Predictive coding and stochastic resonance as fundamental principles of auditory phantom perception.

Summary

This review proposes that auditory phantom perception (tinnitus) can be understood through two linked computational principles: predictive coding, in which the brain constantly compares expected and actual sensory input, and stochastic resonance, in which weak or absent signals are amplified by intrinsic neural noise to become detectable, or, when miscalibrated, perceptible as sound that isn't there. Reduced peripheral input is proposed to shift this balance so that adaptive stochastic resonance overshoots, and top-down predictive machinery interprets the resulting noise as a real signal.

Key findings

• Frames tinnitus as an interaction between bottom-up stochastic resonance (noise amplification) and top-down predictive coding (expectation).
• Offers a computational account for why not everyone with reduced peripheral input (hidden hearing loss) develops tinnitus: it depends on how these two processes are calibrated.
• Connects tinnitus mechanistically to the broader Bayesian/predictive-brain framework also used to explain other perceptual and psychiatric phenomena.