Binaural independence indices and the edges of auditory images

The edge model is uniquely consistent with spatial hearing when spatial cues are broadly distributed over time, prompting auditory images or their edges to ‘split apart’.

Below, we consider four models of spatial hearing: 1) edge, 2) maximal, 3) inter-hemispheric channel, and 4) acoustic cross-correlation, as binaural independence indices are varied.

Listen to examples of varying binaural independence  here or test yourself in an online psychoacoustic experiment.

Edge model.  Top panels show edge model representations in the left and right hemispheres, for various frequency bands, as independence indices are raised. Dark colors indicate half-maximal neural activities (limit-2) and thus the contralateral edges of auditory images. Left and right representations are averaged in the bottom panel to show when the two edges become distinct. Listeners typically begin reporting more than one edge for independence indices exceeding ~3, corresponding to when edges in the left and right hemispheres become distinct after averaging.

Maximal model. Top panels show representations in the left and right hemispheres, for various frequency bands, as independence indices are raised. Dark colors indicate maximal neural activities and thus the auditory images modeled in each hemisphere. Left and right representations are averaged in the bottom panel. Despite noticeable broadening of activity patterns as independence indices are raised, maximal activities remain coincident with frontal ITDs (~0 ms), with no indication as to why listeners might begin to report hearing two images or two edges.

Inter-hemispheric channel model. Neural activities in both hemispheres remain equal as independence indices are raised, with no indication as to why listeners might begin to report hearing two images or two edges.

Acoustic cross-correlation model. Results of acoustic cross-correlation showing that correlation arguments coalesce around zero as the independence indices are raised, with little indication as to why listeners might begin to report hearing two images or two edges.

notes:

 A few citations:

Igor Pro experiment used to obtain modeled results

Igor Pro experiment used obtain correlation arguments & cues versus independence indices

Igor Pro XOP required by both experiments