Speech rhythm as durational marking of prosodic heads and edges

P. Prieto1 – M.M. Vanrell2

1ICREA-UPF, 2UPF-UAB

Workshop on Prosodic Development, April 16, 2010

Introduction: The old ‘isochrony’ hypothesis

• Traditional studies on linguistic rhythm classified languages as being “syllable-timed” (Spanish and French) vs. “stress-timed” (as English and Dutch). Abercrombie 1967, Pike 1945. Spanish English

• It was hypothesized that this perceptual contrast was due to isochrony of speech intervals, that is, syllables tend to be of equal duration in syllable-timed languages and stress-delimited feet tend to be of equal duration in stress-timed languages.

• Yet a variety of instrumental studies showed that stress-based or syllable-based isochrony was not systematic in either rhythm class.

Bolinger 1965, Borzone de Manrique & Signorini 1983, Dauer 1983, among others; see Ramus et al. 1999 for a review).

Introduction

Goals

Methodology

Results

Conclusions

Introduction: The phonological approach

• A radically different view of linguistic rhythm was put forward by Dasher and Bolinger (1982): the perception of rhythm classes is the result of specific phonological phenomena in a given language.

• The phonological approach can be considered to be the ‘standard view’ nowadays. The differences in rhythm are claimed to be the result of the presence vs. absence of two main phonological properties:

• syllable structure : stress-timed languages have a greater variety of syllable types than syllable-timed languages;

• vowel reduction : in stress-timed languages, unstressed syllables usually have a reduced vocalic system;

Dasher and Bolinger 1982, Dauer 1983, 1987, Nespor 1990, Ramus et al. 1999, among others.

Introduction

Goals

Methodology

Results

Conclusions

IntroductionThe phonological approach:

some questions

• Yet, the materials used typically reflect the language-specific phonological properties of the language under study, leading to differences in the rhythm measures.

• Our questions : a. If syllable structure properties are controlled for, do we still observe durational planning differences across languages?

b. Do vowel reduction properties obligatorily entail accentual rhythmic properties?

• In other words, are there any differences of the speech rhythm and timing of successive syllables regardless of the phonological properties found in the language or in the materials?

• If so, what are the rhythm measures that reflect these differences?

Introduction

Goals

Methodology

Results

Conclusions

Goals

The goal of this study is twofold:

• To investigate whether rhythmic differences arise regardless of syllable structure effects.

• Specifically, we analyze the effects of syllable structure composition on different measures of rhythm and find out whether we find a reliable measure of rhythm contrasts.

• To investigate whether vowel reduction differences across languages necessarily entail differences in timing patterns, as predicted by the phonological approach.

Introduction

Goals

Methodology

Results

Conclusions

� If we find top-down durational planning differences, linguistic rhythm cannot just be considered the result of the language-

particular phonological properties making up sentences.

Languages under study

� We compare three languages that have traditionally been classified differently in the rhythmic scale:

� Spanish (a prototypical example of syllable-timed language ): � No vowel reduction;

� Basically CV-syllables;

� English (idem, a stress-timed language );� Vowel reduction;

� Greater variety of syllable types;

� Catalan (a mixed or intermediate language in the rhythmic scale):� Vowel reduction;

� Mixed variety of syllable types;

Introduction

Goals

Methodology

Results

Conclusions

Methodology-Materials

The experimental materials consisted of two types of utterances:

� CONTROLLED: A set of utterances which were controlled for utterance length and syllabic structure

� 10 utterances per language

� MIXED: A set of uncontrolled utterances which reflect the typical phonological properties of the language. The same materials used in Ramus et al. (1999)

� 20 utterances per language

Introduction

Goals

Methodology

Results

Conclusions

Methodology-Examples

A: CONTROLLED MATERIALS:

TYPE 1. OPEN SYLLABLES (Predominantly CV-type Utter ances)Cat: La mare de la Jana és de Badalona.Eng: The mother of Susana is from Badalona.Span: La madre de Susana es de Badalona.

