Set shifting and reversal learning in patients
with bipolar disorder or schizophrenia
J. McKirdy, J. E. D. Sussmann, J. Hall, S. M. Lawrie, E. C. Johnstone and A. M. McIntosh*
Division of Psychiatry, University of Edinburgh, Royal Edinburgh Hospital, Edinburgh, UK
Background. Bipolar disorder and schizophrenia have both been associated with deficits in extra-dimensional set
shifting (EDS). Deficits in reversal learning (RL) have also been shown in schizophrenia but not in bipolar disorder.
This study sought to assess the specificity of these findings in a direct comparison of clinically stable patients with
Method. Theintra-dimensional/extra-dimensional (IDED)set-shiftingtask,partof theCambridgeNeuropsychological
Test Automated Battery (CANTAB), was administered to 30 patients with schizophrenia, 47 with bipolar disorder
and a group of 44 unaffected controls. EDS and RL errors were compared between the groups and related to
measures of current and past psychiatric symptoms and medication.
Results. Both groups of patients with schizophrenia or bipolar disorder made more EDS and RL errors than controls.
Neither measure separated the two disorders, even when the analysis was restricted to euthymic patients. No
relationship was found with prescribed medication.
Conclusions. Patients with bipolar disorder or schizophrenia show common deficits in EDS and RL. These deficits
do not seem to be attributable to current symptoms and are consistent with disrupted networks involving the ventral
Received 15 November 2007; Revised 4 November 2008; Accepted 14 November 2008; First published online 24 December 2008
Key words: Bipolar disorder, cohort study, neuropsychology, schizophrenia.
Executive impairments using the Wisconsin Card Sort
Test (WCST) are one of the most frequently replicated
findings in people with schizophrenia (Nieuwenstein
et al. 2001) and have been shown to persist across
depressed, manic and euthymic phases in bipolar
disorder (Martinez-Aran et al. 2004). The WCST con-
sists of a series of rule changes, involving the learning
of new stimulus–reward associations based on differ-
ent types of stimuli [extra-dimensional set shifting
(EDS)] and shifting attention to a new but previously
irrelevant stimulus [reversal learning (RL)]. The intra-
dimensional/extra-dimensional (IDED) set-shifting
task, part of the Cambridge Neuropsychological Test
Automated Battery (CANTAB; Cambridge Cognition,
Cambridge, UK), has been developed to test each of
these component processes separately. Subsequent
work suggests that these processes are associated with
dissociable patterns of neural activity involving the
ventrolateral prefrontal cortex (VLPFC) and orbito-
frontal cortex (OFC) (Hampshire & Owen, 2006).
Schizophrenia and bipolar disorder have been pre-
viously investigated using the IDED task (Clark et al.
2001, 2002; Clark & Goodwin, 2004; Waltz & Gold,
2007). Patients with schizophrenia show impairments
in both EDS and RL whereas patients with remitted
bipolar disorder have shown deficits in EDS only. The
presence of RL errors in a single study of manic
patients suggests that, where present, these errors are
secondary to residual symptoms (Clark et al. 2001).
As these tasks involve potentially separable neural
systems, the specificity of the findings could have im-
portant implications for the aetiology of each disorder.
As no direct comparison of these disorders has been
made to date, we sought to assess this issue in a study
of patients with schizophrenia, bipolar disorder and
Individuals with bipolar I disorder or schizophrenia
were identified from the case loads of consultant
* Address for correspondence: Dr A. M. McIntosh, Division of
Psychiatry, University of Edinburgh, Kennedy Tower, Royal
Edinburgh Hospital, Edinburgh EH10 5HF, UK.
Psychological Medicine (2009), 39, 1289–1293.
f 2008 Cambridge University Press
Printed in the United Kingdom
psychiatrists across Edinburgh and the Lothians. All
patients had at least one affected first- or second-
degree family member with the same diagnosis. After
informed consent had been given, case-note diagnoses
of bipolar I disorder were established using the
Operational Criteria (OPCRIT) symptom checklist
(McGuffin et al. 1991) and confirmed at face-to-face
interview using the Structured Clinical Interview for
DSM-IV (SCID). Unaffected controls were identified
from the same regions and communities as the
patients themselves and their diagnostic status was
also confirmed using the SCID. To reduce selection
bias, we preferentially chose controls from the non-
genetic relatives and social networks of the patients
themselves. All study procedures were approved by
the Local Research Ethics Committee.
All participants were rated by a trained psychiatrist
(A.M.) using the Young Mania Rating Scale (YMRS),
the Hamilton Depression Rating Scale (HAMD) and
the Positive and Negative Symptoms Scale (PANSS).
