The Morgan Freeman effect: Long-term memory for voices frees up cognitive capacity to enhance speech perception in noise

When masked by competing speech, an utterance is more intelligible when it is spoken by someone familiar than by a stranger (novel voice). If this benefit occurs because listening to familiar voices is less cognitively demanding, then a concurrent task should disrupt perception of speech less if the attended voice is familiar. Participants (N=30) heard a mixture of two sentences in two different voices (familiar-novel or novel-novel) and reported the content of one (target) while ignoring the other (masker). Simultaneously, participants tracked the location of four moving dots on a screen (multiple-object tracking, MOT, dual-task) or ignored the visual input (single-task). Word-report was best when the target voice was familiar, worst when the masker voice was familiar, and intermediate when both voices were novel. Concurrent MOT performance reduced intelligibility of the novel but not the familiar target, suggesting that familiar voices require fewer cognitive resources to process than unfamiliar ones.