The relationship between performance and fMRI signal during
working memory in patients with schizophrenia,
unaffected co-twins, and control subjects
Katherine H. Karlsgodta, David C. Glahnb, Theo G.M. van Erpa, Sebastian Thermanc,
Matti Huttunenc, Marko Manninenc, Jaakko Kaprioc,d, Mark S. Cohene,f,g,
Jouko Lönnqvistc,h, Tyrone D. Cannona,e,i,⁎
aDepartment of Psychology, UCLA
bDepartment of Psychiatry and Research Imaging Center, UTHSCA, USA
cDepartment of Mental Health and Alcohol Research, National Public Health Institute of Finland, Finland
dDepartment of Radiology, Helsinki University Central Hospital, Finland
eDepartment of Psychiatry and Biobehavioral Sciences, UCLA, USA
fDepartment of Neurology, Radiology, Biomedical Physics UCLA, USA
gDivision of Brain Mapping, UCLA, USA
hDepartment of Psychiatry, University of Helsinki, Finland
iDepartment of Human Genetics, UCLA, USA
Received 29 May 2006; received in revised form 9 August 2006; accepted 11 August 2006
Available online 5 October 2006
While behavioral research shows working memory impairments in schizophrenics and their relatives, functional neuroimaging
studies of patients and healthy controls show conflicting findings of hypo- and hyperactivation, possibly indicating different
relationships between physiological activity and performance. In a between-subjects regression analysis of fMRI activation and
ofschizophrenics being intermediate betweenthegroups.Accordingly,thissupportsthe ideathatbothhyper and hypoactivationmay
be possible along a continuum of behavioral performance in a way consistent with a neural inefficiency model. Further, this study
offers preliminary evidence that the relationship between behavior and physiology in schizophrenia may be heritable.
© 2006 Elsevier B.V. All rights reserved.
Keywords: Schizophrenia; Working memory; fMRI; Twin; Genetic; Inefficiency
Functional magnetic resonance imaging (fMRI) of
working memory (WM) in schizophrenia has generated
Cannon et al., 2005; Ragland et al., 1998; Stevens et al.,
1998) and hyperfrontality (Callicott et al., 2000;
Schizophrenia Research 89 (2007) 191–197
⁎Corresponding author. Department of Psychology, UCLA, 1285
Franz Hall, Los Angeles, CA 90095-1563, USA. Tel.: +1 310 206
8765; fax: +1 310 794 9740.
E-mail address: email@example.com (T.D. Cannon).
0920-9964/$ - see front matter © 2006 Elsevier B.V. All rights reserved.
Manoach et al., 2000, 1999). It has been proposed that
patients' and controls' relationships between task
difficulty and brain activation differ. Specifically, while
both groups express an inverted-U shaped function
relating fMRI signal to WM load, the patient curve is
shifted, reflecting lower WM capacity. This difference
may reconcile these discrepant findings (Callicott et al.,
2003a; Manoach, 2003) and indicates that inefficiency
may be characteristic of brain function in schizophrenia.
This model supports previous findings that fMRI
activation in dorsolateral prefrontal cortex (DLPFC)
increases with WM load in controls (Rypma et al., 1999)
but is less consistent in schizophrenics (Callicott et al.,
2000; Perlstein et al., 2001). However, assessing load
interactions with brain activation has garnered mixed
results (Callicott et al., 2000; Manoach et al., 2000,
1999), as has matching for performance (Cannon et al.,
2005; Manoach et al., 2000, 1999; Perlstein et al., 2001;
Thermenos et al., 2005). We posit that behavioral per-
formance may more sensitively relate to variations in
fMRI signal than load or task difficulty, as variance in
behavior is lost when difficulty is categorized by load.
We further propose that overall behavioral performance
may be reflected in the degree of functional activation,
and that relationships between these factors may differ
between patients and controls.
