Presentation is loading. Please wait.

Presentation is loading. Please wait.

Language and Speech, 2000, 43 (2), 183-203 THE BEHAVIOUR OF H* AND L* UNDER VARIATIONS IN PITCH RANGE IN DUTCH RISING CONTOURS Carlos Gussenhoven and Toni.

Similar presentations


Presentation on theme: "Language and Speech, 2000, 43 (2), 183-203 THE BEHAVIOUR OF H* AND L* UNDER VARIATIONS IN PITCH RANGE IN DUTCH RISING CONTOURS Carlos Gussenhoven and Toni."— Presentation transcript:

1 Language and Speech, 2000, 43 (2), 183-203 THE BEHAVIOUR OF H* AND L* UNDER VARIATIONS IN PITCH RANGE IN DUTCH RISING CONTOURS Carlos Gussenhoven and Toni Rietveld University of Nijmegen Presented by: Maria del Mar Vanrell Bosch

2 Are the low rise and the high rise categorically distinct contours of Dutch?

3 1.INTRODUCTION 2.EXPERIMENT 1 3.EXPERIMENT 2 4.EXPERIMENT 3 5.GENERAL DISCUSSION

4 The real question: Is the difference between two phonetically different intonation contours phonological or phonetic? Phonological (the two contours are realizations of different tone strings)/ phonetic (the contours are different realizations of the same phonological structure). 1.INTRODUCTION

5 INTRODUCTION HIGH RISELOW RISE Mid pitch at a point halfway into the accented syllable Low pitch during the first half of the accented syllable/ throughout the accented syllable If there is sufficient space between the accented syllable and the end of the IP: a level stretch after the rise followed by a further rise in the last syllable. With IP –final accented syllables-: the two rising movements form a single, continuous rise.

6 INTRODUCTION HIGH RISE LOW RISE

7 This distinction has not been recognized as a phonological contrast (‘t Hart, Collier, & Cohen, 1990; Gussenhoven, 1988, 1991). There are indications that Dutch has a high rise which is different from a low rise: -Collins and Mees (1981) describe two contour for Dutch  L*+H-H% (“low rise”) and H*H-H% (“high rise”). “Dutch high rise is common where British English would use a low rise”. -The implementation of Gussenhoven’s description in a synthesis program (Gussenhoven & Rietveld, 1992)INTRODUCTION

8 -Judith Haan suggests that the high rise is relatively frequent in speech read aloud. -Liberman and Pierrehumbert (1984): L* is lowered as the pitch range increases (“speak up”). -Gussenhoven and Rietveld (1988): range differences have been widely associated with perceived prominence. They wanted to understand better how pitch range perception varies as a result of variation in the realization of the L* tone (L*H-H%).INTRODUCTION

9 INTRODUCTION It is not obvious what response variable best captures variations in pitch range: -”Speaking up” by Lieberman and Pierrehumbert (1984): pitch span/ pitch level. -Patterson and Ladd (1999): “deepness of the voice” (pitch level), “expresiveness” and “emphasis” (pitch span). They aim to show that increases in pitch span lower the values corresponding to L* targets, while raising H- tones (“expresiveness”, “liveness” and “insistence”).

10 Method: Purpose: to establish how variations in the F 0 of L*, H- and H% affect the perceived prominence on accented words realized with a L* H-H% contour. Stimuli with the accented syllable in final position/ in non final position. Stimuli: from Een nieuwe motor is te DUUR? (A new engine is too expensive?) and Je was al EERder bij hem geweest? (You had been to see him before?). Male native speaker of Dutch. 2. EXPERIMENT 1: THE PROMINENCE OF L* H- H%

11 EXPERIMENT 1/ method

12 Each set of nine stimuli was randomized six times (18 blocks of nine stimuli). An anchor stimulus preceded each stimulus (%L H*L-L% -a peak of 150 Hz on the syllable that carried the accent in the test stimulus-). Two groups of 15 listeners, divided over both sexes, ages between 19-28. They had to indicate the degree of “EMPHASIS” with which the accented word was spoken (placing a tick across a 10 cm long line). EXPERIMENT 1/ method

13 Scale: “little emphasis” on the left and “much emphasis” on the right. Results: Scores: obtained by measuring the distances in millimeters (from the beginning of each line). EXPERIMENT 1/ method

14 EXPERIMENT 1/ results

15 In final accented syllables, only the end point has an effect on the perceived prominence. In non final ones, only the beginning point has an effect on the perceived prominence. Lower realizations of L* and higher realizations of H%  greater perceived prominence. EXPERIMENT 1/ results

16 Conclusion: L*: lower values elicit higher prominence judgments (-final position). H%: higher values elicit greater perceived prominence (final syllable). Perceived prominence is not an appropriate response variable. EXPERIMENT 1

