Figure 1 summarises the general trends. First, the subglottic pressure and the pressure across the glottis may be varied. Heightened subglottic pressure generally assists in making focus and prominence. Second, the length and the mass of the VFs can be directly controlled. Two muscles are mainly involved: the cricothyroid (CT) and the vocalis (VOC). The contraction of CT stretches the VFs. Lengthening results in an increase in the longitudinal tension of the VFs, which increases the rate of vibrations of the VFs (and therefore of the perceived pitch). Pitch variations is mainly used for tone, accent and intonation. The contraction of the VOC assists CT for pitch raising and may favor voicing at high frequency rate of vibration by slackening the cover of the folds [7], [8]. Third, the glottis (airspace between the VFs ) can be finely adjusted. The arytenoids cartilages may be pressed together or abducted, and there may be a medial compression of the vocal fold tissue. Slightly abducted glottis or slightly adducted glottis allows vibration of the VFs, in breathy voice, murmur, or slack voice, and in creaky voice, laryngealized voice, and stiff voice, respectively. The abduction of the glottis leads to air leakage and to the presence of noise. Strongly abducted or adducted VFs hinder voicing. Combined spacing between the VFs and tension are used in phonation types for segments (breathy, creaky vowels and consonants). Voice quality over larger units also carries also affective message, as seen later. Fourth, the whole larynx can be raised or lowered. First, the raising of the larynx decreases the volume of the supraglottic cavities, increases the VFs tension by stretching them, and increases the coupling stiffness in the vertical dimension of the larynx. A lowering movement has the reverse effects. The larynx movements are essential for the production of glottalic consonants. Second, they are used to assist F0 raising and lowering, through their effect on VFs length. Third, a lowered larynx may assist voicing by a 'bunching' of the surface issue (slackening) the VFs and participate to the phonetic implementation of the [+slack] feature for voiced consonants [9], [10]. Vaissière, J., (1997), "Phonological use of the larynx: a tutorial", Larynx 97, Marseille, 115-126. Hindi, for example, contrast between breathy- plain and voiced laterals. The reader is again referred to the book written by Ladefoged and Maddieson for more information. Most of the languages use difference in laryngeal settings to differentiate among stops and fricatives. They generally have a two-way contrast for the stops. The phonetic implementation of the two way contrast differs from one language to the next. French contrasts between voiced and unvoiced stops, while Chinese contrast between voiceless unaspirated and voiceless aspirated stops. Some languages have a three-way contrast. Thai contrasts between voiced unaspirated, voiceless unaspirated and voiceless aspirated. Hindi has a four way contrast: voiced unaspirated, voiced aspirated, voiceless unaspirated, and voiceless aspirated (see figure 2). - Voiced, unaspirated [b,d,g]: A two-way contrast may typically typically consist in opposing voiceless to voiced. The main articulatory difference is the presence versus absence of VFs vibration during the consonant production. According to the model developed by Halle and Stevens [22], the difference in voicing is essentially a function of VFs tension (see also Stevens, [23]). In voiceless consonants, the stiffness of the VFs hinders vibration, while in voiced consonants, their slackness renders vibration possible, even at low transglottal pressure. The terminology [stiff] and [slack] is not unanimously accepted (some prefer the more traditional feature [±voice]). A difference in stiffening of the VFs is difficult to measure and there is no indication that the glottal constriction is looser for /b/ than for /p/. But the commonly observed higher fundamental frequency (FO) in the vowel after voiceless consonants may be interpreted as an progressive assimilation of an increased stiffness. According to Stevens [10], stiffness of the VFs may be indirectly achieved by raising the larynx, and slackness by lowering it. In accordance to the observation that voiceless is considered as the unmarked state for obstruents by phonologists, voicing is indeed difficult to maintain in voiced stops. Vibration of the VFs is maintained for a limited number of pulses through a passive extension of the vocal tract volume, due to heightened supraglottal air pressure. Active maintenance of voicing can be done in a number of ways. First, voicing can be maintained by expansion of the pharyngeal