Low Frequency BOLD Fluctuations During Resting
Wakefulness and Light Sleep: A Simultaneous
Silvina G. Horovitz,1Masaki Fukunaga,1Jacco A. de Zwart,1
Peter van Gelderen,1Susan C. Fulton,1Thomas J. Balkin,2and Jeff H. Duyn1
1Advanced MRI Section, LFMI, NINDS, National Institutes of Health, Bethesda, Maryland
2Department of Behavioral Biology, Walter Reed Army Institute of Research, Silver Spring, Maryland
Abstract: Recent blood oxygenation level dependent functional MRI (BOLD fMRI) studies of the
human brain have shown that in the absence of external stimuli, activity persists in the form of distinct
patterns of temporally correlated signal fluctuations. In this work, we investigated the spontaneous
BOLD signal fluctuations during states of reduced consciousness such as drowsiness and sleep. For
this purpose, we performed BOLD fMRI on normal subjects during varying levels of consciousness,
from resting wakefulness to light (non-slow wave) sleep. Depth of sleep was determined based on con-
currently acquired EEG data. During light sleep, significant increases in the fluctuation level of the
BOLD signal were observed in several cortical areas, among which visual cortex was the most signifi-
cant. Correlations among brain regions involved with the default-mode network persisted during light
sleep. These results suggest that activity in areas such as the default-mode network and primary sen-
sory cortex, as measured from BOLD fMRI fluctuations, does not require a level of consciousness typi-
cal of wakefulness. Hum Brain Mapp 29:671–682, 2008.
C2007 Wiley-Liss, Inc.
Key words: fMRI; EEG; sleep; resting state; functional connectivity; low frequency fluctuations; default-
mode network; visual cortex
involve electrical and chemical signaling in many of the
brain’s cellular networks. The hemodynamic processes
associated with this activity can be detected with positron
emission tomography (PET) [Fox and Mintun, 1989; Min-
tun et al., 1984; Sadato et al., 1996] or blood-oxygen level
dependent functional magnetic resonance imaging (BOLD
fMRI) [Bandettini et al., 1992; Kwong et al., 1992; Ogawa
et al., 1990]. These neuroimaging techniques were first
used to study activity in brain networks by contrasting
their signals during a sensory, motor, or cognitive task
with those during a resting state, characterized by the ab-
sence of external stimuli. Implicit in this strategy is the
assumption that there is an invariantly low level of neuro-
nal activity in the resting state.
This article contains supplementary material available via the Internet
Material has been reviewed by the Walter Reed Army Institute of
Research. There is no objection to its presentation and/or publication.
The opinions or assertions contained herein are the private views of
the author, and are not to be construed as official, or as reflecting true
views of the Department of the Army or the Department of Defense.
This research was funded by National Institute of Neurological Dis-
orders and Stroke (NINDS) Intramural Program.
*Correspondence to: Silvina G. Horovitz PhD, 9000 Rockville Pike,
Building 10, Room B1D722, Bethesda, MD 20814-1065, USA.
Received for publication 7 December 2006; Revised 13 April 2007;
Accepted 26 April 2007
Published online 27 June 2007 in Wiley InterScience (www.
C2007 Wiley-Liss, Inc.
r Human Brain Mapping 29:671–682 (2008) r
However, during resting wakefulness, a substantial level
of electrical and metabolic brain activity exists. In the EEG,
the presence of a regular ‘idling’ rhythm was observed in
scalp voltage as early as the 1920s which, somewhat para-
doxically decreased in amplitude when subjects opened
their eyes or became more engaged in a task [Berger,
1929]. Slower and more irregular fluctuations were also
clearly observable in resting animals, and appear to be
controlled as much by brainstem centers as by sensory
stimulation [Moruzzi and Magoun, 1995]. Moreover, many
individual neurons show a high spontaneous firing rate
that is not significantly attenuated even when an animal
falls into sleep [Evarts et al., 1962].
The influence of the ongoing EEG activity in cortical func-
tion and cognitive processes has recently received new inter-
est following a set of studies by Arieli et al., in the 1990s
[Arieli et al., 1996; Creutzfeldt et al., 1966], with recent work
suggesting that large spontaneous activity fluctuations com-
monly occur over time scales that match the dynamics of
resting state BOLD fluctuations [Leopold et al., 2003]. Most
interestingly, evidence suggests that the ongoing rhythms
contribute to the majority of the brain electrical activity and
the evoked/event-related activity, if present, contributes to
only a small percentage of the variance [Raichle, 2006].
In a meta-analysis of PET studies, Shulman et al. [1997a]
showed two distinctive hemodynamic behaviors. Based on
data from nine studies, it was shown that visual tasks con-
sistently result in blood flow changes confined to visual
cortex. When subjects performed tasks that were cogni-
tively more demanding, a set of deactivated areas was
observed [Shulman et al., 1997b] together with the acti-
vated cognitive areas. These findings led to the study of
the so-called ‘‘default-mode network’’ [Raichle et al., 2001],
defined as the network of brain regions that become more
active when goal-directed cognitive activity ceases. This
network has been hypothesized to facilitate functions such
as the monitoring of the body and environment, mainte-
nance of consciousness, and ongoing conscious processes
including internal information retrieval, and cognitive in-
formation processing [Binder et al., 1999; Mazoyer et al.,
2001; McGuire et al., 1996]. The default-mode network
includes dorsal medial prefrontal cortex, posterior cingu-
late, precuneus, and inferior parietal cortex.
In BOLD fMRI studies, temporally correlated signal fluc-
tuations have been observed during the resting state in
distinct regions including sensory areas [Biswal et al.,
1995; Cordes et al., 2000, 2001; Lowe et al., 2000], language
areas [Hampson et al., 2002], as well as the default-mode
network [Greicius et al., 2003]. The temporal frequencies
present in these fluctuations are concentrated in the 0.005–
0.05 Hz range [Cordes et al., 2001; Fransson, 2006; Fuku-
naga et al., 2006]. Using independent component analysis
(ICA) of resting state data, it has been shown that many
networks exist with apparent functional significance, and
with spatial distributions and signal time courses that are
largely independent [De Luca et al., 2006; Fukunaga et al.,
The origin and role of these BOLD fMRI fluctuations,
unknown. Although they may, in part, reflect purely vas-
cular processes [Mitra et al., 1997] and may partly relate to
changes in cerebral blood flow [Wise et al., 2004], hemody-
namic fluctuations in fMRI, near infrared spectroscopy
(NIRS) and PET have been shown to correlate with under-
lying electroencephalographic activity [Goldman et al.,
2002; Laufs et al., 2003; Moosmann et al., 2003; Obrig et al.,
2000; Sadato et al., 1998]. Furthermore, the multiple fine-
scale, temporally independent, and reproducible patterns
of fluctuations found in the ICA studies of fMRI data
[Arfanakis et al., 2000; De Luca et al., 2006; Fukunaga
et al., 2006; van de Ven et al., 2004] suggest these fluctua-
tions are not systemic and have functional relevance.
