Rhythm, Metrics, and the Link to Phonology

Rhythm, Metrics, and the Link to Phonology Jason Brown Dept. of Applied Language Studies & Linguistics University of Auckland jason.brown@auckland.ac.nz Sam Mandal University of Western Sydney and MARCS Auditory Laboratories s.mandal@uws.edu.au Abstract Since Ramus et al. (1999) a number of statisti- cal metrics have been routinely employed by researchers (Ramus 2003, Grabe & Low 2002 etc.) in an effort to rhythmically classify lan- guages. However, recent studies by Arvaniti (2009), Tilsen & Arvaniti (2013), Arvaniti & Rodriquez (2013) etc., have challenged both the validity of these metrics in reflecting speech rhythm, and the physical measurability of rhythm itself. The present study takes a comparative evaluative approach, and explores the applicability of the proposed metrics to a Papuan language (Urama) with a phonology quite different than traditional Western Euro- pean (W.E.) languages. It is argued here that the statistical underpinning of the existing rhythm metrics is a direct outcome of an overt effort to capture the temporal durational char- acteristics of the phonotactics of W.E. lan- guages. As such, these metrics are only capa- ble of providing a crude measure of timing. 1 Introduction Approaches to rhythm in language have tradi- tionally viewed languages as falling into either strict or less precisely-defined categories that were based primarily on the notion of timing (Abercrombie 1967, Port et al. 1987). Timing, as used in these contexts, is more or less a blanket- term referring to all aspects of durational varia- tion in speech. Originally linked to the notion of isochrony (Abercrombie 1967), the term has since been increasingly used to refer to the rela- tive temporal durational variability of consonan- tal and vocalic intervals in running speech (Ra- mus et al. 1999). The cornerstone of subsequent studies (Grabe & Low 2002, Ramus et al. 2003, Dellwo 2006 etc.) have been a consistent focus on the relative variability of consonantal and vo- calic durations across languages, with an effort to establish generalizations in patterns of durational variability to help classify languages into differ- ent rhythm classes. Metrics were developed as tools of statistical measurement of said duration- al variability in speech. The efficacy of such ef- forts, and the phonological basis of the rationale offered for their methodological choices, have been vigorously disputed in recent works. Arvan- iti (2009, 2012), Tilsen & Arvaniti (2013), Ar- vaniti & Rodriquez (2013) present empirical evi- dence to illustrate that the aforementioned statis- tical metrics can neither classify non-prototypical languages, nor provide correlates of perceptual discrimination. These authors offer perceptual experimental data to prove that not only is timing affected by a multitude of factors (speaking rate, voice quality, stimuli type etc.), but perceptual discrimination is often achieved through attune- ment to duration-independent acoustic factors such as fundamental frequency (f0). This being the case, the present paper aims to investigate the phonological basis for these acoustic metrics, the interrelation between the mathematical formulae they employ and the acoustic correlates of rhythm they are supposed to measure. Our primary hypothesis is that these metrics are simply different statistical measures …

Content truncated to 500 words