Context
Establishing a neurobiological account of delusion formation that links cognitive processes, brain activity, and symptoms is important to furthering our understanding of psychosis.
Objective
To explore a theoretical model of delusion formation that implicates prediction error–dependent associative learning processes in a pharmacological functional magnetic resonance imaging study using the psychotomimetic drug ketamine.
Design
Within-subject, randomized, placebo-controlled study.
Setting
Hospital-based clinical research facility, Addenbrooke's Hospital, Cambridge, England. The work was completed within the Wellcome Trust and Medical Research Council Behavioral and Clinical Neuroscience Institute, Cambridge.
Participants
Fifteen healthy, right-handed volunteers (8 of whom were male) with a mean ± SD age of 29 ± 7 years and a mean ± SD predicted full-scale IQ of 113 ± 4 were recruited from within the local community by advertisement.
Interventions
Subjects were given low-dose ketamine (100 ng/mL of plasma) or placebo while performing a causal associative learning task during functional magnetic resonance imaging. In a separate session outside the scanner, the dose was increased (to 200 ng/mL of plasma) and subjects underwent a structured clinical interview.
Main Outcome Measures
Brain activation, blood plasma levels of ketamine, and scores from psychiatric ratings scales (Brief Psychiatric Ratings Scale, Present State Examination, and Clinician-Administered Dissociative States Scale).
Results
Low-dose ketamine perturbs error-dependent learning activity in the right frontal cortex (P = .03). High-dose ketamine produces perceptual aberrations (P = .01) and delusion-like beliefs (P = .007). Critically, subjects showing the highest degree of frontal activation with placebo show the greatest occurrence of drug-induced perceptual aberrations (P = .03) and ideas or delusions of reference (P = .04).
Conclusions
These findings relate aberrant prediction error–dependent associative learning to referential ideas and delusions via a perturbation of frontal cortical function. They are consistent with a model of delusion formation positing disruptions in error-dependent learning.
In this article, we attempt to understand how delusional beliefs may come about by relating them, in the setting of a drug model, to the neurobiology of associative learning. A number of theories of psychosis suggest aberrant learning processes as the basis for delusions.1-8 According to such theories, very early in delusion formation, subjects report that their attention is drawn to irrelevant environmental stimuli9-15 and that their capacity to form associations (between perceptions, thoughts, stimuli, and events) is markedly heightened.1-3,9,16 What might be the basis for such experiences and associations? One account is that they arise from inappropriate prediction error signals. Prediction error refers to a mismatch between expectancy and outcome. According to formal learning theories, prediction error guides adaptive behavior by diverting attention to critical environmental stimuli and driving causal associations between those stimuli and significant environmental events.17-20 The neural instantiation of prediction error has been widely explored.21-29 The theory posits that noise in the prediction error system leads to inappropriate signaling of a mismatch4,5 resulting in new association formation4-8,17,20 and the diversion of attention to potentially explanatory (but irrelevant) stimuli in the environment.4-8,18,19
Although it is difficult to track the formation of delusions in patients, we nevertheless believe that an experimental analysis of this theory is tractable. Using the N-methyl-D-aspartate receptor antagonist ketamine, which may provide a useful model for psychosis30-32 and has also been increasingly explored in terms of cognitive changes30-46 and brain function,47-57 we sought to evaluate the suggested link between prediction error–dependent associative learning and the emergence of delusional beliefs.
The aim of this study was therefore to link prediction error–dependent causal associative learning to the very early stages of delusion formation using specific brain responses as a marker for such learning. Frontostriatal engagement, tracked using functional magnetic resonance imaging (fMRI), served as an assay of prediction error in healthy individuals, and we explored the behavior of this system in the setting of a causal learning task.29 First, we evaluated the brain response to prediction error–dependent learning with placebo in each subject. This was carried out using an empirically derived, regionally specific analysis.27-29 Next, we explored 3 interrelated questions relevant to the learning-based model of delusions. First, does ketamine, which is known to produce delusion-like beliefs in volunteers, act even at subpsychotic doses as a modulator of the brain response to prediction error? Second, with a high (psychotogenic) dose of ketamine, does the occurrence of delusional beliefs correlate with perceptual changes such as heightening of sensations and salience of environmental stimuli experienced by the subjects? Third, can the behavioral phenomena occurring at the high dose be predicted across individuals by brain responses signaling prediction error–dependent learning with placebo and low-dose ketamine?
Fifteen healthy, right-handed volunteers (8 of whom were male) with a mean ± SD age of 29 ± 7 years and a mean ± SD predicted full-scale IQ58 of 113 ± 4 were recruited from the local community by advertisement, and they were screened using an initial telephone interview and subsequent personal interview. Psychiatric or physical illness, head injury, drug or alcohol dependence, and smoking were excluding factors, as were family history of psychiatric history and alcohol problems. The study was approved by the Cambridge Local Research Ethics Committee, Cambridge, England, and was carried out in accordance with the Declaration of Helsinki. Written informed consent was given by all of the subjects.
A double-blind, placebo-controlled, randomized, within-subjects design was used. On each visit, 2 phases of testing occurred. First, with administration of saline or low-dose ketamine (plasma target, 100 ng/mL), subjects performed the learning task in the fMRI scanner. After scanning, the subject was taken from the scanner and the saline administration was continued or the ketamine dose was increased to 200 ng/mL of plasma. Subjects then underwent a series of clinical interviews exploring the presence, nature, and severity of psychotic phenomena. The order of drug and placebo visits was counterbalanced across subjects and spaced by at least 1 month.
