CoChat
Sign Up
Cited by 133Schizophrenia research and treatmentMental Health and PsychiatryEmbodied and Extended Cognition

Predictive processing, source monitoring, and psychosis

Open access
Griffin, Juliet D., Fletcher, Paul C.·2017·Annual Review of Clinical Psychology
DOI 10.1146/annurev-clinpsy-032816-045145PMC PMC5424073
Open this paper in CoChat — free accountOpen PDF

Abstract

A comprehensive understanding of psychosis requires models that link multiple levels of explanation: the neurobiological, the cognitive, the subjective, and the social. Until we can bridge several explanatory gaps, it is difficult to explain how neurobiological perturbations can manifest in bizarre beliefs or hallucinations, or how trauma or social adversity can perturb lower-level brain processes. We propose that the predictive processing framework has much to offer in this respect. We show how this framework may underpin and complement source monitoring theories of delusions and hallucinations and how, when considered in terms of a dynamic and hierarchical system, it may provide a compelling model of several key clinical features of psychosis. We see little conflict between source monitoring theories and predictive coding. The former act as a higher-level description of a set of capacities, and the latter aims to provide a deeper account of how these and other capacities may emerge.

AI snapshot

Written from the paper’s full text.

TL;DR

This conceptual review argues that predictive processing (PP) can complement and deepen source monitoring accounts of psychosis by explaining source-monitoring deficits as emergent from disturbed prediction-error weighting and inference. It proposes that a hierarchical, dynamic PP framework can account for key clinical features including delusion persistence, social content, negative affect, and bizarreness.

Key findings

  • PP and source monitoring are not competing accounts: source monitoring is a higher-level description of capacities, while PP aims to provide a deeper computational account of how those capacities emerge.
  • Source monitoring attributions are framed as abductive inferences about the most likely cause of an event under noise, ambiguity, and uncertainty; prediction error is proposed as a crucial signal for internal source monitoring.
  • A hierarchical and dynamic PP system may explain the evolution of delusions from early flexibility to later fixity, as well as why delusions often have social content, negative emotional tone, and bizarre content.
  • No primary numerical results are reported; the paper is a theoretical synthesis rather than an empirical study.

Methods

  • Design: conceptual/theoretical review and model-building paper, not an empirical study.
  • Data or sample: no new sample or dataset is reported; the argument draws on prior literature on predictive processing, source monitoring, schizophrenia, delusions, hallucinations, and related cognitive neuroscience.
  • Main analysis: narrative synthesis and theoretical extension, linking PP concepts such as prediction error, precision, hierarchical inference, and adaptive learning to source monitoring and psychotic symptoms.

Limitations

  • The paper states that PP applications have been largely descriptive, have yet to be fully and formally tested by novel predictions, and that internal inconsistencies remain regarding neurobiological implementations of PP.
  • The paper repeatedly describes parts of its account as speculative, including proposed links between prediction-error dysregulation, source monitoring, agency attribution, and delusion formation.
  • Evident from the text: there is no empirical test, systematic review, meta-analysis, or new quantitative data; the claims are conceptual and require future formal testing.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.1146/annurev-clinpsy-032816-045145

Cites

157 of 157 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Claims the paper makes

0 of 117 cited sources checked · 0 supported · 0 flagged

  • Premisecitation

    People with chronic schizophrenia show impairments in the flexible generalisation of correctly learned and accurately remembered associations to novel contexts.

    “People with chronic schizophrenia show impairments in the flexible generalisation of (correctly learned and accurately remembered) associations to novel contexts (Shohamy et al 2010)”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites Shohamy et al 2010

      Not checked in the free run.

  • Interpretationcitation

    Psychosis could be characterised by fundamental disturbances in adapting learning responses to both prediction error variability and environmental volatility.

    “recent theoretical work suggests that psychosis could be characterised by fundamental disturbances in adapting learning responses to both PE variability and environmental volatility (Adams et al 2013).”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites Adams et al 2013

      Not checked in the free run.

  • Premisecitation

    Unmedicated people in their very first episode of psychosis show an increased tendency to update in response to prediction error.

    “Schlagenhauf et al (2010) demonstrated an increased tendency to update in response to prediction error in unmedicated people in their very first episode of psychosis.”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites Schlagenhauf et al (2010)

      Not checked in the free run.

  • Interpretationreasoning

    The aberrant and apparently irrational perceptions and beliefs that characterise psychosis may fruitfully be considered within the predictive processing framework.

    “we argue that the aberrant and apparently irrational perceptions and beliefs that characterise psychosis may fruitfully be considered within this framework.”
    Quote in the paperNot a substantial claim
  • Premisecitation

    Context-dependent adaptive coding of rewards is disrupted in schizophrenia.

    “there is recent evidence that such context-dependent adaptive coding of rewards is disrupted in schizophrenia (Kirschner et al 2016).”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites Kirschner et al 2016

      Not checked in the free run.

  • May combine two claims. 33
  • Classified as probably not a substantial claim. 53

  • May combine two claims.

    Premisecitation

    The successful brain must model its world and generate accurate predictions about upcoming events to marshal preparatory responses.

    “The successful brain must model its world (Conant & Ashby 1970). It needs to be able to generate accurate predictions about upcoming events, so that it can marshal preparatory responses.”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites Conant & Ashby 1970

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S2 (checkable), S4 (based).

    Interpretationnone given

    Source monitoring and the attribution of agency fall naturally within the ambit of predictive processing and prediction error minimisation.

