Exploring a listener's ability to “eavesdrop” on two simultaneous conversations

Abstract

Most speech perception studies assume that the talkers are speaking to the subject, but little is known regarding subject ability to monitor exchanges between interlocutors (i.e., “eavesdrop”). The current study measured subject ability to monitor two concurrent, artificially generated “conversations” using phrases from the Coordinate Response Measure corpus. Individual conversations consisted of a “prompt” phrase of the form: “Ready [call sign] go to [color]-[number] now” and a “response” phrase of the form: “[same call sign] going to [color]-[number]”. A color and/or number prompt-response mismatch occurred within a conversation on 50% of trials. The listener was asked to identify whether a mismatch took place and, if so, identify in which of the two conversations the mismatch occurred. The (virtual) conversations were either co-located or spatially separated (±15°, ±90°). All talkers were of the same gender, or competing prompters and competing responders were of different genders. Group-mean performance ranged from about 65–80% correct in mismatch detection, and about 70%–85% correct in call sign identification (for mistake-ID correct trials). Performance improved when competing talkers were separated to ±15 deg in space, and/or in gender. This suggests that, on average, listeners can follow simultaneous conversations over time, particularly when competing talkers differ in pitch or space.