To the extent that resting state BOLD fMRI activity
relates to brain function, is it representative of a conscious
waking state1? Does it depend on the level of conscious-
ness? Preliminary studies of resting state BOLD activity
during conditions of reduced consciousness, such as sleep
[Fukunaga et al., 2006] and sedation [Kiviniemi et al.,
2005] found continued resting state activity in widespread
brain regions, including sensory areas such as visual cor-
tex. This suggests that BOLD resting state fluctuations in
these regions do not require conscious activity. Here, we
extend our previous study [Fukunaga et al., 2006] by add-
ing simultaneous EEG recordings to allow direct correla-
tion of the resting state BOLD fMRI fluctuation levels (see
below) with the level of consciousness, as reflected by
polysomnography. Furthermore, in the present study, we
aimed at establishing whether correlated BOLD signal fluc-
tuations in the default-mode network persist at the
reduced levels of consciousness characteristic of light
The paradigm used by Fukunaga et al.  was modi-
fied to extend the time available for subjects to sleep. Our
60-min paradigm (Fig. 1) started with a 1-minute long pe-
riod of quiet wakefulness with eyes open, after which the
subjects were instructed to close their eyes, relax and try
to sleep. Subjects were allowed to sleep undisturbed for
48 min, after which they were instructed to open their
eyes. Starting from minute 50, a flickering checkerboard (6
Hz) was presented, stimulating alternately the central 58 of
the visual field or its periphery (58–158), in seven blocks of
48 s each. During the last 4.4 min, the volunteers were
1Throughout this work, we equate consciousness to the waking
state [Zeman, 2001]. Defined this way, consciousness is a matter of
degree that extends from waking through sleep into anesthesia
and coma [see also Laureys, 2005].
r Horovitz et al. r
r 672 r
Experimental paradigm and sample time course data from one
subject. (a) Timing of the scan paradigm. One minute of rest
with eyes open (EO) was followed by 48 min of eyes closed
(EC), during which the subject was allowed to sleep. This was
followed by another minute of EO, and 5.6 min of checkerboard
stimulation and 4.4 min of center dot fixation for a total scan
time of 60 min. (b) EEG spectrogram: time-frequency spectrum
of the data from one subject at electrode C3. Notice elevated
alpha activity (8–12 Hz) -indicating wakefulness- at the beginning,
then slowing towards extended theta (2–7 Hz) - indicating sleep-
and just before minute 30 returning to predominantly alpha ac-
tivity. (c) Inverse index of wakefulness (IIoW): Larger values rep-
resent lower frequencies in the EEG, which are associated with
sleep. (d) Hypnogram: Sleep score, over 30 s intervals, as
assessed by sleep expert (TJB). Higher scores indicate deeper
sleep (Rechtschaffen and Kales, 1968). (e) fMRI time course: av-
erage percentage signal change in the VC ROI, defined from the
response to the visual task. Overlay: relative standard deviation
of the BOLD signal change in the VC ROI, in 2 min (20 images)
intervals, data offset vertically by 3 units for display purposes.
r fMRI During NREM Sleep r
r 673 r
instructed to fixate on a dot in the center of the visual dis-
Simultaneous EEG and fMRI data were collected in
twenty two sessions from fourteen subjects (5 females, 9
males; age range: 21–56, average age: 32) after each provided
written informed consent in accordance with a protocol
approved by the National Institutes of Health (NIH) Intra-
mural Research Board (IRB). Ear-plugs were inserted for
hearing protection. All studies were performed during the
daytime. All subjects reported normal sleep/wake patterns,
and no prior sleep deprivation or any other manipulation
was performed to facilitate sleep onset in the scanner.
EEG was collected using silver-silver chloride sintered
electrode caps via carbon fibers (MagLink) using Syn-
amps2 amplifiers and Scan 4.3 software (all from Neuro-
scan, Compumedics USA Ltd, El Paso, TX). Forty electro-
des, including those in the standard 10–20 International
system, were used. The ground electrode was located fron-
tal to Fz and the reference electrode was between Cz and
Pz, all on the brain midline. Two bipolar electrodes were
used to monitor electro-oculogram (EOG) and cardiac sig-
nal (ECG). Acquisition rate was 10 kHz, with a low pass
filter set to 3 kHz. Data were collected continuously. Respi-
ration was measured using respiratory bellows and cardiac
rate was measured using a pulse oximeter placed on the
left index finger. Both transducers were provided with the
MRI scanner. Signals from respiratory and cardiac sensors
and scanner-generated TTL trigger pulses were collected at
a frequency of 1 kHz to allow adequate sampling of the
cardiac signal [Lund et al., 2006].
fMRI Data Acquisition
BOLD fMRI was acquired on a 3T scanner (GE Signa,
Milwaukee, WI) using a 16-channel receive-only detector
array (Nova Medical, Wakefield, MA) [de Zwart et al.,
2004]. The coil was foam-padded to restrict head motion
and improve subject comfort. After a 3D localizer, single-
shot echo-planar-images (EPI) were collected from 28
oblique-axial slices (1.7 3 1.7 mm2nominal in-plane reso-
lution, 3.0-mm thickness, 0.5-mm gap) covering most of
the brain. Cerebellum and subthalamic nuclei were often
only partially included. The excitation flip angle was set to
908. Repetition time (TR) was 6 s and echo time (TE) was
43 ms. The MRI slice acquisition was grouped (bunched)
in time in the first 3 s of the TR to allow for 3 s of EEG
without contamination from gradient artifacts. A TR of 6 s
is longer than typically used, but it should capture most of
the low frequency BOLD signal fluctuations with only little
contribution of aliased cardiac and respiratory signals [see
frequency distribution in Fukunaga et al., 2006]. 3D T1-
weighted images were collected from all volunteers using
a GE head-coil in a different session, using an IR-prepared
3D SPGR sequence (MPRAGE).