walls and supraglottal cavities [24]. Active extension of the supraglottic volume is easiest for labials (by puffing the cheeks). It is the most difficult for back stops. Accordingly, /g/ is less frequent in languages than /b/ and /d/ [20]. Second, it can be sustained by an active slackening and thickening of the VFs. The slackening gesture may be absent in the phonetic realisation of /b,d,g/ in English, which often lack a voice bar during the closure [21]. In Stevens' description, the tension cannot be neutral for stops; it has to be either slack or stiff [22] [23]. -voiceless aspirated [ph,th,kh] The presence of aspiration versus its absence is often contrastively used in languages. The aspirated stops are generally preaspirated stops, but there may be postaspirated stops. Aspirated means postaspirated. The articulatory manoeuvers to achieve contrast based on aspiration have received a number of interpretation. Halle and Stevens (1971) interpret it essentially in terms of spreading of the glottis at the time of the release or after: [+ spread] for aspirated consonants and [-spread] for unaspirated stops [22]. Lisker and Abramson relate the contrast mainly to a different timing of laryngeal activity relative to the suprasegmental constriction (Voice Onset Time or VOT). VOT is longer in the case of aspirated consonants [25]. The contrast can be interpreted also in terms of greater abduction for aspirated stops. PCA is more active in aspirated voiceless stops than in unaspirated stops, and it is not just therefore a question of timing. The dominant manoeuver does not need to be the same for all languages. Acoustically, aspiration is characterised by the presence of turbulence noise at the time of Vaissière, J., (1997), "Phonological use of the larynx: a tutorial", Larynx 97, Marseille, 115-126. the release and after, with non-harmonic varying components. Figure :3: relative glottal aperture in Hindi consonants (after Kagaka and Hirose (1975) [26] For slack consonants, the larynx is lowered (F1 lowers), the VFs are vibrating more loosely than in the modal voice, there is a slightly increased glottal aperture, and a moderate increases in air flow, FO is lower in the following vowel. There is reduced energy in the upper frequency range of the spectrum. There is a continuum between slack and breathe voice. ejective implosi ves breathy voiced voiceles s aspirate d French b p English b ph Chinese p ph Korean p* p ph Thai b p ph Hindi bH b p ph Breathy voiced and slack consonants do not to contrast [19]. Figure 2: Source: Ladefoged and Maddieson, 1996. - breathy (aspirated) voiced [bH, dH, gH] AND slack consonants - creaky and stiff (laryngealized) consonants In creaky consonants, the arytenoids are closer than in modal voice [+constricted][- spread], and the tension is high [+stiff]. The VFs do not vibrate as a whole, or the parts of the VFs vibrate out of phase. Hausa contrasts between voiced, voiceless, “glottalized” consonants (In the bilabial and alveolar stops, the glottalization affects the following vowel as well). (for more information [19]). Voiced aspirated (also called breathy or murmured) stops are less common than voiceless aspirated stops. The VFs are vibrating during the closure of the stop, and the arytenoids are spread after the release. The VFs are supposed to be slack (by lowering the larynx). A more or less great deal of air is allowed to pass through the slightly open glottis. The figure below illustrates the glottal aperture aspirated and breathy stops, based on fiberoptic data. In stiff consonants, like in Thai, there is a slight degree of laryngealisation. No languages contrast modal voice with no more than one degree of laryngealized voice. t h d h t release-100 10050 -50 ms 2) Vowels breathy modal creaky glottal configuration volume velocity spectral tilt Vaissière, J., (1997), "Phonological use of the larynx: a tutorial", Larynx 97, Marseille, 115-126. acoustic characteristics bandwidth increase noise high frequencies decrease glottal formant bandwidth decrease no noise high frequencies increase Figure 4: phonation types of vowels. Adapted from Stevens, 1977 [10]. Glottal configurations, state of the arytenoids, volume velocity, spectrum of glottal pulse, schematised spectrograms, and acoustic characteristics in three phonation types for vowels: breathy, modal and creaky Most of the languages do not use any differences between glottal states to distinguish among vowels. A number of languages use two types of phonation (modal versus creaky or breathy). A few rare languages, like Mazatec, contrast between modal, breathy and creaky vowels [19]. - creaky and slack vowels The glottis is constricted. The glottal pulse