- Количество слайдов: 83
Intonation and Meaning Doctorat en Ciència Cognitiva i Llenguatge Pilar Prieto, ICREA-UAB
References n n Gussenhoven, C. (2004). The Phonology of Tone and Intonation. Cambridge University Press, Cambridge, ch. 4 and 5. Gussenhoven, C. (2002). “Intonation and Interpretation: Phonetics and Phonology, ” in Proceedings of the Speech Prosody 2002 Conference, pp. 47 -58, 2002.
Intonation and Meaning n n Languages have tonal grammars Speakers of all languages also use the phonetic implementation to signal paralinguistic meanings ‘Grammatical’ intonational meaning will often mirror those form-function relations, and such phonology-morphology relations may be arbitrary ‘Phonetic’ paralinguistic meaning draws on universally understood metaphors of biologically determined form-function relations
Discreteness in tonal patterns n Experimental approaches towards establishing discreteness in intonation (Gussenhoven 2004) n n n The imitation task (Pierrehumbert & Steele 1989) The pitch-range task (Gussenhoven & Rietveld 2000) The semantic task and categorical perception tasks ((Ladd and Morton 1997, Kohler 1987)
Intonation makes the difference A: What types of foods are a good source of vitamins? B 1: Legumes are a good source of vitamins. B 2: Legumes are a good source of vitamins. A: I’d like to fly to Davenport, Iowa on TWA. B: TWA doesn’t fly there. . . B 1: They fly to Des Moines. B 2: They fly to Des Moines. A 1: I met Mary and Elena’s mother at the mall yesterday. A 2: I met Mary and Elena’s mother at the mall yesterday.
[TWA doesn’t fly there. . . ] they fly to Des Moines Rise right at start of stressed syllable cues statement of fact.
Alignment with syllable matters [TWA doesn’t fly there. . . ] they fly to Des Moines Rise which is delayed somewhat cues suggestion, or uncertainty about whether the statement qualifies as relevant.
The imitation task n Pierrehumbert & Steele (1989) undertook an imitation task with the two intonation patterns of the utterance Only a millionaire illustrated below.
n Pierrehumbert & Steele (1989) synthesized many intonation contours with varying degrees of peak delay, and asked speakers to imitate what they heard. n Peak delay of speakers’ responses patterned in two categories: early (‘assertion’) and late (‘suggestion’).
The results of the experiment revealed the existance of two separate phonological categories n See also later experiments by Ward & Hirschberg 1985, Hirschberg & Ward 1992, among others, which confirmed a clear separation between the two). n
Gradient effects of tonal range Free Gradient Hypothesis: variation in pitch range only reflects the speakers’ implication in speech act n Liberman & Pierrehumbert (1984) n Gradient or categorical effects? n
Liberman & Pierrehumbert 1984
Free Gradient Hypothesis n Free Gradient Hypothesis: variation in pitch range only reflects the speakers’ implication in speech act
Interrogative sentences and the locality/range problem n n n Suspension of declination in questions in Hausa (Lindau 1986, Leben 1990). Cf. also Catalan. Higher F 0 peaks in interrogatives in Spanish (Navarro Tomás 1944; Canellada & Madsen 1987; Sosa 1992, 1999), Danish and Swedish (Hadding & Kennedy 1972) and Bengali (Hayes & Lahiri). *Not in French and English (Mettas 1971). Are these differences phonological?
Spanish interrogatives Fo contour of the utterance ¿Le dieron el número de vuelo? (Sosa 1992)
The pitch-range task (Gussenhoven & Rietveld 2000) Difference between the high rise and the low rise in Dutch (the accented syllables is mid vs low respectively). n In the experiment, listeners heard two sets of 9 accent stimuli: n 9 versions of the low rise by combining 3 values from the beginning and end n 9 versions of the high rise by combining 3 values from the beginning and end n
1 %L Gussenhoven & Rietveld 2000 H* H%
Perceived surprise scores High rise set: Perceived surprise is higher with higher values
Low rise set: Perceived surprise is higher with lower values
Explanation n The high rise and the low rise are discretely different contours. Perceived surprise is a function of pitch range Pitch range is perceived in terms of the distance between realizations of L and H H H L
n If the high rise and the low rise were different realizations of the same phonological contour, differing only in pitch range, listeners would have treated any itch range variation in either contour type in the same manner.