Data analysis was performed to characterize resting state
activity during both wakefulness and sleep. EEG data was
scored using Rechtschaffen and Kales  criteria to
determine sleep stage, and further processed to derive an
index of wakefulness as described below. Subsequently,
fMRI data was analyzed to determine the relationship
between this index and the fluctuation level, and to inves-
tigate resting state correlations and fluctuation levels of the
BOLD signal during both waking rest and light sleep. For
this purpose, whole brain analysis was performed, as well
as more focused analyses in selected ROIs, as defined
After low-pass filtering at 250 Hz, artifacts induced by
MRI gradient switching were removed based on template
subtraction [Allen et al., 2000]. Subsequently, band-pass fil-
tering was performed using a frequency range of 0.5–28
Hz. Cardio-ballistic artifact was removed by first detecting
the R-wave and creating an averaged template of the arti-
fact, then extracting its principal components and finally
removing those components from the original data. All
these procedures were performed using Scan 4.3.2 soft-
ware (Neuroscan, Compumedics USA Ltd, El Paso, TX).
Determination of Sleep Stage and Index
EEG datasets were visually inspected by a sleep expert
(TJB). A sleep stage (wakefulness (W), Stage 1, Stage 2,
slow wave sleep (SWS) or rapid eye movement (REM))
was generated for each 30-s interval. Frequency analysis
was performed using IDL 6.2 (ITT visual information solu-
tions, Boulder, CO, USA) for electrodes C3, C4, P3, and P4.
EEG of wakefulness while resting with eyes closed is char-
acterized by a relatively high level of a activity (8–12 Hz),
while lower frequencies are more prominent during non-
REM (NREM) sleep [Rechtschaffen and Kales 1968]. There-
fore, wakefulness was quantified by taking the ratio of the
square root of the power in the 2–7 Hz band to that in the
alpha band, computed over 2-min intervals. In the present
article, we refer to this value as the inverse index of wake-
fulness (IIoW), since a higher index corresponds to a lower
waking consciousness level. We compared the IIoW to the
fluctuation level of the fMRI percentage signal change as
Data sets were processed using custom routines written
in IDL, unless otherwise specified. First, each run was reg-
istered to the last image in the time series. Slice timing cor-
rection and rigid body motion correction were performed
using SPM2 software (Wellcome Department of Cognitive
r Horovitz et al. r
r 674 r
Neurology, London, UK). The following procedures were
performed on the ‘‘resting state’’ data (min 1–49): A high-
pass filter (0.006 Hz) was applied to remove baseline drift.
The global signal change was removed by linear regression
of the time course of the average brain signal, after brain/
skull segmentation [Birn et al., 2006; Fukunaga et al.,
2006]. The cardiac rate and the respiratory depth, derived
from the peak to peak amplitude of the respiratory trace
[Birn et al., 2006]—were also regressed out. The cardiac
rate and respiration regressors were estimated by correlat-
ing the physiological signals with the fMRI BOLD signal
for each voxel at different delays (610 TRs), then averag-
ing the correlation coefficient across whole brain for each
delay and finally selecting the delay with highest correla-
tion (typically 1 or 2 TRs). Images were registered to a
standard MNI brain (2 3 2 3 2 mm3). Data sets were con-
verted to percentage signal change by subtracting the
mean value of each voxel from its time course, then divid-
ing by the mean and multiplying by 100.
In addition to whole brain analysis of the fluctuation
levels, several analyses were also performed on pre-
defined regions of interest. A visual cortex (VC) ROI was
defined functionally for each subject by selecting the vox-
els in the occipital region that correlated significantly (r >
0.5) with the central visual stimulation paradigm. The av-
erage time course of this ROI was used as the seed for the
VC correlation analysis (see below).
ROIs in posterior cingulate, precuneus, anterior cingu-
late, superior occipital, calcarine sulcus, supplementary
motor area, inferior parietal, and inferior frontal gyri were
chosen to explore the default-mode network and the areas
that (anti) correlate with it, sensory processing and atten-
tion areas. These ROIs were prescribed from a MNI tem-
plate [Tzourio-Mazoyer et al., 2002], since no task was per-
formed for their definition (details about the ROIs are
shown in supplementary information: Figs. S1 and Table
SI). The average time course of the posterior cingulate ROI
was used as the seed for the default-mode network corre-
lation analysis (see below). The posterior cingulate ROI
time course was correlated with the time course of a seed
selected using coordinates from Raichle’s work (Talairach
?5, ?49, 40) [Raichle et al., 2001].
Fluctuation Level of the BOLD Signal as a
Function of the Inverse Index of Wakefulness
After fMRI pre-processing, the fluctuation level of the
BOLD signal was estimated for each voxel in the brain
from the standard deviation of the BOLD signal and
expressed as percentage. For a zero mean signal, the stand-
ard deviation approximates the root mean square (r.m.s)
of the fluctuation level, a measure of the amount of energy
in the signal fluctuation.
To allow sufficient accuracy for the determination of the
fluctuation level of the BOLD signal, only subjects that had
a minimum of two uninterrupted minutes each of sleep
and wakefulness were included in this analysis. The fluc-
tuation level was calculated for each 2-min interval during
the resting part of the experiment as indicated above, for
each voxel after smoothing using a 2-pixel Gaussian filter.
The length (2 min) of the intervals was chosen to have
enough time points for the level estimation (20 TRs) but
also enough points for the regression (22 two-minute inter-
vals). For each voxel in the brain, the fluctuation level was
correlated to the IIoW computed over the same 2-min
intervals. Significance was established by permutation
methods [Holmes et al., 1996].
For the VC ROI, the same analysis was repeated without
Gaussian filtering. Data from the VC ROI were further
explored by combining the fluctuation level of all voxels
within the ROI for each condition (wakefulness or sleep)
and compared using Wilcoxon’s test. Measures during the
visual task are reported as a comparison.
We computed the power spectral density of the BOLD
signal over the VC ROI to allow comparison with previous
studies [Fransson, 2006; Fukunaga et al., 2006].
Temporal Correlation During
Sleep and Wakefulness
To allow sufficient accuracy for determination of tempo-
ral correlation, only subjects who had a minimum of 45
contiguous time points (4.5 min) each of sleep and wake-
fulness were included in this analysis. Functional connec-
tivity was computed as the correlation of the seed time
courses with all other voxel time courses. The reference
time course (seed) was created by averaging the time
TABLE I. BOLD fluctuation differences between
wakefulness and sleep in the VC ROI
Wilcoxon test (P-value)
1.35 (6 0.19)
1.48 (6 0.29) 1.65 (6 0.53)
The fluctuation level of the BOLD signal, expressed as relative
standard deviation, was computed for each voxel and averaged
within the ROI (see methods). Values during task are also pre-
sented for comparison.
r fMRI During NREM Sleep r
r 675 r
course signals of all voxels belonging to the selected ROI
(VC or Posterior Cingulate). All preprocessing was per-
formed prior to this analysis, as indicated in fMRI Prepro-
Spatial Extent of Correlated Activity in the
Default-Mode Network and Visual Cortex
The spatial extent of the temporal correlation in default-
mode network and visual cortex was determined by corre-
lation with the time course of the posterior cingulate seed
region and the functional VC time course, respectively.
