What “imprecise predictive coding” actually means
Predictive coding is a way of describing perception as a guessing game. The brain constantly generates models of what it expects to encounter, then updates those models when sensory input confirms or contradicts them [2]. The “precision” of a prediction is how much confidence the brain assigns to it — how strongly it trusts one expectation over the noisy stream of incoming signals.
Imprecise predictive coding means this confidence weighting goes wrong. Predictions may be held too loosely, or prediction errors — the mismatches between expectation and input — may be given too much or too little weight. The result is aberrant perceptual inference: the brain misjudges what is real versus what it expected, and in extreme cases that misjudgment could surface as a hallucination [2][4]. This is a framework for explaining symptoms, not a directly observed cause.
The strongest evidence comes from hearing voices in schizophrenia
Most of the support for the imprecision idea comes from auditory verbal hallucinations — hearing voices — in schizophrenia. One framework links the temporal precision of predictive coding to the spatial wiring of sensorimotor circuits, proposing that disrupted motor-to-sensory transformation produces the imprecision behind these hallucinations [1].
Two related models place imprecise predictive coding at the core of classical schizophrenia. They argue that electrophysiology and functional imaging show aberrant activity at brain network hubs where predictions are encoded, and that frequent prediction errors might drive excess dopamine release, which in turn could trigger episodes of florid psychosis including hallucinations [2]. A later model extends this to a circuit-level account: unbalanced inputs to the striatum disturb how precisely subcortical signals modulate cortical activity, and this imprecision may precipitate acute reality distortion at times of stress [3]. These are proposed mechanisms, and the authors themselves note that environmental factors such as chronic inflammation may also contribute to the imprecision [2].
For vision, the picture is suggestive but unresolved
Visual disturbances short of full hallucinations are increasingly studied as early signs of psychosis, and the predictive coding hypothesis has been applied to them too: perceptual distortions are thought to arise from imprecise integration of internal predictions and incoming sensory data [4]. But it remains inconclusive whether these visual disturbances come from abnormal perceptual sensitivity, heightened sensory expectations, or both [4].
A virtual-reality study aims to separate a person's sensitivity to psychosis-like visual distortions from their bias toward seeing things that are not there, testing healthy adults across the general population [4]. That design is a plan for evidence rather than a finished answer, and it measures distortions and psychosis-like experiences rather than clinical hallucinations. So the visual side of the question is an active hypothesis, not a demonstrated shared mechanism.
A direct test of prediction-error signaling found no link to hallucinations
The cleanest challenge comes from mismatch negativity, a brain response treated as a prediction-error signal. In a study of 56 people with schizophrenia, 34 nonclinical voice hearers, and 48 healthy comparison subjects, people with schizophrenia showed the expected reduced mismatch negativity, but nonclinical voice hearers looked statistically similar to healthy controls [5]. If reduced prediction-error signaling were what produces hallucinations, the voice hearers should have shown it too.
The authors conclude that mismatch negativity as a prediction-error signal does not reflect processes relevant to the manifestation of hallucinations and delusions [5]. That does not dismantle the broader imprecision framework, which concerns the weighting of predictions rather than this one signal, but it does mean the auditory evidence is not uniform. The honest answer to the question is that imprecision in predictive coding is a promising and widely discussed explanation, best supported for auditory verbal hallucinations in schizophrenia, still speculative for visual disturbances, and not yet confirmed as a single mechanism spanning both modalities.
Sources used in this answer
Impaired functional double dissociation in motor-to-sensory transformation mediates aberrant predictive coding in auditory verbal hallucinations
Proposes a framework linking the temporal precision of predictive coding to the spatial architecture of sensorimotor circuits in auditory verbal hallucinations.
Imprecise Predictive Coding Is at the Core of Classical Schizophrenia
Argues that imprecise predictive coding is the core pathological process in classical schizophrenia, with frequent prediction errors possibly promoting dopamine release that triggers hallucinations and delusions.
The Mechanisms of Persisting Disability in Schizophrenia: Imprecise Predictive Coding via Corticostriatothalamic-Cortical Loop Dysfunction
Proposes a model in which unbalanced striatal inputs disturb the precision of subcortical modulation of cortical activity, potentially precipitating acute psychosis including hallucinations.
Using virtual reality to characterize perceptual factors underlying visual disturbances and their relation to psychosis-like experiences in the general population
Describes a virtual-reality study designed to test whether visual disturbances arise from abnormal perceptual sensitivity, heightened sensory expectations, or both, in relation to psychosis-like experiences.
Differing Pattern of Mismatch Negativity Responses in Clinical and Nonclinical Voice Hearers Challenge Predictive Coding Accounts of Psychosis
Finds that nonclinical voice hearers did not show the reduced mismatch negativity seen in schizophrenia, concluding that mismatch negativity as a prediction-error signal does not reflect processes relevant to hallucinations.