A multidimensional scaling study of normal voice quality in females

Abstract

Little is currently known about the perceptual dimensions listeners use in discriminating female voices and how these dimensions relate to perceptual rating‐scale judgements. The purpose of this study was to examine these issues for normal female voices. In addition, the judgments of 20 naive listeners and 20 experienced listeners (speech pathologists) were compared. Listeners provided dissimilarity judgments of sentences produced by 20 females (24–25 years). The listeners also rated each voice on 17 perceptual rating scales (e.g., high pitch‐low pitch, clear‐hoarse). Results of the multidimensional scaling (MDS) analysis produced a five‐dimensional solution (RSQ = 0.509). A stepwise linear regression related the perceptual attributes of pitch, age, loudness, “liveliness,” and pitch variability/quality to the five MDS dimensions. When listener groups were compared, the experienced listeners relied on pitch somewhat more than the naive listeners, while the naive listeners utilized the pitch variability/quality dimension to a greater degree. This study demonstrates that listeners' perceptual ratings can be significantly correlated with perceptual dimensions derived from MDS analysis. [Research supported by BRSG Grant S07 RR07031, DRR, NIH.]