Specificity of adaptation to non-individualized head-related transfer functions

Abstract

Initial accuracy is poor when listeners are asked to localize virtual sound sources that have been low-pass filtered or rendered with non-individualized head-related transfer functions (HRTFs). However, Majdak et al. (2013) showed that, with training, localization performance improved when virtual sounds were rendered using HRTFs that were low-pass filtered at 8.5 kHz. This result suggests that previous outcomes showing training-induced improvement in localization performance with non-individualized HRTFs may merely be the result of listeners attending to low-frequency spectral information that is consistent with their own HRTFs, and not attending to the information from high frequencies, where HRTFs differ widely across individuals. That being the case, one would expect training to generalize to all non-individualized HRTFs, not just the non-individualized HRTF used during training. The current study investigated this hypothesis by performing tests of localization accuracy before and after three weeks of auditory localization training with a single non-individualized HRTF. For all subjects, localization performance with non-individualized HRTFs improved to a level at or near their performance with individualized HRTFs, and no generalization to the other non-trained HRTFs was found, suggesting subjects do learn to utilize an alternative set of high-frequency spectral information.