    “source monitoring and the attribution of agency (itself an important consideration in psychosis) fall naturally within the ambit of PP and prediction error minimisation.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Premisecitation

    Culbreth et al (2016) demonstrated a similar increased tendency to update in response to prediction error in schizophrenia, which was mediated entirely by impaired activation in cingulate, parietal and frontal regions.

    “Consistent with this, Culbreth et al (2016) demonstrated a similar tendency in schizophrenia, which they attributed to ‘unstable’ representations of value (but which they might equally have been attributed to ‘noisy’ or ‘imprecise’ prior representations of value) and which was mediated entirely by impaired activation in cingulate, parietal and frontal regions.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites Culbreth et al (2016)

      Not checked in the free run.

  • May combine two claims.

    Premisecitation

    People with chronic schizophrenia classically have a perseverative, inflexible response pattern on probabilistic reversal learning tasks, consistent with a tendency to underutilise prediction errors.

    “they classically have a perseverative, inflexible response pattern (consistent with a tendency to underutilise prediction errors) on probabilistic reversal learning (PRL) tasks (Elliott et al 1995, Kerns & Berenbaum 2002, Pantellis 1997 – although see also Waltz et al 2013).”
    Quote in the paperSubstantial claim0 of 3 sources checked
    • cites Elliott et al 1995

      Not checked in the free run.

    • cites Kerns & Berenbaum 2002

      Not checked in the free run.

    • cites Pantellis 1997

      Not checked in the free run.

  • Classified as probably not a substantial claim. May combine two claims.

    Interpretationnone given

    Source monitoring acts as a higher-level description of a set of capacities, while predictive coding aims to provide a deeper account of how these and other capacities may emerge.

    “The former acts as a higher level description of a set of capacities, and the latter aims to provide a deeper account of how these and other capacities may emerge.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S2 (checkable), S4 (based). May combine two claims.

    Interpretationnone given

    Inferences must be flexible in the face of change but robust in the face of noise.

    “Put simply, inferences must be flexible in the face of change but robust in the face of noise.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S2 (checkable), S4 (based).

    Interpretationreasoning

    A persistent underlying disturbance could, as a result of ongoing adaptive processes, manifest in different ways across time.

    “This is all speculative, but the key point is that a persistent underlying disturbance could, as a result of ongoing adaptive processes, manifest in different ways across time.”
    Quote in the paperNot a substantial claim

  • Classified as probably not a substantial claim. May combine two claims.

    Interpretationreasoning

    A small but persistent disruption in the ability to optimally weight each prediction error could create sub-optimal adaptation, leading to initially high levels of prediction-error-driven updating followed by later downregulation of learning.

    “One can envisage how a small but persistent disruption in the ability to optimally weight each PE could create sub-optimal adaptation, leading to evolving consequences such as initially high levels of PE-driven updating being followed by later downregulation of learning.”
    Quote in the paperSubstantial claim

  • Classified as probably not a substantial claim.

    Premisecitation

    The proactive control system's ability to facilitate flexible, rapid updating in response to relevant information while preventing interference from irrelevant information depends on temporally precise dopaminergic neuromodulation of this network.

    “This proactive control system’s ability to facilitate flexible, rapid updating in response to relevant information, while preventing interference from irrelevant information, has been suggested to depend on temporally precise dopaminergic neuromodulation of this network (Braver 2012).”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites Braver 2012

      Not checked in the free run.

  • May combine two claims.

    Premisecitation

    The brain is foremost attempting to model the world and updates inferences by iteratively reducing prediction errors.

    “the brain is foremost attempting to model the world (Conant & Ashby 1970) and that it does so by updating inferences through iteratively reducing ‘prediction errors’”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites Conant & Ashby 1970

      Not checked in the free run.

  • May combine two claims.

    Premisecitation

    Dopamine signals the magnitude of reward prediction errors and has been suggested to be critical in signalling the precision or epistemic value of incoming information, setting the gain on cortical prediction error signals.

    “dopamine, in addition to signalling the magnitude of reward PEs (Schultz et al 1997) has been suggested to be critical in signalling the precision (Fiorillo et al 2003), or epistemic value (Schwartenbeck et al 2016) of incoming information, setting the gain on cortical PE signals.”
    Quote in the paperSubstantial claim0 of 3 sources checked
    • cites Schultz et al 1997

      Not checked in the free run.

    • cites Fiorillo et al 2003

      Not checked in the free run.

    • cites Schwartenbeck et al 2016

      Not checked in the free run.

  • Classified as probably not a substantial claim.

    Premisecitation

    The persistent experience of a noisy, unreliable prediction error signal should cause an information-processing system to adapt by down-regulating updating in response to a given magnitude of prediction error signal.

    “Preuschoff and Bossaerts (2007) consider, in computational terms, how the persistent experience of a noisy, unreliable (and hence informationally-devalued) PE signal should cause an information-processing system to adapt by down-regulating the updating that occurs in response to a given magnitude of PE signal.”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites Preuschoff and Bossaerts (2007)

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S2 (checkable).

    Interpretationreasoning

    The predictive processing framework has much to offer for understanding psychosis by linking multiple levels of explanation.

    “We propose that the predictive processing framework has much to offer in this respect.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S2 (checkable).

    Interpretationreasoning

    A hierarchical and dynamic perspective on prediction error signalling in predictive processing has much to offer for understanding psychosis.