is sharper as compared to the modal vowels, leading a very low intensity of the fundamental. There is more energy in higher frequencies, and the bandwidth of the formants are reduced. - breathy and stiff vowels The arytenoids are spread. Voice source pulses have a larger open coefficient and a nonabrupt glottal closure. There is a relative decrease of the amplitude of the harmonics in the middle and upper part of the spectrum and a consequent increase in very low frequencies: there is a higher amplitude level of the fundamental compared to the second partial [27]. There is a greater random component. Due to open glottis and subsequent loss due to the subglottal system, there is widening of the bandwidth, and the first formant, in particular, is less well defined (see figure below). - voiceless vowels In some languages (North America, see [19]) , voiceless vowels seem to contrast with modal vowels. Devoicing of generally is not contrastive; it is mostly due to assimilation with the surrounding voiceless context and concerns mainly close vowels, where the heightened supraglottal pressure results in a delay or a suppression of voicing. This process is regular in Japanese, and occasional in French. The next figure illustrates the features used by Stevens for describing the different types of phonation in vowels. The dimensions are the same than for the consonants. The different degrees of tension of the VFs (stiff, neutral, slack) contrast the vowels with high mid and low tones, respectively. B) The glottal consonants [h ], [H ] [? ] The larynx serves as the primary place of articulation for the stops [h ], [H ] [? ]. [h ], [H ] are voiceless and voiced, respectively. The glottal stop [? ] (coup de glotte in French) requires a very constricted glottis, and there is no voicing contrast possible. It is a speech sound in many languages (Arabic). In German, it is a regular characteristic of the beginning of the stressed initial vowels. In English, it is a device for marking morpheme boundaries in pair of words such as "an ice man" and "a nice man". It may also be a substitute for [t]. [Let us note a very interesting sound change: [H ] (or [h] and [? ] in coda position may disappear and be replaced by a falling tone on the preceding syllable and a raising tone, respectively [12]. stiff VFs neutral VFs stiffness slack VFs spread arytenoids voiceless vowels ë∞ breathy vowels A5 low-pitched breathy vowels neutral glottal position high pitched vowel A' vowel in the mid-range A low-pitched vowel A' constricted glottis glottalized vowels creaky vowels A£ low-pitched creaky vowels? Fig 5 Phonetic categories associated with different ways of producing vocal-fold stiffness-slackness and with various types of abduction- adduction of the glottis for the vowel (Stevens, 1977). Vaissière, J., (1997), "Phonological use of the larynx: a tutorial", Larynx 97, Marseille, 115-126. position of emphatic stress allows to draw a clear distinction between stress and boundary languages. When emphasised, the same (marked) stressed syllable is made more prominent in stress languages. In a boundary languages, emphatic stress falls in word beginning (while boundary tones are bounded to the final syllable). Word stress and word boundaries are realised by a combination of factors, not necessarily FO. The physiological correlates of stress are very complex and involves an higher voice excitation strength (calculated as the derivative of the glottal flow closing). Final lengthening mark the right boundary. The number of pitch patterns that the whole word may receive is rather large; ([35] for English). In neutral, isolated sentences, the choice of pitch patterns for the successive words are in close relation with the syntactic structure [36]. Other factors play a greater role in other contexts. Rather simple tests allow to separate languages in different type(see for example [37] for ten languages, etc... D) Intonation Pitch has an intonational function when it is used in domain larger than the word. There seem to be a number of similarities between languages in the use of FO at the supralexical levels. Two basic patterns can be recognised. The rising means non finality, openness, arousal of interest, intentness, non definiteness, unfinishness, continuation, questions. The falling pattern involves semantic finality [38], [39]. E) Positional effect on the larynx - The suprasegmental features [±tense] Supraglottic and laryngeal configurations for phonemes are different, depending in their position (initial, internal or final) in larger units. I have proposed to add systematically the positional feature [±tense] to the phonological specification of each phoneme for