Auditory Attention Based on Differences in Median Vertical Plane Position

The first experiment described here asks whether listeners are able to selectively attend to one of two sentences differing in median vertical plane (MVP) location using a paradigm developed for azimuthal attention [1]. It also asks whether their ability to use MVP cues improves with a difference in fundamental frequency (Fo) between the two sentences. Listeners attend to one of two simultaneous sametalker utterances and report which of two target words ("speech" or "phrase") occur in the attended sentence. The sentences are played from matched loudspeakers in an anechoic chamber from different MVP positions. When the sentences are both played on the same constant Fo, listeners report the target word almost perfectly with a 31° vertical separation. For smaller separations, performance is worse but improves with increasing difference in Fo between the sentences. This improvement is not due to the Fo difference improving listeners' ability to use MVP cues, but rather to their using continuity of Fo difference to track the target sentence. When the sentences are low-pass filtered, listeners are less able to use MVP cues and so there is a greater relative use of Fo continuity. The second experiment showed that both the ability to selectively attend to one of two sentences and the ability to localise a single sentence were worse with headphone presentation after convolution with individually-optimised library HRTFs than with free-field presentation. However, the low-pass filtering at 5 kHz gave relatively little additional degradation in either task.