Patients with a score of f8 on both the YMRS and the
HAMD were considered euthymic, in accordance
with previous papers using the IDED. Neuropsycho-
logical testing took place on the day of the clinical
All participants were first assessed using the National
Adult Reading Test (NART) as a measure of pre-
morbid intellectual function. The IDED task from the
CANTAB was then administered to all participants.
The task consists of nine stages, including four reversal
stages (simple discrimination reversal, compound dis-
crimination reversal, intra-dimensional reversal, and
extra-dimensional reversal) and an EDS stage. At each
stage, two sets of stimuli are presented, and subjects
must acquire a stimulus–reward association. After six
consecutive correct responses, the task proceeds to the
next stage. Should the participant be unable to reach
this standard after 50 trials, on any stage, the test
comes to an end. Reversal stages involve a reversal
of contingencies, whereby the previously rewarded
stimulus becomes irrelevant and the previously irrel-
evant stimulus is rewarded. Extra-dimension shifts
involve shifting attention from one type of stimuli to
another (e.g. from a solid shape to a line).
The total number of reversal errors was analysed
from the first three reversal stages combined (RL
errors). The results of the final (fourth) reversal stage
were not used because it is preceded by the EDS
phase, which several people failed (see Fig. 1, the
extra-dimensional stage of the task). EDS errors were
also used in the subsequent analysis.
Total RL and total EDS errors were the primary
measures of interest and were compared between
groups. NART-estimated full-scale IQ was included as
a covariate to ensure that the differences could not be
simply attributed to differences in general intellectual
ability. Euthymic patients with bipolar disorder were
then compared to controls and to patients with schizo-
phrenia to ensure that the results were not driven by
residual affective symptoms in the bipolar patients.
SDSRCD CD2 CDRIDS IDREDSEDR
Fig. 1. Graph showing the percentage of subjects in each group passing subsequent stages of the intra-dimensional/
extra-dimensional (IDED) set-shifting task. CTR, control group (--:--); BPD, bipolar disorder group (––%––);
SCZ, schizophrenia group (—n—); SD, simple discrimination; SR, simple reversal; CD, compound discrimination 1;
CD2, compound discrimination 2; CDR, compound discrimination reversal; IDS, intra-dimensional shift;
IDR, intra-dimensional reversal; EDS, extra-dimensional shift; EDR, extra-dimensional reversal.
1290 J. McKirdy et al.
The association between EDS and RL transformed er-
ror scores and rating scale measures of current psy-
chotic and affected symptoms was examined using
Pearson’s bivariate correlation coefficient in each
group. Pairwise group contrasts were conducted using
Tukey’s HSD test. Finally, the relationship of EDS and
RL error scores to medication was addressed compre-
hensively by (a) comparing individuals prescribed
and not prescribed each medication class (lithium,
antidepressants, antipsychotics) within each patient
group, (b) comparing individuals prescribed second-
versus first-generation antipsychotics and (c) examin-
ing the relationship between antipsychotic dose and
error scores within each group.
The distribution of residuals from each analysis was
inspected at each stage to ensure that it conformed to
an approximately normal distribution. Where this
assumption was not met, transformations were ap-
plied to the data until a suitable approximation could
be found. Subsequently, the negative inverse of RL
errors and the square root of EDS errors were used in
the analyses as their residuals most closely resembled
a normal distribution.
Forty-seven bipolar I patients, 30 schizophrenia
patients and 44 control participants completed the
IDED subtest of the CANTAB and the NART (Table 1).
The groups were matched closely on age, gender
and pre-morbid IQ. NART IQ was imputed for five
missing values using the appropriate mean for the
A significant difference in RL errors was found be-
tween the groups [F(2,118)=6.68, p<0.05]. Pairwise
comparisons showed significant differences between
control and schizophrenia participants (p<0.05), and
between control and bipolar participants (p<0.05). A
similar pattern was also found for EDS errors [overall
F test: F(2,117)=4.21, p<0.05]. Pairwise comparisons
demonstrated a significant difference between control
and schizophrenia participants (p<0.05) and between
control and bipolar participants (p<0.05). There was
no significant difference between schizophrenia and
bipolar participants for either RL or EDS errors and
neither measure was associated with rating scale
measures of current psychotic or affective symptoms.
Significant between-group differences in EDS and
RL errors remained when the bipolar group was
reduced to the 40 (40/47) individuals meeting criteria
for euthymia (controlling for IQ, EDS errors: F=5.93,
p=0.004; reversal errors: F=7.25, p=0.001). Further-
more, when euthymic and non-euthymic bipolar
subjects were compared directly, there were no sig-
nificant differences in either variable (EDS errors:
t=0.62, p=0.54; reversal errors: t=–1.03, p=0.31).