TYPE 2. CLOSED SYLLABLES (Predominantly CVC-type Ut terances)Cat: Els donuts d’Amsterdam són realment internacionalsEng: These doughnuts from Amsterdam taste almost exceptionalSpan: Los donuts de Ámsterdam són realmente internacionales

B. UNCONTROLLED MATERIALS:

MIXEDCat: Ell mai va tenir la possibilitat d'expressar-se. Eng: A hurricane was announced this afternoon on the TV. Span: Se enteraron de la noticia en este diario.

Introduction

Goals

Methodology

Results

Conclusions

Methodology-Speakers

� 24 speakers read the 30 target utterances at a normal speech rate:

� 8 Southern English speakers (Cambridge);

� 8 Central Peninsular Spanish speakers (Madrid)

� 8 Central Catalan speakers (Barcelona).

� Reading task performed as a part of a long recording session, which included other tasks which investigated the rhythmic properties of children’s speech and adult speech (see other contributions in this panel).

� Total number of utterances analyzed = 720 (24 speakers x 30 utterances).

� Total number of syllables analyzed = 12086.

� Total number of segments analyzed = 29151.

Introduction

Goals

Methodology

Results

Conclusions

Introduction

Goals

Methodology

Results

Conclusions

Segmentation with Praat

Time (s)

100

150

200

250

300

Fundamental frequency (Hz)

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5

ssabssa s s ssa s sa ss ssa s s nsa s sef

cv ccvvcvccccvcvccccvc v c cvvcccvccvvccvc c vc ccvc

a a a a a na

b ef

Introduction

Goals

Methodology

Results

Conclusions

Measuring rhythm: Metrics

� Several metrics relatively successful in distinguishing languages which fall in different rhythmic categories (see White & Mattys 2007 for a review).

� We calculated full set of measures, taking into account White & Mattys (2007), who find most effective measures to be %V, VarcoV and nPVI-V.

� Interval measures (Ramus et al 1999):

� %V, or proportion of vocalic intervals within the sentence (sum of vocalic intervals divided by total duration of sentence).

� ∆V: the standard deviation of the duration of vocalic intervals within each sentence.

� ∆C: the standard deviation of the duration of consonantal intervals within each sentence.

Introduction

Goals

Methodology

Results

Conclusions

Measuring rhythm: PVIs and rate-normalized measures

� Pairwise variability indices (PVIs): Grabe & Low 2002, Low, Grabe & Nolan 2000)

� nPVI-V, or normalized PVI: the mean of the duration differences between successive intervals (Vs) divided by the sum of the same intervals.

� rPVI-V, or non-normalized Pairwise Variability Index.

� Rate-normalized measures VarcoV and VarcoC(Dellwo & Wagner 2003, Dellwo 2006):

� VarcoV: standard deviation of vocalic interval duration divided by mean vocalic duration (and multiplied by 100).

� VarcoC: standard deviation of consonantal interval duration divided by mean consonantal duration (and multipled by 100).

Effects of syllable structure on interval measures, %V, ∆V, ∆C)

� Strong effects of syllable structure on %V: the simpler the syllable structure, the higher %V is.

� English utterances with mostly CV syllables have similar %V values to mixed Catalan and Spanish utterances.

Results

�A one-way ANOVA shows a significant main effect of language and syllable type (at p<0.001) on %V (p=0.295) and no significant interaction between lang*sylltype

(p=0.396).

�Same effects for ∆V and ∆C.

%V

Effects of syllable structure on interval measures (%V, ∆V, ∆C)

� Ramus et al’s (1999) graph representing the distribution of their data (Catalan, English, Spanish) over the ∆C and %V plane.

Results

Error bars represent standard errors.

Effects of syllable structure on interval measures (%V, ∆V, ∆C)

� Ramus et al’s (1999) graph representing the distribution both our data and their data for Catalan, English, and Spanish.