17 Method: Purpose: investigating effect of variation in the F0 of L*, H-, and H% on the perception of SURPRISE, INSISTENCE and INDIGNATION (depended on pitch range variation?). Hypothesis: for each of these attributes, lower values of L* and higher values of H- and H% would result in higher scores. 3. EXPERIMENT 2: LOOKING FOR A NEW RESPONSE VARIABLE

18 Two new utterances: Is honderd gulden niet VEEL? (Isn’t a hundred guilders a lot?), Zijn er niet MEER mensen op afgekomen? (Haven’t more people come?). Four test tapes: two contained the stimuli of Experiment 1 (27 stimuli randomized four times= 54 test stimuli), preceded on each tape by a set of seven practice stimuli and followed by two filler stimuli (54 + 2 + 7). EXPERIMENT 2/ method

19 Nine groups of seven each tape. In the same way, two tapes with the stimuli from uterances 3 and 4. Three scales were presented: the speaker expresses his surprise/ the speaker insists on receiving an answer/ the speaker vents his indignation (by circling one of the numerals of the corresponding scale). 32 judges split into four groups of eight, divided over both sexes, ages between 19-32. EXPERIMENT 2/ method

20 Results: EXPERIMENT 2

21 SURPRISE: significant effects for L* and H% (in all cases). INSISTENCE: a significant effect for H% in set 1. INDIGNATION: a significant effect for L* in set 2. Conclusion: The attributes INSISTENCE and INDIGNATION are not perceived in accordance with the hypothesis that these dependent variables correlate with variations in pitch range. SURPRISE produced scores which were in agreement with their hypothesis. EXPERIMENT 2/ results

22 Method: Purpose: to show that high rise contours, analyzed as H*H-H%, behave in the same way as low rise contours for this dependent variable (exception: the lower beginnings express less surprise, instead of more, as found for L* in Experiment 2). A different male speaker of Dutch, the four sentences of Experiment 2 (unemphatic, weakly rising pitch accent). (not variation in H-) 4.EXPERIMENT 3: PITTING L* AGAINST H*

23 EXPERIMENT 3

24 Non final accented L*H-H% Final accented L*H-H% Non final accented H*H-H% Final accented H*H-H%

25 EXPERIMENT 3/ method 72 stimuli (2 contours x 9 versions x 4 sentences) + 24 fillers %L H* L- H% (for increasing the intonational variety in the test). 96 stimuli were randomized twice + ten practice stimuli + two closing stimuli (seven blocks of 15 and one block of 3). Anchor stimulus: the appropiate procedure. Two groups of subjects (15 and 14 subjects), age between 19 and 30. Two tapes (with a different random order of the stimuli).

26 Subjects were asked to rate the degree of surprise (placing a mark on a 100 mm scale). Results: Factors: ACCENT-LOCATION (early/late), CONTOUR (L*H- H% vs. H*H-H%), HEIGHT-OF-T* (3 levels). Important result: interaction between CONTOUR and HEIGHT T* (raising the beginning of the rise does not have the same effect in low rise as it has in the high rise). EXPERIMENT 3/ method

27 EXPERIMENT 3/ results H* and H% L* and H%

28 Discussion: Conclusion: “high rises” and “low rises” form separate phonological categories in Dutch. “Low rises” begin with L* and “high rises” begin with H*. EXPERIMENT 3

29 Interaction between ACCENT-LOCATION, CONTOUR and HEIGHT-OF-H%: scores for the high rise are considerably higher for final accents than for nonfinal accents. Scores for the low rises are less affected by accent location (less extreme but opposite pattern). Interaction between ACCENT-LOCATION and HEIGHT-of- H%: greater differentiation of the scores for H% in final- accented rising stimuli, as compared with nonfinal accents. EXPERIMENT 3/ discussion

30 EXPERIMENT 3/ duscussion

31 1.It may not be useful to consider perceived prominence to be an interesting attribute of a nuclear contour like L*H-H% or of the nuclear accent L*. 2.One way in which speakers can convey surprise in Dutch is by increasing the pitch range as conceived by Lieberman- Pierrehumbert: lowering L* and raising H- tones. 3.L*+H-H% and H*H-H% are categorically distinct contours of Dutch. 5. GENERAL DISCUSSION

32 General conclusion: the consideration of pitch range variation appears not only to be relevant for studying models of phonetic implementation but also for models of phonological structure. GENERAL DISCUSSION


Download ppt "Language and Speech, 2000, 43 (2), 183-203 THE BEHAVIOUR OF H* AND L* UNDER VARIATIONS IN PITCH RANGE IN DUTCH RISING CONTOURS Carlos Gussenhoven and Toni."

Similar presentations


Ads by Google