Abstract:

Categorical judgments of vocal quality are often applied in studies seeking to synthesize particular voice types, and in studies of the acoustic correlates of particular qualities. However, listeners' ability to reliably classify voices as ``breathy'' or ``rough'' has never been demonstrated. In this study, 15 listeners decided whether each of 160 voice samples was or was not primarily rough in quality; in a separate experiment, a second group of listeners judged whether each voice was primarily breathy. Although listener agreement exceeded chance levels, only a few voices were consistently categorized as breathy or rough. Many voices with a high probability of being judged ``breathy'' also had a high probability of being judged ``rough.'' Finally, the likelihood of a ``rough'' or ``breathy'' response correlated highly with the rated severity of vocal pathology. These results confirm that breathiness and roughness are not independent concepts, and that neither captures information apart from overall severity of pathology. Thus categorical descriptions of a pathological voice are probably not useful clinical or research tools. [Work supported by NIH.]