interpreting their allophonic realisation [40]. What are the consequences of the feature [±tense] on the larynx behavior in speech? - the initial position is characterised by a general tensening of all articulators. The stiffening of the adductory muscles laryngeal muscle results in less voicing or a delay of voicing, for voiced consonants. It results in glottalization or in glottal stop in words beginning by a vowel (for German [41], for English: [42], [43]). It leads to a larger glottal width for voiceless consonants (like in English) and then to aspiration. Diachronically and synchronically initial phonemes are resistant to change and less likely to disappear. Great care should be taken when taking data: the synchronisation between the activation of all articulators during the speech ready gesture is random to a certain extend [44]. - In medial, unstressed position, there is a tendency to diminish the amplitude or even suppress some of the gestures. The omission of the glottal opening gesture for the voiceless stops leads to voicing; /t/ is often transformed into a glottal stop ("butter" in American English). Synchronically and diachronically (sounds changes), intervocalic voiceless consonants are proned to be voiced. - Final position is marked by relaxation of all articulators, a slow down of the movements (related to final) lengthening, a decrease of the degree of stricture and even disappearance. At the laryngeal level, it is often marked early devoicing and breathiness. In a quite large number of languages (like German), the voicing distinction for stops and fricatives is neutralised in word final position and all final consonants are voiceless. Final position in sentences may also be marked by creakiness, due to irregular vibrations of the vocal fold (probably concomitant to a very low transglottal pressure). PART III: PARALINGUISTICS Finely coded, attitudinal tones of voice are used by adult speakers to shed nuances on the spoken material. Controlled Vaissière, J., (1997), "Phonological use of the larynx: a tutorial", Larynx 97, Marseille, 115-126. tones of voice inform on expressiveness, incredulousness, sarcasm, annoyance, bored resignation, apology, supplication, etc. The realisation of the tones of voice includes manipulation of mean FO (and intensity), FO range, voice quality, etc [45]. For example, whispery voice signals confidentiality, breathy voice intimacy, harsh voice dissatisfaction. The outward expression of attitudes directly through tones of voice may be more or less standard, depending on the cultures, and on the degree of formality (attitudes may be expressed by lexical and grammatical means).The study of the culturally specific contribution of the setting of the tone of voice is an area still largely unexplored (with the exception of Fonagy [46] and Laver [17]). There may be a link between the emotional aspects of speech, where low F0 is bounded to dominance and high FO to submission and the linguistic use of high FO, trough the frequency code [47], [18], [38]. It is not easy to draw the separation between strictly linguistic and paralinguistic aspects, since the two aspects are intimately joined in every day conversations. The paralinguistic information is very important in human relations since the tune of voice is often as important as the literal sense of the speech form (as in the speech of the human mothers addressed to the new-born). CONCLUSION We may conclude that the larynx play a very important role in speech. Its most important function is to provide phonation. Phonation does not require a sophisticated laryngeal mechanisms (voice is possible without larynx). The possibility of a fine control of the tension of the VFs, the aperture of the glottis, the movements of the larynx, have been largely exploited by the languages, for distinguishing phonemes, words and sentences, etc... The degree of involvement of pitch at the lexical level differ greatly from one language to the next (such as in tone and stress languages). There are cross-languages similarities in the use of the laryngeal configurations at supralexical levels. The linguistic functions of the larynx are very important, but no less important are the paralinguistic functions in every day exchanges. Human listeners are trained since they are born to pay attention to nuances of voices. Human mothers exaggerate their melodies in their speech addressed to their new-born, to draw their attention and to pacify them. The "tunes" of voice convey information and they are used prelinguistically by the infants, before they understand speech. Much more work is needed to elaborate a model of the interaction between segmental, linguistic, para- and