Different frequencies for different scales of cortical integration: from

Transcription

International Journal of Psychophysiology 38 Ž2000. 301᎐313
Different frequencies for different scales of cortical
integration: from local gamma to long range
alphartheta synchronization
Astrid von SteinU , Johannes Sarnthein
Institute of Neuroinformatics, ETHr Uni¨ ersity Zurich, Winterthurerstr 190, CH 8057 Zurich, Switzerland
Received 6 October 1999; accepted 29 June 2000
Abstract
Cortical activity and perception are not driven by the external stimulus alone; rather sensory information has to be
integrated with various other internal constraints such as expectations, recent memories, planned actions, etc. The
question is how large scale integration over many remote and size-varying processes might be performed by the
brain. We have conducted a series of EEG recordings during processes thought to involve neuronal assemblies of
varying complexity. While local synchronization during visual processing evolved in the gamma frequency range,
synchronization between neighboring temporal and parietal cortex during multimodal semantic processing evolved in
a lower, the beta1 Ž12᎐18 Hz. frequency range, and long range fronto-parietal interactions during working memory
retention and mental imagery evolved in the theta and alpha Ž4᎐8 Hz, 8᎐12 Hz. frequency range. Thus, a
relationship seems to exist between the extent of functional integration and the synchronization-frequency. In
particular, long-range interactions in the alpha and theta ranges seem specifically involved in processing of internal
mental context, i.e. for top-down processing. We propose that large scale integration is performed by synchronization
among neurons and neuronal assemblies evolving in different frequency ranges. 䊚 2000 Elsevier Science B.V. All
rights reserved.
Keywords: Synchronization; Oscillation; Interaction; Top-down; Attention; Imagery; Memory; Semantics
1. Introduction
Sensory information is known to be propagated
U
Corresponding author. Tel.: q41-255-2179; fax: q41-2558946.
E-mail address: [email protected] ŽA. von Stein..
along the various input channels to the cortex. To
be used efficiently, however, this information has
to be integrated with various other sensory information and, more important, with internal constraints such as the behavioral goal pursued in
each particular situation. Evidently, this integration cannot be a sequential processing, as the
0167-8760r00r$ - see front matter 䊚 2000 Elsevier Science B.V. All rights reserved.
PII: S 0 1 6 7 - 8 7 6 0 Ž 0 0 . 0 0 1 7 2 - 0
302
A. ¨ on Stein, J. Sarnthein r International Journal of Psychophysiology 38 (2000) 301᎐313
behavioral goal substantially influences the selection of stimuli and thus the perceptual process
per se Ž‘top-down’ processes.. Rather, the multiple brain processes have to mutually influence
each other in any moment in time to produce a
coherent and behaviorally adequate outcome. Local stimulus properties have been shown to induce synchronous activity among visual sensory
cortical neurons Žfor review see Singer and Gray,
1995; Konig
¨ and Engel, 1995.. Similar processes
have been shown in other sensory domains ŽdeCharms and Merzenich, 1996; Murthy and Fetz,
1992.. Also, neuronal activity related to motor
planning, expectancy and decision making has
been established ŽRiehle et al., 1998; Shadlen and
Newsome, 1996.. It is, however, not known how
the different processes interact to obtain a multiple constraint satisfaction, and in particular, how
such large scale integration might be performed
by the brain.
We describe a series of EEG recordings during
processes thought to involve neuronal assemblies
of varying complexity. While local sensory integration evolved with a fast, gamma Ž25᎐70 Hz.
dynamics, multisensory integration evolved with
an intermediate, beta Ž12᎐18 Hz. dynamics, and
long-range integration during top-down processing evolved with a temporal dynamics in a low,
thetaralpha Ž4᎐12 Hz. frequency range. Inspired
by these results we propose that large-scale integration is performed by synchronization among
neurons and neuronal assemblies evolving with
varying temporal dynamics.
2. Relationship between scalp EEG and cortical
processing
Since synchronization of cortical activity was
found to play a role in cortical processing, it
seems of great benefit to investigate fast temporal
dynamics also in humans. Traces of intracortical
potentials can be measured even at distant electrodes, on the scalp ŽEEG., allowing to investigate processes with fast temporal dynamics ᎏ
albeit with low spatial resolution ᎏ also in human subjects. This allows us to extend the study
of synchronization processes onto humans, which
is desirable since in particular demanding topdown processes are only accessible in human subjects.