Results

0,02

0,025

0,03

0,035

0,04

0,045

30 35 40 45 50 55 60

%V

∆C

CA CV

CA CVC

CA mixed

SP CV

SP CVC

SP mixed

EN CV

EN CVC

EN mixed

�The degree of overlap between languages demonstrates that interval measures are strongly influenced by syllable structure.

Effects of syllable structure on pairwise variability indices (nPVI-V)

� Mean nPVI-V is higher in English than in Catalan and Spanish, across all syllable types.

� nPVI-V is a stable rhythm measure across languages.

Results

�A one-way ANOVA shows a significant main effect of

language (p<0.001) and no significant main effects of syllable type on nPVI-V

(p=0.295) and no significant interaction between

lang*sylltype (p=0.921).

� Post-hoc analyses groupEnglish vs. Catalan/Spanish

nPVI-V

Effects of syllable structure on pairwise variability indices (rPVI-V)

� Similar results for rPVI-V: higher in English than in Catalan and Spanish, across all syllable types.

� rPVI-V is a somewhat less stable rhythm measure across languages.

Results

� A one-way ANOVA shows significant effects of syllable type and language on rPVI-V

(p<0.05) and a significant interaction between

lang*sylltype.�Group differences : English

vs. Catalan/Spanish

rPVI-V

Varco-V and Varco-C

� Varco-V (but not Varco-C) is higher in English than in Catalan and Spanish, across all syllable types, even though in some cases the effects are not significant.

� Varco-V is another stable rhythm measure across languages.

Results

�A one-way ANOVA shows a significant effect of language on VarcoV (p<0.001), no significant effects of syllable type (p=0.270) and no significant interaction lang*sylltype (p=0.227).

�No significant effects of language nor of syllable type on VarcoC (p=0.486 and p=0.141).

Varco-V Varco-C

Thus, what are the phonetic correlates cueing rhythmic differences between English vs. Catalan/Spanish?

Hypotheses:� Durational marking of prominence (prosodic

heads);� Durational marking of prosodic edges;

Timing patterns. Effects of levels of stresson syllable duration

� English has a larger difference between unstressed vs. stressed syllables vs. nuclear stressed syllables than Catalan and Spanish.

Results

�A one-way ANOVA shows a significant main effect of language (p<0.001) and

stress level (p<0.001) and a significant interaction

between lang*sylltype.

Timing patterns. Effects of final lengtheningon vowel duration

� English has longer syllables at the end of prosodic domains than Catalan and Spanish.

Results

�A one-way ANOVA shows a significant main effect of language (p<0.001) and

boundary type on syllable duration (p<0.001) and a

significant interaction between lang*sylltype.

Conclusions

� Strong effects of syllable structure on interval measures, namely, %V, ∆V, and ∆C.

� Important rhythmic differences arise using the nPVI-V and the Varco-V indexes between English vs. Catalan/Spanish, regardless of syllable structure.

� Thus pairwise variability measures (PVIs) and normalized speech rate measures (VarcoV) are more robust and independent of syllable structure.

Introduction

Goals

Background

Methodology

Results

Conclusions

Conclusions� Even though the phonology of the language

can help in increasing the rhythmic differences, there are important differences in durational planning too.

Spanish English

� Final claim: an important correlate of rhythmic differences across languages is the durational marking of prosodic heads and edges on timing patterns across languages.

� This can represent a tool to analyze in a more fine-grained way rhythmic differences across languages, both in adult and developmental systems.

Introduction

Goals

Background

Methodology

Results

Conclusions

Gràcies!

Acknowledgments

We would like to thank Núria Argemí, Anna Barberà, Melanie Jean Bell, Ana Estrella, and Francesc Torres-Tamarit for recording the data in the three

languages, and to Naomi Hilton for performing the segmentation and coding of the data.

This research has been funded by a Batista i Roca research project entitled The acquisition of rhythm in Catalan, Spanish and English (Ref.: 2007 PBR 29) awarded

by the Generalitat de Catalunya, and by the projects HUM2006-01758/FILO and HUM2006-01758/FILO and CONSOLIDER-INGENIO 2010 “Bilingüismo y

Neurociencia Cognitiva CSD2007-00012” awarded by the Spanish Ministerio de Ciencia e Innovación.

