Perceiving the speech of multiple concurrent talkers in a combined divided and selective attention task.

Detection and identification of concurrently spoken key words was investigated using the Coordinate Response Measure corpus. On every trial, listeners first had to explicitly detect callsign keywords in a multi-talker stimulus (divided attention), and, if all callsigns were present, identify the color and number words produced by one of the talkers (selective attention). Increasing the number of concurrent talkers and the number of callsigns to be detected each had a marked negative effect on detection and identification performance. These findings indicate that, when memory involvement is limited, listeners cannot reliably detect more than two concurrently spoken words in diotic listening.