This was done in both sleep and wakefulness. Correlation
coefficients were converted to z-scores using Fisher’s trans-
formation [Howell, 1997] to account for individual vari-
ability [Hampson et al., 2002] and combined across sub-
jects. Significance level (P-value) was established by a
randomization method in which phase data was scrambled
[Bullmore et al., 1999].
Strength of the Correlation of the fMRI Signal
The strength of the correlation within each ROI was
computed separately for sleep and wakefulness by corre-
lating the voxel-averaged time course of each ROI with
time courses of the individual voxels in that ROI. The
resulting correlation values were then averaged, and a t-
test was performed to determined significance of differen-
ces between sleep and wake conditions.
Data Quality and Subject Inclusion
Despite the adverse conditions of MRI scanning (gradi-
ent acoustic noise, fixed head and body position), most
subjects were able to fall asleep during at least one of the
sessions, displaying mostly light sleep. In all but one sub-
ject no deep sleep was observed, which is attributed to the
nonsleep-conducive experimental conditions, and the fact
that scans were performed during daytime hours without
prior sleep deprivation.
For sixteen of the twenty two-data sets, the EEG data
was of sufficient quality to discriminate between sleep and
wakefulness, both by IIoW and by sleep scoring done by
the sleep expert (TJB). The six remaining data sets were
excluded from further analysis either due to apparent
motion artifacts (5 mm or more in estimation of the motion
correction during the image registration procedure) or due
to incomplete data sets resulting from technical problems.
Three data sets did not include any sleep. Thirteen of
the sixteen data sets had at least 2 min of wakefulness and
2 min of sleep. They were collected from eleven subjects,
so two more studies were excluded to have only one data
set per subject. Those data sets that presented more contin-
uous segments of sleep and wakefulness were included in
For the eleven subjects included in this study, the mean
sleep efficiency index, defined as the percentage of time
scored as sleep over the total time allowed for sleep, was
22.28%, ranging from 2.0 to 23.0 min. Most of the sleep
data were scored as Stages 1 and 2, with only one subject
reaching SWS for 2 min.
Only six subjects had more than 4.5 consecutive minutes
of sleep and more than 4.5 consecutive minutes of wakeful-
ness, and were thus included in the temporal correlation
analysis. For these six subjects, the average time included in
the analysis for each condition was 12.6 6 2.6 min.
Consistent with an earlier report from our laboratory
[Fukunaga et al., 2006], the 11 subjects included in this
study showed, during the 48-min ‘‘eyes-closed’’ portion of
the study, an fMRI signal fluctuation level in grey matter2
(1.14% of the baseline signal intensity) that was substan-
tially above thermal noise level3[de Zwart et al., 2002]
(0.44% of the baseline signal intensity) and that was com-
parable to the signal changes during task performance
(Table I and Fig. 3a).
Fluctuation Level of the BOLD Signal as a
Function of the Inverse Index of Wakefulness
This analysis was performed on the 11 subjects matching
the ‘‘2 min sleep and 2 min wakefulness’’ criterion. The
power spectral density is shown in Figure 2. Most of the
energy was concentrated below 0.03 Hz, and some
expanding to 0.05 Hz, similar to the range observed in pre-
vious work [Fransson, 2006; Fukunaga et al., 2006].
The BOLD signal fluctuation level observed in the VC
during sleep periods was larger than during resting wake-
fulness and was generally comparable to the amplitude of
activation evident during the performance of the visual
task (see example in Figs. 1 and 3a). Correlation of the
fluctuation level with IIoW and comparison with sleep
scoring showed that, as subjects fell asleep, the BOLD fluc-
tuation level in the VC generally increased (Fig. 3b and
Table I). The correlation between the relative standard
deviation of the BOLD signal in the VC and the IIoW
reached r ¼ 0.37 (F(11,223)¼ 12.139, P < 0.025) (Fig. 3b).
The whole brain maps were surface rendered with
SUMA [Saad et al., 2006] for display purposes. Areas that
showed fluctuation levels significantly correlated to the
IIoW are shown in Figure 3c. As subjects fall asleep, an
increase in fluctuation level is seen in widespread cortical
regions, especially in the sensory areas (different parts of
visual, motor and primary auditory cortices), in precuneus,
and to a lesser extent in inferior parietal, supramarginal,
2Signal was computed from an ROI defined as the sum of all grey
matter areas from an MNI template [Tzourio-Mazoyer et al., 2002].
3Thermal noise is random noise that can be calculated from an
EPI acquisition without RF excitation [de Zwart, 2002].
r Horovitz et al. r
r 676 r
and temporal cortices as well as in parts of frontal gyri,
paracentral areas, and also in some deep structures as thal-
amus, caudate, and putamen. Decreases were seen in the
medial frontal gyri, although they did not reach signifi-
In the VC ROI the fluctuation level of the BOLD signal
was significantly larger (P < 0.0029) during sleep com-
pared to wakefulness (data reported in Table I).
Temporal Correlations During Sleep
This analysis was performed in the six subjects who
obtained at least 4.5 uninterrupted min of both sleep and
Analysis of the temporal correlation with the average
time course in the VC showed a similar spatial pattern
during both sleep and wakefulness (Fig. 4). Figure 5 shows
the composite maps for wakefulness and sleep detected by
correlation with the average time course in the posterior
The default-mode network was present both during
wakefulness and light sleep and had a similar spatial
extent during these conditions. It was also detected during
both conditions in each of the individual subjects (data
shown in supplementary Fig. S2). The correlation between
the averaged ROI time course and the time course of the
seed reported by Raichle et al.  was significant (r ¼
0.46) and the correlation maps using that seed provide
similar results (supplementary Fig. S3). The nonoverlap-
ping spatial distribution of the networks detected by corre-
lation with VC and posterior cingulate seeds suggest these
represent two independent processes present in the resting
state (supplementary Fig. S4).
