Phonetic patterns in Oto-Manguean tonal systems Christian T. DiCanio [email protected] Haskins Laboratories http://linguistics.berkeley.edu/∼dicanio 5/14/14 Christian DiCanio (((Haskins))) Oto-Manguean tone 5/14/14 1 Introduction Typology and tonal systems How useful is a typological perspective for the study of tonal phonetics? 1 Structural diversity is abundant. Structural differences among languages contribute to phonetic variation in tone production/perception, even across well-known languages. 2 The phonetic timing of tones differs dramatically. There is substantial cross-linguistic variation in how tones are coordinated in larger utterances/units. 3 Models of speech production should be inclusive with respect to such cross-linguistic variability. Christian DiCanio (((Haskins))) Oto-Manguean tone 5/14/14 2 Introduction Typology and tonal systems How useful is a typological perspective for the study of tonal phonetics? 1 Structural diversity is abundant. Structural differences among languages contribute to phonetic variation in tone production/perception, even across well-known languages. 2 The phonetic timing of tones differs dramatically. There is substantial cross-linguistic variation in how tones are coordinated in larger utterances/units. 3 Models of speech production should be inclusive with respect to such cross-linguistic variability. Christian DiCanio (((Haskins))) Oto-Manguean tone 5/14/14 2 Introduction Typology and tonal systems How useful is a typological perspective for the study of tonal phonetics? 1 Structural diversity is abundant. Structural differences among languages contribute to phonetic variation in tone production/perception, even across well-known languages. 2 The phonetic timing of tones differs dramatically. There is substantial cross-linguistic variation in how tones are coordinated in larger utterances/units. 3 Models of speech production should be inclusive with respect to such cross-linguistic variability. Christian DiCanio (((Haskins))) Oto-Manguean tone 5/14/14 2 Introduction Typology and tonal systems How useful is a typological perspective for the study of tonal phonetics? 1 Structural diversity is abundant. Structural differences among languages contribute to phonetic variation in tone production/perception, even across well-known languages. 2 The phonetic timing of tones differs dramatically. There is substantial cross-linguistic variation in how tones are coordinated in larger utterances/units. 3 Models of speech production should be inclusive with respect to such cross-linguistic variability. Christian DiCanio (((Haskins))) Oto-Manguean tone 5/14/14 2 Introduction Typology and tonal systems How useful is a typological perspective for the study of tonal phonetics? 1 Structural diversity is abundant. Structural differences among languages contribute to phonetic variation in tone production/perception, even across well-known languages. 2 The phonetic timing of tones differs dramatically. There is substantial cross-linguistic variation in how tones are coordinated in larger utterances/units. 3 Models of speech production should be inclusive with respect to such cross-linguistic variability. Christian DiCanio (((Haskins))) Oto-Manguean tone 5/14/14 2 Introduction E. Zsiga, R. Nitisaroj 347 Structural differences? Mandarinfalling slightly and vs. remaining Thai as low or lower than the mid tone for the first half of the syllable, then rising steeply in the second half. The low tone falls steadily, reaching the bottom of the pitch range at the end of the syllable. The falling tone is realized as a rise-fall contour, and the rising tone as a fall-rise contour. Contextual tonal ariations 67 &## 160 Figure 1 Tone 1 Contour shapes of Thai tones Tone 2 %$" ! in citation form. Representative examples from one speaker 140 % Tone 3 %"# Tone 4 120 (Hz) 0 f %%" " &'()*+%, 100 %## # !$" 80 $ 0 25 50 75 100 !"# # #'! #'% #'& #'( #'" Normalized time (%) )*+,-.) Figure 2 . Mean f 0 contours (averaged over speakers and tokens ; n 48) of four Mandarin tones in the monosyllable / m a / produced in isolation . The time is normalized , with all tones plotted with their average duration Creating these complex contours from simple H and L associated to the syllable proportional to the average duration of Tone 3 . is certainly possible. Such phonetic mapping rules would be complex, however. A What does a falling tone look like? Whatsingle accounts H associated for to the asyllable delayed would havefall to be mapped in into a level-rising scooped contour, while an H linked as part of a falling tone would correspond to a quick Thai? 3 . Results rise to the top of the pitch range. The complexity of attempting to map H and L 3 . 