Perception of pitch range not like this… but like this:
%L H* H%
2 %L L*H H%
Semantic Tasks Categorical Perception Tasks n n n This task relies on the perception of semantic differences between phonetically different forms. Categorical perception: listeners interpret the phonetic continuum as belonging to two phonological categories and do not perceive differences between stimuly belonging to the same category. Problem: in intonation speakers tend to associate phonetically different forms with different meanings (Ladd & Morton 1997).
Categorical Perception Paradigm n Two separate (and complementary) tests are required to establish categorical perception. n n Identification task, in which listeners are asked to assign randomly presented stimuli from the continuum to either of two categories (results should show an abrupt perceptual shift at a given point) Discrimination task, in which listeners hear pairs of stimuli that differ by one acoustic step on the continuum, and are asked to say whether the members of each pair are the same of different. The results of this experiment should reveal that discrimination is good at the point where listener’s perception shifts from one category to the other.
The a. LARM went off A: “This was an everyday occurrence” B: “This was an unusual experience” Ladd & Morton 1997
Late peak Early peak Ladd & Morton 1997
n n Listeners are more likely to choose the “unexpected” interpretation as the F 0 peak was higher. “Although the shift in interpretation is clearly brought about by the increased F 0 peak, it could be due to a gradient form-function relationship (higher peaks signalling greater significance; Gussenhoven 2004: 66)”.
Substitute pitch height n Late f 0 peaks can substitute for high f 0 peaks Hz Raised and delayed tim
Neapolitan Italian (D’Imperio 1999) Mamma ballava da Lalla H*+L Li Mamma ballava da Lalla ? H*+L L*+H Li ‘Mum used to dance at Lalla’s’
With in broad focus Mamma ballava da Lalla H*+L !H*+L Li Mamma ballava da Lalla ? H*+L L*+H Li ‘Mum used to dance at Lalla’s’
Narrow focus statement Vedrai mamma domani ‘You will see your mum tomorrow’ Yes-no question Vedrai mamma domani ‘Will you see your mum tomorrow? ’
(D’Imperio et House, 1997; D’Imperio, 2000) H original H L L D’Imperio (2002) 40 ms stylized A 40. ms. difference in peak alignment statistically increased question responses in perception experiments (D’Imperio 2000) statement CONCLUSION: A small alignment H difference is crucial for identification
Methods 300 Alignment: A. Linear stylisation 280 F 0 (Hz) 250 H 210 200 210 180 L 1 150 L 2 Time Alignment: B. Temporal shift A 2 A 3 A 4 200 250 H L 1 L 2 150 F 0 (Hz) 300 A 1 -15 Vocalic offset 0 Shift (msec) 15 30
• Pragmatic contrast: (D’Imperio, 2000) In Neapolitan Italian, later synchronization of F 0 peaks differentiates questions from statements coupled with raising the F 0 range and slower descent. (D’Imperio, 2000, 2001, 2002; D’Imperio&House, 1997) ces in tonal alignment are employed by listeners to distinguish
Universality of Intonational Meaning n n A discussion of intonational meaning typically raises the issue of whether such meaning is universal or language-specific. The position defended here is that both the universal and the language-specific perspectives are true, simultaneously, for any language, but that the universal part is exercised in the phonetic implementation, while the language-specific meaning is located in the intonational morphology.
Ladd 1981 n n Universalist Hypothesis: biologically determined pattern of rising questions and falling statements Language-specific Hypothesis: grammatically determined, arbitrary relationship between form and function
View defended by Gussenhoven n n Universal meaning is omnipresent There is a statistical bias towards ‘natural’ phonological interpretations There are degrees of ‘naturalness’ in phonological interpretations ‘Unnatural’ representations may arise in many ways through language change
Biological codes n Gussenhoven suggests that the intonation of any language involves universal and language-specific components. n n n The Frequency Code: variation in larynx size is associated with frequency, height of pitch. The Effort Code: variation in effort is associated with the excursion size of pitch movements The Production Code: high initial pitch with beginnings and final low pitch with endings of speech events.
