Predictive Coding and Unusual Experiences

Persistent hallucination and delusions occur in 30-50% of patients with psychosis, despite adequate antipsychotic treatment. These patients experience significant distress and impaired daily functioning. In this study, we are exploring the emergence and maintenance of these symptoms. We will use predictive coding to account for overlapping and independent features of positive symptoms. Predictive coding is a theoretical framework that addresses how the brain actively predicts likely sensory input as people engage with their everyday environment. Thus, what we see and believe represents a balance between our prior expectations and incoming sensory information. If this balance is disrupted, people may perceive stimuli that are not actually present in the environment (a hallucination) or may form inaccurate beliefs about the causes of events in the environment (a  delusion). The goal of this study is to better understand how problems balancing prior beliefs and new information may result in a range of unusual experiences and beliefs. We plan to examine this question both in people who have these symptoms and seek psychiatric care to help them cope with these experiences as well as people who have similar experiences but do not seek care.

We propose that hierarchical predictive coding (PC) offers an explanatory framework relating constructs of perception, expectancy, and agency, from the NIMH Research Domain Criteria (RDoC) initiative, to hallucinations and delusions. In PC, the brain attempts to predict future events and respond adaptively to them by minimizing errors and uncertainty in belief, action, and perception. The framework is embedded in the hierarchical neuroanatomy of the brain such that predictions or priors are specified top-down, while new inputs are introduced bottom-up. Any mismatch between predictions and new inputs results in a prediction error (PE) that may be used to update future predictions. We theorize that hallucinations and delusions originate, and are maintained, by alterations in prediction error minimization during perception, action, and belief. The glutamate system may play a critical role in PE processing. We will test whether alterations in glutamate signaling underlie perturbations in PE that give rise to hallucinations and delusions. While other investigators have used predictive coding to conceptualize psychosis, we propose a broader theoretical and empirical approach, whereby specific abnormalities within the PC hierarchy differentially relate to the origin and maintenance of psychotic symptoms over time. We will use computational modeling of task behavior to map brain and behavior to specific model parameters including PE, learning rate, and priors for each cognitive task, and we will test whether particular model parameters correlate with positive symptom severity.

Across two sites, Yale University and Maryland Psychiatric Research Center (MPRC), we will gather a suite of RDoC-relevant PC measures in a large sample with a broad range of psychosis severity and duration:

• 50 psychosis patients within 3 years of illness onset,
• 50 chronically-psychotic patients, and
• 100 age-matched controls, including

• 50 with attenuated psychosis-like symptoms (APS) and 50 without.

We will use functional magnetic resonance imaging (fMRI) during tasks of perception, belief formation, and action; electroencephalography (EEG) to measure mismatch negativity (MMN), an electrophysiological response to unexpected perceptual stimuli); and proton magnetic resonance spectroscopy (MRS) to measure glutamate. MRS glutamate levels correlate with MMN9 and perturbed fMRI signals in people with psychosis. However, the associations between glutamate, behavior, brain responses, symptoms have yet to be delineated. We will use computational modeling to describe task performance and ascertain whether and how particular model parameters will relate to specific behaviors, brain responses, and symptoms. 

Inclusion Criteria

 General inclusion criteria:

1. Participants 18-55 years of age.
2. Fluent English-speakers.

Group 1 inclusion criteria:

1. Participants with schizophrenia or schizoaffective disorder in the first 3 years of their illness, identified by the clinical interview.

Group 2 inclusion criteria:

1. Participant with schizophrenia or schizoaffective who have been ill longer than 3 years, identified by the clinical interview.

Group 3 inclusion criteria:

1. Controls with similar demographic features as Group 1 and Group 2.

Group 4 inclusion criteria:

1. Participants with schizophrenia or schizoaffective who have been ill longer than 3 years, identified by the clinical interview.

Exclusion Criteria

 General exclusion criteria:

1. Participants who have a contradiction for MRI scanning, such as metallic implants of any kind, pacemakers, history of accidents involving metal, or claustrophobia.
2. Participants with substance use disorder within the last 12 months or abuse within the past month.
3. Participants with a history of intellectual disability according to DSM criteria.
4. Participants with medical or neurological conditions likely to impact cognitive function, such as serious traumatic brain injury (prolonged loss of consciousness), epilepsy, or multiple sclerosis.
5. Participants who are pregnant.

Group 3 exclusion criteria:

1. Participants who have previously been diagnosed with a clinical psychiatric disorder.
2. Participants who have been treated with antipsychotic medication.
3. Participants who have been treated with antidepressants in the past 6 months.
4. Participants who have a first-degree relative with psychosis.

Group 4 exclusion criteria:

1. Participants who have been previously diagnosed with a psychotic disorder.
2. Participants who have been treated with antipsychotic medication.