Within-ROI correlation coefficient values were signifi-
cant for all ROIs, and similar for light sleep and wakeful-
ness for most of the defined areas (Table II). Significantly
larger coefficients during sleep were seen bilaterally in
supplementary motor area (t > 2.5, P < 0.0272) and in
right calcarine sulcus (t ¼ 2.13, P ¼ 0.043). None of the
selected areas showed larger correlations during wakeful-
ness compared to sleep. These results indicate that signifi-
cant functional connectivity within the selected ROIs is
present during both sleep and wakefulness.
In this work, EEG and fMRI were performed simultane-
ously to study the effect of sleep on resting state BOLD ac-
tivity throughout the brain. Collecting simultaneous EEG/
fMRI sleep data in the scanner during daytime hours with-
out any sleep deprivation proved challenging. While most
volunteers slept fitfully for parts of the study, only a mod-
est number (six) slept uninterrupted for at least 4.5 min,
limiting the amount of data available for the analyses
involving correlations. However, results were obtained
both at individual subject and group level. The use of a
relatively long TR of 6 s provided results similar to our
previous data collected at 3 s TR, and the spectral distribu-
tion was similar to that reported previously [Fransson,
2006; Fukunaga et al., 2006] suggesting a minimal contri-
bution of aliased cardiac and respiratory signals in the
present results [see also Lowe et al., 1998].
Presence of Resting State Activity During Sleep
Consistent with a preliminary earlier report from our
laboratory [Fukunaga et al., 2006], widespread fMRI rest-
ing state activity was observed during sleep. This was evi-
dent from both fluctuation levels and temporal correlation
analysis. It is intriguing that some areas, including primary
visual cortex, showed an increase in resting state fluctua-
tion levels with sleep and reduced levels of consciousness.
The continued presence of correlated fluctuations during
sleep suggests that BOLD fMRI resting state activity does
not require conscious wakefulness but rather, in most
brain areas, persists during the reduced levels of con-
sciousness characteristic of light sleep. This notion of a
continuation of resting state activity with reduced con-
sciousness is consistent with a previous study that found
and increase in BOLD fMRI signal correlation in visual
cortex after sedation [Kiviniemi et al., 2005]. The fluctua-
tion level increases found in selected brain regions with
sleep and sedation are intriguing and require further
Another important finding of the current study is the
continued presence of correlated BOLD signal fluctuations
in the default-mode network during light sleep. This new
finding suggests that the default-mode network is active
Power spectral density of the BOLD time course in the VC ROI
computed for 11 subjects over the entire (48 min) resting period.
r fMRI During NREM Sleep r
r 677 r
BOLD fluctuation differences between sleep and wakefulness (n
¼ 11). (a) Boxplot of the relative BOLD signal fluctuation levels
in VC during resting wakefulness, sleep and task. The boxes have
lines at the lower quartile, median, and upper quartile values.
The whiskers are lines extending from each end of the box to
show the extent of the rest of the data. These data have no out-
liers. (b) Relationship between fluctuation level of the BOLD sig-
nal and IIoW in the visual cortex ROI. Scatter plot shows rela-
tive standard deviation of the BOLD percentage signal change
averaged over the VC ROI, versus IIoW. Each point is obtained
as an average over 2 min intervals (that is over 20 TRs). Two
hundred and thirty-five points derived from eleven subjects are
presented (22 intervals per subject, two subjects contributed 17
and 19 points respectively due to shorter scans). Each color rep-
resents a volunteer. The regression curve is shown by the dot-
ted line. Correlation coefficient ¼ 0.37, F(11,223)¼ 12.139, P <
0.025). (c) Relationship between fluctuation level of the BOLD
signal and IIoW across the entire brain. Correlations between
relative standard deviation of the BOLD percentage signal
change in each voxel and IIoW converted to p-values and shown
on inflated brain. Yellow-red tones indicate areas where fluctua-
tions are larger during sleep compared to wakefulness (larger
IIoW). A widespread distribution of this effect is seen in visual
cortex, primary auditory cortex, and precuneus among other
areas. Blue tones (not seen) indicate areas where fluctuations
during wakefulness were larger than during sleep.
and functionally connected in the absence of goal-directed
mental activity, not only during awake rest [Gusnard and
Raichle, 2001; Raichle et al., 2001], but also during light
sleep. This network has previously been hypothesized to
subserve monitoring of the environment, conscious aware-
ness, sustained information processing or retrieval, and
manipulation of episodic memory [Greicius et al., 2003;
Gusnard and Raichle, 2001; Raichle et al., 2001]. Further-
more, its activity has been shown to increase during brief
periods of impaired performance after sleep deprivation
[Drummond et al., 2005]. The observed connectivity within
this network during sleep in the current study suggests
that it does not require or reflect the level of consciousness
that is typical of wakefulness.
Comparison With Previous Neuroimaging Studies
The results of the current study are based on analysis of
signal fluctuation levels, and do not provide information
about absolute BOLD signal at specific sleep stages. This
was not attempted because absolute activity is difficult to
extract reliably form BOLD signals. This makes it difficult
to compare the current findings directly to many earlier
PET neuroimaging studies of sleep, which report on steady
state levels of cerebral blood flow (CBF) or glucose utiliza-
tion PET [Andersson et al., 1998; Braun et al., 1997; Buchs-
baum et al., 1984; Hofle et al., 1997; Kajimura et al., 1999,
2004; Kjaer et al., 2002; Maquet, 2000; Maquet et al., 1992;
Nofzinger et al., 2002; Peigneux et al., 2001]. In general,
these studies observed CBF and metabolism decreases dur-
ing NREM sleep. However, only few of these studies have
reported on differences between deep sleep (Stages 3 and
4) and light sleep (Stages 1 and 2) [Kjaer et al., 2002;
Maquet et al., 1992]. Maquet et al.  found non-signifi-
cant changes in most areas, whereas Kjaer et al. 
found (during Stage 1) increases in extrastriate areas and
decreases in posterior parietal cortex, cerebellum, premotor
cortex and thalamus.
Similarly, comparison with previous EEG/fMRI studies
that have investigated resting state activity is difficult as
these studies did not look at BOLD fluctuation levels, as
the current study did, but rather directly correlated EEG
signals with BOLD signal levels. Additionally, level of
wakefulness was not reported. In general, these studies
found a negative correlation of BOLD signal with alpha
power [Goldman et al., 2002; Laufs et al., 2003; Moosmann
et al., 2003] in the visual cortex. Laufs et al.  reported
two distinctive networks correlating with alpha power and
speculated that vigilance levels could affect the relative
network activity. We find increases in signal fluctuations
in the visual cortex and other cortical areas with increased
levels of IIoW, suggesting that the vigilance level does
indeed affect the resting BOLD signal fluctuations.