    “arguing that a hierarchical and dynamic perspective has much to offer.”
    Quote in the paperNot a substantial claim

  • Classified as probably not a substantial claim.

    Premisecitation

    Theoretical models of associative learning invoke simple error minimisation: the model is updated as a function of its failure to predict accurately.

    “Theoretical models of associative learning invoke simple error minimisation to achieve this: the model is updated as a function of its failure to predict accurately (Rescorla & Wagner 1972).”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites Rescorla & Wagner 1972

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S2 (checkable).

    Premisecitation

    There is a need for a formal computational account that could bridge the explanatory gaps between brain, cognition and symptoms.

    “There is a need for a formal computational account that could bridge the explanatory gaps between brain, cognition and symptoms (Frith 2015).”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites Frith 2015

      Not checked in the free run.

  • Classified as probably not a substantial claim. May combine two claims.

    Premisecitation

    A failure to correctly identify the origin of a movement, thought or experience may capture the essential problem at the heart of other features of schizophrenia beyond passivity phenomena, so the source monitoring account could offer a truly fundamental description of the condition.

    “since it has been suggested that a failure to correctly identify the origin of a movement, thought or experience may capture the essential problem that lies at the heart of other features of schizophrenia beyond those formally described as passivity phenomena, the source monitoring account could offer a truly fundamental description of the condition (Nelson et al 2014b).”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites Nelson et al 2014b

      Not checked in the free run.

  • May combine two claims.

    Interpretationreasoning

    The predictive processing framework points towards a bridge between brain, cognition and symptoms; it does not replace source monitoring ideas but offers them a deeper, computational foundation.

    “The PP framework points towards this bridge. It does not replace source monitoring ideas but rather offers them a deeper, computational foundation.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S2 (checkable).

    Interpretationreasoning

    These are precisely the sorts of inferences with which predictive processing is concerned.

    “These are precisely the sorts of inferences with which PP is concerned.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S2 (checkable). May combine two claims.

    Premisecitation

    Source monitoring is a conclusion or inference to best account for the overall data; it is an abductive inference from the data back to the most likely cause.

    “Ultimately, we see source monitoring as a conclusion or inference to best account for the overall data. That is, it is an abductive inference: an inference from the data back to the most likely cause (see Coltheart 2010).”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites Coltheart 2010

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S2 (checkable), S4 (based).

    Premisenone given

    Source monitoring attributions can be thought of as inferences about what or who was the most likely cause of an event, made under conditions of noise, ambiguity and uncertainty.

    “we can think of source monitoring attributions as inferences about what (or who) was the most likely cause of an event, made under conditions of noise, ambiguity and uncertainty.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing), S4 (based).

    Interpretationnone given

    Source information may be critical insofar as it could provide a marker for the reliability of information.

    “this source information may be critical especially insofar as it could provide a marker for the reliability of this information.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S2 (checkable).

    Premisecitation

    Source monitoring is a high level description of a cognitive capability, encompassing other processes, rather than being a specific cognitive process itself.

    “Source monitoring is a high level description of a cognitive capability, one that encompasses other processes, rather than being a specific cognitive process itself.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S2 (checkable).

    Interpretationnone given

    Source monitoring failure provides a compelling descriptive account of a core problem in psychosis: a failure to discriminate internal from external, reality from imagination.

    “In short, source monitoring failure provides a compelling descriptive account of what is intuitively a core problem in psychosis: a failure to discriminate internal from external, reality from imagination.”
    Quote in the paperNot a substantial claim

  • Classified as probably not a substantial claim.

    Premisecitation

    According to the source monitoring account, delusions and hallucinations result from incorrect attribution of imagined or self-generated representations to external causal sources.

    “According to the source monitoring account, delusions and hallucinations result from incorrect attribution of imagined or self-generated representations to external causal sources.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S2 (checkable), S4 (based).

    Interpretationreasoning

    Even a subtle impairment in source monitoring ability could profoundly change one's world model.

    “Clearly, it is critical to be able to do this, and even a subtle impairment in source monitoring ability could profoundly change one’s world model.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Interpretationreasoning

    The predictive processing account can be naturally extended to encompass source attributions more broadly, offering a computational explanation for passivity experiences involving attribution to an external agent.

    “the PP account can go further. It can be naturally extended to encompass source attributions more broadly, and in so doing, it has the potential to offer a computational explanation for a hitherto poorly understood characteristic of passivitiy experiences in schizophrenia: they involve not just the failure to recognise oneself as the agent of an action, but also its attribution to an external ag”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Premisecitation

    Sensory events for which I am the source are predictable in advance, while sensory events arising from external sources are not.

    “sensory events for which I am the source are predictable in advance (Frith et al 2000, Blakemore et al 2002, Wolpert & Flanagan 2001) while sensory events arising from external sources are not.”
    Quote in the paperSubstantial claim0 of 3 sources checked
    • cites Frith et al 2000

      Not checked in the free run.

    • cites Blakemore et al 2002

      Not checked in the free run.

    • cites Wolpert & Flanagan 2001

      Not checked in the free run.

  • May combine two claims.

    Premisecitation

    Events caused by agentic sources usually have features that distinguish them from those caused by external physical forces: they start and stop, speed up and slow down, take non-linear trajectories; they are relatively unpredictable without recourse to a higher-level model possibly involving goals and intentions.

    “Events caused by agentic sources usually have features that distinguish them from those caused by external physical forces: they start and stop, they speed up and slow down, they take non-linear trajectories. They are, in short, relatively unpredictable without recourse to some higher-level model possibly involving goals and intentions.”
    Quote in the paperSubstantial claim0 of 2 sources checked
    • cites Biro et al 2007

      Not checked in the free run.

