Psychological Medicine, 2001, 31, 189–195. Printed in the United Kingdom
" 2001 Cambridge University Press
THEORETICAL PAPER
A cognitive model of the positive symptoms of psychosis
P. A. GARETY,
"
E. KUIPERS, D. FOWLER, D. FREEMAN P. E. BEBBINGTON
From the Academic Department of Psychiatry and Psychology, Guy’s, King’s and St Thomas’s Medical
School, King’s College London, Department of Psychology, Institute of Psychiatry and Royal Free and
University College Medical School, London; and University of East Anglia, Norwich
INTRODUCTION
In the last 10 years a consensus has developed
that the symptoms of psychosis may be better
understood by linking the steps between the
phenomenological experiences and social,
psychological and neurobiological levels of
explanation. Cognitive models of psychosis are
an important link in this chain. They provide a
psychological description of the phenomena
from which hypotheses concerning causal pro-
cesses can be derived and tested ; social, in-
dividual, and neurobiological factors can then
be integrated via their impact on these cognitive
processes. In this paper, we set out the cognitive
processes that we think lead to the formation
and maintenance of the positive symptoms of
psychosis and we attempt to integrate into our
model research in social factors. If this model
proves useful, a fuller integration with the
findings of biological research will be required
(Frith, 1992).
Our cognitive model builds on the work of
other researchers (e.g. Maher, 1988; Frith, 1992;
Hemsley, 1993; Bentall et al. 1994; Chadwick &
Birchwood, 1994; Morrison et al. 1995) and our
own clinical and theoretical studies. The model
is new in that it incorporates both disruptions in
automatic cognitive processes and maladaptive
conscious appraisals; it covers delusions and
hallucinations in one framework; it posits a
central role for emotion; and it considers how
social factors may contribute to the origins,
"
Address for correspondence: Professor P. A. Garety, Academic
Department of Psychiatry and Psychology, Guy’s, King’s and St
Thomas’s Medical School, King’s College London, Adamson Centre
for Mental Health, St Thomas’s Hospital, Lambeth Palace Road,
London SE1 7EH.
maintenance or recurrence of symptoms. The
resulting cognitive model of psychosis is con-
sistent with existing theoretical and therapeutic
studies. It generates testable hypotheses and
should lead to theoretical and therapeutic
advances.
A cognitive model of psychosis
There is widespread agreement that psychosis
occurs in people with a vulnerable predisposition
(of biopsychosocial origin); that onset often
follows life events, adverse environments, illicit
drug use, or periods of isolation; that there are
emotional changes, and disruptions in cognitive
processes of attention, perception, or judgement ;
and that, at onset, its most prominent symptoms
are delusional beliefs and hallucinations (Garety
et al. 2000).
We postulate two proximal routes to the
development of the positive symptoms of psy-
chosis: one proceeds through cognitive and
affective changes ; the other proceeds through
affective disturbance alone. We think the first
route is the more common : a triggering event
gives rise in a predisposed person to a disruption
of cognitive processes (Garety & Hemsley, 1994).
The basic automatic cognitive disturbance may
be conceptualized in two ways. First, it may be
viewed as a ‘weakening of the influences of
stored memories of regularities of previous input
on current perception ’, which leads to am-
biguous, unstructured sensory input and the
subsequent intrusion into consciousness of unin-
tended material from memory (Hemsley, 1993).