Table 1. Descriptive statistics and IDED results for each group
IDED, Intra-dimensional/extra-dimensional; NART, National Adult Reading
Test; PANSS, Positive and Negative Symptoms Scale; YMRS, Young Mania Rating
Scale; HAMD, Hamilton Depression Rating Scale; GAF, Global Assessment
of Functioning; SGA, second-generation ‘atypical’ antipsychotics; FGA,
first-generation antipsychotics; ‘both’ refers to subjects regularly prescribed both
SGA and FGA; CPZ, chlorpromazine; EDS, extra-dimensional set-shifting;
RL, reversal learning; N.A., not applicable.
Impaired set shifting in remitted bipolar disorder1291
Within the bipolar group alone, EDS and RL errors
showed no relationship with residual affective or
psychotic symptoms and no differences were found
in either measure when bipolar patients with and
without a past history of psychotic symptoms were
compared (32 previously psychotic versus 15 non-
Finally, no relationship was found between either
measure of set shifting and the prescription of any
class of medication (lithium, antidepressants, anti-
psychotics). Furthermore, no differences were found
between individuals prescribed first- and second-
generation agents, and no relationship was found with
antipsychotic dose, defined in antipsychotic equiv-
alents, within either group.
Patients with schizophrenia or bipolar disorder show
common deficits in both EDS and RL. These deficits
could not be accounted for by differences in either
IQ or residual symptoms. Furthermore, they did not
differentiate bipolar patients with and without psy-
chotic symptoms and remained significant when the
bipolar group was restricted to only euthymic in-
RL and EDS have been shown in both animal
models and human functional magnetic resonance
imaging (fMRI) experiments to involve dissociable
neural systems (Dias et al. 1996; Hampshire & Owen,
2006). In humans, RL is associated with activity in the
OFC and ventral striatum, whereas EDS is associated
with activity in the VLPFC (Hampshire & Owen,
2006).The common neuropsychological
shown in the current study are also supported by
functional imaging studies showing involvement of
these regions across both conditions (Lawrence et al.
2004; Ragland et al. 2004; Malhi et al. 2005; Schneider
et al. 2007). Further work is required to clarify whether
the underlying deficits are secondary to structural
and/or neurochemical abnormalities and the role of
medication and duration of illness.
Previous studies have demonstrated deficits in EDS
but not RL in patients with bipolar disorder (Clark
et al. 2002), and one study found no differences in EDS
errors (Sweeney et al. 2000). Although our findings
initially seem to be at odds with this earlier work, bi-
polar patients in the study by Clark et al. (2002) also
showed a greater number of RL errors, albeit to a non-
significant degree. The observed number of EDS errors
in the study by Sweeney et al. (2000) were also in the
same direction as our results (mean=4.4 EDS errors in
controls, mean=5 to 8 in symptomatic bipolar sub-
jects), but did not meet statistical significance. The
apparent inconsistency may be due to differences in
statistical power or the clinical characteristics of bi-
polar subjects (e.g. family history, medication or
severity of illness). The current study broadly re-
plicates previous studies of individuals with schizo-
phrenia (Waltz & Gold, 2007) and children with
bipolar disorder (Gorrindo et al. 2005; Dickstein et al.
2007), but significantly extends these findings by con-
firming that they are present in adults with bipolar
disorder and are not diagnostically specific. The
current investigation provides further evidence of
common neuropsychological impairments in both
disorders (McIntosh et al. 2005). Unlike the majority
of neuropsychological tasks, however, each task com-
ponent can be dissociated anatomically and neuro-
chemically. This task may therefore provide a means
of identifying the precise neural mechanisms common
to each disorder.
Several limitations to the current work should be
considered. First, the study is not sufficiently large to
exclude a differential impairment between patients
with schizophrenia and those with bipolar disorder.
Nevertheless, it is currently one of the largest using the
IDED task, and the only study to compare these dis-
orders directly. Second, the profile of prescribed psy-
chotropic medication was clearly different between
each patient group and may have confounded the re-
sults. Although it is not possible to rule out an effect of
medication, no relationship of error score with drug
class, second- versus first-generation antipsychotic
medication or dose equivalence was found in the cur-
rent study. Third, although common impairments
suggest common neural mechanisms, this hypothesis
cannot be confirmed without functional imaging.
Studies of these disorders using existing tasks based
on the IDED may provide further fruitful avenues for
Here, in a relatively large sample of people with
schizophrenia and bipolar disorder, we have shown
that patients with both disorders show common defi-
cits in RL and EDS that transcend conventional diag-
nostic boundaries. These findings may provide a basis
for investigating the functional correlates of psychosis
common to both conditions.
