Psychological Medicine, 1997, 27, 565–578. Copyright # 1997 Cambridge University Press
The interaction between mood and cognitive function
studied with PET
S. C. BAKER, C. D. FRITH R. J. DOLAN
"
From the Wellcome Department of Cognitive Neurology, Institute of Neurology; MRC Cyclotron Unit,
Hammersmith Hospital; Royal Free Hospital School of Medicine; National Hospital for Neurology and
Neurology and Neurosurgery; and University College London
ABSTRACT
Background. Experimentally induced depressed mood is a suggested model for retarded depres-
sion. We describe the neural response associated with induced mood and the locus of the interaction
between systems mediating mood and cognitive function.
Methods. Normal subjects performed a verbal fluency task during induced elated and depressed
mood states. Regional cerebral blood flow (rCBF) was measured as an index of neural activity
using Positron Emission Tomography (PET).
Results. In both elated and depressed mood state rCBF was increased in lateral orbitofrontal cor-
tex, rCBF was also increased in the midbrain in elated mood. In the depressed condition rCBF
was decreased in rostral medial prefrontal cortex. Verbal fluency produced an expected increase
of rCBF in left dorsolateral prefrontal, inferior frontal and premotor cortex, anterior cingulate
and insula cortex bilaterally, the left supramarginal gyrus posteriorly and the thalamus. Activa-
tion in the verbal fluency task was attenuated throughout the left prefrontal, premotor and cingu-
late cortex and thalamus in both elated and depressed mood conditions. An attenuation of anterior
cingulate activation was specific to depressed mood.
Conclusions. Alteration of mood is associated with activation of orbitofrontal cortex which may be
critical to the experience of emotion. The mood induced modulation of verbal fluency induced
activations is consistent with resting state findings of decreased function in these regions in
depressed patients. The present data suggest that resting state rCBF profile may represent the
modulation of spontaneous activity in this network by a core system that is dysfunctional in
depression.
INTRODUCTION
Depressed mood and lack of emotional reactivity
is pathognomic of depression. Cognitive and
motivational abnormalities are also integral to
depressive disorders (Weingartner et al. 1981).
Functional imaging studies have revealed dis-
tinct patterns of abnormal resting regional
cerebral blood flow (rCBF) in depression that
correlate with the principal symptom profiles
(Dolan et al. 1992; Bench et al. 1993).
"
Address for correspondence : Professor R. J. Dolan, Wellcome
Department of Cognitive Neurology, Institute of Neurology, 12
Queen Square, London WC1 3BG.
Cognitive techniques have been widely used
to alter mood in normal subjects (Gerrards-
Hesse et al. 1994). The distinction between
normal and pathological mood states has been
emphasized by psychiatrists. However, a mood
state characterized by negative self-evaluation
and psychomotor changes analogous to those
associated with clinical depression can be in-
duced in normal subjects (Velten, 1968) and is
referred to as ‘ depressed mood’ in contrast to
normal sadness. Psychologically induced de-
pressed mood is also regarded as a valid model
of retarded depression (Clarke, 1983 ; Riskind &
Rholes, 1985). The validity of this model is
565
566 S. C. Baker and others
reinforced by the associated low self-esteem
(Brown & Mankowski, 1993), psychomotor
retardation (Velten, 1968; Teasdale & Fogarty,
1979; Natale & Boylan, 1980) and cortisol
hypersecretion (Brown et al. 1993). Current
psychological theories of the pathogenesis of
depressive disorders have emphasized the role of
deviant cognitive style, characterized by self-
deprecating thoughts, analogous to the some of
the Velten statements, and the mood state
induced by this procedure has been invoked to
support cognitive models of the pathogenesis of
depression (Riskind & Rholes, 1985).
An understanding of the pathophysiology of
mood disorders requires specification of the
neural systems that mediate normal emotional
response and the neurophysiological mech-
anisms by which these systems influence cog-
nitive function. Animal studies suggest a critical
involvement of the ventral prefrontal cortex in
the regulation of emotion (Rolls, 1994). Con-
sistent with this, activation of the orbitofrontal
cortex has been observed during recall or
imagination of sad events (Pardo et al. 1993,
George et al. 1995), during evoked anxiety in
phobic and obsessive–compulsive disorders
(Rauch et al. 1994, 1995) and with the dysphoria
of angina pectoris (Rosen et al. 1994). Lesions to
the orbitofrontal cortex disrupt social and
emotional regulation of behaviour in humans
and monkeys (Butter et al. 1968 ; Damasio et al.