Secondly, the basic cognitive dysfunction may
be recently developed difficulties with the self-
189
190 P. A. Garety and others
monitoring of intentions and actions, which
would lead to individuals’ own intentions to act
not being recognized and therefore being ex-
perienced as alien (Frith, 1992). (There are
similarities between these accounts, as Hemsley
(1998) has noted. Both imply a breakdown in
‘willed intention’ activity, particularly in the
later stages of the disorder, albeit for different
reasons.) What is important for our model is
that at onset both accounts emphasize that the
basic cognitive disturbance leads to anomalous
conscious experiences (e.g. heightened percep-
tion, actions experienced as unintended, racing
thoughts, thoughts appearing to be broadcast,
thoughts experienced as voices, two unconnected
events appearing to be causally linked). Many
sufferers report of these cognitive and perceptual
changes of the psychotic prodrome that they are
experienced as unfamiliar and not like ordinary
self-generated cognitions – they feel external and
potentially threatening. At this point, however,
these experiences have not been transformed
into psychotic symptoms. Emotional changes
also occur, in direct response to the triggering
event and in response to the anomalous
experiences. The generation of arousal is also
implicit in the model of cognitive disturbance
proposed by Hemsley (1993). Such emotional
changes feed back into the moment-by-moment
processing of anomalous experiences and in-
fluence their content. For example, if anxiety
and depression result from a job loss, and
further anxiety directly from the experience of
voices, the person’s voices may develop a
threatening and critical content: ‘You’re useless,
you won’t get another job now. We’re after you,
we’ve got you marked’.
Furthermore, the anomalous experiences, be-
ing puzzling and associated with emotional
changes, seem personally significant and trigger
a search for explanation as to their cause (Maher,
1988). Here biased conscious appraisal processes
are crucial: they contribute to a judgement that
these confusing experiences (which feel external
in any case) are in fact externally caused. Garety
& Freeman (1999) have reviewed the evidence
for biases in cognitive processes and found
empirical support for an information gathering
cognitive style characterized by jumping to
conclusions, externalizing attributional biases,
and deficits in understanding social situations
and the intentions of others. It is likely that these
biased appraisal processes are made worse by
negative emotional states (e.g. anxiety,
depression, anger).
These immediate processes occur against a
conducive social–cognitive background. Inner
city birth and rearing appear to be major risk
factors for psychosis (Mortensen et al. 1999),
and there are also variations in incidence of
psychosis in ethnic groups. Bhugra and his
colleagues (1997) in particular have argued that
these factors are linked to social adversity and
deprivation. We suggest that earlier adverse
experience, such as social marginalization, child-
hood loss (Agid et al. 1999), or severe childhood
trauma, may create an enduring cognitive
vulnerability, characterized by negative sche-
matic models of the self and the world (e.g.
beliefs about the self as vulnerable to threat, or
about others as dangerous) that facilitate ex-
ternal attributions and low self-esteem.
This would be one plausible reason for the
finding of Myhrman et al. (1996) from the 1966
Finland birth cohort that unwanted pregnancies
resulted in children with twice the risk of
developing schizophrenia. Evidence consistent
with a role for negative schemas in the de-
velopment of psychosis has recently been
reported by Van Os (2000). In a large epidemio-
logical study in the Netherlands over 7000
people were screened for symptoms and psy-
chiatric status and followed up for 3 years.
Those who subsequently developed psychosis
were found more likely to have low self-esteem
and depressive schemas. Furthermore,
Birchwood et al. (2000), in developing a model
of auditory hallucinations, have suggested that a
childhood experience of social adversity leads to
the development of negative schemas involving
social humiliation and subordination, which in
turn fuel voices and paranoia.
These pre-existing negative schemas also
provide content to the psychotic attribution
(Bowins & Shugar, 1998 ; Fowler et al. 1998).
Thus, a person with religious beliefs about
innate wickedness concludes that the external
threat is caused by a punishing God. Some
triggering events may have attributes that render
externalizing appraisals particularly compelling;
we have pilot data linking life events categorized
as intrusive to first episodes of psychosis
characterized by persecutory themes (D. Raune,
personal communication). Finally, we suggest
Cognitive model of psychosis 191
that social isolation contributes to the accept-
ance of the psychotic appraisal by reducing
access to alternative more normalizing expla-
nations (White et al. 2000).
