Sources of listener disagreement in voice quality assessment.

摘要: Traditional interval or ordinal rating scale protocols appear to be poorly suited measuring vocal quality. To investigate why this might so, listeners were asked classify pathological voices as having not different voice qualities. It was reasoned that simple task would allow focus on the kind of quality a had, rather than how much it possessed, and thus provide evidence for validity traditional In experiment 1, judged whether natural samples primarily breathy rough. Listener agreement in both tasks above chance, but agreed individual belonged particular perceptual classes. determine these results reflect listeners’ difficulty agreeing about single attributes complex stimuli, 2 classified synthetic stimuli (varying f0 only) low pitched pitched. If disagreements derive from difficulties dividing an auditory continuum consistently, then patterns should similar kinds stimuli. fact, listener significantly better voices. Difficulty isolating dimensions appears one reason unidimensional are unsuited pathologic Listeners did agree few aphonic breathy, with prominent fry and/or interharmonics These cases may have occurred because acoustic characteristics question corresponded limiting case being judged. Values generated more extreme consistent interpretation.