1990). However the relationship between affec-
tive disorders and orbitofrontal dysfunction
remains unclear. The surgical procedure of
subcaudate tractotomy, resulting in discon-
nection of the orbitofrontal cortex, is reported
as effective in the treatment of severe resistant
depression (Bridges et al. 1994). Curiously,
functional abnormalities are only rarely reported
in this region in imaging studies of depression
(e.g. Mayberg et al. 1990 ; Goodwin et al. 1993).
We used a combined mood induction and
cognitive activation paradigm to investigate the
functional anatomy of elated and depressed
mood in normal subjects. Combining these tasks
provides a powerful tool to determine the
interaction between mood and cognitive func-
tion.
METHOD
Subjects
Ten male volunteers aged between 18 and 35
years took part in the study, which was approved
by the Hammersmith Hospital Ethics Committee
and the Advisory Committee on the Admin-
istration of Radioactive Substances (ARSAC)
UK. Subjects were recruited from the staff and
students at London teaching hospitals, they
were screened to exclude previous psychiatric
disorder and drug use and were strongly right-
handed as assessed by the Edinburgh inventory.
Informed consent was obtained from all subjects.
PET scanning techniques
Regional cerebral blood flow was measured with
a CTI model 953B PET Scanner (CTI, Knoxville,
TN, USA), with the interplane septa retracted
(Spinks et al. 1992). Following a ‘slow bolus ’
infusion of H
#
"&
O, integrated counts per pixel
during the 90 s acquisition frame, corrected for
background counts, provided an index of rCBF.
Then 11±2 mCi of H
#
"&
O was flushed with
normal saline through a cannula in an ante-
cubital vein over 20 s at 10 ml}min by an
automatic pump. After a constant delay a rise in
counts at the head was detected, which peaked
between 30–40 s in individual subjects. Although
counts were collected for 90 s, only the activity
occurring during the rising phase of the head
count curve contributes significantly to the signal
in the final image (Silbersweig et al. 1993). Each
subject underwent 12 scans, with a 10 min
interval between scans. Correction for attenu-
ation was made by performing a transmission
scan with an exposed
')
Ge}
')
Ga external ring
source before each session and a 30 s frame for
background activity correction was acquired
before each infusion.
Study design
During each scan subjects performed either
paced orthographic verbal fluency or word
repetition tasks, generating or repeating one
word every 5 s, in a repeated ABAB … design
(Frith et al. 1991). In the verbal fluency
condition subjects continued to generate words
beginning with a given letter until they failed to
respond within the time limit after which they
were given a new letter. Subjects were studied in
elated, neutral and depressed mood states. Mood
Interaction between mood and cognitive function 567
Verbal fluency
Mood
11
9
7
5
3
1
12
10
8
6
4
2
Elation
Neutral
Depression
Fluency Repetition
F. 1. Diagrammatic illustration of the study design. There are six
scans in the verbal fluency condition and six in the repetition
condition. These are subdivided into three blocks of four scans each
in elated, neutral and depressed mood. The order of induced mood
states was counterbalanced across subjects to avoid systematic time
effects.
induction material was presented for 7±5 min
prior to each scan. The study was divided into
three blocks of four scans specific to each
induced mood state (Fig. 1). This factorial
design allows examination of the main effect of
the verbal fluency task in the comparison of the
verbal fluency and control conditions across all
mood states, the main effect of mood state in the
comparison of the elated and depressed mood
and neutral mood states across task conditions
and the interaction between verbal fluency and
mood; i.e. the difference in verbal fluency
activation in elated and depressed mood com-
pared to neutral mood. The order of the mood
states was counterbalanced across subjects to
accommodate time dependent effects. All sub-
jects were debriefed after the study.
Mood induction paradigm
Subjects were studied in the presence of low
background noise and dimmed ambient lighting.