    • cites Premack 1999

      Not checked in the free run.

  • May combine two claims.

    Interpretationcitation

    In psychosis, as suggested for autism, loss of capacity or confidence in the social domain could lead to highly selective sampling of information, avoidance of especially uncertain sources of evidence, and a tendency to cut oneself off from information and beliefs passed on by other people.

    “In psychosis, just as has been suggested for autism (Elsabbagh and Johnson 2016), the sometimes- profound loss of capacity (or confidence) in the social domain could lead to highly selective sampling of information, avoidance of especially uncertain sources of evidence, and a tendency to cut oneself off from a very valuable source of uncertainty resolution: the information and beliefs passed on to”
    Quote in the paperSubstantial claim0 of 2 sources checked
    • cites Elsabbagh and Johnson 2016

      Not checked in the free run.

    • cites Ratcliffe 2015

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S4 (based).

    Interpretationnone given

    Source monitoring underpins the kind of context-dependent learning necessary to deal with complex, higher-order worldly contingencies, and with cues that have differential reliability and predictive power.

    “It underpins the kind of context-dependent learning necessary to deal with complex, higher-order worldly contingencies, and with cues that have differential reliability and predictive power.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S4 (based).

    Premisenone given

    A notion central to predictive processing frameworks is that sensory evidence has the capacity to drive inference only insofar as it has sufficient precision or reliability to usefully update inference in the context of existing prior expectations.

    “We suggest this because a notion central to PP frameworks is that sensory evidence has the capacity to drive inference only insofar as it has sufficient precision or reliability to usefully update inference in the context of existing prior expectations, and”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Interpretationreasoning

    Information about the source of an event may be useful in guiding decision-making when the event is later recalled, and may be useful in ways that are not always predictable in advance; hence source memory is important for decision flexibility.

    “Even if currently irrelevant for action choice, information about the source of an event may be useful in guiding decision-making when the event is later recalled – and importantly, it may be useful in ways that are not always predictable in advance: hence the importance of source memory for decision flexibility, in particular.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Interpretationreasoning

    The abductive inference to the best explanation-the one that minimises prediction error at all levels of the hierarchy-entails not merely that the movement was caused by something other than myself, but also that it was caused by an agent.

    “The abductive inference to the best explanation – that is, the one that minimises prediction error at all levels of the hierarchy – therefore entails not merely that the movement was caused by something other than myself, but also that it was caused by an agent”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S2 (checkable), S4 (based). May combine two claims.

    Interpretationreasoning

    The predictive processing framework provides a deeper understanding of source monitoring and offers a compelling account of how a hypothesised impairment in processing prediction errors could lead to agency disturbances and, via subtle changes in how sources of evidence are weighted, a profound change in how a person models and interprets their world.

    “We suggest that the PP framework provides a deeper understanding of source monitoring and offers a compelling account of how the hypothesised impairment in the processing of prediction errors could lead to agency disturbances and, via subtle changes in how sources of evidence are weighted, a profound change in how a person models and interprets their world.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S2 (checkable), S4 (based). May combine two claims.

    Interpretationreasoning

    Source monitoring itself has an extremely important and far-reaching role in allowing us to make sense of the world and to ensure that we update our model only when necessary.

    “This arises, we argue, because source monitoring itself has an extremely important and far-reaching role to play in allowing us to make sense of the world and to ensure that we update our model only when necessary.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Interpretationcitation

    Reduced discriminability between social sources could account for the peculiar social content of delusions and for their social form.

    “This could account not only for the peculiar social content of delusions - widespread conspiracies involving a growing list of suspicious people - but also for their social form: delusions involve a kind of ‘Global social estrangement…[and] a failure to treat others as reliable sources of information’ (Ratcliffe 2015).”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites Ratcliffe 2015

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S4 (based).

    Interpretationnone given

    Reduced discriminability in prediction error signalling could lead to a consistent sense of unease or surprise.

    “Just as reduced discriminability in PE signalling could lead to a consistent sense of unease or surprise, so too could reduced discriminability between social sources make everything (and everyone) seem uniformly unreliable - even suspicious.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S4 (based). May combine two claims.

    Interpretationnone given

    The ability to identify the causal sources of current inputs is a key concern of predictive processing, and may be crucial for flexible learning and behaviour.

    “The ability to identify the causal sources of current inputs is a key concern of PP, and may be crucial for flexible learning and behaviour.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S4 (based).

    Interpretationnone given

    An accurate representation of the source of some evidence will very often be an important clue to estimating its reliability.

    “and because an accurate representation of the source of some evidence will very often be an important clue to estimating its reliability.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Interpretationreasoning

    Prediction error is useful not only in updating one's prior belief about the reliability of the information itself but also in updating a meta-level prior as to the general reliability of its source.

    “PE is useful not only in updating one’s prior belief about the reliability of the information itself (‘how reliably does this signal produce the outcome?’) but also in updating a meta-level prior as to the general reliability of its source (‘how reliable are signals from that source?).”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Premisecitation

    A key point is not just what the evidence is, but the optimising meta-level judgments about how reliable that evidence is expected to be and, by extension, the degree to which it should be weighted.

    “and a key point is not just what the evidence is, but the optimising meta-level judgments about how reliable that evidence is expected to be (Mathys 2011) and, by extension, the degree to which it should be weighted.”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites Mathys 2011

      Not checked in the free run.