Statistical composite maps (n ¼ 6) showing the temporal corre-
lation of the percentage BOLD signal change with the mean time
course in the VC ROI during wakefulness (top) and sleep
(bottom). Color scale represents p-values, thresholded at P ¼
60.05 (corrected). Positive (red-yellow) and negative (blue) cor-
relations are shown.
Statistical composite maps (n ¼ 6) showing the temporal corre-
lation of the percentage BOLD signal change with the seed in
the posterior cingulate ROI during wakefulness (top) and sleep
(bottom). Four representative axial slices show similar default-
mode network correlations during wakefulness and light sleep.
Color scale represents p-values, thresholded at P ¼ 60.05 (cor-
rected). Positive (red-yellow) and negative (blue) correlations
r fMRI During NREM Sleep r
r 679 r
An increase in BOLD fMRI fluctuation levels during
light sleep in many cortical regions may be due, at least in
part, to changes in perfusion unrelated to neuronal activ-
ity. Some studies have reported an increase in blood flow
fluctuations after reduction in perfusion pressure [Fujii
et al., 1990; Hudetz et al., 1992; Jones et al., 1995]. If sleep
also reduces perfusion pressure, the possibly resultant
blood flow fluctuations would be observable in the BOLD
signal. However, this mechanism would not explain the
fine scale, while widespread, spatial patterns of correlated
activity seen during sleep, wakefulness [Fukunaga et al.,
2006] and during visual fixation [Nir et al., 2006]. Nor
does it explain the correlation observed between BOLD
signal amplitude and EEG a power [Goldman et al., 2002;
Laufs et al., 2006; Moosmann et al., 2003]. Simultaneous
perfusion/BOLD fMRI studies might clarify this issue.
Alternative Origins of Resting
State Activity During Sleep
Among the possible alternative explanations for the
presence of resting state activity during sleep is the exis-
tence of dream-like reverie activity, which sometimes
occurs during light sleep. Alternatively, or concurrently, it
might also be that during sleep, fluctuations in the level of
consciousness increase, and that they are reflected in local
fluctuation in BOLD fMRI signal. For example, brief
arousals could occur during light sleep that are too short
to affect sleep score but nevetheless resulted in BOLD
fMRI signal fluctuations. This is suggested by Laufs et al.
 findings and the current findings that IIoW corre-
lates with BOLD fluctuation level in the visual cortex and
other sensory areas (Fig. 3c). The correlation between
BOLD fluctuation level in visual cortex and the EEG-
derived IIoW suggests that BOLD fMRI could serve as a
surrogate indicator of wakefulness when EEG is not avail-
able. Finally, it could be that the homeostatic processes
that increase during sleep cause spatially distinct fluctua-
tions in activity.
The continuation of resting-state BOLD fMRI signal fluc-
tuations during sleep in several brain regions, including
the hypothesized ‘‘default-mode network’’ of brain func-
tion suggests that this activity is not specific to the waking
state. The fluctuation levels increase, particularly in pri-
mary sensory cortex, as the consciousness level decreases
from wakefulness to light sleep. Further research is needed
to establish a direct link between BOLD resting activity
and electrical brain signals.
The authors thank David A. Leopold (NIMH, NIH) for
TABLE II. Strength of the within-region correlation during wakefulness and sleep
No. of voxels
(2 3 2 3 2 mm3)
Average r (6s.d.) z-score
xyz WakefulnessSleep WakefulnessSleep
Mean r and standard deviation are reported. Talairach–Tourneaux (T-T) coordinates and ROI volume are provided for reference.
z-scores were computed to test the difference in strength of the correlations between wakefulness and sleep.
Correlation values within ROI reached P < 0.001 in all subjects (n ¼ 11) and ROIs with the exception of two ROIs of one subject that
were significant only to P < 0.01 (marked with#).
*Denotes significantly higher correlation (P < 0.05) during sleep compared to wakefulness.
r Horovitz et al. r
r 680 r
Allen PJ, Josephs O, Turner R (2000): A method for removing
imaging artifact from continuous EEG recorded during func-
tional MRI. Neuroimage 12:230–239.
Andersson JL, Onoe H, Hetta J, Lidstrom K, Valind S, Lilja A,
Sundin A, Fasth KJ, Westerberg G, Broman JE, Watanabe Y,
Langstrom B. (1998): Brain networks affected by synchronized
sleep visualized by positron emission tomography. J Cereb
Blood Flow Metab 18:701–715.
Arfanakis K, Cordes D, Haughton VM, Moritz CH, Quigley MA,
Meyerand ME (2000): Combining independent component
analysis and correlation analysis to probe interregional connec-
tivity in fMRI task activation datasets. Magn Reson Imaging
Arieli A, Sterkin A, Grinvald A, Aertsen A (1996): Dynamics of
ongoing activity: Explanation of the large variability in evoked
cortical responses. Science 273:1868–1871.
Bandettini PA, Wong EC, Hinks RS, Tikofsky RS, Hyde JS (1992):
Time course EPI of human brain function during task activa-
tion. Magn Reson Med 25:390–397.
Berger H (1929): U¨ber das Elektroenkephalogramm des Menschen.
Archiv fu ¨r Psychiatrie und Nervenkrankheiten 87:527–570.
Binder JR, Frost JA, Hammeke TA, Bellgowan PS, Rao SM, Cox
RW (1999): Conceptual processing during the conscious resting
state. A functional MRI study. J Cogn Neurosci 11:80–95.
Birn RM, Diamond JB, Smith MA, Bandettini PA (2006): Separat-
ing respiratory-variation-related fluctuations from neuronal-
activity-related fluctuations in fMRI. Neuroimage 31:1536–1548.
Biswal B, Yetkin FZ, Haughton VM, Hyde JS (1995): Functional
connectivity in the motor cortex of resting human brain using
echo-planar MRI. Magn Reson Med 34:537–541.
Braun AR, Balkin TJ, Wesenten NJ, Carson RE, Varga M, Baldwin
P, Selbie S, Belenky G, Herscovitch P (1997): Regional cerebral
blood flow throughout the sleep-wake cycle. An H2(15)O PET
study. Brain 120 (Part 7):1173–1197.
Buchsbaum MS, Mirsky AF, DeLisi LE, Morihisa J, Karson CN,
Mendelson WB, King AC, Johnson J, Kessler R (1984): The
Genain Quadruplets: Electrophysiological, positron emission,
and X-ray tomographic studies. Psychiatry Res 13:95–108.
