Nuclear Accent Shape and the Perception of Syllable Pitch

1. Nuclear Accent Shape and the Perception of Syllable Pitch Rachael-Anne Knight rak34@cam.ac.uk LAGB 16 April 2003

2. Outline • Factors affecting perception of pitch • Contour shape • Experiment • Design • Results • Theoretical and psychoacoustic explanations • Conclusions and implications

3. The Perception of Pitch • The perception of the pitch of a complex sound is related to the fundamental frequency • Sounds that have a higher fundamental frequency sound higher in pitch • And in speech are associated with greater prominence

4. In an utterance the position of a syllable also affects the perception of pitch If two syllables have identical F0, the one later in the utterance will sound higher in pitch Explained as the listener ‘normalising for declination’ Syllable Position Frequency Time

5. Plateau Peak Contour Shape • Real contours are not stylised peaks and troughs • The majority of falling nuclear accents are realised as more of a flat stretch of contour

6. Peak 4% range Plateau Definition of the Plateau • Plateaux are defined as being 4% down from any absolute peak in F0 • 4% is the range of perceptual equality

7. Segmental and Prosodic Effects on the Production of the Plateau • Plateaux take up more of syllables that have sonorant onsets and codas • Plateaux are aligned later in the syllable in polysyllabic than monosyllabic feet • Some speakers align the end of the plateau earlier in the syllable before a word boundary

8. The End of the Plateau • The end of the plateau is stably aligned within the syllable regardless of pitch span • More errors made with incorrect EP alignment in a true/false judgment task • The end of the plateau seems to be the real target (rather than the peak)

9. Experiment • How does a plateau affect the perception of pitch? • 3 possible hypotheses • No effect • A longer plateau makes a syllable sound higher in pitch • A longer plateau makes a syllable sound lower in pitch

10. Stimuli • Test sentence • “…came with Manny” taken from “Anna came with Manny” • Resynthesised nuclear accent (12 versions) • Frequency of contour • 160, 180, 200, 210 Hz • Shape of contour • Peak, 50ms or 100 ms plateau

11. Examples of Different Contours Anna Manny

12. Procedure • 7 subjects heard pairs of sentences • ‘Manny’ differed only in shape never in pitch • Question: “In which version is ‘Manny’ higher in pitch?” • Responded by pressing 1 or 2 on a keyboard 160 Hz plateau, peak 210 Hz peak, plateau

13. Percentage of times the longer stretch of contour sounds higher in pitch than the shorter stretch Overall 50ms vs. peak 100ms vs. peak 100ms vs. 50ms By frequency 160Hz 180Hz 200Hz 210Hz By Position Longer stretch in utt. 1 Longer stretch in utt. 2 By shape 1 50ms vs. peak 100ms vs. peak By shape 2 100ms vs. 50ms Results

14. Results Overall • The shape of the contour does affect the perception of pitch • 73% of responses were for “longer stretch of contour sounds higher”

15. Results by Frequency • The significant result holds at each of the 4 frequencies • 160 Hz 67% • 180 Hz 72% • 200 Hz 83% • 210 Hz 67%

16. Results by Position • The result is not significant when the plateau occurs in the first token • The result is significant when the plateau occurs in the second position

17. Results by Shape 1 • Both lengths of plateau are perceived as longer when compared to a peak • 50ms 77% • 100ms 77%

18. Results by Shape 2 • There is no significant difference between the two lengths of plateau

19. Results Summary • Overall plateaux are perceived as higher in pitch than peaks • Regardless of the frequency • This result is only significant when the plateau is in second position • Suggesting an interaction with position • There is no significant difference between the perception of the 2 plateau lengths • Suggesting a categorical rather than gradient difference

20. Possible Explanations • There are two possible explanations for why longer stretches of contour may sound higher in pitch • Integration • Temporal Smoothing

21. Integration • The listener may be integrating the area under the curve • There is a larger area under the curve of a plateau

22. Smoothing • The listener may be smoothing the curve • So peaks will sound lower than they actually are

23. Predictions • Integration • As the plateau gets longer pitch will be perceived as increasingly higher • Smoothing • There will be a cut-off point at which a longer plateau no longer sounds higher in pitch

24. Psychoacoustic Explanation • “Stability-sensitive weighting” • In pitch perception less weight is given to portions of the signal where frequency is changing rapidly • Due to the sluggishness of the auditory system

25. Pitch Perception • There are 2 mechanisms for extracting pitch from a signal • Place mechanism • Perceived pitch corresponds to the place of maximum excitation on the basilar membrane • Temporal mechanism • Pitch is derived from the frequency of nerve firings • Phase Locking

26. The Inner Ear

27. The Cochlea

28. Cochlea Cross Section

29. The Organ of Corti

30. Phase Locking • Neural firings firings occur at the same phase of the waveform each time • The intervals between firings will be integral multiples of the period of the wave

31. Sluggishness • If the pitch is changing too rapidly • The auditory system does not have enough time to phase lock • Pitch may not be perceived accurately

32. Conclusions • The shape of the contour does affect perceptions of pitch • Longer stretches of pitch sound higher than single turning points • There is a categorical difference between peaks and plateau • Suggesting an explanation based on smoothing • This effect can be explained as the sluggishness of the phase locking mechanism

33. Implications • My PhD • Plateaux may occur to enhance the prominence of the nucleus • Models of pitch span • May need to take account of contour shape • Intonational phonology • The situation is more complicated than peaks and troughs in the contour • A more perceptual approach is necessary