2023/11/01 by Benjamin Cook, Juliet D. Griffin · 1 voice
Medicine · Neuroscience · Psychology · #Mental Health Research Topics #Psychology of Moral and Emotional Judgment #Psychosomatic Disorders and Their Treatments
paper · pdf · doi:10.1093/schbul/sbad124
openalex publication_date 2023/11/01 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/01
As my delusional system expanded and elaborated, it was as if I was not “thinking the delusion,” the delusion was “thinking me!” I was totally enslaved by the belief system. Chadwick1 Predictive processing (PP) provides a framework for understanding disordered cognition and experiences in terms of altered Bayesian inference. Petrovic and Sterzer2 re-examine its application to delusions (a project in which their own work has been central), highlighting an important paradox. If delusions emerge due to underweighting of priors relative to sensory evidence: why do they persist rather than dissolve in the face of contradictory evidence? Though Petrovic and Sterzer note that all established beliefs resist change, they view this as inadequately explaining delusional persistence. Their view is supported by Chadwick’s phenomenological description above, which illustrates that a delusion’s form crucially differs from that of an ordinary belief—implying an explanation of delusional persistence must go beyond invoking ordinary mechanisms of belief maintenance. For Petrovic and Sterzer, the solution lies in PP’s hierarchical structure. Delusional thinking emerges due to underweighted priors at lower levels, whose failure to precisely predict sensory inputs allows perceptual uncertainty to percolate up the hierarchy, where higher-level priors become overweighted to compensate for uncertainty that could not be explained away within lower-level perceptual circuits. As the authors nicely put it, this compensatory overweighting acts to “sculpt perceptual processing into conformity with delusions and foster their resistance to contradictory evidence”. For example, ambiguous visual inputs consistent with a friendly smile, interpreted in light of a patient’s higher-order belief that others are malevolent, may instead be perceived as a sinister leer—which experientially reinforces the delusional prior. A dissociation between hierarchical levels has been suggested,3,4 but Petrovic and Sterzer make a stronger empirical case for it, additionally proposing lateral orbitofrontal cortex (lOFC) as a key hub for higher-level priors’ overweighting. Though previous work has examined frontal cortical links to delusionality,5 their hypothesis is more precisely specified. To support it, they marshal evidence that lOFC is centrally involved in the effects of verbal instructions on subjective analgesic, gustatory, and olfactory experiences. Across these domains, the effect of instructions on perceptual experience was mediated by lOFC’s functional connectivity with the relevant lower-level affective and sensory circuits: indicating that lOFC dysfunction would alter how beliefs interact with perceptual information. Evidence for the dissociation between hierarchical levels is cited, from a visual perception task in which the effects of high-level and low-level priors were anti-correlated.6 The evidence presented that lOFC’s aberrant engagement contributes to persistence of delusional inferences is drawn largely from general population studies of self-reported non-clinical experiences and beliefs—an approach rooted in the continuum model of psychosis. As with any model, such studies carry certain interpretive pitfalls, but these are deftly circumvented by the authors’ careful analysis of the experiments reviewed. For example, Sulik et al7 demonstrated potential spurious associations between Peters’ Delusion Inventory (PDI) and certain classic schizophreniform “cognitive biases” (since the tasks assaying them are sensitive to “careless responding”, as are questionnaires like the PDI). Though cautionary, this observation does not invalidate the principal findings supporting Petrovic and Sterzer’s argument that lOFC mediates increased use of high-level beliefs among individuals with high PDI scores. Neither enhanced instruction effects on affective fear discrimination learning,8 nor the enhanced behavioral and neural signatures of instruction effects on motion perception,6 are easily explained in “deficit” terms like “careless responding”. This is a major strength of both paradigms, and of their interpretation here. How directly these findings support the reviewers’ solution to the “delusion paradox”, though, depends how well “delusion-related phenotypes” capture the key processes underpinning delusions as a clinical phenomenon. Non-clinical delusion-like ideation is sometimes related to “delusion-proneness”, which might be taken to imply those scoring higher on such measures are likelier to develop full-blown clinical delusions in future. These scales’ predictive validity with respect to psychotic illness is not yet established, and the more common construal of their usefulness is as modeling (rather than predicting) delusions proper. While non-clinical models of complex clinical phenomena are invaluable, we must ask how convincingly such scales capture the full experiential force of a delusion, by which the sufferer may be “totally enslaved”. The PDI elicits ratings of “conviction”, “distress”, and “preoccupation” pertaining to unusual beliefs. However, a gulf remains between the applicability of these terms to the experience of being immersed in the grip of a delusional conviction, as opposed to how they might relate to the experiences of a healthy person endorsing PDI items couched in non-literal terms (“Do you ever feel as if…” “…you are a very special or unusual person?”; “…some people are not what they seem to be?”). None of this negates the potential of such scales to illuminate how subtle perturbations in thought patterns relate to neurobiological and computational factors, which this review amply demonstrates. However, we highlight two further points of caution. First, the clinical data presented appears sometimes to undermine the predictions of the reviewers’ neurocomputational hypothesis. This discrepancy highlights general limitations in how reliably findings from non-clinical models can support confident inferences about analogous mechanisms within the clinical target system. Patients with schizophrenia demonstrated a reduced behavioral effect of instructed beliefs, alongside enhanced connectivity of lOFC with visual areas9: thus, the outcome measure of priors’ overweighting seems to diverge from its putative “neural basis” in these patients. There may also be some contradiction between enhanced instruction effects on affective learning in high-PDI (vs low-PDI) healthy subjects, and the opposite tendency in psychosis—which is associated with blunted fear discrimination learning.10 Of course, patient studies bring their own interpretative challenges, including medication and comorbidity confounds, so these inconsistencies do not necessarily undermine Petrovic’s and Sterzer’s argument. One intriguing suggestion the authors make (and one, indeed, that would obviate the above concerns) is that high scores on non-clinical delusional scales are of direct interest in themselves: that high-scoring healthy individuals are not the model but the target. Such a perspectival shift seems to be advocated in the authors’ recommendation that “future studies should focus on…why delusion-related phenotypes more readily acquire delusion-like ideas including conspiracy narratives outside the experimental setting”. Focusing on delusion-like ideation as an intrinsic object of psychological investigation would be interesting, and it is certainly true that conspiracy beliefs in the general population are attracting growing attention across the behavioral sciences. However, we would point out some important ways in which conspiracy beliefs are decidedly unlike delusional ones. Though conspiracy theories centrally involve inferences about the hidden objectives of powerful agents, the agendas attributed to alleged conspiracists are relatively impersonal. Persecutory, erotomanic, and referential delusions revolve around a similar thematic motif, but center the patient themselves as the direct target of malicious, amorous, or communicatory intentions. Conspiracy beliefs and delusions also differ in their form, notably in relation to evidence. Proponents of conspiracy theories generally cite evidence that is discernibly pertinent to establishing the conspiracy’s veracity, and drawn from publicly-verifiable external sources. Meanwhile, “Deluded patients often cite as evidence pieces of information that seem totally disconnected from the belief”11 and moreover are often sourced from private, subjective experience. Finally, conspiracy beliefs are to a significant extent socially shared within the relevant social-subcultural epistemic communities, and indeed distress associated with them is mitigated by their perceived credibility within the believer’s social network.12 Leaving this aside, we see great promise in developing Petrovic and Sterzer’s hypothesis in other models of delusions, with a view to evaluating it in the target system of clinical psychosis. The tasks reviewed may be translated to animal studies manipulating activity in homologs of lateral and medial OFC (to establish their precise computational involvements), and to neurological studies of patients with circumscribed lesions to orbitofrontal subregions (to clarify necessity and sufficiency of their causal contributions). Compensatory alterations at different timescales and levels may interactively explain temporal shifts in the form of delusions’ “evidence-resistance”: from initial “living” flexibility, to “inert” rigidity—which in chronic schizophrenia may be characterized by a “double orientation” wherein dogged conviction and detached ambivalence about the delusion paradoxically co-exist. As PP is both dynamic and hierarchical, a consideration of changes across time as well as levels may prove useful in confronting this paradox.