References (1)

Introduction

Goals

Background

Methodology

Results

Conclusions

References

Abercrombie, D. (1967). Elements of general phonetics. Edinburgh: Edinburgh University Press.

Asu, E. & Nolan, F. (2006) Estonian and English rhythm: a two-dimensional quantification based on syllables and feet. In Hoffmann, R. i H. Mixdorf (eds). Proceedings of Speech Prosody 1, 249-252. Dresden: TUDpress.

Dasher, R. & Bolinger, D. (1982). On pre-accentual lengthening, Journal of the International Phonetic Association, 12, pp. 58-69.

Dauer, R. M. (1983). Stress-timing and syllable-timing reanalyzed. Journal of Phonetics, 11, 51–62.

Dauer, R. M. (1987). Phonetic and phonological components of language rhythm. In Proceedings of the 11th International Congress of Phonetic Sciences (pp. 447–450). Talinn.

Dellwo, V. (2006). Rhythm and speech rate: A variation coefficient for deltaC. In P. Karnowski, & I. Szigeti (Eds.), Language and language processing: Proceedings of the 38th linguistic colloquium (pp. 231–241). Piliscsaba 2003. Frankfurt: Peter Lang.

Dellwo, V., & Wagner, P. (2003). Relations between language rhythm and speech rate. In Proceedings of the 15th international congress of phonetics sciences (pp. 471–474). Barcelona.

Gavaldà-Ferré, N. (2007). Vowel reduction and Catalan speech rhythm. Unpublished MA Thesis, University College London.

Low, E. L., Grabe, E., & Nolan, F. (2000). Quantitative characterisations of speech rhythm: ‘Syllable-timing’ in Singapore English. Language and Speech, 43, 377–401.

Grabe, E., & Low, E. L. (2002). Durational variability in speech and the rhythm class hypothesis. In N. Warner, & C. Gussenhoven (Eds.), Papers in laboratory phonology 7 (pp. 515–546). Berlin: Mouton de Gruyter.

References (2)

Nazzi, T., Bertoncini, J., & Mehler, J. (1998). Language discrimination by newborns: Towards an understanding of the role of rhythm. Journal of Experimental Psychology: Human Perception and Performance, 24, 756–766.

Nespor, M. (1990). On the rhythm parameter in phonology. In I. M. Roca, Logical issues in language acquisition, (pp. 157-175). Dordrecht: Foris

Payne, E., Post, B., Astruc, L., Prieto, P., Vanrell, M.M. (submitted). “Measuring Rhythm in Child-Mother Interaction”. Language and Speech.

Peterson, G. E., & Lehiste, I. (1960). Duration of syllable nuclei in English. Journal of the Acoustical Society of America, 32, 693–703.

Pike, K. (1945). The intonation of American English. Ann Arbor: University of Michigan Press.

Ramus, F., Nespor, M., & Mehler, J. (1999). Correlates of linguistic rhythm in the speech signal. Cognition 73, 265-292.

Ramus, F. (2002). Acoustic correlates of linguistic rhythm: Perspectives. In Proceedings of speech prosody 2002 (pp. 115–120). Aix-en-Provence.

Ramus, F., Dupoux, E., & Mehler, J. (2003). The psychological reality of rhythm classes: Perceptual studies. In Proceedings of the 15th International Congress of Phonetic Sciences (pp. 337–342). Barcelona.

White, L. & Mattys, S.L. (2007). Rhythmic typology and variation in first and second languages. In P. Prieto, J. Mascaró, & M.-J. Solé (eds.), Segmental and Prosodic Issues in Romance Phonology. Current Issues in Linguistic Theory series. Amsterdam: John Benjamins.

White, L. & Mattys, S.L. (2007). Calibrating rhythm: First language and second language studies. Journal of Phonetics, 35, 501-522.

Introduction

Goals

Background

Methodology

Results

Conclusions

References