Studying dynamic sound source localization in a virtual maze with EEG and psychophysics

Auditory scene analysis benefits from spatial separation of the sound sources to be disentangled. Studies characterizing the ability to localize sound sources have revealed that spatial ambiguities can be resolved by slightly moving the head. This implies that the auditory system must take into account how head movements displace the sensors (ears) relative to the sound source, in order to separate the consequences of own movement from actual movement of the sound source. Outside of the typical lab, not only the head can be turned, but also the whole body might move towards or away from the sound source. With modern virtual-reality (VR) technology including near-real-time auditory feedback, it has now become feasible to study the relations between body movement, head movement, and sound localization in a controlled manner. Here we present a VR study in which participants move through a virtual maze to locate a repetitive alarm signal. Participants' turning decisions provide a behavioral measure of their ability to localize the sound source. In addition, we introduce occasional task-unrelated location deviations in the alarm signal to record an unobtrusive neural measure of localization ability: the Mismatch Negativity (MMN) component extracted from participants' continuous electroencephalogram (EEG) informs us about participants' deviance detection at the sensory level. We examine these neural and behavioral correlates of sound localization during simulated locomotion and real head movements. Our results show high behavioral accuracy and the elicitation of a robust MMN component. This indicates faithful extraction of source location although participants move relative to the sound source. Our findings are compatible with a predictive-coding framework, where the effects caused by own movement are taken into account for the interpretation of sensory signals. Yet because the range of physical variation by own movement was narrower than intended, we cannot exclude the possibility that deviance detection rested on simpler sensory mechanisms. We discuss how this constraint can be overcome in future studies to fully exploit our new paradigm for naturalistic, yet controlled studies on sound localization and auditory scene analysis under dynamic listening requirements.