What is the relationship between EEG and
cortical processing? The EEG measured at a site
on the scalp consists of the summed electrical
field potentials from cortical neurons in a certain
volume of tissue under the electrode. Cortical
field potentials are generated by postsynaptic
dendritic currents. If several dendrites are arranged in parallel they generate potentials that
can be measured at the scalp, at sites which are
perpendicular to these dendrites. Effective contribution comes mostly from the synchronized
components of cortical activity ŽCooper et al.,
1965; Elul, 1972; Nunez, 1995, 2000.. Besides
synchronous input to the cortical input layers
which produces the dipolar field of the evoked
potential Že.g. Creutzfeldt et al., 1969., there is a
continuous change in cooperativity within the cortical network even without external stimulation
Žintrinsic potentials, Bullock, 1995; Lopes da Silva,
1991.. This changing pattern of synchronization
and desynchronization within different neuronal
groups of the network generates the macroscopic
potentials of the ongoing EEG ŽElul, 1972.. Since
intrinsic potentials stem from the network interactions, they generate low amplitude fields which
are effective only in close vicinity, thus influencing the potentials mostly at individual scalp electrodes. It has been estimated that one electrode
integrates cortical input under a scalp surface of
the order of 10 cm2 ŽNunez, 1995.. Changes in
the degree of synchronization of this intrinsic
activity are reflected in changes in EEG amplitude of different frequencies at the scalp. Investigation of induced changes of EEG spectral
parameters Žsee contributions to this volume.,
thus, is a suitable method to detect stimulusinduced changes of synchronization of this intrinsic activity ŽRappelsberger and Petsche, 1988;
Makeig, 1993; von Stein et al., 1993; Tallon et al.,
1995; Lutzenberger et al., 1995; Muller
et al.,
¨
1996..
While EEG amplitude predominantly reflects
the amount of synchronized activity in the underlying tissue, the degree of interactions between
two signals can be measured by coherence ŽRap-
pelsberger and Petsche, 1988.. Coherence is a
statistical measure: the value of coherence depends on the amount of repeated correlations
between events in the frequency domain. The
phase relationship between the two signals is irrelevant; however, it must be stable. Since the
signal at each electrode-site mostly reflects the
network activity under the electrode, coherence
between two electrodes should measure interactions between two neuronal populations; while
cortico-cortical correlations occupy only a minute
part of the power of the EEG-signal at each site,
the statistical nature of coherence helps to unravel them from noise if they repeat consistently.
Despite the success of cross-correlation analysis
in intracortical recording, analysis of inter-areal
interactions has only been rarely applied to the
human EEG ŽSarnthein et al., 1998; von Stein et
al., 1999a; Rodriguez et al., 1999.. A techniqual
question is whether functional coherence can be
distinguished from coherence arising because of
volume conduction between two electrodes. As
the spatial resolution of EEG has been estimated
to be approximately 5 cm ŽNunez, 1995; Srinivasan et al., in press., with an electrode spacing of
approximately 7 cm as in the 10r20 System
volume conduction becomes less likely. Also, in
volume conducted coherence, the amplitudes of
the signals at the concerned electrodes should
increase, and in both electrodes. To consider an
increase in coherence as functionally relevant, we
always check whether this is the case.
We aimed to investigate synchronization
processes in humans using both amplitude and
coherence analysis of the EEG.
3. Local, visual feature-dependent synchronization
in the gamma frequency range
As a first step in using the scalp EEG for
detection of synchronization processes in humans
Žvon Stein et al., 1995., we had to relate to the
results that were found in cats. In these experiments Žreview: Singer and Gray, 1995. it was
shown that Gestalt criteria of coherency such as
stimulus contingency and common motion are
translated into coherent neuronal firing. The
303
difficulty to be dealt with in reproducing these
results with scalp recordings was the large amount
of averaging that lies between the single neuronal
potential and the scalp EEG. The following argument was used to develop an appropriate experimental design: if the coherency within a stimulus
is translated into coherent neuronal firing, then
by increasing the coherency within a stimulus we
should be able to increase the amount of coherent neuronal activity, and in this way neuronal
synchronization should become large enough to
be macroscopically visible on the scalp. We therefore presented parallel grating stimuli ŽFig. 1a.
containing an increasing number of bars per texture and expected cortical synchronization to increase. Scalp EEG ŽFig. 1b. was recorded during
presentation of 16 such grating stimuli Ž n s 40
each. presented in a randomized order. Average
power was determined in five frequency ranges.
The hypothesis was that power in the gamma
frequency range would increase with the number
of bars per grating.