extra-linguistic factors on pitch and voice quality. REFERENCES [1] Sapir, A, (1921), Language: an introduction to the study of speech, New York, Harcourt, Brace and World. [2] Catford, C. (1977), Fundamental problems in Phonetics", Edinburgh University Press. [3] Camperio Ciani, A., and B. Chiarelli. , (1988), “A systematic relationship between brain size increase and larynx transformation during hominization.” in Language Origin: a multidisciplinary approach. Wind et al (Eds.), Dordrecht, Boston, London: Kluwer Academic Publishers, 51-66. [4] Honda, K. and Fujimura, O., (1991), "Intrinsic vowel FO and phrase-final lowering; phonological Vs biological explanations", in Gauffin, J., and Hammarberg; B., (1991), Vocal Fold Physiology: acoustic, perpetual and physiological aspects of Voice Mechanisms, Singular Publishing Group, San Diego, 149-158. [5] Hirose, H., (1976), "Posterior cricoarytenoid as a speech muscle", Ann. Otology, Rhinology and Laryngology, 85, 334-343. [6] Titze, I.R., (1991), "Mechanisms underlying the control of fundamental frequency" in Gauffin and Hammarberg (Eds), 129-148. Vaissière, J., (1997), "Phonological use of the larynx: a tutorial", Larynx 97, Marseille, 115-126. [7] Hirano, M., (1981), Clinical Examination of Voice, Vienna: Springer-Verlag. [8] Fujimura, O., (1981), "Body-cover Theory of the vocal fold and its phonetic implication", in Stevens, K.N., and Hirano, M. (Eds), (1981), Vocal Fold Physiology, Tokyo University Press, Tokyo. [9] Ohala, J., (1973), "The physiology of tone" in Hyman, .L. (Ed.), Consonant Type and Tone, Sth. Calif. Occ; Papers, 1, 1-14. [10] Stevens, K.N., (1977), "Physics of laryngeal behavior and larynx modes" , Phonetica, 34, 264-279. [11] Pierrehumbert, J., (1989), "A preliminary study of the consequences of intonation for the voice source", KTH, Speech Transmission Laboratory, Quarterly Progress Status Report, 4, 23-36. [12] Hagège, C., and A. Haudricourt. , (1978), La Phonologie Panchronique. PUF, Paris. [13] Hombert, J-M, Ohala, J., and Ewan, G.., (1979), "Phonetic explanations for the development of tones", Language, 55, 1, 37-58. [14] Ananthapadmanabha, T., (1995), "Acoustic factors determining perceived voice quality", in Fujimura and Hirano (Eds), 113-126). [15] Laver, J., (1980), The Phonetic Description of Voice Quality, Cambridge University Press. [16] Ladefoged, P., (1988), "Discussion of Phonetics: a note on some terms for phonation types", in Fujimura, O. (Ed), (1988), Vocal physiology: Voice Production, Mechanisms and Functions, Raven Press, Ltd., New York, 373-375. [17] Laver, J., (1991) , The gift of speech, Edinburgh University Press. [18] Ladefoged, P., (1983), "The linguistic use of different phonation types", in Bless, D., and Abbs, J., (Eds), Vocal Fold Physiology: Contemporary Research and Clinical Issues, San Diego: College-Hill Press., 351-360. [19] Ladefoged. P. et Maddieson, I., (1996), The Sounds of the World's Languages, Blackwell Publishers. [20] Maddieson, I., (1984), Patterns of Sounds, Cambridge: Cambridge University Press. [21] Kenstowicz, M, (1994), Phonology in Generative Grammar, Blackwell). [22] Halle, M., and Stevens, K., (1971), "A note on laryngeal features", Quarterly Progress Report 101, 198-212, Cambridge, mass.: Research Laboratory of Electronics, MIT. [23] Stevens, K.N., (1991), "Vocal-fold vibration for obstruent consonants", in Gauffin and Hammarberg (Eds), 29-36. [24] Westbury, J.R., (1983), "Enlargement of the supraglottic cavity and its relation to stop consonant devoicing", JASA, 73:1322-1336. [25] Lisker , L. and Abramson, A.S., (1964), "A cross-language study of voicing in initial stops: acoustical measurements", Word 20, 384-422. [26] Kagaya, R., and Hirose, H., "Fiberoptic, electromyographic and acoustic analyses of Hindi stop consonants", Ann. Bull. Res. Inst. Logopedics and Phoniatrics, 9, 27-46. [27] Bickley, C., (1982), "Acoustic analysis and perception of breathy vowels", MIT Working Papers, 1, 71-81. [28] Ladefoged, P., Maddieson, I. and Jackson, M. (1988), "Investigating phonation types in different languages", in Fujimura, O. (Ed), 297-318. [29] Ladefoged, P., (1973), "The features of the larynx", J. Phon., 1, 73-83. [30] House, D., (1990), Tonal Perception in Speech, Trav. Inst. Ling. Lund, Lund University Press. [31] Trubetskoy, (1957), Principes de Phonologie, Paris. [32] Vaissière, J., (1983), "Language independent prosodic features", in Cutler, A. and Ladd, D.R. (Eds), Prosody: Models and Measurements", Berlin: Springer- Verlag., 53-66. [33] Pike, K., (1948), Tone languages, university of Michigan Press, Ann Arbor, Michigan. [34] Gärding, E., (1983), "A generative model of intonation", in Cutler and Ladd (Eds.), 11-21.