Natural ITD statistics predict human auditory spatial perception

A neural code adapted to the statistical structure of sensory cues may optimize perception. We investigated whether interaural time difference (ITD) statistics inherent in natural acoustic scenes are parameters determining spatial discriminability. The natural ITD rate of change across azimuth (ITDrc) and ITD variability over time (ITDv) were combined in a Fisher information statistic for assessing the amount of azimuthal information conveyed by this sensory cue. We hypothesized that natural ITD statistics drive the neural code for ITD and thus influence spatial perception. Human spatial discriminability and spatial novelty detection of sounds with invariant statistics correlated with natural ITD statistics. The lack of natural statistics in the stimuli suggests that this correlation results from natural statistics driving the neural code. Additionally, the density distribution of ITD tuning matching natural statistics is consistent with classic models of ITD coding and can explain the ITD tuning distribution observed in the mammalian brainstem.