  • May combine two claims.

    Interpretationcitation

    Retrospectively inferring 'who said what' is likely to be a highly challenging task because a great deal of potentially-identifying perceptual information in the speech signal is filtered out during encoding and there may be a very long delay before a piece of information is re-evoked and updated.

    “Retrospectively inferring ‘who said what’ is likely to be a highly challenging task: a great deal of potentially-identifying perceptual information in the speech signal is filtered out during encoding in order to prioritise semantic processing (Johnson 2005), and there may be a very long delay before a piece of information is re-evoked and updated.”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites Johnson 2005

      Not checked in the free run.

  • May combine two claims.

    Interpretationreasoning

    Predictive processing offers a framework for understanding source monitoring capabilities and a parsimonious explanation for how specific perturbations generate characteristic oddities of both source and agency attribution.

    “In short, PP offers a framework for understanding source monitoring capabilities as well as a parsimonious explanation for how specific perturbations might generate characteristic oddities of both source and agency attribution.”
    Quote in the paperNot a substantial claim

  • Classified as probably not a substantial claim. May combine two claims.

    Premisecitation

    Elsabbagh and Johnson (2016) suggested that a 'domain general and diffuse' abnormality in synaptic connectivity and transmission has partial and mild effects on multiple systems but these effects are only clearly revealed when more complex multisensory integration and high temporal resolution processing becomes important in the increasingly complex social environment.

    “Elsabbagh and Johnson (2016) suggested that a ‘domain general and diffuse’ abnormality in synaptic connectivity and transmission has ‘partial and mild effects on multiple systems…but [these] effects are only clearly revealed when more complex multisensory integration and high temporal resolution processing becomes important…[in] the increasingly complex social environment’.”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites Elsabbagh and Johnson (2016)

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S2 (checkable).

    Interpretationreasoning

    The predictive processing explanation of a reduction in associative memory could have further explanatory power, accounting for difficulty in using source and context information to allow flexibility of thinking and decision-making.

    “Given the importance of associative memory (for source and spatiotemporal context) in facilitating flexible behaviour, we advance the possibility that the PP explanation of a reduction in this capacity could have further explanatory power, accounting for difficulty in using source and context information to allow flexibility of thinking and decision-making.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    McKay (2012) takes this further in relation to a Bayesian account of Capgras delusion proposed by Coltheart et al (2010).

    “McKay (2012) takes this further in relation to a Bayesian account of Capgras delusion (the belief that a loved one has been replaced by an imposter) proposed by Coltheart et al (2010).”
    Quote in the paperNot a substantial claim0 of 2 sources checked
    • cites McKay (2012)

      Not checked in the free run.

    • cites Coltheart et al (2010)

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S1 (asserted).

    Premisecitation

    From a Bayesian perspective, the imposter hypothesis has such a low prior probability that it is hard to understand how it could be selected as the most likely cause of the experience.

    “From a Bayesian perspective, the imposter hypothesis has such a low prior probability that it is hard to understand how it could be selected as the most likely cause of the experience.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites McKay 2012

      Not checked in the free run.

  • May combine two claims.

    Interpretationreasoning

    Mental images whose content is for some reason poorly predicted by the current cognitive context will be experienced as more vivid, and thus particularly vulnerable to external misattribution.

    “Extending the analogy with the agency case, mentalmages whose content is for some reason poorly predicted by the current cognitive context will be experienced as more vivid, and thus particularly vulnerable to external misattribution.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Interpretationcitation

    The precision of self-generated images may be 'dampened down' because they contain nothing newsworthy, and their consequently reduced salience or vividness will be a powerful cue that they are indeed self-generated rather than externally caused.

    “By analogy with self-generated movement, the precision of self-generated images may be ‘dampened down’ because they contain nothing newsworthy -their content is already highly predictable given the system’s antecedent state – and their consequently reduced salience or vividness will be a powerful cue that they are indeed self-generated rather than externally caused (Frith 2005).”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites Frith 2005

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S2 (checkable).

    Interpretationreasoning

    The predictive processing account has the potential to offer a unified explanation for a remarkably wide array of the disorder's symptoms, features and risk factors.

    “This is admittedly speculative, but it highlights the potential of the PP account to offer a unified explanation for a remarkably wide array of the disorder’s symptoms, features and risk factors.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    The idea that hallucinations result from vivid mental imagery being misattributed to an external cause is a long-standing one.

    “The idea that hallucinations result from vivid mental imagery being misattributed to an external cause is a long-standing one (Mintz & Alpert 1972), and indeed, increased vividness of imagery does seem to be a trait-like feature of schizophrenia (Sack et al 2005, Oertel et al 2009).”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites Mintz & Alpert 1972

      Not checked in the free run.

  • May combine two claims.

    Interpretationcitation

    Highly schizotypal individuals are particularly prone to the self-reinforcing effect of high-level knowledge on perceptual processing, which perhaps contributes to their elevated risk of psychotic disorder.

    “Indeed, Schmack and colleagues (2013) demonstrated that the effect of high-level knowledge on perceptual processing, in healthy subjects, increases with their tendency towards delusional ideation – suggesting that highly schizotypal individuals are particularly prone to this self-reinforcing effect, which perhaps contributes to their elevated risk of psychotic disorder.”
    Quote in the paperSubstantial claim

  • Classified as probably not a substantial claim.