A combination of the Velten, musical, social
interaction and gift mood induction procedures
were employed (Velten, 1968; Clark & Teasdale,
1985; Gerrards-Hesse et al. 1994). Subjects were
given explicit instructions to enter into the
suggested emotional state. Elated, neutral and
depressed mood induction statements, modified
from Velten (1968), were presented on an Apple
Macintosh microcomputer, at a rate of one
statement every 30 s. Subjects were played
extracts of ‘ Russia under the Mongolian Yoke ’,
from Prokofiev’s music for the film ‘Alexander
Nevsky’ (Clark, 1983), at half speed in the
depressed condition ; ‘Stressbusters ’, an ano-
dyne recording of popular classics, in the neutral
condition; and Delibes’ ‘Coppelia’ in the elated
condition. The investigator adopted an appro-
priately cheerful, neutral or solemn manner.
Subjects were presented with an unexpected gift
of £30 at the beginning of the elated mood
condition.
Data analysis
Image analysis was performed on a SPARC 10
workstation (Sun Microsystems Inc., Surrey,
UK) using interactive image display software
(ANALYZE, Biodynamic Research Unit, Mayo
Clinic; Robb & Hanson, 1990) and statistical
parametric mapping (SPM software, MRC
Cyclotron Unit, London, UK). Calculations
and image matrix manipulations were performed
in PRO MATLAB (Mathworks Inc., New York,
USA).
Image reconstruction
Images were reconstructed into 31 slices by
three-dimensional back projection using a
Hanning filter with a cut-off frequency of 0±5
cycles per pixel. The resulting images consisted
of 128¬128 pixels of 2±006¬2±006 mm having a
resolution of 8±5¬8±5¬4±3 mm full width at half
maximum (FWHM).
Image analysis
The 31 original slices were interpolated to 43
planes in order to render the voxels approxi-
mately cubic. Images were automatically re-
aligned to correct for head movement between
scans (Woods et al. 1993) and transformed into
a standard stereotactic space (Friston et al.
1991a). The stereotactically normalized images,
consisting of 26 planes, correspond to the
horizontal sections of the standard stereotactic
atlas (Talairach & Tournoux, 1988), each pixel
represents 2¬2 mm with an interplanar distance
of 4 mm. The field of view in all subjects
extended from 8 mm below to 56 mm above a
line joining the anterior and posterior com-
missures. Images were smoothed with a Gaus-
sian filter 10 pixels wide in order to suppress
high frequency noise in the images and accomo-
date normal variability in functional and gyral
anatomy for group analysis.
568 S. C. Baker and others
Stastistical analysis
Differences in global activity within and between
subjects were removed by analysis of covariance
(ANCOVA) on a pixel by pixel basis with global
counts as covariate and regional activity in each
condition across subjects as treatment (Friston
et al. 1990). The ANCOVA generated a mean
rCBF value, normalized to 50 ml}100 ml}min,
and associated error variance for every pixel in
each condition. This adjusted rCBF represents a
weighted mean over a sphere of approximately
20 mm. Differences between the adjusted mean
pixel values across conditions were assessed
using the t statistic (Friston et al. 1991b), the
resulting images of pixel t values constitute a
statistical parametric map (SPM(t)). The om-
nibus significance of the SPMs was assessed by
comparing the expected and observed distri-
bution of the t statistic under the null hypothesis
of no treatment effect. The SPM(t)s were
additionally displayed as volume images of the
highest t values in three orthogonal projections
and as surface renderings onto a standard
cerebral cortex. SPM(t)s were transformed to
the Unit Gaussian distribution using a prob-
ability integral transform so that changes could
be reported as Z scores.
RESULTS
Mood induction
Nine of the 10 subjects reported subjective mood
change during the mood induction procedure,
reflected in highly significant change in scores on
the PANAS (Watson et al. 1988) and ratings of
elation and depression (Table 1). One subject
reported no subjective change in mood and
there was no change in his ratings ; he was
excluded from the subsequent analysis. There
was no significant difference in the number of
words generated in the verbal fluency task in
different mood states, there were a total of six
omissions in neutral and depressed mood and
seven in elated mood.
Comparison of rCBF in verbal fluency
compared to repetition conditions
The main effect of the verbal fluency task was
examined by comparison of all scans in the
verbal fluency and control conditions. Foci of
maximal change in rCBF are displayed in Fig. 2
Table 1. Mean positive and negative affect
scores on the Positive and Negative Affect Scale
(PANAS; Watson et al. 1988) and mood ratings
in depressed, neutral and elated mood conditions
Induced PANAS scores Elation depression
mood
ve (..) ®ve (..) Score (..) Score (..)
