Fundamental frequency stability in speech and singing

Fundamental frequency (f0) stability is generally assumed to be an attribute of the singing voice, though it can also be found in speech. In a previous work, it was found that sung sentences performed with relative freedom within MPB (Musica Popular Brasileira) style show less f0 variability in comparison with their spoken counterparts and, for this reason, we account for f0 stability to be an important acoustic feature that distinguishes singing from speech. The current study is complementary to the previous study whose results statistically confirmed the hypothesis that singing has a more stable f0 than speech. Our aim with the perpetual experiment is to verify whether the stability feature is perceived by naïve listeners. 36 listeners were asked to classify sentences spoken by actresses and sung by singers from four available answers, choosing only one of them. The dataset used for the perceptual stimuli was formed by four textual units comprising 25 spoken and 25 sung sentences uttered respectively by an actress and a singer and was a representative part of the previous study full dataset. Listeners' classification obtained for each sentence was compared to the f0 variation of this same sentence. Statistical analysis was done by means of a discrete choice model (DCM) aiming to verify whether f0 variance could be an attribute for listeners to choose between one of the four responses. The result indicates a statistically significant (Pr (> |z| 2.22e-16) relationship between f0 stability and the singing classification. The discussion takes listeners' impressions about the test and the conclusion indicates that the current finding is empirical evidence of cognitive aspects of stability in music.