We thank all of the participants without whom this
study would not have been possible. This study and
two of the investigators (A.M.M. and S.M.L.) were
supported by the Dr Mortimer and Theresa Sackler
Foundation during the period of data collection.
Dr Hall is supported by the Medical Research Council
and Dr McIntosh is supported by the Health
1292 J. McKirdy et al.
Declaration of Interest
Clark L, Goodwin GM (2004). State- and trait-related deficits
in sustained attention in bipolar disorder. European
Archives of Psychiatry and Clinical Neuroscience 254, 61–68.
Clark L, Iversen SD, Goodwin GM (2001). A
neuropsychological investigation of prefrontal cortex
involvement in acute mania. American Journal of Psychiatry
Clark L, Iversen SD, Goodwin GM (2002). Sustained
attention deficit in bipolar disorder. British Journal of
Psychiatry 180, 313–319.
Dias R, Robbins TW, Roberts AC (1996). Dissociation
in prefrontal cortex of affective and attentional shifts.
Nature 380, 69–72.
Dickstein DP, Nelson EE, McClure EB, Grimley ME,
Knopf L, Brotman MA, Rich BA, Pine DS, Leibenluft E
(2007). Cognitive flexibility in phenotypes of pediatric
bipolar disorder. Journal of the American Academy of Child
and Adolescent Psychiatry 46, 341–355.
Gorrindo T, Blair RJ, Budhani S, Dickstein DP, Pine DS,
Leibenluft E (2005). Deficits on a probabilistic
response-reversal task in patients with pediatric bipolar
disorder. American Journal of Psychiatry 162, 1975–1977.
Hampshire A, Owen AM (2006). Fractionating attentional
control using event-related fMRI. Cerebral Cortex 16,
Lawrence NS, Williams AM, Surguladze S, Giampietro V,
Brammer MJ, Andrew C, Frangou S, Ecker C, Phillips ML
(2004). Subcortical and ventral prefrontal cortical neural
responses to facial expressions distinguish patients with
bipolar disorder and major depression. Biological Psychiatry
Malhi GS, Lagopoulos J, Sachdev PS, Ivanovski B, Shnier R
(2005). An emotional Stroop functional MRI study of
euthymic bipolar disorder. Bipolar Disorders 7 (Suppl. 5),
Martinez-Aran A, Vieta E, Reinares M, Colom F, Torrent C,
Sanchez-Moreno J, Benabarre A, Goikolea JM, Comes M,
Salamero M (2004). Cognitive function across manic
or hypomanic, depressed and euthymic states in
bipolar disorder. American Journal of Psychiatry 161,
McGuffin P, Farmer A, Harvey I (1991). A polydiagnostic
application of operational criteria in studies of psychotic
illness. Development and reliability of the OPCRIT system.
Archives of General Psychiatry 48, 764–770.
McIntosh AM, Harrison LK, Forrester K, Lawrie SM,
Johnstone EC (2005). Neuropsychological impairments
in people with schizophrenia or bipolar disorder and
their unaffected relatives. British Journal of Psychiatry 186,
Nieuwenstein MR, Aleman A, de Haan EHF (2001).
Relationship between symptom dimensions and
neurocognitive functioning in schizophrenia: a
meta-analysis of WCST and CPT studies. Journal of
Psychiatric Research 35, 119–125.
Ragland JD, Gur RC, Valdez J, Turetsky BI, Elliott M,
Kohler C, Siegel S, Kanes S, Gur RE (2004). Event-related
fMRI of frontotemporal activity during word encoding and
recognition in schizophrenia. American Journal of Psychiatry
Schneider F, Habel U, Reske M, Kellermann T, Stocker T,
Shah NJ, Zilles K, Braus DF, Schmitt A, Schlosser R,
Wagner M, Frommann I, Kircher T, Rapp A,
Meisenzahl E, Ufer S, Ruhrmann S, Thienel R, Sauer H,
Henn FA, Gaebel W (2007). Neural correlates of working
memory dysfunction in first-episode schizophrenia
patients: an fMRI multi-center study. Schizophrenia Research
Sweeney JA, Kmiec JA, Kupfer D (2000). Neuropsychologic
impairments in bipolar and unipolar mood disorders on
the CANTAB neurocognitive battery. Biological Psychiatry
Waltz JA, Gold JM (2007). Probabilistic reversal learning
impairments in schizophrenia: further evidence of
orbitofrontal dysfunction. Schizophrenia Research 93,
Impaired set shifting in remitted bipolar disorder 1293