Although we also ran the learning task with high-dose ketamine, performance was inadequate and we will not report this part of the study further. Subjects also performed 3 other cognitive tasks while in the fMRI scanner. These will be reported in a separate article.
Racemic ketamine (1 mg/mL) was administered intravenously by initial bolus and subsequent continuous target-controlled infusion using a computerized pump (Graseby 3500; Graseby Medical Ltd, Watford, United Kingdom) to achieve plasma concentrations of 100 or 200 ng/mL using the pharmacokinetic parameters of a 3-compartment model described by Domino et al.59
Associative learning task with placebo and low-dose ketamine
We used the task reported by Corlett et al,29 who used a retrospective revaluation paradigm in which engendered expectations are violated to produce a prediction error. Subjects were asked to imagine themselves working as an allergist confronted with a new patient, “Mr X,” who has allergic reactions following some meals but not others. Their task was to work out which foods caused allergic reactions by observing the consequences of eating various foods. The task consisted of a series of trials, each of which had the general structure summarized in Figure 1. Trials comprised presentation of a food picture (representing a meal eaten by Mr X), a predictive response by the subject, and then an outcome. Responses allowed for both prediction and confidence outcome measures for each trial.
The experimental structure was identical to that used previously.29 The key manipulations relevant to the question under study are summarized in Figure 2. Each subject was trained concurrently on a number of different contingencies between foods and allergic reactions. Learning occurred over 3 stages: training, unovershadowing, and violation. This design is clarified with examples in Figure 2. In summary, expectancies were set up during a training phase, unovershadowing occurred, and we explored the impact of violation of the consequent expectancies on brain activity at the critical stage. We were then able to determine the impact of ketamine on these brain responses. We used parallel versions of the task so that no subject was exposed to the same foods on separate occasions. The same food was also not repeated across different conditions within the learning session.
Clinical interview with placebo and high-dose ketamine
Subjects underwent a shortened form of the Present State Examination (PSE)60 and the Clinician-Administered Dissociative States Scale (CADSS).61 This form comprises 19 self-reported ratings and 4 clinician-rated items (rated from 0 [not at all] to 4 [extremely]). The interviews were videotaped, and the videotapes were rated by 2 of us (E.P.-C. and P.J.M.). The videotapes of subjects' responses were also rated on the 24-item version of the Brief Psychiatric Ratings Scale.62
The CADSS incorporates 5 subscales61: body perception, environmental perception, feelings of unreality, memory impairment, and time perception. Given the specific theoretical predictions regarding perceptual aberrations during the very early stages of psychosis precipitating psychotic phenomena,5,9-15 this investigation made use of the environmental perception subscale of the CADSS. We assessed the strength of correlation between ratings on this subscale and scores for delusional ideation on the PSE, a link that has been alluded to theoretically5,9-15 but to our knowledge not formally tested.
fMRI DATA ACQUISITION AND ANALYSIS
The fMRI data were acquired using a Bruker MedSpec 30/100 (Bruker Optics, Ettlingen, Germany) operating at 3 T. Gradient-echo echo-planar T2*-weighted images depicting blood oxygenation level–dependent contrast were acquired from 21 noncontiguous near-axial planes with the following parameters: repeat time, 1.1 seconds; echo time, 27.5 milliseconds; flip angle, 66°; in-plane resolution, 3.1 × 3.1 mm; matrix size, 64 × 64; field of view, 20 × 20 cm; bandwidth, 100 kHz. A total of 1608 volumes (21 slices each of 4-mm thickness with an interslice gap of 1 mm) per subject were acquired in 2 scans at each visit. The first 6 volumes from each scan were discarded.
The software SPM2 (available at http://www.fil.ion.ucl.ac.uk; Wellcome Department of Cognitive Neurology, London, England) was used. After realignment, spatial normalization, and spatially smoothing (gaussian kernel, 8 mm), the time series in each session were high-pass filtered (maximum of 1/120 Hz) and serial autocorrelations were estimated using an autoregression 1 model.
The average hemodynamic response to each event type was designated as occurring at the presentation of the outcome stimulus (ie, when subjects were informed whether an allergic reaction had occurred for that trial) and modeled using a canonical, synthetic hemodynamic response function.63 This function was used as a covariate in a general linear model, generating parameter estimates for each condition. Individuals' contrast images for prespecified comparisons were produced by comparing parameter estimates across relevant conditions and were entered into group analyses treating intersubject variability as a random effect.
Identification of basic learning effects
Behavioral effects are summarized in Figure 3. A groupwise analysis was carried out to identify key regions responding to basic associative learning processes (all of the learning trials vs all of the fixation trials) during stage 1. In keeping with our regionally specific a priori hypotheses from previous normative data25,28,29 concerning frontostriatal systems, this analysis was confined to anatomically defined regions of interest (Figure 4A) identified by our previous study.29 These regions of interest were the bilateral substantia nigra, bilateral striatum (dorsal and ventral), and lateral right prefrontal cortex (rPFC). In a subsequent analysis, we extended our exploration to include bilateral hippocampi, particularly in light of the fact that mismatch theories of psychosis4,5 suggest that the hippocampus serves as the comparator, computing mismatches between expectation and experience. Brain loci showing learning effects within this predefined system (false discovery rate–corrected P<.05)64 were recorded and used to generate spheres (radius, 10 mm) that were tested in subsequent analyses. This use of analysis of regions of interest boosts sensitivity while protecting against type I error.
Prediction error and ketamine effects
Parameter estimates for each of the 4 conditions (violation events and well-learned control items for both drug and placebo) were extracted from the spheres centered on each region of interest. A repeated-measures analysis of variance was used to identify regions showing task and prediction error effects (violation vs control) and drug × task interactions at stage 3.
