How prediction errors update presence/absence beliefs

Determine how the brain uses prediction error signals to update probabilistic beliefs about a stimulus’s presence or absence when the expected stimulus does not occur (i.e., during omission or absence of sensory input).

Background

The paper situates this problem within predictive coding frameworks, which traditionally model beliefs about stimulus features (e.g., pitch) using Gaussian representations and feature prediction errors. However, these approaches do not directly address how the brain updates beliefs regarding whether a stimulus occurred at all, particularly in omission scenarios where sensory input is absent.

To address this gap, the authors introduce the concept of an "absence prediction error" specifically for updating beliefs about stimulus occurrence. Their experimental paradigm in the human auditory system is designed to dissociate absence prediction error from feature prediction error and from passive entrainment effects.