Our goal was to investigate performance/activation
interactions in chronic schizophrenic patients, unaffect-
ed co-twins of schizophrenics, and healthy twins during
verbal WM. In particular, we assessed the DLPFC and
posterior parietal cortex (PPC), with the occipital lobe
(OCC) as a control region. We hypothesized that in a
between-subjects analysis, patients would be relatively
hyperactive at high performance levels (needing greater
activation to perform easy task) and hypoactive with
lower performance (demonstrating an inability to
activate WM circuitry sufficiently for a difficult task).
Because WM deficits (Cannon et al., 2000; Conklin
et al., 2005; Shaw et al., 2002) and frontal lobe changes
(Cannon et al., 2002) may be heritable, and fMRI fin-
dings in siblings of patients with schizophrenia are also
inconsistent (Callicott et al., 2003b; Keshavan et al.,
2002; Thermenos et al., 2004) we predicted that
unaffected co-twins would show a similar physiologic-
performance pattern to probands.
2. Methods and materials
Institutional review boards of the University of
California, Los Angeles and the National Public Health
Institute of Finland approved this study. Eight schizo-
phrenic patients with an unaffected co-twin, 10 unaf-
fected co-twins of schizophrenics [4 monozygotic (MZ),
6 dizygotic (DZ)], and 13 healthy twins participated
(Table 1). This group included 7 intact pairs [4 dis-
cordant (2MZ/2DZ); 3 healthy (2MZ/1DZ)] and 17
subjects without their matched co-twin[4 schizophrenic,
6 co-twins (2MZ/4DZ), and 8 controls]. Participants
were drawn from a cohortof Finnish twinsborn between
1940–1957 and gave informed consent. All subjects
were assessed using standardized clinical and medical
history measures (Cannon et al., 1998) including
Structured Clinical Interview for DSM-IV (Spitzer
et al., 1979) Patient or Non-Patient edition, Scale for
Assessment of Positive Symptoms (SAPS; (Andreasen
et al., 1990)) and Scale for Assessment of Negative
Symptoms (SANS; (Andreasen, 1982)) which were
given to any subjects with psychotic conditions, the
Personality Disorders Examination (Loranger et al.,
1985), and a medical-record coding form (Cannon et al.,
1998). No unaffected co-twins or control subjects had
any psychotic disorder or any cluster A personality
disorder. However, two controls had histories of anxiety
disorders (lifetime but not current social phobia and
panic disorder, respectively), and one control and one
proband had alcohol abuse. Subjects were excluded for
poor or missing behavioral data, scanner artifacts, or
2.2. Behavioral task
We employed a modified Sternberg item recognition
task (Sternberg,1966). A target set of 3, 5, 7, or 9 yellow
uppercase consonants was displayed for 2 s, followed by
a 3-s fixation. A green uppercase probe then appeared
for 3 s, followed by 1 s of fixation before the next trial.
Age of onset (yrs)
Duration of Illness (yrs)
aGender does not significantly differ across groups c2(2, n=31)=
bHandedness does not significantly differ across groups c2(2,
n=31)=3.27, p N.05.
cRauhala scale, range 1–9 (Rauhala, 1970).
192 K.H. Karlsgodt et al. / Schizophrenia Research 89 (2007) 191–197
Subjects indicated whether the probe matched the target
set. The task included 12 trials per load, which were
clustered by load into two-trial long blocks; presentation
order of the blocks was counterbalanced, but remained
the same for all subjects.
2.3. Scanning parameters
Imaging was performed on a Siemens (Erlagen,
Germany) 1.5-T Vision scanner in the University of
Helsinki. A high-resolution T1 image for anatomical
reference (TR/TE=720/14 ms, 24 axial slices, 256×256
matrix) and twenty-four contiguous AC-PC aligned
4 mm gradient echo EPI slices were acquired (TR/
TE=3 s/64 ms, flip angle 90°, 64×64 matrix, FOV
256 mm). Stimuli were projected onto a screen and
viewed through a head-coil mounted mirror. Accuracy
and response time were recorded via button-box. The
scanning battery included two other WM paradigms and
one hemodynamic response task, which were counter-
balanced across subjects; this task's functional run
lasted 7 min and 12 s.
