# Predictive Coding as a Framework for Understanding Psychosis

> **NIH NIH R01** · UNIVERSITY OF MARYLAND BALTIMORE · 2020 · $701,897

## Abstract

7. Project Summary
 This application responds to the NIMH PAR-16-136, “Using the NIMH Research Domain Criteria (RDoC)
Approach to Understand Psychosis.” Psychotic symptoms, such as delusions and hallucinations, are
treatment-resistant in many patients and are associated with high levels of distress and impairment. Treatment
advances have been slowed by the lack of a model of how these symptoms arise and persist. Adopting the
RDoC approach, we suggest that these symptoms may result from abnormalities in the neural and cognitive
processes that underlie perception, action, and belief formation. Hierarchical predictive coding represents an
explanatory framework that unites function and dysfunction in perception action and belief formation. We
perceive, act, and believe based on our prior experiences, and we update those priors in light of new data and
the prediction errors they elicit. We suggest that hallucinations and delusions form, and are maintained, via
aberrant predictive coding mechanisms that vitiate perception, action and belief.
 We will test these hypotheses with a suite of predictive coding measures in a large sample, capturing
variability in symptom severity and duration. We will use functional magnetic resonance imaging (fMRI) during
tasks of perception, action, and belief, electroencephalography to measure mismatch negativity (MMN) to
unexpected perceptual stimuli, and magnetic resonance spectroscopy (MRS) to measure glutamate
concentrations, which may underlie the perturbed MMN and fMRI signals in people with psychosis. We will
bring together behavioral and brain data with formal computational modeling that will allow us to estimate, from
each individual subject's data, the strength of their priors and prediction errors across a hierarchy of
representational richness from simple stimuli through more complex percepts, action choices, and beliefs.
 We propose four specific aims: (1) testing whether inappropriately strong top-down perceptual priors cause
hallucinations; (2) testing if delusions are caused by aberrant prediction error signaling; (3) examining whether
psychotic symptoms result from a failure to attribute outcomes to one's own actions appropriately; (4) and
assessing whether glutamate levels are related to predictive coding phenomena assayed in Aims 1-3. In a fifth
exploratory aim, we will examine whether predictive coding abnormalities change over course of illness.
 Our overall goal is to provide a computationally rigorous test of the predictive coding account of delusions
and hallucinations. Depending on the outcome, we will either discard the theory, or use it to design and test
treatment approaches more tailored to the specific, and this far unmet, needs of individuals with psychosis.

## Key facts

- **NIH application ID:** 9825560
- **Project number:** 5R01MH112887-03
- **Recipient organization:** UNIVERSITY OF MARYLAND BALTIMORE
- **Principal Investigator:** PHILIP Robert CORLETT
- **Activity code:** R01 (R01, R21, SBIR, etc.)
- **Funding institute:** NIH
- **Fiscal year:** 2020
- **Award amount:** $701,897
- **Award type:** 5
- **Project period:** 2017-12-01 → 2022-10-31

## Primary source

NIH RePORTER: https://reporter.nih.gov/project-details/9825560

## Citation

> US National Institutes of Health, RePORTER application 9825560, Predictive Coding as a Framework for Understanding Psychosis (5R01MH112887-03). Retrieved via AI Analytics 2026-05-23 from https://api.ai-analytics.org/grant/nih/9825560. Licensed CC0.

---

*[NIH grants dataset](/datasets/nih-grants) · CC0 1.0*