Linking cognition, brain activity, and psychosis
To assess the relationship between behavior, brain activity, and symptoms, we evaluated correlations (across subjects) between brain responses to prediction error and ratings on perceptual aberration and delusional ideation. The perceptual aberration score was taken from the environmental perception subscale of the CADSS. This included ratings on items pertaining to the qualities of auditory and visual perceptions, particularly their clarity. Referential ideas were rated using the PSE. Each of these scores was entered into a second-order regression model to determine whether psychopathological responses to a high dose of ketamine could be predicted by the extent of prediction error–dependent activation. Analyses were confined to regions of interest identified as described earlier. Regression analyses were thresholded using the false discovery rate at P<.05.64
The mean ± SD observed ketamine plasma levels for targets of 100 ng/mL and 200 ng/mL were 88 ± 10 ng/mL and 210 ± 13 ng/mL, respectively. (These data are based on 14 of the 15 subjects since we were unable to draw blood samples from 1 subject.)
Behavior results are summarized in Figure 3. With both placebo and ketamine, subjects quickly learned to make correct positive and negative predictions during stages 1 and 2. Repeated-measures analyses of variance on the behavioral data from stages 1 and 2 revealed main effects of learning in both cases (stage 1: F11,110 = 4.04, P<.001; stage 2: F5,49 = 28.39, P<.001). There was no main effect of drug on either stage (stage 1: F1,120 = 0.68, P = .41; stage 2: F1,53 = 2.62, P = .14), and there was no drug × task interaction (stage 1: F11,110 = 0.97, P = .48; stage 2: F5,49 = 1.12, P = .36), suggesting no measurable impact of ketamine on behavioral performance.
Figure 3C shows the subjects' first predictive responses to the unovershadowed items with placebo and ketamine. This can be taken as an index of the extent to which they have indeed revalued the items that were absent at stage 2. There was a nonsignificant trend toward a difference in predictions about the unovershadowed item (banana in Figures 1, 2, and 3) between drug and placebo (2-tailed paired t test, df = 1, 14, P = .09). This trend was seen as a reduced tendency for subjects to predict a positive link between food and allergy with low-dose ketamine, ie, a reduction in the magnitude of unovershadowing.
Activations During Stage 1 With Placebo
Acquisition of associative relationships significantly activated rPFC, bilateral caudate, ventral striatum, substantia nigra, and in the subsequent analysis, medial temporal cortex (Figure 4 and Table). As described earlier, these activations defined the loci for planned comparisons pertaining to the violation stage.
Brain Activations Accompanying Violation of Expectancies (Stage 3)
With placebo, prediction error was associated with significant activation in the rPFC as expected.25,28,29 That is, subjects activated this region relatively more when their expectancies about the unovershadowed item were violated than when confirmed. There was a significant drug × condition interaction (P = .03). Post hoc t tests (P<.05) revealed an overactivation of rPFC in response to nonsurprising control items on low-dose ketamine compared with placebo (P = .03) and a strong trend toward underactivation in response to expectancy violation on low-dose ketamine compared with placebo (P = .08) (Figure 4B).
Consistent with previous work,29,65-69 we found a significant main effect of task in the right striatum (caudate) (P = .01). Ketamine did not have a significant impact on the blood oxygenation level–dependent response in this region (P = .43) (Figure 4C).
There was increased activity in the left substantia nigra for violation of learned expectancy relative to the well-learned control item (P = .047) (Figure 4D). There was no significant effect of ketamine on prediction error processing in this region (P = .30).
Although bilateral hippocampi showed activation in response to initial learning, only the left medial temporal lobe showed a significant activation in response to prediction error (P = .048). No significant effects of ketamine were found (P = .60).
In addition, the left caudate (P = .06) and right substantia nigra (P = .06) showed a trend toward a main effect of expectancy violation. We will not discuss these trends further.
Symptoms produced by high-dose ketamine
The higher dose of ketamine precipitated a typical profile of symptoms. Relevant here were changes in the clarity of thoughts and percepts (CADSS perceptual subscale, placebo vs drug, 2-tailed, paired t test, P = .01) and referential ideation (PSE simple ideas of reference subscale score, placebo vs drug, 2-tailed, paired t test, P = .007). Importantly, there was a significant correlation across subjects between profundity of perceptual aberration and strength of referential ideation (Pearson r = 0.7, 2-tailed P<.001). This is experimental evidence for a theoretical prediction of the model we are testing, that perceptual aberrations are a prelude to delusional ideas.4-8 Ketamine also induced subjectively inefficient thinking, depressed mood, expansive mood, and anxiety. However, there was no indication of correlation between these scores on the PSE and CADSS (inefficient thinking: Pearson r = −0.060, 2-tailed P = .56; depressed mood: Pearson r = 0.237, 2-tailed P = .40; expansive mood: Pearson r = −0.333, 2-tailed P = .16; and anxiety: Pearson r = −0.049, 2-tailed P = .86).
Linking brain activity to subsequent ketamine-induced symptoms
Using regression analysis, we observed that both CADSS scores and PSE ideas of references subscale scores were predicted by the magnitude of rPFC response to prediction error with placebo (Figure 5). That is, those subjects who showed the greatest PFC response to violation of their expectancies experienced the greatest perceptual illusions and delusional ideation on the high dose of the drug. The relationship between the rPFC activity and perceptual aberration remained significant (r10 = 0.8, 2-tailed P = .002) when plasma levels were included in the analysis. Individual-specific attenuations by ketamine, relative to placebo, of PFC responses to prediction error did not predict CADSS or PSE scores. There was, however, a trend (uncorrected P = .002 for PSE scores, .007 for CADSS scores) for subjects showing the least prediction error–related response with low-dose ketamine to show higher CADSS and PSE scores.