2.4. Image processing
Since the brains of the patients with schizophrenia
are expected to be morphologically different from those
of the control subjects, as well as from subjects who's
brains have been used to create traditional standard
space templates, rather than using a pre-existing
standard space a group-average template brain was
created out of the subjects included in the analysis using
the Automated Image Registration (AIR) package
(Woods et al., 1998). The purpose of using a study
specific standard brain is to minimize the distortion of
the functional data during spatial normalization and, in
particular, to avoid creating group differences by
causing relatively greater distortion in the patient
group than in the control group. EPI Data were motion
corrected, registered to the individual's T1, and then to
the study specific standard brain using FSL (FMRIB's
Software Library, (Jenkinson et al., 2002; Smith et al.,
Individual subject analyses were performed using
FEAT (FMRI Expert Analysis Tool v5.4). Data were
spatiallysmoothed (8 mm FWHM Gaussian Kernel) and
high-pass filtered. In the individual first-level analyses,
loads 5, 7 and 9 were modeled with load 3 as baseline.
Results were fed into a second-level group analysis in
which probands, co-twins, and controls were separately
modeled so that variance was calculated for each group
individually. The second-level design matrix was
applied to all contrasts from the first-level analysis.
Motion parameters were analyzed and found to not
differ between groups.
Functional regions of interest (ROI's) were created in
the space of the study-specific standard brain. Active
regions were defined using activations from the all
subjects all loads contrast to allow both groups to
contribute to definition of the ROI. The t-statistic map
for this contrast was thresholded at T30N2.042(2 tailed,
pb.05). From this thresholded map, clusters in the
DLPFC, PPC, and OCC were identified using Brod-
mann's and anatomical landmarks. In each of these
regions all contiguous voxels above the threshold were
included in the ROI (see Fig. 1). The Featquery (fmrib.
ox.ac.uk/fsl/feat5/featquery.html) program applied the
inverse of the transformation matrix from individual to
standard space that was generated during the initial
registration to warp the ROI's back into each subject's
individual space where the statistics were performed.
The motion corrected, smoothed, and filtered data
across each entire ROI were probed for their percent
signal change from baseline (load 3). ROIs were moved
into MNI space for regional localization (see Fig. 2 for
2.5. Function-behavior analysis
To assess the relationship of performance and fMRI
data (percent correct) for each subject were collapsed
across load. We performed a robust linear regression,
predicting fMRI signal with performance as a factor
using Stata (v8.1, Stata Corporation). Group differences
(proband vs. co-twin vs. control) were examined using
the interaction between the slopes of the relationship of
performance and BOLD signal for the groups and tested
for significance using a Wald test. To account for rela-
tedness between co-twins, the pair IDs were clustered to
adjust standard errors for intragroup correlation of twin
Fig. 1. Regions of interest.
193 K.H. Karlsgodt et al. / Schizophrenia Research 89 (2007) 191–197
3.1. Behavioral data
in reaction time as load increased, there were no group
differences in the analysis by load. When performance
had lower performance than cotwins who had lower
performance thancontrolswas observed(seeFig.3).This
pattern is generally consistent with previous work in
which co-twins have shown intermediate performance
Fig. 2. Regressionanalysispredictingpercentsignalchangewithbehavioralperformance,betweensubjects,⁎⁎indicatessignificance atpN.01(corrected
for multiple comparisons, .05/5=.01),⁎indicates trend level significance.
194 K.H. Karlsgodt et al. / Schizophrenia Research 89 (2007) 191–197
showed only trend level significance (p=.062, 1-tailed),
however the difference between probands and control
subjects was significant when assessed with a t-test
(p=.0476, 1 tailed).
3.2. Functional imaging data
Given the small sample, MZ and DZ co-twin groups
no significant differences in the group×load whole brain
voxelwise analysis or in an ROI analysis comparing
activation within the ROIs across groups. In the ROI
regression analysis the slope of the relationship between
fMRI signal and behavior differed between probands and
right DLPFC and right PPC. Unaffected co-twins did not
significantly differ from probands in any region exam-
ined. Unaffected co-twins did not differ from controls in
left or right DLPFC or right parietal lobe, although they
in the OCC control region (see Fig. 2).