We argue that people with anomalous quasi-
psychotic experiences do not develop full-blown
psychotic symptoms if they are able to reject the
hypothesis of externality, leading to a protective
self-correcting decision, e.g. ‘I thought I was
hearing the voice of God, but more likely my
mind is playing tricks’; ‘ Things look different,
somehow, I must be stressed with all that’s
going on’. It is thus quite possible to have
hallucinatory experiences without becoming de-
luded, as has been reported (Peters et al. 1999).
The externalizing appraisal is thus a defining
decision. Psychosis is recognized as occurring
when the individual appraises experiences as
externally caused and personally significant.
Such appraisals are formally identified as
delusions and hallucinations : e.g. ‘I am being
poisoned’; ‘God is giving me special powers ’;
‘my voices are coming from persecutors who
want to kill me ’ ; ‘a transmitter is beaming my
thoughts worldwide’.
In a small proportion of cases (e.g. some cases
of delusional disorder), it appears there is a
second route to psychosis: the triggering event
does not appear to cause a basic information
processing disruption, leading to anomalous
experiences. Here, life events trigger only dis-
turbed affect, which in turn directly activates
biased appraisal processes and maladaptive
self\other schemas leading to an externalizing
appraisal (i.e. the delusion) for the life event or
the disturbed affect. In such cases, delusions
occur independently of hallucinations and other
psychotic symptoms.
Central to our model are the factors re-
sponsible for the maintenance\recurrence of the
psychotic appraisal. Why does it not correct
itself when the evidence for it is not forthcoming?
We hypothesize that a number of different
factors maintain the psychotic appraisal, as
follows.
1 Reasoning processes
The biased cognitive processes we think con-
tribute to symptom formation are also likely to
contribute to symptom maintenance, since these
biases have been found in people with persistent
symptoms, but may not be present after recovery
(Garety & Freeman, 1999). These are : a
‘jumping to conclusions’ data gathering bias, an
externalizing attributional style, and poor social
understanding or theory of mind. This abnormal
reasoning may in turn be maintained by social
isolation. We would also highlight two other
factors that have received less experimental
attention. First, we have obtained evidence that
a lack of belief flexibility (the willingness to
consider alternatives to delusional beliefs) is
associated with poorer outcome, independent of
the severity of the delusion (Garety et al. 1997).
Secondly, the normal belief confirmation bias is
likely to maintain psychotic beliefs (Maher,
1974). These two factors suggest that delusions
are more likely to be maintained in individuals
who have a dichotomous thinking style and who
cannot tolerate ambiguity.
2 Dysfunctional schemas and adverse social
environments
There is a growing literature on the poor self-
concept and self-esteem of people with psychosis
(e.g. Trower & Chadwick, 1995; Kinderman &
Bentall, 1996). Freeman et al. (1998) found that
self-esteem was poor in many people with
psychosis. Close & Garety (1998) found that
hallucinations and delusions that have negative
content are associated with negative self-
concepts. Psychotic beliefs may be more firmly
held if they are consistent with firmly-held
distorted beliefs about the self (e.g. that one is
bad), others (e.g. that others are hostile) and the
world (e.g. the world is dangerous). Moreover,
once formed, the delusion is likely to be
considered as further confirmation of the nega-
tive beliefs, leading to further strengthening of
the delusion.
Low self-esteem almost certainly develops in
specific social contexts (Brown et al. 1990;
Harris et al. 1990). Aversive social environments,
such as living with high expressed emotion
families, are a robust predictor of poor outcome
(Bebbington & Kuipers, 1994 ; Butzlaff &
Hooley, 1998). Mueser et al. (1998) report that
the lifetime prevalence of traumatic events
among a sample of 275 patients with schizo-
phrenia and bipolar disorder was 98%. We have
preliminary evidence, from a consecutive series
of 77 first episode in-patients with psychosis,
that severe trauma histories are more common
in those with symptoms unresponsive to medi-
192 P. A. Garety and others
cation than in those whose symptoms were
responsive (Fowler, 1999) while Doering and
colleagues (1998) have found that traumatic
experiences and adverse circumstances in child-
hood were related to relapse and re-
hospitalization in schizophrenia. Furthermore,
as discussed earlier, differences in the incidence
of psychosis in the inner city and in certain
ethnic groups has been linked to social adversity.