Bullmore ET, Suckling J, Overmeyer S, Rabe-Hesketh S, Taylor E,
Brammer MJ (1999): Global, voxel, and cluster tests, by theory
and permutation, for a difference between two groups of struc-
tural MR images of the brain. IEEE Trans Med Imaging 18:32–
Cordes D, Haughton VM, Arfanakis K, Wendt GJ, Turski PA,
Moritz CH, Quigley MA, Meyerand ME (2000): Mapping func-
tionally related regions of brain with functional connectivity
MR imaging. AJNR Am J Neuroradiol 21:1636–1644.
Cordes D, Haughton VM, Arfanakis K, Carew JD, Turski PA,
Moritz CH, Quigley MA, Meyerand ME (2001): Frequencies
contributing to functional connectivity in the cerebral cortex in
‘‘resting-state’’ data. AJNR Am J Neuroradiol 22:1326–1333.
Creutzfeldt OD, Watanabe S, Lux HD (1966): Relations between
EEG phenomena and potentials of single cortical cells. II. Spon-
taneous and convulsoid activity. Electroencephalogr Clin Neu-
De Luca M, Beckmann CF, De Stefano N, Matthews PM, Smith
SM (2006): fMRI resting state networks define distinct modes
of long-distance interactions in the human brain. NeuroImage
de Zwart JA, van Gelderen P, Kellman P, Duyn JH (2002): Appli-
cation of sensitivity-encoded echo-planar imaging for blood ox-
ygen level-dependent functional brain imaging. Magn Reson
de Zwart JA, Ledden PJ, van Gelderen P, Bodurka J, Chu R, Duyn
JH (2004): Signal-to-noise ratio and parallel imaging perform-
ance of a 16-channel receive-only brain coil array at 3.0 Tesla.
Magn Reson Med 51:22–26.
Drummond SP, Bischoff-Grethe A, Dinges DF, Ayalon L, Mednick
SC, Meloy MJ (2005): The neural basis of the psychomotor vigi-
lance task. Sleep 28:1059–1068.
Evarts EV, Bental E, Bihari B, Huttenlocher PR (1962): Spontane-
ous discharge of single neurons during sleep and waking. Sci-
Fox PT, Mintun MA (1989): Noninvasive functional brain mapping
by change-distribution analysis of averaged PET images of
H215O tissue activity. J Nucl Med 30:141–149.
Fransson P (2006): How default is the default mode of brain func-
tion? Further evidence from intrinsic BOLD signal fluctuations.
Fujii K, Heistad DD, Faraci F (1990): Vasomotion of basilar arteries
in vivo. Am J Physiol Heart Circ Physiol 258:1829.
Fukunaga M, Horovitz SG, van Gelderen P, de Zwart JA, Jansma
JM, Ikonomidou VN, Chu R, Deckers RHR, Leopold DA, Duyn
JH (2006): Large-amplitude, spatially correlated fluctuations in
BOLD fMRI signals during extended rest and early sleep
stages. Magn Reson Imaging 24:979–992.
Goldman RI, Stern JM, Engel J Jr, Cohen MS (2002): Simultaneous
EEG and fMRI of the alpha rhythm. Neuroreport 13:2487–2492.
Greicius MD, Krasnow B, Reiss AL, Menon V (2003): Functional
connectivity in the resting brain: A network analysis of the
default mode hypothesis. Proc Natl Acad Sci USA 100:253–258.
Gusnard DA, Raichle ME (2001): Searching for a baseline: Func-
tional imaging and the resting human brain. Nat Rev Neurosci
Hampson M, Peterson BS, Skudlarski P, Gatenby JC, Gore JC
(2002): Detection of functional connectivity using temporal cor-
relations in MR images. Hum Brain Mapp 15:247–262.
Hofle N, Paus T, Reutens D, Fiset P, Gotman J, Evans AC, Jones
BE (1997): Regional cerebral blood flow changes as a function
of delta and spindle activity during slow wave sleep in humans.
J Neurosci 17:4800–4808.
Holmes AP, Blair RC, Watson G, Ford I (1996): Nonparametric
analysis of statistic images from functional mapping experi-
ments. J Cereb Blood Flow Metab 16:7.
Howell DC (1997): Statistical Methods for Psychology. Belmont,
CA: Wadworth Publishing. 724 p.
Hudetz AG, Roman RJ, Harder DR (1992): Spontaneous flow oscil-
lations in the cerebral cortex during acute changes in mean ar-
terial pressure. J Cereb Blood Flow Metab 12:491–499.
Jones CJH, Kuo L, Davis MJ, Chilian WM (1995): Regulation of
coronary blood flow: Coordination of heterogeneous control
mechanisms in vascular microdomains. Cardiovasc Res 29:585–
Kajimura N, Uchiyama M, Takayama Y, Uchida S, Uema T, Kato
M, Sekimoto M, Watanabe T, Nakajima T, Horikoshi S, et al.
(1999): Activity of midbrain reticular formation and neocortex
during the progression of human non-rapid eye movement
sleep. J Neurosci 19:10065–10073.
Kajimura N, Nishikawa M, Uchiyama M, Kato M, Watanabe T,
Nakajima T, Hori T, Nakabayashi T, Sekimoto M, Ogawa K,
et al. (2004): Deactivation by benzodiazepine of the basal fore-
brain and amygdala in normal humans during sleep: A pla-
cebo-controlled [15O]H2O PET study. Am J Psychiatry 161:
r fMRI During NREM Sleep r
r 681 r
Kiviniemi VJ, Haanpaa H, Kantola JH, Jauhiainen J, Vainionpaa V, Download full-text
Alahuhta S, Tervonen O (2005): Midazolam sedation increases
fluctuation and synchrony of the resting brain BOLD signal.
Magn Reson Imaging 23:531–537.
Kjaer TW, Law I, Wiltschiotz G, Paulson OB, Madsen PL (2002):
Regional cerebral blood flow during light sleep–a H(2)(15)O-
PET study. J Sleep Res 11:201–207.
Kwong KK, Belliveau JW, Chesler DA, Goldberg IE, Weisskoff
RM, Poncelet BP, Kennedy DN, Hoppel BE, Cohen MS, Turner
R, et al. (1992): Dynamic magnetic resonance imaging of
human brain activity during primary sensory stimulation. Proc
Natl Acad Sci USA 89:5675–5679.
Laufs H, Kleinschmidt A, Beyerle A, Eger E, Salek-Haddadi A,
Preibisch C, Krakow K (2003): EEG-correlated fMRI of human
alpha activity. Neuroimage 19:1463–1476.