The finding that probands showed decreased BOLD
activation relative to controls with decreasing perfor-
mance in DLPFC and PPC, and increased relative
activation with increasing performance, supports our
hypothesis that the activation-performance relationship
is disrupted in schizophrenia in a way consistent with an
inefficiency model. Low performing patients were
“hypoactive” relative to controls and high performing
patients were “hyperactive”. Previous work has catego-
rized patients, with low performers demonstrating
hypofrontality and higher performers hyperfrontality
a continuum of performance. We believe that while per-
formance matching can provide a snapshot of activation
differences ata specific performance level,itmaybe less
informative than testing across a range of behavior. The
finding that the effects are more significant on the left
thantheright supportsthe ideathattheleft hemisphere is
with performance on this verbal task.
subject investigations and that are hypothesized in the
shifted inverted U model of inefficiency. The pattern seen
and low performers hypoactive overall, which does not
preclude the possibility that these subjects' activations
might also change as their performance changes, in a way
consistent with a double-U function, due to the fact that
between-subjects analysis requires averaging across
within-subjects variance and within subjects perfor-
mance-activation curves. It is possible that in individual
set point for the curve than higher performing controls,
patients have an overall lower set point reflecting an
overall inability to activate. What this indicates is that in
patients with schizophrenia to assess whether there is
some overall change in neural function above and beyond
a shift in the averaged inverted-U curves.
The pattern of results for unaffected co-twins was
intermediate between patients and controls, suggesting
that the inefficiency pattern may be related to genetic
indications that DLPFC-related cognitive functions are
disordered in unaffected relatives (Conklin et al., 2005;
Egan et al., 2001), and that brain regions supporting WM
are altered as well (Breier et al., 1992; Cannon et al.,
2002). Further, previous fMRI studies in relatives of
schizophrenics have been subject to the same hypo-
(Keshavan et al., 2002) and hyper-activation (Callicott
et al., 2003b; Thermenos et al., 2004) discrepancies as
have case-control studies. While the analysis of the co-
twins should be considered preliminary, given the small
sample size, it potentially indicates that the differences in
may stem from the same alteration in the performance-
activation relationship found in probands.
inOCCindicates thatthe groupdifferencesare specificto
WM circuitry and are task-related, they are not global
activation differences. Additionally, as this task is
sufficiently difficult to show a range of performance in
Fig. 3. Behavioral working memory performance.
195K.H. Karlsgodt et al. / Schizophrenia Research 89 (2007) 191–197
all groups, not just in patients, we were able to assess
functional activation at many levels of performance
within each group.
Our study did have some limitations. First, the sample
size, which necessitated combining MZ and DZ co-twins
for analysis. The small sample size also limits our
conclusions about heritability, and future work addressing
this in larger samples with matched twin pairs is necessary
to fully address this issue. Secondly, analyzing individual
loads (within-subjects) would allow fuller assessment of
the functional/behavioral relationship and would allow the
comparison of between and within subjects effects. Our
study was also limited by the lack of a non-task baseline,
While the use of a task-based baseline does circumvent
some issues relating to subtraction, which may arise with a
non-task baseline, future studies with resting baselines
would also be informative. And finally, our patients had a
relatively long duration of illness, and long history of
medication. The question of whether this phenomena is
also present in the early stages of the illness can be
addressed by doing similar studies in recent onset patients.
Overall, these findings lend empirical support to the
neural inefficiency model of working memory in schi-
zophrenia, which has implications for the previously
described hypo/hyperfrontality discrepancies in the liter-
ature. Further, this study can give preliminary evidence
may be a heritable aspect of the disorder, and may be
present in unaffected relatives even without a behavioral
deficit, which may help inform our conceptualization of
the nature and origin of WM deficits in schizophrenia.
Thanks to Lauren Ellman, Jacquie Horwitz, Vikas
Rao, and David Shirinyan for help with data analysis,
Russell Poldrack, Robert Bilder, and Joaquin Fuster for
input on the project, Carl-Gustaf Standertskjold-Nor-
denstam, Ulla Mustonen and the participants. This work
was supported in part by National Institutes of Mental
Health Grant MH52857 to T.D.C. and National Re-
search Service Award 1F31MH068111-01A2 to K.H.K.