The experiment was performed in four different sessions. A positive correlation was indeed
found ŽFig. 1d. in all four experiments. The maximum of the correlation was at lateral occipital
electrode sites, roughly corresponding to the horizontal median. No correlation was found over
temporal cortex Želectrodes T5rT6.. This is compatible with the interpretation that the increased
power reflects increased synchrony among the
columns of the first stages of visual cortex. In all
subjects, two frequency ranges showed such a
positive correlation with the number of bars per
degree: gamma, and alpha. The increase of alpha
was also present at temporal electrodes, indicating the involvement of alpha in spatially extended
processes Žsee below.. To control for a simple
increase of the number of responding neurons to
the higher density of contrasts, we presented a
stimulus with the same amount of bars, with,
however, disturbed coherency ŽFig. 1c.. If the
increase in EEG power would be due to an increase in the mere number of generators and not
their synchrony, then in the case of disturbed
coherency, EEG power should remain equal Žor
increase due to the slight increase of contrast
borders.. This, however, was not the case. As
304
Fig. 1. Ža. Stimuli: 16 types of parallel grating stimuli with different numbers of bars per degree of visual field were presented for 2
s, interstimulus interval was 2᎐3 s. The gratings were presented in randomized order, each n s 40 times. Ža. In a control stimulus,
subparts of the gratings were flipped to disturb coherency Žc. . Žb. Recording and data analysis: Scalp EEG was recorded referred to
the digitally averaged signal of both ear lobes from an electrode array consisting of five electrode rows above the occipital cortex
plus two temporal cortical electrodes ŽT5, T6.. ŽSampling rate 256 Hz, time constant of 0.3 s, low-pass filtering at 70 Hz..
Non-overlapping artifact-free 2-s epochs during stimulus presentation Ž n s 40 for each type. were Fourier transformed, and
averaged power spectra were computed for the six frequency ranges theta Ž4᎐7 Hz., alpha Ž8᎐12 Hz., beta1 Ž12᎐18 Hz., beta2
Ž19᎐24 Hz., gamma1 Ž24᎐35 Hz. and gamma 2 Ž35᎐70 Hz.. Žd. Results: The amplitudes in the gamma Ž24᎐35 Hz. frequency range
are plotted for the 16 different grating stimuli. An increasing EEG gamma amplitude is observed with increasing number of parallel
bars per stimulus, i.e. with increasing coherent elements in the stimulus Ž P- 0.00001.. Že. If coherency within the stimulus is
disturbed by flipping subparts of the parallel gratings Žc.. EEG amplitudes drop, indicating that the different amounts of gamma
amplitude do not just reflect the amount of contrast in the stimulus. Žflip 13:13 subparts flipped; e 74:74 subparts flipped. Žvon Stein
et al., 1995..
predicted by our hypothesis, gamma power during
the uncoherent stimulus dropped considerably
ŽFig. 1e ). Thus, the crucial feature playing a role
was indeed the coherence in the stimulus and in
the neuronal activity. These results indicate that,
similar to what has been found in the cat, cortical
synchronization in the gamma frequency range
can be measured by EEG power in human subjects and reflects visual Gestalt criteria of
coherency.
4. Supramodal processing induces temporoparietal interactions that evolve with a lower
(12–18 Hz) temporal dynamics than gamma
As a next step to expand the knowledge about
cortical synchronization, we wanted to investigate
a related but more global process Žvon Stein et
al., 1999a.. Objects are more than visual features.
Only if sensory experience through all sensory
channels is consistent and if sensory᎐motor interactions lead to repeatedly similar results, then the
organism will represent such a sensory constellation as an ‘object’. Thus, objects are multimodal,
semantical entities. Until now, synchronization
has been found locally, between neurons from
different visual areas ŽFrien, 1994. and across the
two visual hemispheres ŽEngel et al., 1991.. Also,
the relevance of synchronization has been extended to other sensory domains such as the
auditory cortex or the motor cortex ŽdeCharms
and Merzenich, 1996; Murthy et al., 1992; Riehle
et al., 1997.. Few studies have shown perception-
related correlations between neurons from different modality-specific areas. If coherent activity
between the different feature-representations is a
possible mechanism to guarantee their binding
into a unified percept Žfor review see Konig
¨ et al.,
1995., however, then coherent activity should also
be observed between neuronal populations from
305
areas serving different modalities, e.g. during representation of multi-modal semantical entities,