Laufs H, Holt JL, Elfont R, Krams M, Paul JS, Krakow K, Kleinsch-
midt A (2006): Where the BOLD signal goes when alpha EEG
leaves. Neuroimage 31:1408–1418.
Laureys S (2005): The neural correlate of (un)awareness: Lessons
from the vegetative state. Trends Cogn Sci 9:556–559.
Leopold DA, Murayama Y, Logothetis NK (2003): Very slow activ-
ity fluctuations in monkey visual cortex: Implications for func-
tional brain imaging. Cereb Cortex 13:422–433.
Lowe MJ, Mock BJ, Sorenson JA (1998): Functional connectivity in
single and multislice echoplanar imaging using resting-state
fluctuations. Neuroimage 7:119–132.
Lowe MJ, Dzemidzic M, Lurito JT, Mathews VP, Phillips MD
(2000): Correlations in low-frequency BOLD fluctuations reflect
cortico-cortical connections. Neuroimage 12:582–587.
Lund TE, Madsen KH, Sidaros K, Luo WL, Nichols TE (2006):
Non-white noise in fMRI: Does modelling have an impact?
Maquet P (2000): Functional neuroimaging of normal human sleep
by positron emission tomography. J Sleep Res 9:207–231.
Maquet P, Dive D, Salmon E, Sadzot B, Franco G, Poirrier R,
Franck G (1992): Cerebral glucose utilization during stage 2
sleep in man. Brain Res 571:149–153.
Mazoyer B, Zago L, Mellet E, Bricogne S, Etard O, Houde O, Cri-
vello F, Joliot M, Petit L, Tzourio-Mazoyer N (2001): Cortical
networks for working memory and executive functions sustain
the conscious resting state in man. Brain Res Bull 54:287–298.
McGuire PK, Paulesu E, Frackowiak RS, Frith CD (1996): Brain ac-
tivity during stimulus independent thought.
Mintun MA, Raichle ME, Martin WR, Herscovitch P (1984): Brain
oxygen utilization measured with O-15 radiotracers and posi-
tron emission tomography. J Nucl Med 25:177–187.
Mitra PP, Ogawa S, Hu X, Ugurbil K (1997): The nature of spatio-
temporal changes in cerebral hemodynamics as manifested in
functional magnetic resonance imaging. Magn Reson Med
Moosmann M, Ritter P, Krastel I, Brink A, Thees S, Blankenburg
F, Taskin B, Obrig H, Villringer A (2003): Correlates of alpha
rhythm in functional magnetic resonance imaging and near
infrared spectroscopy. Neuroimage 20:145–158.
Moruzzi G, Magoun HW (1995): Brain stem reticular formation
and activation of the EEG. 1949. J Neuropsychiatry Clin Neuro-
Nir Y, Hasson U, Levy I, Yeshurun Y, Malach R (2006): Wide-
spread functional connectivity and fMRI fluctuations in human
visual cortex in the absence of visual stimulation. Neuroimage
Nofzinger EA, Buysse DJ, Miewald JM, Meltzer CC, Price JC, Sem-
brat RC, Ombao H, Reynolds CF, Monk TH, Hall M, et al.
(2002): Human regional cerebral glucose metabolism during
non-rapid eye movement sleep in relation to waking. Brain 125
Obrig H, Neufang M, Wenzel R, Kohl M, Steinbrink J, Einhaupl
K, Villringer A (2000): Spontaneous low frequency oscillations
of cerebral hemodynamics and metabolism in human adults.
Ogawa S, Lee TM, Kay AR, Tank DW (1990): Brain magnetic reso-
nance imaging with contrast dependent on blood oxygenation.
Proc Natl Acad Sci USA 87:9868–9872.
Peigneux P, Laureys S, Delbeuck X, Maquet P (2001): Sleeping
brain, learning brain. The role of sleep for memory systems.
Raichle ME (2006): Neuroscience. The brain’s dark energy. Science
Raichle ME, MacLeod AM, Snyder AZ, Powers WJ, Gusnard DA,
Shulman GL (2001): A default mode of brain function. Proc
Natl Acad Sci USA 98:676–682.
Rechtschaffen A,Kales A, editors (1968): A Manual of Standarized
Terminology, Techniques and Scoring System for Sleep Stage
of Human Subjects. Bethesda, MD: U.S. Department of Health,
Education and Welfare. Public Health Service. National Insti-
tutes of Health, NINDS.
Saad ZS, Chen G, Reynolds RC, Christidis PP, Hammett KR, Bell-
gowan PS, Cox RW (2006): Functional imaging analysis contest
(FIAC) analysis according to AFNI and SUMA. Hum Brain
Sadato N, Pascual-Leone A, Grafman J, Ibanez V, Deiber MP,
Dold G, Hallett M (1996): Activation of the primary visual cor-
tex by Braille reading in blind subjects. Nature 380:526–528.
Sadato N, Nakamura S, Oohashi T, Nishina E, Fuwamoto Y, Waki
A, Yonekura Y (1998): Neural networks for generation and
suppression of alpha rhythm: A PET study. Neuroreport
Shulman GL, Corbetta M, Buckner RL, Fiez JA, Miezin FM,
Raichle ME, Petersen SE (1997a): Common blood flow changes
across visual tasks. I. Increases in subcortical structures and
cerebellum but not in nonvisual cortex. J Cogn Neurosci 9:624–
Shulman GL, Fiez JA, Corbetta M, Buckner RL, Miezin FM,
Raichle ME, Petersen SE (1997b): Common blood flow changes
across visual tasks. II. Decreases in cerebral cortex. J Cogn
Tzourio-Mazoyer N, Landeau B, Papathanassiou D, Crivello F,
Etard O, Delcroix N, Mazoyer B, Joliot M (2002): Automated
anatomical labeling of activations in SPM using a macroscopic
anatomical parcellation of the MNI MRI single-subject brain.
van de Ven VG, Formisano E, Prvulovic D, Roeder CH, Linden
DE (2004): Functional connectivity as revealed by spatial inde-
pendent component analysis of fMRI measurements during
rest. Hum Brain Mapp 22:165–178.
Wise RG, Ide K, Poulin MJ, Tracey I (2004): Resting fluctuations in
arterial carbon dioxide induce significant low frequency varia-
tions in BOLD signal. Neuroimage 21:1652–1664.
Zeman A (2001): Consciousness. Brain 124 (Part 7):1263–1289.
r Horovitz et al. r
r 682 r