Perceptual judgments of breathy voice quality using rating scale, magnitude estimation, and matching tasks

Abstract

Experiments to study voice quality have typically used rating scales to obtain listener judgments. Recent research has questioned the validity of this data due to poor agreement between listeners [Kreiman et al., J. Acoust. Soc. Am. 87, 820–857 (1993)]. However, Shrivastav et al., J. Acoust. Soc. Am. 117, 2622 (2005) showed that such variability was partly related to the experimental methodology and could be minimized by multiple presentations of each stimulus to each listener. The nature of this averaged data, whether ordinal or interval, needs to be determined since it is necessary to develop appropriate perceptual scales for voice quality. The present experiment used three different psychophysical methods to obtain perceptual estimates of voice quality. These include a rating scale (with single and multiple presentations of stimuli to each listener), a direct magnitude estimation task, and a matching task. The experiment sought to determine if these three methods provided equivalent information. Perceptual judgments of breathiness were obtained for ten synthetic vowel continua that varied in terms of their aspiration noise, open quotient, or both. Preliminary results show that perceptual judgments using the magnitude estimation task and rating scale task (with multiple presentations of each stimulus) give very similar information. Results of the matching task will also be discussed. [Research supported by NIH/R21DC006690.]