Background: Speech-in-speech perception is challenging, but can be facilitated by prior knowledge. For example, advance cueing of the semantic context (“what”) and spatial location (“where”) of an upcoming target sentence each improves speech-in-speech intelligibility. Additionally, speech-in-speech intelligibility is better when the target sentence is spoken by a familiar rather than unfamiliar voice (known as the familiar-voice benefit), although explicit cueing of talker identity (“who”) has not been tested. Here, we aimed to compare how advance cueing of talker identity, semantic context, and spatial location affects speech-in-speech perception.

Experiment 1: We recruited 24 participants without hearing loss, aged 18–45 years. Participants were first trained to identify the voices of three male talkers (for 166 sentences each). Then, in the speech-in-speech intelligibility task, participants heard two concurrent sentences (target and masker) amongst background noise on each trial, and repeated the target sentence aloud. The two concurrent sentences were spoken by different (male) talkers, had different topics, and were spatially separated. Before the sentences began, participants were visually cued to the talker identity (e.g., “John”), topic (e.g., “Animals”), or spatial location (e.g., “Left”) of the upcoming target sentence, or saw an uninformative visual cue (baseline condition).

We found that semantic and spatial cues produced significantly better intelligibility than uninformative cues, with no significant difference in intelligibility between semantic and spatial cues. In contrast, talker-identity cues produced no significant benefit compared to uninformative cues.

Experiment 2: To test whether there is a talker-cue benefit to intelligibility under certain conditions, Experiment 2 focused on the talker- and uninformative-cue conditions. Twenty naïve participants were trained to identify three male talkers, but were only cued to one talker (trained for 166 trials) during the speech-in-speech intelligibility task.

We found a significant familiar-target intelligibility benefit for the trained voice compared to novel voices. However, participants gained no additional benefit from receiving a cue that signalled the identity of the upcoming target talker compared to an uninformative cue.

Conclusions: Our findings suggest that advance cues about semantic context and spatial location improve speech-in-speech intelligibility. Whereas, across two experiments, we found no evidence that talker-identity cues improve intelligibility, even when participants are sufficiently familiar with a voice to gain a familiar-voice benefit over novel voices.