January 16, 2020 / by Giorgia Cantisani / SONG

Neural signatures of musical and linguistic interactions during natural song listening

Song, poetry, nursery rhymes, and other expressive modalities that combine music and speech help us convey meaning and emotion beyond mere linguistic content. Such lyrics-tune synergy hints at a common neural basis beyond a shared sensorimotor system. Yet it remains unknown how the neural processing of music and speech interacts to form a cohesive perceptual whole.

In this work, we probed music and speech predictive mechanisms as they co-occur in song. The neural encoding of melodic expectations was compared when listeners were presented with songs or the corresponding hummed melodies. Similarly, the encoding of phonemic expectations was studied in song and the corresponding spoken lyrics. Here are some example stimuli:

We selected lyrics from the NHSS Speech and Singing Parallel Database, where the same singer recorded the same material both spoken and sung. This allowed us to compare speech, song, and melody while minimizing acoustic mismatches. We then created a hummed version by extracting the melody from each song and resynthesizing it with a voice-like sound.

We found that the two predictive processes occur in parallel and influence each other. We conjecture that such interaction is grounded in a competition for shared resources, which are allocated according to listeners’ preferences (endogenous bias), and in the balance of salience between the lyrics and the tunes, as established by the composer/performer (exogenous bias).