    Premisecitation

    Fonagy and Allison (2014) have offered a mechanistic account for how childhood trauma could lead to an inability to appropriately assign epistemic trust.

    “Moreover, in relation to the discussion above about the social nature of delusions, Fonagy and Allison (2014) have offered a mechanistic account for how childhood trauma could lead to an inability to appropriately assign epistemic trust.”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites Fonagy and Allison (2014)

      Not checked in the free run.

  • Classified as probably not a substantial claim.

    Premisecitation

    The idea that an altered physiological response to a familiar person confers a strange sense of unfamiliarity, in the context of preserved conscious recognition of the person, has been argued to provide the germ of the ultimate belief that the person has been replaced by an impostor.

    “The idea that an altered physiological response to a familiar person confers a strange sense of unfamiliarity, in the context of preserved conscious recognition of the person, has been argued to provide the germ of the ultimate belief that the person has been replaced by an impostor (Ellis et al 1997).”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites Ellis et al 1997

      Not checked in the free run.

  • Classified as probably not a substantial claim. May combine two claims.

    Interpretationcitation

    People who suffered neglect and abuse may have formed a persisting meta-level prior that 'who said it' makes little difference: social source is simply not a very useful guide to epistemic reliability, and therefore not worth carefully monitoring.

    “Since neglect and abuse prevent the formation of specific attachment relations that differentially pick out particular adults who can be relied upon as sources of valuable, generalizable information, people who suffered neglect and abuse may have thereby formed a persisting meta-level prior that ‘who said it’ makes little difference: in other words, that social source is simply not a very useful”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites Fonagy and Allison (2014)

      Not checked in the free run.

  • Classified as probably not a substantial claim.

    Premisecitation

    Coltheart (2007) makes a powerful case that a separate factor, in addition to that which generates the strange experiences, must be invoked to account for why such an unlikely belief is not rejected.

    “This objection has been eloquently outlined by Coltheart (2007) who make a powerful case that a separate factor, in addition to that which generates the strange experiences,must be invoked to account for why such an unlikely belief is not rejected.”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites Coltheart (2007)

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing).

    Interpretationreasoning

    It would be misguided to compare directly the predictive processing and source monitoring accounts, because they are pitched at different levels.

    “As above, it would be misguided to compare directly the predictive processing and source monitoring accounts, because they are pitched at different levels.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Premisereasoning

    Although there are reports of "autochthonous" delusions which appear to arise suddenly and fully-formed, the more characteristic pattern is that delusion formation is preceded by a prodromal period of 'delusional mood', involving a vague sense that the world has subtly changed somehow to become more sinister or foreboding.

    “Consistent with this overall picture, although there are reports of “autochthonous” delusions which appear to arise suddenly and fully-formed, the more characteristic pattern is that delusion formation is preceded by a prodromal period of ‘delusional mood’, involving a vague sense that the world has subtly changed somehow to become more sinister or foreboding.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    The temporally extended predictive processing account relates to a cognitive model proposed by Fleminger (1992), who refers to 'a morbid cycle of misperception and delusion reinforcing one another' occurring iteratively across time.

    “Note that the temporally extended PP account we have adumbrated in this section relates to a cognitive model proposed by Fleminger (1992), who refers to ‘a morbid cycle of misperception and delusion reinforcing one another’ occurring iteratively across time.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites Fleminger (1992)

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S2 (checkable).

    Interpretationreasoning

    Different theoretical accounts of psychosis can often seem to be in competition or conflict, when in fact they may show themselves to be complementary when we identify the particular level(s) at which their claims are intended to apply, and the particular characteristic(s) of symptoms with which they concern themselves.

    “Different accounts can often seem to be in competition or conflict, when in fact they may show themselves to be complementary when we identify the particular level(s) at which their claims are intended to apply, and the particular characteristic(s) of symptoms with which they concern themselves.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S2 (checkable).

    Interpretationnone given

    Without intervening levels of understanding, correlative links between brain activity and source monitoring capacities will remain frustratingly vague.

    “The brain basis of source monitoring has been explored (Mitchell & Johnson 2009) but, without intervening levels of understanding, correlative links between brain activity and source monitoring capacities will remain frustratingly vague.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S2 (checkable), S4 (based).

    Interpretationnone given

    The key components of a hierarchical and dynamic predictive processing framework - prediction error, reliability, uncertainty, adaptivity - readily translate to the subjective experiences that are key to source monitoring ideas: self versus other; real versus imagined; agentic versus non-agentic.

    “Moreover, the key components that are the concern of a hierarchical and dynamic PP framework – prediction error, reliability, uncertainty, adaptivity – readily translate to the subjective experiences that are key to source monitoring ideas: self versus other; real versus imagined; agentic versus non-agentic.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S2 (checkable).

    Interpretationreasoning

    Source information is valuable in optimally modelling the world.

    “Moreover, by scrutinising source monitoring afresh within the PP context, we are encouraged to appreciate more fully the value of source information in optimally modelling the world”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing).

    Interpretationreasoning

    The computational parameters invoked by predictive processing models appear, prima facie at least, to lend themselves readily to formulations in terms of brain processes.

    “The computational parameters invoked by predictive processing models appear, prima facie at least, to lend themselves readily to formulations in terms of brain processes.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S2 (checkable).

    Interpretationcitation

    The predictive processing framework offers opportunities for bridging levels of understanding that extend beyond merely tabulating the neural correlates of a capacity or behaviour.