Interestingly, the PFC responses with placebo during the training stage (stage 1) also predicted scores on the CADSS environmental perception subscale (uncorrected P<.001). Furthermore, subjects' basic behavioral learning rates during this stage (assessed by individuals' slopes of change in predictive responses to successive trials) also predicted the rPFC response to prediction error (uncorrected P = .004).
Finally, other symptoms (negative symptoms and thought disorder, neither of which were theoretically implicated in the model) did not correlate with response to prediction errors in the rPFC system identified.
Our results provide behavioral and neurobiological evidence for the existence of a relationship between prediction error–dependent associative learning and the development of delusional beliefs with ketamine. First, perceptual anomalies such as heightened salience correlated with ideas or delusions of reference, consistent with the link between perception of environmental changes and delusions posited by the model. Second, lateral rPFC, as predicted by our data24,28,29 and indeed by related fMRI work,70 indexes prediction error, and this response is disrupted by low-dose ketamine. The nature of this disruption was consistent with the model in that the region no longer distinguished between predicted and unpredicted occurrences. Finally, subject-specific activity in precisely the same region during placebo administration was predictive of individuals' likelihood of experiencing perceptual changes and ideas or delusions of reference (Figure 5). Thus, the observed links between brain activity, cognition, and subjective experience are consistent with the model under scrutiny.
We also draw attention to a degree of internal consistency whereby responses with placebo in precisely the same frontal region during the initial stage of basic associative learning were predictive of high-dose–associated symptoms. Moreover, frontal responses to learning at each stage were specific in their predictive power; while reliably predicting CADSS scores and referential beliefs, they did not predict other symptoms engendered by ketamine (negative symptoms, thought disorder, mood changes, or other positive symptoms).
It is interesting to note the nature of frontal modulation produced by low-dose ketamine (Figure 4). The reduction in the extent to which rPFC distinguished between violation and control stimuli was driven both by an attenuation of response to violations and an augmentation in response to stimuli, which should be highly predictable. This disturbance points to a 2-faceted change in prediction error signaling. First, the signal to true expectancy-outcome mismatches is diminished. This is perhaps unsurprising given that low-dose ketamine produces a trend toward a behavioral attenuation of the unovershadowing effect. The other phenomenon driving this interaction, and one that is perhaps more directly relevant to the model under study, is that there is a ketamine-related augmentation of the rPFC response to well-learned events (Figure 4B). Taken together, our observations suggest that the perturbation of prediction error signaling may manifest as both decreases in appropriate signaling and increases in inappropriate signaling. The observations are consistent with the idea that delusions arise in the context of a failure to attend to and learn about salient stimuli while attention is increased toward stimuli that should be peripheral, familiar, or predictable.
There are 2 contrasting explanations for correlation between individuals' “healthy” (placebo) responses to prediction error and subsequent psychopathological scores. One is that greater levels of brain response signal some form of physiological inefficiency71 and that those subjects showing this are the ones most likely to show a prediction error disturbance when exposed to the ketamine challenge. The other possibility is that the magnitude of right prefrontal response indexes individual sensitivity to prediction error, with more sensitive individuals being most vulnerable to perturbation of the system. Ultimately, our key finding lies in the existence of a relationship, and we are in a position only to speculate on the source of individual vulnerability. However, relevant to this question is our observation that rPFC responses at the violation stage were significantly predicted by subjects' baseline speed of learning estimated at stage 1. This admittedly post hoc observation counsels in favor of the latter view, that rPFC response indicates subjects' sensitivity to prediction error–driven associative learning and that this sensitivity may be penalized with high-dose ketamine.
Another question is whether it is prediction error signal or perceptual change that has primacy in accounting for the fundamental change that ultimately engenders delusional beliefs. That is, does disrupted prediction error produce the perceptual change or vice versa? Again, we can only speculate on this, but although ketamine is well known to produce perceptual changes at higher doses,31,32 these changes in our study were largely expressed in terms of changes in the ways in which subjects felt about external stimuli rather than changes in perceptions per se. We argue that these changes are engendered by anomalies of prediction error signaling rather than that prediction error arises out of a change in perceptual processing. Further evidence in favor of this view comes from the observation that low-dose ketamine, which produced very little perceptual change, was associated with significant modulation of rPFC prediction error signal.
The idea that delusions may arise from disruptions in prediction error–dependent associative learning would, of course, predict disrupted associative learning in schizophrenia. Indeed, the early development of such models derived from observations of disturbed latent inhibition and blocking in people with schizophrenia.4,5 Importantly, reductions in blocking and latent inhibition suggest inappropriate persistence of prediction error firing in response to situations that should be predictable, an observation in keeping with the model. In schizophrenia, there is also evidence of disrupted prediction error indexed by mismatch negativity signal72 induced by inserting an unpredictable stimulus into a train of predictable stimuli. The reduced mismatch negativity signal in schizophrenia may point to reduced prediction error firing in response to deviant tones, increased firing to nondeviant tones, or both (any of these possibilities would lead to a reduction in the electrophysiological distinction between standard and deviant tones). It is important to consider, too, that Umbricht et al38 have demonstrated a relationship between reduced mismatch negativity signal with placebo and positive symptoms induced by ketamine. This appears to be at odds with our finding. The discrepancy may arise because Umbricht and colleagues used a stereotyped pattern of tones (every 10th tone was the deviant). Therefore, perhaps subjects who were most sensitive to this stereotyped pattern would show the least prediction error and hence an attenuated mismatch negativity signal. However, this is highly speculative, and there are a number of major psychological and methodological differences between the studies that could explain the apparent discrepancy.
