Research Committee | Current Projects
The evaluation of vocal roughness by different listener groups
Angelika Braun, University of Marburg, Fachbereich Phonetik
In 2002, a study was carried out focusing on cultural differences between members of different language groups with respect to vocal "roughness" (Braun/Wagner 2002).
A total of 145 speakers speaking three different languages were studied. The languages covered were German, Italian, and Polish. Subjects read a reference text ("The North Wind and the Sun") for F0 and its SD measurements and phonated the vowel /a/ for a minimum of 3 sec. for the remainder of the measurements. The parameters studied were F0 and factors derived from it , i.e. SD, shimmer, jitter, and HNR.
The results of the statistical analyses (two-way ANOVA) indicate a significant difference in F0 between the Polish and the German group, thus confirming earlier findings about comparatively high F0 in speakers of Polish. With respect to HNR, the Polish and the Italian groups differ significantly, the Italian voices exhibiting the higher noise level. Furthermore, the Italian group exhibits significantly higher values of shimmer than the other two. We conclude that the differences found are caused by cultural stereotypes in the languages involved.
The ideal complement for this production study would be a perception study which would again involve subjects from the three countries. They will be asked (a) to decide whether or not the sustained vowels from the production study were spoken by a native speaker of their own language and (b) to rate the samples from "bright" to "rough" on a seven-point scale. Two groups of listeners will be interviewed: naïve and (forensically) trained. The results will be correlated with the production results. We hope to find that (a) trained listeners do better than naïve ones in that the per cent correct rate is higher for them than for naïve listeners and a higher correlation with production results is found and (b) the study will shed additional light on the complex relationship between the perception of "roughness" on the one hand and the measurable voice quality parameters on the other.
The forensic implications of the study are manifold: the production study has provided experimental proof for the fact that voice quality is indeed language-dependent. A "rough" voice would thus be regarded as "marked" in one language but as "unmarked" in another. The study which is proposed here will shed light on the important question of whether - and to which degree - human judgement coincides with acoustic measurements of jitter, shimmer, and HNR, the background being that in forensic samples measurements are usually impossible due to technical constraints.
Angelika Braun / Anita Wagner (2002): Is voice quality language-dependent? In: Braun, A. / Masthoff, H.R. (ed.) Phonetics and its applications. Festschrift for Jens-Peter Köster on the occasion of his 60th birthday. Stuttgart: Steiner, S. 298-312.