It seems likely that social marginalization,
difficult or traumatic experiences or un-
supportive family environments contribute to
the development of negative schemas. We
speculate from this literature that early trauma
and chronic stress can create dysfunctional
negative schemas, which in turn contribute to
treatment resistance and a vulnerability to
relapse.
3 Emotion (e.g. anxiety, depression, anger,
mania) and cognitive processes associated
with emotion
Clearly, dysfunctional negative schemas will be
closely associated with levels of emotional
distress. The presence of emotional distress will
also contribute to the maintenance of the
psychotic appraisal through other processes.
Birchwood and colleagues have demonstrated
the importance of depression in psychosis (e.g.
see Birchwood & Iqbal, 1998), They report that
residual symptoms of both hallucinations and
delusions are more common in depressed people
with psychosis and propose that feelings of
hopelessness and uncontrollability contribute to
symptom maintenance. Consistent with this,
hopelessness is a predictor of poor outcome in
early schizophrenia (Aguilar et al. 1997). We
ourselves have focused on anxiety. We argue
that three processes traditionally associated with
anxiety disorders may be of particular import-
ance. Information processing biases will provide
evidence or substantiation for psychotic beliefs
and hence maintain them (e.g. deployment of
attention; Freeman et al. 2000). Safety-
behaviours will prevent the receipt of discon-
firmatory evidence and hence prevent change in
psychotic beliefs (Freeman & Garety, 2000).
Meta-cognitive beliefs, such as beliefs concerning
the uncontrollability of one’s thoughts, will
increase the distress caused by psychotic
experiences (Freeman & Garety, 1999).
Clinically, it has also long been observed that
anxiety triggers hallucinations and increases in
delusional thoughts, i.e. affects cognitive pro-
cessing (Slade, 1972). Finally, the experience of
emotion will drive a search for meaning and
understanding that is consistent with affect-
associated beliefs: for example, anxiety will
increase the probability that a threatening
explanation is sought and accepted.
4 The secondary appraisal
The secondary appraisal of the experience of
psychosis itself (‘illness perception’ or insight).
We suggest appraisals of illness influence en-
gagement with treatment and adaptive behav-
iour. Standard assessments of insight are only
modestly correlated with outcome (David, 1998).
However, insight is correlated with the ‘belief
maintenance’ subscale of the Maudsley As-
sessment of Delusions, and improved outcome
for delusions in response to CBT is associated
with and predicted by changes in the MADS
subscale (Garety et al. 1997). Also important are
appraisals of the experience of chronic mental
illness as stigmatizing and humiliating,
appraisals which may influence the development
of depression (Birchwood & Iqbal, 1998). These
appraisals probably have some basis in reality,
given evidence that significant others readily
apply negative labels to people developing first
episodes of schizophrenia (Bean et al. 1996).
The cognitive model and psychological
treatments
We think theory and therapy in this area are
mutually enhancing. Therapy provides a context
for theoretical developments and the empirical
investigation of our model, which in turn should
clarify the targets of psychological treatment,
refine its techniques, and enhance its efficacy.
Two sorts of psychological intervention seem
to be effective in diminishing positive symptoms
and their re-emergence. These are cognitive
behaviour therapy (CBT) and interventions with
the families of people with psychosis (FI)
(Haddock et al. 1998 ; Birchwood & Spencer,
1999; Kuipers et al. 1999). It is unlikely that
these work in the same way.
We hypothesize that CBT acts directly on
psychological processes (this is after all its aim).
In particular, it is most effective in treating
psychosis when the key appraisal, of inner
Cognitive model of psychosis 193
mental disturbance as externally caused, is re-
appraised as inner. In traditional terminology,
‘good insight’ is developed. This new schema is
incompatible with its psychotic predecessor.