Andreasen, N.C., 1982. Negative symptoms in schizophrenia.
Definition and reliability. Arch. Gen. Psychiatry 39, 784–788.
Andreasen, N.C., Flaum, M., Swayze II, V.W., Tyrrell, G., Arndt, S.,
1990. Positive and negative symptoms in schizophrenia. A critical
reappraisal. Arch. Gen. Psychiatry 47, 615–621.
Barch, D.M., Sheline, Y.I., Csernansky, J.G., Snyder, A.Z., 2003.
Working memory and prefrontal cortex dysfunction: specificity to
schizophrenia compared with major depression. Biol. Psychiatry 53,
Breier, A., Buchanan, R.W., Elkashef, A., Munson, R.C., Kirkpatrick,
B., Gellad, F., 1992. Brain morphology and schizophrenia. A
magnetic resonanceimaging study of limbic, prefrontal cortex, and
caudate structures. Arch. Gen. Psychiatry 49, 921–926.
Callicott, J.H., Bertolino, A., Mattay, V.S., Langheim, F.J., Duyn, J.,
Coppola, R., et al., 2000. Physiological dysfunction of the dorso-
lateral prefrontal cortex in schizophrenia revisited. Cereb. Cortex
Weinberger, D.R., 2003a. Complexity of prefrontal cortical
dysfunction in schizophrenia: more than up or down. Am. J.
Psychiatry 160, 2209–2215.
Callicott, J.H., Egan, M.F., Mattay, V.S., Bertolino, A., Bone, A.D.,
Verchinksi, B., et al., 2003b. Abnormal fMRI response of the
dorsolateral prefrontal cortex in cognitively intact siblings of
patients with schizophrenia. Am. J. Psychiatry 160, 709–719.
Cannon, T.D., Kaprio, J., Lonnqvist, J., Huttunen, M., Koskenvuo, M.,
1998. The genetic epidemiology of schizophrenia in a Finnish twin
cohort. A population-based modeling study. Arch. Gen. Psychiatry
Cannon, T.D., Huttunen, M.O., Lonnqvist, J., Tuulio-Henriksson, A.,
Pirkola, T., Glahn, D., et al., 2000. The inheritance of neuro-
psychological dysfunction in twins discordant for schizophrenia.
Am. J. Hum. Genet. 67, 369–382.
Cannon, T.D., Thompson, P.M., van Erp, T.G., Toga, A.W., Poutanen,
V.P., Huttunen, M., et al., 2002. Cortex mapping reveals regionally
specific patterns of genetic and disease-specific gray-matter
deficits in twins discordant for schizophrenia. Proc. Natl. Acad.
Sci. U. S. A. 99, 3228–3233.
Cannon, T.G., Glahn, D.C., Kim, J., Van Erp, T.G.M., Karlsgodt, K.,
Cohen, M.S., Nuechterlein, K.H., Bava, S., Shirinyan, D., 2005.
Dorsolateral prefrontal cortex activity during maintenance and
manipulation of information in working memory in patients with
schizophrenia. Arch. Gen. Psychiatry 62, 1071–1080.
Conklin, H.M., Curtis, C.E., Calkins, M.E., Iacono, W.G., 2005.
Working memory functioning in schizophrenia patients and their
first-degree relatives: cognitive functioning shedding light on
etiology. Neuropsychologia 43, 930–942.
Egan, M.F., Goldberg, T.E., Gscheidle, T., Weirich, M., Rawlings, R.,
Hyde, T.M., et al., 2001. Relative risk for cognitive impairments in
siblings of patients with schizophrenia. Biol. Psychiatry 50, 98–107.
Glahn, D.C., Therman, S., Manninen, M., Huttunen, M., Kaprio, J.,
Lonnqvist, J., et al., 2003. Spatial working memory as an
endophenotype for schizophrenia. Biol. Psychiatry 53, 624–626.
