Influences of word predictability and type of masker noise on intelligibility of sung text in live concerts

Vocal music is often intended to convey meaning, but how effectively this is achieved is poorly understood. This study systematically assessed the influence of three non-phonetic factors on the intelligibility of sung words in six public concerts in different venues: word predictability from sentence context, type of masker noise (spoken babble, sung vowels, [∫(w)]), and signal-to-noise ratio (SNR). Stimuli were sung live by a professional a cappella ensemble with one male singing target sentences and five others (two female) producing the masker sounds. The concert audiences (N = 319) reported the final word of each sentence using a handheld voting device, from four phonetically- and semantically-controlled written alternatives projected onto a screen after the sentence was sung. Although overall accuracy differed between performances, intelligibility patterns were robust across concerts. They included predicted main effects of masker noise type ([∫(w)] masking least disruptive, babble most), SNR (high > low), semantic predictability (high > low), listener age (young > old), and listener language status (native > non-native), and some strong interactions. These results suggest that, despite acoustic differences between sung and spoken words and the unusual and varied experimental venues, key findings from traditional speech research apply to sung words, given appropriate musical composition.