Both symptoms and the risk of relapse are
thereby reduced. Therapy achieves this by
changing the appraisals and, where possible
negative self-schemata, and by compensating for
the biased reasoning processes (Fowler et al.
1995). We would also propose that clinical
improvements and reduced risks of relapse can
still occur when the externalizing psychotic
appraisal remains unaltered (i.e. insight remains
‘poor’), provided some of the hypothesized
maintenance factors are changed (e.g. safety
behaviours), disrupting the vicious cycle.
Family intervention (FI), in contrast, pri-
marily targets the behaviour of family members.
It improves social function (Barrowclough &
Tarrier, 1990) and reduces exacerbations of
positive symptoms (Pharoah et al. 1999; Kuipers
et al. 1999). We offer competing hypotheses
about how it affects the cognitive processing of
sufferers.
One possibility is that it operates by reducing
environmental stress and improving the
associated affect (anxiety or depression). Anom-
alous experiences would be improved mainly by
way of these affective changes. We know that
conflictual relationships predict poor outcome
in social and psychiatric patholology, and that
affirmative ones can improve outcome even in
those with persistent emotional damage (Lewis,
1998). People with schizophrenia are well able to
perceive criticism in their relatives (Tompson et
al. 1995; Scazufca et al. 2000). Such relationships
must also affect mood : while critical or intrusive
behaviour might increase anxiety and de-
pression, supportive relationships would reduce
them.
It is of interest that supportive counselling
and befriending had some effect in reducing
symptoms in two recent trials of CBT in
schizophrenia (Tarrier et al. 1998; Sensky et al.
2000). However, the benefits were limited to the
duration of the intervention. There is now
evidence from a meta-analysis of 19 randomized
controlled trials (reported in Kuipers et al. 1999)
that FI also needs to be continued for positive
effects to be maintained. This analysis found
that for single family therapy, the number needed
to treat (NNT) to prevent relapse in the first
year of treatment was 6n3; in the second year of
treatment this fell to 3n9. Once treatment ended
the NNT went back up to 7n1 to prevent relapse,
and 20n8 to prevent readmission. Thus, although
individual studies have found that treatment
effects can continue for as long as 8 years after
FI finishes (e.g. Tarrier et al. 1994) this was not
confirmed by the meta-analysis. This suggests
that both FI and supportive therapies have
effects which continue only for as long as therapy
is offered. In contrast, evidence is emerging that
changes from CBT may continue (and may even
be enhanced) when therapy finishes (Kuipers et
al. 1998; Tarrier et al. 1999; Sensky et al. 2000).
In families (and possibly also the supportive
therapies), it seems likely that therapeutic im-
provement occurs via the reduction of negative
affect through the provision of a less stressful
environment, and the improvement of social
functioning by appropriate prompting and re-
inforcement of small changes in behaviour.
There might also be improvements in thinking
processes (e.g. relative to negative schemas or
cognitive flexibility), but these would be sec-
ondary to the other changes. Thus, under this
hypothesis, the cognitive changes brought about
by FI would be less specific than with CBT and
the maintenance of effects would be less secure.
The contrasting hypothesis is that FI helps
family members to discuss directly with sufferers
their psychotic experiences and to provide
alternative explanations. FI would then operate
through cognitive change like CBT. No evidence
currently exists to test between these hypotheses
in terms of change in individual affect, schemas
or appraisals. The second hypothesis requires
that carers effectively perform as therapists, but
it seems unlikely that most would have the
specific skills to do this without training.
Conclusion
We have put forward a new cognitive model of
the positive symptoms of psychosis. While our
model is speculative, we think it has plausibility
in the context of what we currently know about
the various processes involved in delusional
thinking and anomalous experience. Moreover,
it integrates much of this information in a way
that will contribute to the further expansion of
knowledge, and may have important practical
benefits for people suffering from psychosis.