Jenkinson, M., Bannister, P., Brady, M., Smith, S., 2002. Improved
optimization for the robust and accurate linear registration and
motion correction of brain images. Neuroimage 17, 825–841.
Keshavan, M.S., Diwadkar, V.A., Spencer, S.M., Harenski, K.A.,
Luna, B., Sweeney, J.A., 2002. A preliminary functional magnetic
resonance imaging study in offspring of schizophrenic parents.
Prog. Neuropsychopharmacol. Biol. Psychiatry 26, 1143–1149.
Loranger, A.W., Sussman, V.L., Oldham, J.M., Russakoff, L.M., 1985.
Personality Disorder Examination: A Structured Interview for
Making Diagnosis of DSM-III-R Personality Disorders. Cornell
Medical College, White Plains, NY.
Manoach, D.S., 2003. Prefrontal cortex dysfunction during working
memory performance in schizophrenia: reconciling discrepant
findings. Schizophr. Res. 60, 285–298.
Manoach, D.S., Press, D.Z., Thangaraj, V., Searl, M.M., Goff, D.C.,
Halpern, E., etal., 1999. Schizophrenic subjects activate dorsolateral
196 K.H. Karlsgodt et al. / Schizophrenia Research 89 (2007) 191–197
prefrontal cortex during a working memory task, as measured by
fMRI. Biol. Psychiatry 45, 1128–1137.
Manoach, D.S., Gollub, R.L., Benson, E.S., Searl, M.M., Goff, D.C.,
activation of dorsolateral prefrontal cortex and basal ganglia during
working memory performance. Biol. Psychiatry 48, 99–109.
schizophrenia. Am. J. Psychiatry 158, 1105–1113.
Ragland, J.D., Gur, R.C., Glahn, D.C., Censits, D.M., Smith, R.J.,
Lazarev, M.G., et al., 1998. Frontotemporal cerebral blood flow
change during executive and declarative memory tasks in schizo-
phrenia: a positron emission tomography study. Neuropsychology
Rauhala, U., 1970. The quantitative strength of the social strata of
Finnish society (Finnish) Sos. Aikak. 63, 347–362.
1999. Load-dependent roles of frontal brain regions in the
maintenance of working memory. Neuroimage 9, 216–226.
Shaw, M.E., Strother, S.C., McFarlane, A.C., Morris, P., Anderson, J.,
Clark, C.R., et al., 2002. Abnormal functional connectivity in
posttraumatic stress disorder. Neuroimage 15, 661–674.
Smith, S.M., Jenkinson, M., Woolrich, M.W., Beckmann, C.F.,
Behrens, T.E., Johansen-Berg, H., et al., 2004. Advances in
FSL. Neuroimage 23 (Suppl 1), S208–S219.
Spitzer, R., Endicott, J., Williams, J., 1979. Research diagnostic
criteria. Arch. Gen. Psychiatry 36, 29–50.
Sternberg, S., 1966. High-speed scanning in human memory. Science
Stevens, A.A., Goldman-Rakic, P.S., Gore, J.C., Fullbright, R.K.,
Wexler, B., 1998. Cortical dysfunction in schizophrenia during
auditory word and tone working memory demonstrated by func-
tional magnetic resonance imaging. Arch. Gen. Psychiatry 55,
Thermenos, H.W., Seidman, L.J., Breiter, H., Goldstein, J.M.,
Goodman, J.M., Poldrack, R., et al., 2004. Functional magnetic
resonance imaging during auditory verbal working memory in
nonpsychoticrelatives of persons with schizophrenia: a pilot study.
Biol. Psychiatry 55, 490–500.
Thermenos, H.W., Goldstein, J.M., Buka, S.L., Poldrack, R.A., Koch,
J.K., Tsuang, M.T., et al., 2005. The effect of working memory
performance on functional MRI in schizophrenia. Schizophr. Res.
1998. Automated image registration: I. General methods and
intrasubject, intramodality validation. J. Comput. Assist. Tomogr.
22 (1), 139–152.
197K.H. Karlsgodt et al. / Schizophrenia Research 89 (2007) 191–197