    “Third, it offers opportunities for bridging levels of understanding that extend beyond merely tabulating the neural correlates of a capacity or behaviour.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites Teufel and Fletcher 2016

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Currie (2000) has already proposed a 'metarepresentational' account of delusions as being due to the individual imagining some state of affairs and misidentifying this imaging as a belief.

    “Indeed, Currie (2000) has already proposed a ‘metarepresentational’ account of delusions as being due to the individual imagining some state of affairs (e.g., that his wife has been replaced by an imposter) and misidentifying this imaging as a belief.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites Currie (2000)

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing).

    Premisecitation

    There are internal inconsistencies with respect to neurobiological implementations of predictive processing that demand careful scrutiny.

    “Moreover, as has been discussed elsewhere (Teufel and Fletcher, 2016), there are internal inconsistencies with respect to neurobiological implementations of PP that demand careful scrutiny.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites Teufel and Fletcher, 2016

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing).

    Interpretationcitation

    The question of which aspects of reality a model should represent, and which it should leave out, is unanswerable without a consideration of the model's purpose and intended level of explanation or description.

    “The question of which aspects of reality a model should represent, and which it should it leave out, is unanswerable without a consideration of the model’s purpose and intended level of explanation or description (see Teufel and Fletcher, Brain 2016 for full discussion).”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites Teufel and Fletcher, Brain 2016

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S2 (checkable).

    Interpretationreasoning

    Predictive processing provides a starting point for a computational understanding of the links between brain activity and source monitoring capacities, and of how they may be changed in psychosis.

    “Predictive processing, we suggest, provides a starting point for a computational understanding of these links, and of how they may be changed in psychosis.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Interpretationreasoning

    Misattributions of imaginings as beliefs could have severe consequences since imaginings could become able to influence perception which would in turn support developing beliefs, via the process of gradual self-reinforcement proposed.

    “Misattributions of this nature could have severe consequences since imaginings could become able to influence perception which would in turn support developing beliefs, via the process of gradual self-reinforcement that we have proposed.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S2 (checkable), S4 (based).

    Interpretationnone given

    One ultimate consequence of a failure to create, judge, compare and develop models of external reality and be clear about their aims and scope is psychosis.

    “One ultimate consequence of a failure to do this is psychosis.”
    Quote in the paperNot a substantial claim
    • Rests on an assumption: The brain must create, judge, compare and develop models of external reality and must be clear about the aims and scope of each model.

  • May combine two claims.

    Premisecitation

    Belief is not an all-or-nothing phenomenon, but varies along a continuum from more to less; if we implicitly interpret our thoughts as evidence indicative of the presence in the world of that to which the thought corresponds, then the stronger the presence of some information in our minds, the more likely we are to accept the validity of the situation posited by the thought.

    “Schwartz (1998) frames these overall ideas eloquently: ”belief is not an all-or-nothing phenomenon, but, rather, varies along a continuum from more to less…If we implicitly interpret our thoughts as evidence indicative of the presence in the world of that to which the thought corresponds, then it seems reasonable to assume that the stronger the presence of some information in our minds, i.e., the”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites Schwartz (1998)

      Not checked in the free run.

  • Classified as probably not a substantial claim.

    Interpretationcitation

    The predictive processing framework demands an explicitness and precision that conceptual/descriptive models do not, by translating conceptualisations to a mathematical framework.

    “First, in translating conceptualisations to a mathematical framework, it demands an explicitness and precision that conceptual/descriptive models do not.”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites Teufel and Fletcher 2016

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S2 (checkable), S4 (based).

    Interpretationreasoning

    A source monitoring impairment profoundly and pervasively affects the modelling process itself, leading to models that are suboptimal and harmful in very specific ways.

    “and the implications of a source monitoring impairment in profoundly and pervasively affecting the modelling process itself, leading to models that are suboptimal and harmful in very specific ways.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Interpretationcitation

    The greater imagination inflation for emotional interpersonal events may contribute to the tendency for delusions to have a negative emotional tone.

    “Interestingly, the ‘imagination inflation’ effect is greater for emotional interpersonal events than it is for neutral ones (Szpunar & Schachter 2013), perhaps contributing to the tendency for delusions to have a negative emotional tone.”
    Quote in the paperNot a substantial claim
76 of 76

  • The paper relies on this without support.

    The social world is highly volatile, and our access to it is extremely noisy.

    “The social world is highly volatile, and our access to it extremely noisy.”

  • The paper relies on this without support.

    Inferences about other people’s intentions are inferences about deeply hidden states.

    “Inferences about other people’s intentions are inferences about deeply hidden states.”

  • The paper relies on this without support.

    The more characteristic pattern is that delusion formation is preceded by a prodromal period of ‘delusional mood’, involving a vague sense that the world has subtly changed somehow to become more sinister or foreboding.

    “the more characteristic pattern is that delusion formation is preceded by a prodromal period of ‘delusional mood’, involving a vague sense that the world has subtly changed somehow to become more sinister or foreboding.”

  • The paper relies on this without support.

    Psychotic experiences are often negative, frightening, unpleasant and depressing.

    “people often suffer from these experiences. They are negative, frightening, unpleasant and depressing.”

  • The paper relies on this without support.

    Delusions tend to have a negative emotional colouring.

    “Third, we ask why delusions tend to have a negative emotional colouring.”

  • The paper relies on this without support.

    The application of predictive processing models has been largely descriptive.

    “As yet, the application of PP models has been largely descriptive.”

  • The paper relies on this without support.

