Mary Kohn and Melissa Farr


            Quantitative analysis of speech rhythm was first made possible by interval-based methods (Ramus, Nespor, & Mehler 1999; Lowe, Grabe & Nolan 2000).  For example, the Pairwise Variability Index (PVI) averages differences in consecutive phonological segments to quantify rhythm (Lowe, et al 2000).  Johnson and Tilsen (2008) question these methods as they impose phonemic boundaries on phonetic output and compare the segments contained by the rhythm of the phrase, instead of measuring the rhythmic structure of the phrase.  They suggest that studies should analyze amplitude fluctuations in the sonorant energy of speech by performing a low-frequency Fourier analysis (LFFA) on the amplitude wave of the band-pass filtered signal.  This study compares PVI and LFFA methods to identify if the methods produce similar distributions among speakers.  If LFFA results differ from PVI results, linguists should explore which method most accurately describes perception of speech rhythm. 
            To compare PVI and LFFA, both methods were used to analyze the speech of six English speakers and two Spanish speakers.  Two of these English speakers are Spanish language heritage speakers and are expected have more syllable-timed speech, while the other four should demonstrate more stress-timing following previous research (Carter 2005).  100 PVI values were calculated per speaker.  The sound sample was low band pass filtered to capture sonorant energy. A spectrum of the amplitude envelope from the filtered signal was analyzed to find the periodicity of the changes in amplitude.  The speakers were compared using each metric in order to identify whether the distributions emerges as similar. 
            The results show, surprisingly, that the speakers do not separate into the iconic groups expected by either method:  the WAV speakers do not categorically appear more stress-timed than the Latino English speakers. Both the LFFA and PVI accounts confirm these results indicating that these techniques allow for inter-speaker comparison of speech rhythm.  The LFFA improves upon PVI, however, by revealing intra-speaker variation as each phrase’s periodicity is quantified.

References

            Carter, P. (2005). Prosodic variation in SLA: Rhythm in an urban NC Hispanic
Community. Penn Working Papers in Linguistics, 11.2.
            Lowe., E. Grabe and F.  Nolan (2000). Quantitative characterizations of speech rhythm: syllable-timing in Singapore English. Language and Speech 43 (3):229 – 259.
            Ramus, F., M. Nespor, and J. Mehler. 1999. Correlates of linguistic rhythm in the
speech signal. Cognition 73:265-92.