Knowledge of vocal sound production n n Size: Frequency Code Energy: (a) degree: Effort Code (b) phasing: Production (Phase) Code
Three biological codes FC High – Small PC EC Wide excursion – More effort High beginning – start of production Low beginning – continuation of production and Low ending – end of production Low- Big Small excursion – Less effort high ending – continuation of production
Many meanings are derived from these codes, by social agreement. n Some influence of the native language (Aoju Chen’s studies) n The three codes are accidental : the speech production mechanism originally arose for different reasons (exaptation). n Grammaticalization of these codes n
Two meaning classes n n Affective meanings: refer to the speaker Informational meaning: refer to the message Affective vs. informational interpretations of the biological codes
Frequency Code n n n Smaller larynxes contain lighter and smaller vocal cords, with which faster vibration rates are achieved for a given amount of energy. The correlation between larynx size and rate of vocal cord vibration is exploited for the expression of power relations. The many ramifications of this latter connection were dealt with by Ohala. The term for this formfunction relation is his, and my labels for the next two relations are by analogy with his term.
Meanings of the Frequency Code High ~ Low, Affective: Vulnerable - protective Submissive - Authoritative High ~ Low, Informational: Uncertain - Certain
Grammaticalizations of the Frequency Code High ~ Low, Informational: Uncertain - Certain Grammaticalization: Statement - Question
Ohala proposes a biological basis for the correlation between high pitch and questions: smaller larynxes produce higher pitch and smaller creatures are often less powerful than larger creatures, high pitch can be used to signal submissiveness or a willingness to cooperate. This, in turn, explains the use of high pitch in questions: questions require cooperation from other speakers may reflect cross-linguistic differences in grammatical
Two substitute forms n n Late peaks to suggest high peaks High register to suggest wide excursion
Substitute pitch height n Late f 0 peaks can substitute for high f 0 peaks Hz Raised and delayed
Substitute pitch span n High pitch can suggest wide pitch span (register for span) wide narrow substitute for wide
Influence of the native language Different choices in the case of conflicting meanings
Influence of the native language n n n Hadding & Studdert-Kennedy 1964 Gussenhoven & Chen 2000 Chen, Rietveld & Gussenhoven 1999
Hadding & Studdert-Kennendy Phonetica 1964. Reprinted in Bolinger’s Intonation: A Book of Readings For Jane/För Jane S H three values
American English and Swedish listeners “Is this a question or a statement? ”
1) Interrupted lines (stimulus with higher peaks) => higher peaks lead to more question responses; 2) higher endpoints too.
S % Question H English Swedish F 0 of end pitch
English % Question Swedish H S
Explanation n n Swedish listeners are more sensitive to peak height differences than English listeners… because Swedish has no final rise to express interrogativity, while English does
Gussenhoven & Chen 2000 n n Universal cues of interrogativity peak height, end pitch, peak alignment
Dutch, Hungarian and Chinese listeners Stimuli presented paired with standard stimulus “These are sentences from a language spoken on a South Sea island. Which of the two utterances is the question? ”
Chen, Rietveld & Gussenhoven 1999 H*L L% L*H H% Female bilingual Du-Eng speaker Confident – Not confident
Explanation n n For some difference in Hz, Dutch listeners get more meaning out of the Frequency Code than Br. E listeners This is because Dutch has a narrower pitch range
Meanings of the Effort Code Wide ~ Narrow, Affective: Excited - Dull Surprised - Unimpressed Helpful - Not helpful Wide ~ Narrow, Informational: Significant - Not significant
Meaning of the Production Code, Informational Beginning of utterance, High ~ Low: New topic - continuation End of utterance, High ~ Low: Continuation - End of turn
Effort Code: Emphatic – non-emphatic Chen, Gussenhoven & Rietveld 2000
380 260 180 Register H*L L%, L*H H%
Explanation n Dutch: Substitute use of register for pitch span to express emphasis (Effort Code) English: Use of register to express friendliness (Frequency Code) So don’t perceive Dutch listeners friendliness in higher registers?
Testing the Production Code: English and Dutch
D K The D is in section 3, the F is in section 5, the K is in section 7 F
Initiality 320 Hz 300 280 260 240 The D is in section. . .
Finality 250 Hz 235 220 205 190 . . . is in section 3 with five source utterances
“Do you think this sentence fragment is from the first, second or third sentence in the passage? ”