Perceptual reality monitoring as higher-order inference on sensory precision

Perception is a constructive process, relying on prior beliefs and expectations to shape what we experience. Besides facilitating more efficient perceptual inference, this generative architecture also enables the internal generation of sensory activity in the absence of input, in the form of imagination. Currently, it remains unclear how perceptual inferences reflecting reality can be kept apart from those that are imagined. Here, we propose a computational framework which formalizes reality inferences as higher-order evaluations of sensory precision. We demonstrate how this simple mechanism can explain a wide range of empirical observations regarding the relationships between perception, expectation, imagery and reality monitoring. Through simulations, we generate novel, testable hypotheses which highlight directions for future research. Our approach offers a simple yet powerful framework for developing a computational neuroscience of reality monitoring.