Errors of their (path)ways: hallucinations and delusions – Nature Mental Health

Hallucinations and delusions hold fascinating and potentially integral places on the continuum of human consciousness and experience. References to these phenomena are embedded in contemporary discourse, from descriptions of ‘hallucinations’ in artificial intelligence systems to the popular slang ‘delulu’, denoting a state of willful self‑deception, which entered the dictionary in 2025.

Reflective of such resonance, epidemiological data also support the fact that having such experiences is not particularly unusual: 6–15% of individuals in the general population report having at least one hallucination — a sensory perception or bodily sensation perceived as real in the absence of an external stimulus. That number increases considerably (38%) in the context of visual hallucinations, such as hypnagogic images, which are experienced in the transition period between wakefulness and sleeping.

The alternative text for this image may have been generated using AI.

Hallucinations are perceptual experiences, such as hearing voices or seeing objects that are not present. Delusions, in contrast, are beliefs that persist despite contradictory evidence. Endorsing irrational beliefs is not uncommon, particularly when such beliefs are reinforced culturally or socially, or even validated through technologies such as chatbots that prioritize agreement. Conceptualizing delusions as existing along a spectrum is therefore useful. What renders delusions clinically relevant is the strength of conviction and the extent to which they disrupt functioning and resist revision.

Within the clinical realm, hallucinations and delusions are conceptually distinct but frequently co-occur. In psychosis, they form part of a broader syndrome that includes positive symptoms such as disorganized thinking and abnormal motor behavior, alongside negative symptoms such as reduced motivation and social withdrawal. Together, these features contribute to an episodic state in which the ability to distinguish between internal reality and external reality is compromised. A reinforcing cycle often emerges: delusions can generate anxiety and lead to social withdrawal, and reduced social interaction limits opportunities for corrective feedback, which allows beliefs to persist and intensify.

In this issue, a Review by Chapman and coauthors considers that much of the work previously done on delusions focused on how the formation of these beliefs occurs, but surprisingly little research has been devoted to understanding why delusions become so entrenched. The authors discuss evidence indicating that delusions are far less stable during an episode and across episodes and that a substantial proportion of people with delusions may be open to receiving evidence to the contrary. Presenting a putative set of dynamic constructs spanning psychodynamic, social, cognitive and neurobiological processes, this Review provides a compelling case for their model of delusion fixity.

Hallucinations and delusions are most often associated with psychiatric disorders, including schizophrenia, bipolar disorder and depression. However, they also occur across a wide range of medical conditions, including autoimmune, endocrine, metabolic and neurological disorders, and in the context of medication use or withdrawal. Despite the heterogeneity of the possible underlying or precipitating conditions, the similarities in the manifestation of hallucinations and the development of delusions offer some intriguing insights into the possible shared mechanistic and computational mechanisms involved.

Understanding these mechanisms increasingly draws on predictive coding frameworks. In this view, the brain operates as an inference system that continuously generates predictions about incoming sensory input and updates those predictions on the basis of error signals. Dopamine is implicated in encoding the precision or salience assigned to these signals. When dopaminergic signaling becomes dysregulated, the system may assign excessive importance to irrelevant or ambiguous stimuli. Neutral events or internally generated thoughts may then be experienced as highly meaningful or threatening, which contributes to the formation of delusions.

Metacognitive processes can serve to further reinforce these effects. Individuals experiencing delusions often show overconfidence in false memories and incorrect judgments, alongside reduced confidence in accurate ones. In this way, memories associated with highly salient but erroneous interpretations may be repeatedly reactivated and strengthened through reconsolidation, which contributes to their persistence. Delusions may arise not only from abnormal perception but also from distortions in how confidence and meaning are assigned to experience.

The occurrence of hallucinations in individuals who are not experiencing psychosis, and the fact that delusional disorder can present in the absence of other psychotic features, underscores where there is still more work to do to clarify the risk factors for progression to psychosis and to identify appropriate and effective early intervention. Individuals who are at clinical high risk for psychosis tend to be younger (14–35 years of age), have a family history of psychosis, and have previously experienced attenuated or intermittent periods of positive symptoms. The phenotype of clinical high risk for psychosis is not determinant for psychosis, but it does confer an elevated risk for conversion, and about a quarter of those who meet criteria will develop a psychotic disorder within 2 years.

Substance use represents an important and modifiable contributor to this vulnerability. Both cannabis use and tobacco use are elevated in clinical high-risk populations and among individuals with psychosis, and both can induce paranoia or perceptual disturbances. In an Article in this issue, Ward and colleagues examine the co-use of cannabis and tobacco, a pattern that is increasingly common among young adults. Using longitudinal data, the authors show that more frequent co-use is associated with greater symptom severity and a nearly threefold increase in the risk of conversion to psychosis. These findings suggest a synergistic effect that potentially arises from shared influences on dopaminergic and endocannabinoid systems, which may sensitize neural circuits involved in salience attribution and predictive processing.

There is also a growing recognition of areas of overlap between psychosis and neurodegenerative disorders. Individuals with psychosis are at a substantially increased risk of developing dementia. Hallucinations are common in conditions such as Parkinson’s disease, in which up to 60% of individuals develop psychosis (specifically Parkinson’s disease psychosis) over the course of this illness. In this context, hallucinations are associated with cognitive decline and increased mortality, which highlights their clinical importance and suggests contributions from underlying brain pathology, as well as treatment side effects from dopaminergic medication.

Emerging work also emphasizes the importance of early and subtle phenomena. In Parkinson’s disease, minor hallucinations such as transient visual illusions or sensed presences may precede more-complex symptoms. In this issue, an Article by Stampacchia and colleagues reports on a connectome-based brain-fingerprinting approach to identifying patient-specific functional networks associated with these minor hallucinations. Their findings show that these experiences are linked to early cognitive alterations and can be distinguished from motor symptoms at the level of individualized network architecture. They also demonstrate associations between these network features and the cortical distribution of neurotransmitter systems linked to hallucinations.

Complementing that work, in another Article, Vignando and colleagues investigate sensory processing in Parkinson’s disease with visual hallucinations. Using dynamic causal modeling of electroencephalogram data, they show that individuals with hallucinations exhibit altered responses to prediction errors during visual processing. These deviations in effective connectivity indicate that hallucinations may arise from disruptions in how sensory input is processed and integrated with prior expectations.

These studies support a convergent perspective, yet from different vantage points. Although there are crucial distinctions between Parkinson’s disease psychosis and primary psychosis associated with psychiatric diagnoses, there are notable parallels that may emerge from altered sensory input, impaired predictive processing and cognitive mechanisms. Hallucinations may arise from aberrant sensory signals or mis-weighted prediction errors, while delusions may reflect the brain’s attempt to interpret and impose coherence on these anomalous experiences.

Ultimately, these lines of inquiry do more than explain symptoms. They elucidate the potential fallible computations and pathways that underlie perception, belief, how we make sense of the world around us — and when we do not.