Our data should be considered in light of a recent formulation of development of delusions in terms of altered salience of environmental stimuli arising from disturbed firing in the mesolimbic dopamine system.4-8 In fact, although the salience model has a different anatomical emphasis (mesolimbic rather than prefrontal) and a different neurochemical substrate (dopaminergic rather than glutamatergic), there is much resonance between the 2 ways of understanding how delusional beliefs arise. There is good evidence that dopaminergic firing lies at the core of prediction error signaling,21-24 and while the evidence is strong that mesolimbic firing is critical to this,21-24,26,27 recent work73 has shown that prefrontal responses to changes in mesocortical dopaminergic firing are mediated in part by glutamate. Thus, glutamate firing in PFC appears to be critical to mediating reward or salience, suggesting important similarities between these 2 accounts.
It is important to note that our data and the model under test address only the very early stage of delusion formation. They suggest that there is a relationship between changes in the sensory experience of environmental stimuli and the development of referential beliefs and that the specific cognitive disruption underlying this may lie in prediction error signaling. These links do not, however, address a key aspect of delusions: their fixity. Ultimately, a delusion consists of an altered and entrenched view of the world. Although the model does not attempt to account for this quality, we suggest that repeated experience of abnormal prediction error signal will be an insidious process12 in which patients are frequently and repeatedly surprised by their experiences. The fully formed delusions may be the ultimate elaborate explanations that attempt to account for the uncertainty of their world.3-5 The insight provided by constructing and supporting an explanatory delusional system that accounts for such bizarre experiences may be so relieving or rewarding as to render the delusional system impervious to contradictory evidence.3
Another possible limitation is that our approach prevents us from determining the behavior of PFC while delusional ideas are current. That is, we did not scan during high-dose ketamine administration. This was for both practical and theoretical reasons. Primarily among the latter was that we do not aim to localize delusions themselves, believing fMRI to be extremely limited in telling us about the functional neuroanatomy of symptoms.74 Rather, we use the fMRI signal as a brain marker for the core cognitive processes suspected to underpin the symptoms, and we explore the behavior of this signal both as a predictor of the symptoms and under a drug challenge that produces subbehavioral effect. Thus, we contend that this approach speaks much more directly to the cognitive model in question.
We should also draw attention to the fact that the reported interaction in rPFC between ketamine and condition (violation vs nonviolation control) did not significantly predict changes in the CADSS or PSE scores. We suggest that this is probably owing to the subtlety of such an interaction (ie, a difference in a difference of activity as a predictor of a behavioral score). Whatever the explanation, we limit ourselves to stating that there is an individual pattern of vulnerability to ketamine predicted by learning-dependent PFC responses with placebo and that precisely the same region shows a groupwise attenuation by low-dose ketamine.
In summary, our study was motivated by the convictions that a satisfactory model of delusion formation must begin with an account of belief formation since delusions are aberrant beliefs and that the development of associations is fundamental to the formation of beliefs. For this reason, we sought to identify links between association formation, brain responses, drug effects, and symptoms. By inducing strong ideas of reference using ketamine, we used the learning of causal beliefs as a setting in which to explore a cognitive model of delusion formation. We demonstrate that a critical process in belief formation (prediction error) and its brain basis (rPFC) are disrupted by low-dose ketamine and that individual variations in task-related activity in this region correlate across individuals with their experience of perceptual aberrations and delusional ideation with a high dose of the drug. We believe that the combination of functional neuroimaging and psychopharmacology may be key in linking the 3 levels of explanation that would be critical to a truer understanding of psychosis: brain, cognition, and symptoms.
Correspondence: Paul C. Fletcher, PhD, MRCPsych, Box 255, University of Cambridge, Department of Psychiatry, Addenbrooke's Hospital, Hills Road, Cambridge CB2 2QQ, England (pcf22@cam.ac.uk).
Submitted for Publication: September 7, 2005; final revision received October 28, 2005; accepted December 12, 2005.
Funding/Support: This work was supported by the Wellcome Trust and was carried out within the University of Cambridge Behavioural and Clinical Neurosciences Institute supported by a joint award from the Medical Research Council and the Wellcome Trust.
1.Miller
R Schizophrenic psychology, associative learning and forebrain dopamine.
Med Hypotheses 1976;2203- 211
PubMedGoogle Scholar 2.Miller
R Hyperactivity of associations in psychosis.
Aust N Z J Psychiatry 1989;23241- 248
PubMedGoogle Scholar 3.Miller
R Striatal dopamine in reward and attention: a system for understanding the symptomatology of acute schizophrenia and mania.
Int Rev Neurobiol 1993;35161- 278
PubMedGoogle Scholar 4.Gray
JAFeldon
JRawlins
JNPHemsley
DRSmith
AD The neuropsychology of schizophrenia.
Behav Brain Sci 1991;141- 20
Google Scholar 5.Hemsley
DR Perceptual and cognitive abnormalities as the bases for schizophrenic symptoms. In:David
ACutting
Jeds. The Neuropsychology of Schizophrenia. Hove, United Kingdom Psychology Press1994;97- 116
6.Kapur
S Psychosis as a state of aberrant salience: a framework linking biology, phenomenology and pharmacology in schizophrenia.
Am J Psychiatry 2003;16013- 23
PubMedGoogle Scholar 7.Kapur
S How antipsychotics become anti-”psychotic”: from dopamine to salience to psychosis.