Depressed 17±3(5±6)*** 18±2(5±7)*** 1±3(0±6)NS 2±7(1±0)***
Neutral 22±5(7±0) 11±7(2±6) 1±9(0±7) 1±1(0±4)
Elated 32±3(7±6)*** 11±0(1±4)NS 3±4(0±9)*** 1±0(0±0)NS
Significance of change in elated and depressed mood compared to
neutral mood: *** P ! 0±0001; NS Not significant.
PANAS norms for college students rating current mood under
standard conditions (Watson et al. 1988) : ve mean 29±7 ..7±9,
®ve mean 14±8 ..5±4. Depressed and elated mood induction were
both associated with significant change in rated mood from the
neutral condition.
and summarized in Table 2. Highly significant
increases in rCBF (P ! 0±001) were observed in
the verbal fluency condition in the left inferior
frontal gyrus and dorsolateral prefrontal cortex,
bilateral anterior cingulate and insula cortex
anteriorly, the left angular gyrus posteriorly and
the thalamus. There were relative decreases in
rCBF (P ! 0±001) in right inferior and superior
frontal gyri, bilaterally in superior and middle
temporal gyri and in medial temporal, inferior
parietal and occipital cortex.
rCBF changes in elated and depressed mood
The main effect of induced mood was examined
by comparison of all scans in elated and
depressed mood compared to the neutral mood.
Bilateral orbitofrontal rCBF increases (P !
0±001) were seen in both elated and depressed
mood compared to the neutral condition (Fig.
3a and b ; Tables 3 and 4), and were significantly
greater in elated mood than in depressed mood
(P ! 0±0005). rCBF increases were also observed
in the superior region of the left dorsolateral
prefrontal cortex and right lateral premotor area
in both elated and depressed mood. In depressed
mood, rCBF increases were apparent in SMA
and posterior cingulate cortex. In elated mood a
focal rCBF increase was also present in the
region of the posterior hypothalamus and
midbrain and also in left superior frontal gyrus.
Posteriorly rCBF was increased in the right
lateral parietal cortex.
Decreased rCBF (P ! 0±001) was observed in
the right caudate nucleus in both elated and
Interaction between mood and cognitive function 569
Table 2. Comparison of verbal fluency and repetition conditions: foci of significant rCBF change
Talairach coordinates
Left} Brodmann’s
Location Right area xyzZvalue
rCBF increases
Anterior cingulate L 32 ®41836 10±1
gyrus
Middle frontal L 46 ®32 44 20 6±36
gyrus
Middle frontal L 10 ®24 36 ®84±02
gyrus R 10 20 34 ®43±47
Inferior frontal L 44 ®36 4 28 6±94
gyrus
Anterior insula L ®30 18 4 7±71
R 22164 4±23
Medial premotor L 6 ®10 2 56 7±85
cortex (SMA)
Angular gyrus L 39 ®28 ®56 36 3±28
Thalamus 2 ®80 7±93
rCBF decreases
Superior frontal R 10 6 58 4 3±89
gyrus R 9 5 50 32 4±07
Inferior frontal R 44}45 48 12 20 3±74
gyrus
Superior temporal R 22 46 ®12 4 7±25
gyrus L 22 ®46 ®16 8 6±54
R2246®48 16 7±84
L22®52 ®32 12 7±07
Middle temporal L 39 ®48 ®60 12 8±40
gyrus R 37 44 ®56 4 7±40
Inferior parietal L 40 ®56 ®42 28 4±04
cortex R 40 50 ®30 32 6±31
Medial temporal R 36 22 ®40 ®84±73
cortex L 19 ®32 ®44 0 3±53
Occipital cortex L 18 ®4 ®92 12 3±16
L19®34 ®80 24 3±64
The co-ordinates of the foci of maximal significant change of rCBF (P ! 0±001) in the standard stereotaxic space of Talairach & Tournoux
(1988) are given in millimeters.
(b)(a)
F. 2. Regional brain activity associated with the verbal fluency task. Statistical parametric maps (SPMs) show regional cerebral
blood flow (rCBF) changes in word generation compared to word repetition scans. The SPMs illustrate: (a) increases ; and (b)
decreases of rCBF. Pixels exceeding an uncorrected threshold of significance P ! 0±001 are displayed on sagittal, coronal and
transverse projections of a standard brain (Talairach & Tournoux, 1988), the left side of the brain is on the left side of the projections.
The co-ordinates of the foci of maximal change of rCBF are given in Table 1.