1 . Tones produced in isolation auto segments onto the actual contours of the tones of Thai leads Abramson (1979) to reject a compositional analysis of the contours. He argues (p.7) that the data “lend To establish the canonical forms of the four Mandarin tones produced byno the phonetic plausibility” to arguments for the specification of rising and falling tones (Figures speakers in this from study , tones Xu produced (1997); in isolation Zsiga were analyzed and first Nitisaroj . Fig . 2 showsas sequences (2007)) of H and L autosegments. the average f 0 contour of the syllable / m a / in the four tones , obtained by averaging over all tokens produced by all eight speakers (48 utterances for eachFor phonetic support of the argument one would expect to be able to devise a tone) and plotted as functions of average relative time of each segment , with theformula by which the dynamic tones were obviously to be derived from the shape average duration of Tone 3 as the reference duration (100%) . The cross symbol ‘‘ ’’of the static tones. Even the citation forms, let alone the F0 curves of running indicates boundaries between m and a . Tone 1 starts with a high f valuespeech, provide no acoustic basis for such a claim. It seems psychologically far Christian DiCanio (((Haskins))) / / / / Oto-Manguean 0 tone 5/14/14 3 (near 130 Hz) and stays around that level throughout the syllable . Tone 2 starts withmore reasonable to suppose that the speaker of Thai stores a suitable tonal shape as part of his internal representation of each monosyllabic lexical item. (p.7) a low f 0 (near 110 Hz) , then falls slightly before rising (starting at 20% into the vowel) throughout the remainder of the syllable . Tone 3 starts with an f 0 valueMorén and Zsiga argue, however, that a mora-based representation as in (2) slightly lower than the onset of Tone 2 , falls to the lowest f 0 of all the four tonesdoes allow a straightforward mapping from autosegments to the acoustics of Thai (about 90 Hz) right at the vowel midpoint , then rises sharply to the end oftones. the During moras with no phonological tone, pitch falls gradually to or within syllable . Tone 4 starts with the highest f 0 value of the four tones (140 Hz) , continuesthe mid range. Phonologically-specified moras reach a high or low pitch inflection to rise before reaching the maximum about one fifth of the way into the vowel , at then their right edge. Thus mid tones, with no phonological specification, remain fairly falls sharply to the end of the syllable . level throughout their duration. High tones are specified only on the second mora, In terms of duration , Tone 4 is the shortest tone (214 ms on average) , Toneand 3 is thus remain in the mid range during the first half of the syllable, begin to turn the longest (349 ms) , while Tones 1 and 2 have intermediate durations , with Toneupward 2 at syllable midpoint, and reach their high point at the rightmost edge of the (273 ms) longer than Tone 1 (247 ms) . These duration patterns match those found in previous studies (e . g ., Lin , 1988) . Language and Speech Downloaded from http://las.sagepub.com at Bibliotheques de l'Universite Lumiere Lyon 2 on April 20, 2009 The f 0 patterns of tones produced in isolation reflect relatively directly the canonical forms of the tones . In multisyllabic utterances , the canonical forms will be distorted by various factors , including the adjacent onset and of fset values of the neighboring tones . As suggested earlier and demonstrated by Xu (1994 a ) , it is High-level tone (55) Mid-level tone (33) Low-rising tone (24) Low-falling tone (21) High-falling tone (51) 330 F2 F2 F2 F2 F2 300 Introduction 270 Coordination differences? Mandarin vs. Taiwanese (Hz) 240 0 f 210 Mandarin falling tones (left, (Xu, 1994)) undergo greater coarticulation 180 than Taiwanese falling tones (right, (Peng, 1997)). (a) (b) (c) (d) (e) 150 Riang Fa • 330 180- 180 F2 F2 F2 F2 F2 x _ / x / _ / x 300/ x Compatible _ _ Context •, / / X_ _ / X X 270! 160 240 (Hz) 0 /_ / / / _ x f x 210x x Conflicting _ _ Context / ! /! X X X 180X _ _ } (f) (g) (h) (i) (j) 150 120 120 –20 30 80 130 180 230 280 330 –20 30 80 130 180 230 280 330 –20 30 80 130 180 230 280 330 –20 30 80 130 180 230 280 330 –20 30 80 130 180 230 280 330 Duration (ms) Duration (ms) Duration (ms) Duration Duration (ms) Duration (ms) Panels (a) – (e) Panels (f ) – ( j) FIG. 2. Tone patternsof trisy!labi½words and phrasesused for F0 mea- — — I phrase-initial — — 55 surements. Compelhie, Faster — — M phrase-medial Christian DiCanio — — (((Haskins))) 33 Oto-Manguean tone 5/14/14 4 – – – – F phrase-final -- Compatible, – – – –Slowe• 21 - - - - - - U utterance-final •-• Conlhcfng.Faster C. Recording Figure 5 . Subject F2 : Mean duration and f 0 of each target tone followed by dif ferent tones (panels f – j) in dif ferent prosodic positions (panels a – e) . ......... Conf•ct'ng, Slower For eachof the 16 patternslisted in Fig. 2, two words or phraseswere found. They were repeatedfive timesand FIG. 3. Mean F0 slopes,mean F0 heightand mean durationfor the printedin Chinesein randomorder.