Trends Pharmacol Sci 2004;25402- 406
PubMedGoogle Scholar 8.Kapur
SMizrahi
RLi
M From dopamine to salience to psychosis: linking biology, pharmacology and phenomenology of psychosis.
Schizophr Res 2005;7959- 68
PubMedGoogle Scholar 9.Matussek
P Untersuchungen uber die wahnwahrnehmung.
Psychiatri Z Neurol 1952;189279- 318
Google Scholar 10.McGhie
AChapman
J Disorders of attention and perception in early schizophrenia.
Br J Med Psychol 1961;34103- 116
PubMedGoogle Scholar 11.Bowers
MB
JrFreedman
DX “Psychedelic” experiences in acute psychoses.
Arch Gen Psychiatry 1966;15240- 248
PubMedGoogle Scholar 13.Bowers
MB
Jr Pathogenesis of acute schizophrenia: an experimental approach.
Arch Gen Psychiatry 1968;19348- 355
PubMedGoogle Scholar 14.Freedman
BChapman
LJ Early subjective experience of schizophrenic episodes.
J Abnorm Psychol 1973;8246- 54
PubMedGoogle Scholar 15.Freedman
BJ The subjective experience of perceptual and cognitive abnormalities in schizophrenia: a review of autobiographical accounts.
Arch Gen Psychiatry 1974;30333- 340
PubMedGoogle Scholar 16.Schneider
C Die Psychologie der Schizophrenen. Leipzig, Germany Germany Thieme1930;
17.Rescorla
RAWagner
AR A theory of Pavlovian conditioning: variations in the effectiveness of reinforcement and non-reinforcement. In:Black
AHProkasy
WFeds. Classical Conditioning II: Current Research and Theory. New York, NY Appleton-Century-Crofts1972;64- 99
18.Mackintosh
NJ A theory of attention: variations in the associability of stimuli with reinforcement.
Psychol Rev 1975;82276- 298
Google Scholar 19.Pearce
JMHall
G A model for Pavlovian learning: variations in the effectiveness of conditioned but not of unconditioned stimuli.
Psychol Rev 1980;87532- 552
PubMedGoogle Scholar 20.Dickinson
A The 28th Bartlett Memorial Lecture: causal learning: an associative analysis.
Q J Exp Psychol B 2001;543- 25
PubMedGoogle Scholar 21.Schultz
W The phasic reward signal of primate dopamine neurons.
Adv Pharmacol 1998;42686- 690
PubMedGoogle Scholar 22.Hollerman
JRSchultz
W Dopamine neurons report an error in the temporal prediction of reward during learning.
Nat Neurosci 1998;1304- 309
PubMedGoogle Scholar 24.Waelti
PDickinson
ASchultz
W Dopamine responses comply with basic assumptions of formal learning theory.
Nature 2001;41243- 48
PubMedGoogle Scholar 25.Fletcher
PCAnderson
JMShanks
DRHoney
RCarpenter
TAPapadakis
NBullmore
ET Responses of human frontal cortex to surprising events are predicted by formal associative learning theory.
Nat Neurosci 2001;41043- 1048
PubMedGoogle Scholar 26.O'Doherty
JPDayan
PFriston
KCritchley
HDolan
RJ Temporal difference models and reward-related learning in the human brain.
Neuron 2003;38329- 337
PubMedGoogle Scholar 27.McClure
SMBerns
G SMontague
PR Temporal prediction errors in a passive learning task activate human striatum.
Neuron 2003;38339- 346
PubMedGoogle Scholar 28.Turner
DCAitken
MRShanks
DRSahakian
BJRobbins
TWSchwarzbauer
CFletcher
PC The role of lateral frontal cortex in causal associative learning: exploring preventative and super-learning.
Cereb Cortex 2004;14872- 880
PubMedGoogle Scholar 29.Corlett
PRAitken
MRDickinson
AShanks
DRHoney
GDHoney
RARobbins
TWBullmore
ETFletcher
PC Prediction error during retrospective revaluation of causal associations in humans: fMRI evidence in favor of an associative model of learning.
Neuron 2004;44877- 888
PubMedGoogle Scholar 30.Ghoneim
MMHinrichs
JVMewaldt
SPPetersen
RC Ketamine: behavioral effects of subanesthetic doses.
J Clin Psychopharmacol 1985;570- 77
PubMedGoogle Scholar 31.Krystal
JHKarper
LPSeibyl
JPFreeman
JKDelaney
RBremner
JDHeninger
GRBowers
MB
JrCharney
DS Subanesthetic effects of the non-competitive NMDA antagonist, ketamine, in humans: psychotomimetic, perceptual, cognitive, and neuroendocrine responses.
Arch Gen Psychiatry 1994;51199- 214
PubMedGoogle Scholar 32.Bowdle
TARadant
ADCowley
DSKarasch
EDRoy-Byrne
PP Psychedelic effects of ketamine in healthy volunteers: relationship to steady-state plasma concentrations.
Anesthesiology 1998;8882- 88
PubMedGoogle Scholar 33.Krystal
JHBennet
AAbi-Saab
DBelger
AKarper
LPD’Souza
DCLipschitz
DAbi-Dargham
ACharney
DS Dissociation of ketamine effects on rule acquisition and rule implementation: possible relevance to NMDA receptor contributions to executive cognitive functions.
Biol Psychiatry 2000;47137- 143
PubMedGoogle Scholar 34.Malhotra
AKPinals
DAWeingartner
HSirocco
KMissar
CDPickar
DBreier
A NMDA receptor function and human cognition: the effects of ketamine in healthy volunteers.