    Predictive processing has yet to be fully and formally tested by examining novel predictions that it may make.

    “While this research has established, as we hope has been shown here, an impressive degree of explanatory power, PP has yet to be fully and formally tested by examining novel predictions that it may make.”

  • The paper relies on this without support.

    The brain must create, judge, compare and develop models of external reality and must be clear about the aims and scope of each model.

    “The brain is no different. It must create, judge, compare and develop models of external reality and must be clear too about the aims and, importantly, the scope of each model.”
8 of 8

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

  • Recent evidence relates delusion-proneness to reduced usage of prior beliefs in perceptual but not cognitive decision making.

    “Differential roles of priors at low and high levels of the hierarchy are also suggested by recent evidence relating delusion-proneness to reduced usage of prior beliefs in perceptual but not cognitive decision making (109).”

    From The Predictive Coding Account of Psychosis (2018) · cites this paper as [109]

    In “Why are Delusions Bizarre?” (page unavailable), the review reports the opposite perceptual association: the effect of high-level knowledge on perception increased with delusional ideation in healthy subjects. The cited passage does not make the claimed perceptual-versus-cognitive decision-making comparison. (LLM, conf=0.86)

    Checked by a reasoning model — may be wrong.

    Experimental — accuracy not yet measured

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

Each contradiction between two passages listed here was confirmed by an independent check.

  • Premack citation year differs from reference

    Minor

    The in-text citation says 1999; the listed Premack reference says 1990.

  • Cicchino citation year differs from reference

    Minor

    The text cites Cicchino et al. as 2011, whereas the corresponding reference gives 2005.

  • LeDoux citation year differs from reference

    Minor

    The text cites LeDoux et al. as 1988, while the corresponding reference prints 1998.

  • Behrens citation year differs from reference

    Minor

    The text cites Behrens et al. as 2008, but the corresponding Behrens-led reference is dated 2007.

  • Moore citation year differs from reference

    Minor

    The text cites Moore as 2004 for auditory cues, while the listed Moore hearing reference is dated 2012.

  • Schlagenhauf citation year differs from reference

    Minor

    The text cites the Schlagenhauf study as 2010, while the listed study is dated 2014.

  • Roediger and McDermott citation year differs from reference

    Minor

    The in-text citation gives 1978; the listed Roediger and McDermott reference gives 1995.

  • Pezzulo citation year differs from reference

    Minor

    The text cites Pezzulo as 2014, but the listed Pezzulo reference is dated 2013.

  • Haug citation year differs from reference

    Minor

    The text cites Haug et al. as 2013, but the listed Haug paper on social dysfunction is dated 2014.

  • Friston and Stephan citation year differs from reference

    Minor

    The text cites Friston and Stephan as 2007, but the corresponding reference lists 2009.

10 of 10

Cites

Flagged first, then by citations

Research works printed but not linked: 10

  1. Adams RA, 2015, J. Neurol. Neurosurg. Psychiatry, 87, 53
  2. Adams RA, 2013, Front. Psychiatry, 4
  3. Ceaser AE, 2015, Front. Hum. Neurosci., 9, 687
  4. Johnson MK, 1993, Theories of Memory, 241
  5. Maher BA, 1974, J. Indiv. Psychol., 30, 98
  6. Mathys C, 2011, Front. Hum. Neurosci., 5
  7. Parnas J, 2003, Acta Psychiatr. Scand., 137, 434
  8. Rescorla RA, 1972, Classical Conditioning II, 64
  9. Roessler J, 2013, Oxford Handbook of the Philosophy of Psychiatry, 658
  10. Vinckier F, 2015, Mol. Psychiatry, 21, 945

Cited by

Sorted by citations

Related

Authors
Griffin, J. D., & Fletcher, P. C.
Affiliations
University of Cambridge (GB)
ORCID
Venue
Annual Review of Clinical Psychology
Publisher
Annual Reviews
Published
2017-05-08
Volume
13
Issue
1
Pages
265-289
ISSN
  • 1548-5943 (ISSN-L)
  • 1548-5951
DOI
10.1146/annurev-clinpsy-032816-045145
PMC
PMC5424073
OpenAlex
W2396823339
Cited by
133
References
157
Citations by year
2026: 23 · 2025: 8 · 2024: 26 · 2023: 14 · 2022: 15 · 2021: 15 · 2020: 8 · 2019: 10 · 2018: 12 · 2017: 2
FWCI
1.32
Type
journal article
Language
en
Open access
Locations
  • Annual Review of Clinical Psychology (publishedVersion)
  • PubMed (publishedVersion)
  • PubMed Central (acceptedVersion)
Funders
Wellcome Trust (095844/Z/11/Z, 095844, 095692)
Topics
  • Health Sciences › Medicine › Psychiatry and Mental health › Schizophrenia research and treatment
  • Social Sciences › Arts and Humanities › Philosophy › Mental Health and Psychiatry
  • Life Sciences › Neuroscience › Cognitive Neuroscience › Embodied and Extended Cognition
Keywords
predictive processing, predictive coding, source monitoring, psychosis, delusions, hallucinations, hierarchical models, trauma, social adversity, explanatory gap
MeSH terms
Delusions, Hallucinations, Humans, Psychotic Disorders, Schizophrenia, Cognitive Dysfunction

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Claims read by deepseek-v4.1-flash under definition version 3 (extract-v3) on 2026-10-10.

Classified by jev-1.13-20260917 (gold-guideline-v0.2) on 2026-10-10.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review