Long-lasting recalibration to auditory listening conditions

N Rabinowitz, M Schemitsch, O Brimijoin and E P Simoncelli

Published in Annual Meeting, Neuroscience, Oct 2015.
Perception can be described as a process of inferring the properties of objects in the world from incoming sensory signals [AlHazen 1040, Helmholtz 1867, Kersten et al 2004]. This inverse process is particularly difficult as object properties are entangled with those of contextual "viewing conditions". For example, before emitted sounds reach the ears, they undergo spectro- and spatio-temporal filtering by the room and head. The details of these transformations depend on the complex geometry and physics of a given environment, and can change considerably with head position. Thus, in order to make inferences about source objects, the brain must separate an incoming signal into "source" and "listening condition" components.

We hypothesized that the brain accomplishes this feat by factoring out systematic regularities that are learned from experience. To test this, we built a novel psychoacoustic chamber, wherein features of sound sources were systematically varied according to human subjects' head positions. When we coupled the carrier-frequency of a pure tone to subjects' head angle, they passively learn to compensate for this in perceptual judgments, without being aware of the perturbations. Curiously, this does not occur for other sound features such as modulation rate or depth [Rabinowitz et al, ARO 2014].

Based on these findings, we developed a model for perceptual recalibration based on known physiological properties of the peripheral auditory system, and used it to predict how listeners would be biased by altered environmental statistics. We hypothesized that spectral recalibration involves a rescaling of the gain of peripheral frequency channels to partially compensate for systematic discrepancies across head angles, and that these gain changes have a fundamental resolution limit imposed by the peripheral representation. From this model, we predicted that listeners would not only compensate for spectral discrepancies that they had been exposed to, but that they would maladaptively "anti-compensate" at nearby frequencies which were not part of the exposure regime. Indeed, we find that this effect manifests precisely as predicted. Moreover, the spectral resolution of perceptual recalibration appears to be comparable to that of peripheral critical bands.

The effects we report last for tens of minutes after the end of the exposure period. Control experiments in the closed-field rule out the possibility that they result from peripheral adaptation. Thus listeners' inferences about sound sources can be biased by systematic relationships experienced over extended time scales, but only through means allowed by physiology.

  • Related Publications: Rabinowitz14a
  • Listing of all publications