Neuropsychopharmacology 1996;14301- 307
PubMedGoogle Scholar 35.Adler
CMGoldberg
TEMalhotra
AKPickar
DBrier
A Effects of ketamine on thought disorder, working memory, and semantic memory in healthy volunteers.
Biol Psychiatry 1998;43811- 816
PubMedGoogle Scholar 36.Newcomer
JWFarber
NBJetovic-Terodovic
VSelke
GMelson
AKHershey
TCraft
SOlney
JW Ketamine-induced NMDA receptor hypofunction as a model of memory impairment and psychosis.
Neuropsychopharmacology 1999;20106- 118
PubMedGoogle Scholar 37.Umbricht
DSchmidt
LKoller
RVollenweider
FXHell
DJavitt
DC Ketamine-induced deficits in auditory and visual context-dependent processing in healthy volunteers: implications for models of cognitive deficits in schizophrenia.
Arch Gen Psychiatry 2000;571139- 1147
PubMedGoogle Scholar 38.Umbricht
DKoller
RVollenweider
FXSchmid
L Mismatch negativity predicts psychotic experiences induced by NMDA receptor antagonist in healthy volunteers.
Biol Psychiatry 2002;51400- 406
PubMedGoogle Scholar 39.Duncan
EJMadonick
SHParwani
AAngrist
BRajan
RChakravorty
SEfferen
TRSziliagy
SStephanides
MChappell
PBGonzenbach
SKo
GNRotrosen
JP Clinical and sensorimotor gating effects of ketamine in normals.
Neuropsychopharmacology 2001;2572- 83
PubMedGoogle Scholar 40.Lahti
ACWeiler
MATamara-Michaelidis
BAParwani
ATamminga
CA Effects of ketamine in normal and schizophrenic volunteers.
Neuropsychopharmacology 2001;25455- 467
PubMedGoogle Scholar 41.Abel
KMAllin
MPHemsley
DRGeyer
MA Low dose ketamine increases prepulse inhibition in healthy men.
Neuropharmacology 2003;44729- 737
PubMedGoogle Scholar 42.Honey
RATurner
DCHoney
GDSharar
SRKumaran
DPomarol-Clotet
EMcKenna
PSahakian
BJRobbins
TWFletcher
PC Subdissociative dose ketamine produces a deficit in manipulation but not maintenance of the contents of working memory.
Neuropsychopharmacology 2003;282037- 2044
PubMedGoogle Scholar 43.Morgan
CJMofeez
ABradner
BBromley
LCurran
HV Acute effects of ketamine on memory systems and psychotic symptoms in healthy volunteers.
Neuropsychopharmacology 2004;29208- 218
PubMedGoogle Scholar 44.Morgan
CJMofeez
ABradner
BBromley
LCurran
HV Ketamine impairs response inhibition and is positively reinforcing in healthy volunteers: a dose-response study.
Psychopharmacology (Berl) 2004;172298- 308
PubMedGoogle Scholar 45.Honey
GDHoney
RASharar
SRTurner
DCPomarol-Clotet
EKumaran
DSimons
JSHu
XRugg
MDBullmore
ETFletcher
PC Impairment of specific episodic memory processes by sub-psychotic doses of ketamine: the effects of levels of processing at encoding and of the subsequent retrieval task [published online ahead of print October 12, 2005].
Psychopharmacology 2005;181445- 457.
PubMed10.1007/s00213-005-0001-z
Google Scholar 46.Honey
GDO’Loughlin
CTurner
DCPomarol-Clotet
ECorlett
PRFletcher
PC The effects of a subpsychotic dose of ketamine on recognition and source memory for agency: implications for pharmacological modelling of core symptoms of schizophrenia [published online ahead of print August 3, 2005].
Neuropsychopharmacology 2006;31413- 423
PubMeddoi:10.1038/sj.npp.1300846
Google Scholar 47.Vollenweider
FXLeenders
KLScharfetter
CAntonini
AMaguire
PMissimer
JAngst
J Metabolic hyperfrontality and psychopathology in the ketamine model of psychosis using positron emission tomography (PET) and [18F]fluorodeoxyglucose (FDG).
Eur Neuropsychopharmacol 1997;79- 24
PubMedGoogle Scholar 48.Vollenweider
FXLeenders
KLOye
IHell
DAngst
J Differential psychopathology and patterns of cerebral glucose utilisation produced by (S)- and (R)-ketamine in healthy volunteers using positron emission tomography (PET).
Eur Neuropsychopharmacol 1997;725- 38
PubMedGoogle Scholar 49.Breier
AMalhotra
AKPinals
DAWeisenfeld
NIPickar
D Association of ketamine-induced psychosis with focal activation of the prefrontal cortex in healthy volunteers.
Am J Psychiatry 1997;154805- 811
PubMedGoogle Scholar 50.Breier
AAdler
CMWeisenfeld
NSu
TPElman
IPicken
LMalhotra
AKPickar
D Effects of NMDA antagonism on striatal dopamine release in healthy subjects: application of a novel PET approach.
Synapse 1998;29142- 147
PubMedGoogle Scholar 51.Holcomb
HHLahti
ACMedoff
DRWeiler
MTamminga
CA Sequential regional cerebral blood flow brain scans using PET with H2(15)O demonstrate ketamine actions in CNS dynamically.
Neuropsychopharmacology 2001;25165- 172
PubMedGoogle Scholar 52.Kegeles
LSMartinez
DKochan
LDHwang
DRHuang
YMalawi
OSuckow
RFVan Heertum
RLLaruelle
M NMDA antagonist effects on striatal dopamine release: positron emission tomography studies in humans.
