What Does It Mean for a Voice to Sound “Normal”?

Abstract

It is rather unclear what is meant by ``normal'' voice quality, just as it is often unclear what is meant by ``voice quality'' in general. To shed light on this matter, listeners heard 1-sec sustained vowels produced by 100 female speakers, half of whom were recorded as part of a clinical voice evaluation and half of whom were undergraduate students who reported no vocal disorder. Listeners compared 20 voices at a time in a series of sort-and-rate trials, ordering the samples on a line according to the severity of perceived pathology. Any voices perceived as normal were placed in a box at one end of the line. Judgments of ``normal'' versus ``not-normal'' status were at chance. Listeners were relatively self-consistent, but disagreed with one another, especially about what counts as normal. Agreement was better, but still limited, about what counts as ``not normal.'' Strategies for separating ``normal'' from ``not normal'' differed widely across individual listeners, as did strategies for determining how much a given voice deviated from normal. However, acoustic modeling of listeners' responses showed that several acoustic measures---F0, F1 and F2, and F0 coefficient of variation---appeared more often than others as significant predictors of both categorical judgments and of scalar normalness ratings. These variables did not account for most of the variance in these analyses, and did not appear together in the perceptual models for even half of the listeners, but they did appear individually in most analyses, suggesting that in practice the concept of ``normal'' may have some small core of meaning based on F0 and vowel quality. Thus, the answer to our initial question of what it means for a voice to sound normal is a complex one that depends on the listener, the context, the purpose of the judgment, and other factors as well as on the voice.