Synapse 2002;4319- 29
PubMedGoogle Scholar 53.Abel
KMAllin
MPKucharska-Pietura
KDavid
ASAndrew
CWilliams
SBrammer
MJPhillips
ML Ketamine alters neural processing of facial emotion recognition in healthy men: an fMRI study.
Neuroreport 2003;14387- 391
PubMedGoogle Scholar 54.Abel
KMAllin
MPKucharska-Pietura
KAndrew
CWilliams
SDavid
ASPhillips
ML Ketamine and fMRI BOLD signal: distinguishing between effects mediated by change in blood flow versus change in cognitive state.
Hum Brain Mapp 2003;18135- 145
PubMedGoogle Scholar 55.Honey
RAHoney
GDO’Loughlin
CSharar
SRKumaran
DBullmore
ETMenon
DKDonovan
TLupson
VCBisbrowne-Chippendale
RFletcher
PC Acute ketamine administration alters the brain responses to executive demands in a verbal working memory task: an FMRI study.
Neuropsychopharmacology 2004;291203- 1214
PubMedGoogle Scholar 56.Northoff
GRichter
ABermpohl
FGrimm
SMartin
EMarcar
VLWahl
CHell
DBoeker
H NMDA hypofunction in the posterior cingulate as a model for schizophrenia: an exploratory ketamine administration study in fMRI.
Schizophr Res 2005;72235- 248
PubMedGoogle Scholar 57.Honey
GDHoney
RAO’Loughlin
CSharar
SRSuckling
JMenon
DKSleator
CBullmore
ETFletcher
PC Ketamine disrupts frontal and hippocampal contribution to encoding and retrieval of episodic memory: an fMRI study.
Cereb Cortex 2005;15749- 759
PubMedGoogle Scholar 58.Nelson
H National Adult Reading Test Manual. Windsor, United Kingdom nfer Nelson1982;
59.Domino
EFZsigmond
EKDomino
LEDomino
KEKothary
SPDomino
SE Plasma levels of ketamine and two of its metabolites in surgical patients using a gas chromatographic mass fragmentographic assay.
Anesth Analg 1982;6187- 92
PubMedGoogle Scholar 60.Wing
JCooper
JSartorius
N Measurement and Classification of Psychiatric Symptoms: An Instruction Manual for the PSE and CATEGO Program. New York, NY Cambridge University Press1974;
61.Bremner
JDKrystal
JHPutnam
FWSouthwick
SMMarmar
CCharney
DSMazure
CM Measurement of dissociative states with the Clinician-Administered Dissociative States Scale (CADSS).
J Trauma Stress 1998;11125- 136
PubMedGoogle Scholar 62.Ventura
JLukoff
DNuechterlein
KHLieberman
RPGreen
MShaner
A Appendix 1: brief Psychiatric Ratings Scale (BPRS) Expanded Version (4.0) scales, anchor points and administration manual.
Int J Methods Psychiatr Res 1993;3227- 243
Google Scholar 63.Friston
KJFletcher
PJosephs
OHolmes
ARugg
MDTurner
R Event-related fMRI: characterizing differential responses.
Neuroimage 1998;730- 40
PubMedGoogle Scholar 64.Genovese
CRLazar
NANichols
T Thresholding of statistical maps in functional neuroimaging using the false discovery rate.
Neuroimage 2002;15870- 878
PubMedGoogle Scholar 65.Knutson
BBjork
JMFong
GWHommer
DMattay
VSWeinberger
DR Amphetamine modulates human incentive processing.
Neuron 2004;43261- 269
PubMedGoogle Scholar 67.Zink
CFPagnoni
GMartin
MEDhamala
MBerns
GS Human striatal response to salient nonrewarding stimuli.
J Neurosci 2003;238092- 8097
PubMedGoogle Scholar 68.Aron
ARShohamy
DClark
JMyers
CGluck
MAPoldrack
RA Human midbrain sensitivity to cognitive feedback and uncertainty during classification learning.
J Neurophysiol 2004;921144- 1152
PubMedGoogle Scholar 69.Rodriguez
PFAron
ARPoldrack
RA Ventral-striatal/nucleus-accumbens sensitivity to prediction errors during classification learning [published online ahead of print August 9, 2005].
Hum Brain Mapp 2006;27306- 313.
PubMed10.1002/hbm.20186
Google Scholar 70.Huettel
SASong
AWMcCarthy
G Decisions under uncertainty: probabilistic context influences activity of prefrontal and parietal cortices.
J Neurosci 2005;253304- 3311
PubMedGoogle Scholar 71.Callicott
JHBertolino
AMattay
VSLangheim
FJDuyn
JCoppola
RGoldberg
TEWeinberger
DR Physiological dysfunction of the dorsolateral prefrontal cortex in schizophrenia revisited.
Cereb Cortex 2000;101078- 1082
PubMedGoogle Scholar 72.Umbricht
DKrljes
S Mismatch negativity in schizophrenia: a meta-analysis.
Schizophr Res 2005;761- 23
PubMedGoogle Scholar 73.Lavin
ANogueira
LLapish
CCWightman
RMPhillips
PESeamans
JK Mesocortical dopamine neurons operate in distinct temporal domains using multimodal signalling.
J Neurosci 2005;255013- 5023
PubMedGoogle Scholar 74.Fletcher
PC Functional neuroimaging of psychiatric disorders: exploring hidden behaviour.
Psychol Med 2004;34577- 581
PubMedGoogle Scholar