i.e. objects. Therefore, we investigated cortical
Fig. 2. Stimuli: 40 concrete objects of daily life were presented
in three modalities, as written word Ž written presentation., as
spoken word Ž auditory presentation., and as picture Ž pictorial
presentation. . Additionally Žnot shown., non-sense items were
presented in the same three modalities Žnon-sense pictures
were line drawings consisting of the same elements as the
pictorial presentations of objects, but not denoting known
objects.. Stimulus presentation 2 s, interstimulus intervals 2᎐3
s in all conditions. Stimuli of a category were presented in
blocks; within each block, stimuli were randomized. As a
baseline condition Ž‘EO’, eyes open. subjects has to relax and
fixate a spot on the screen positioned within the same distance
as the features of the stimulus conditions.Recording and data
analysis: EEG was recorded from 19 electrodes distributed
over the scalp according to the 10᎐20 system Žsampling rate
128 Hz, time constant of 0.3 s and low-pass filtering at 35 Hz..
Averaged power spectra, C x x , and cross-power spectra, C x y ,
for all neighboring electrode pairs were computed over the 2 s
epochs of stimulus presentation for all six stimulus conditions
and the control conditions. Spectral bins were averaged to
obtain mean values for the six frequency ranges delta Ž2᎐4
Hz., theta Ž4᎐8 Hz., alpha Ž8᎐12 Hz., beta1 Ž13᎐18 Hz., beta2
Ž18᎐24 Hz. and gamma Ž24᎐32 Hz.. Coherence K x y for two
signals, x and y, is equal to the average cross-power spectrum
normalized by the averaged powers of the compared signals:
K x y sN C x y N 2 rŽ C x x C y y .. Coherence is the frequency domain
equivalent to the cross-covariance function, it estimates the
degree to which phases at the frequency of interest are
dispersed. Statistically significant differences between the stimulus and the baseline condition were determined using a
paired Wilcoxon test and were interpreted as induced
powerrcoherence. Žb. Results: Inter-electrode sites where
coherence changed significantly with respect to the baseline
condition are mapped, for all three presentation modes Žwritten, auditory, pictorial.. Solid lines indicate an increase, dashed
lines indicate a decrease of coherence. Only changes that
where significant with an error probability of P- 0.05 or
better are depicted Ž n s 19 subjects.. For a better anatomical
orientation, electrode positions are depicted in relation to
cortical structure as determined by an MRI study ŽHoman et
al., 1987.. The shaded areas indicate the variance of electrode
positions obtained with the 10r20 system with respect to the
cortical surface. Common to all three conditions dealing with
objects is an increased coherence between electrodes above
temporal and above parietal cortex Žthick line. in the 13᎐18
Hz range, bilateral. Žc. Same as Žb. for the non-sense conditions. Common to the non-sense conditions is a similar increase in coherent activity between temporal and parietal
electrodes, however mostly at the right hemisphere Žthick line.
Žfrom: von Stein et al., 1999..
306
activity related to supramodal semantical processing of objects.
Our experimental design was to present the
same object in three different modalities and to
search for a correlate of modality independent
processing, i.e. a common electrophysiological
pattern underlying all three ways of presentation.
In this study we extended the analysis of EEG
power to the computation of spectral coherence,
a measure for inter-areal cross-correlation. Forty
daily life objects were presented as spoken words,
as written words, and as pictures Ž n s 19 subjects.
ŽFig. 2a.. EEG was measured over 19 electrodes
covering the scalp Ž10r20 system.. Power and
inter-electrode coherence were computed in six
different frequency ranges for each condition and
for a baseline condition.
The question was whether any induced change
of power and coherence appeared consistently in
all three presentation modes. Such a consistently
induced pattern was a change in inter-areal synchronization between temporal and parietal electrodes ŽFig. 2b, black lines.. Thus, modality independent processing of objects seems to induce
synchronization of activity between temporal cortex and parietal cortex. Interestingly, there was
no such consistent change of power at single
electrodes , implying that there are no local ensembles that seem to be responsible for supramodal representations. The fact that the increase
in coherence was not accompanied by an increase
in power might be interpreted as increased phase
locking between the neuronal activity of two cortical sites without an increase in the number of
active neurons. Thus, what was commonly induced seems not a global activation of one specific area but rather a phase-locking between neurons of different cortical areas, i.e. members of a
distributed ensemble. The frequency range of this
synchronized activity was not gamma as seen during visually induced synchronization, but the lower
beta1 range Ž12᎐18 Hz..
Similar changes of temporo-parietal coherence
but with a tendency of right hemisphere lateralization were found during a control condition, the
presentation of pronounceable non-sense words
ŽFig. 2c., not, however, during non-pronounceable
letter strings Ždata not shown.. This confirms that
the dominant process behind the temporo-parietal
synchronization during object perception is the
inter-modal binding. Furthermore, it stresses the
importance of only the left temporo-parietal cortex for semantic processing. The leftrright lateralization fits to common knowledge about hemispheric specialization ŽKimura, 1973..
We did not only observe induced coherence in
the beta1 frequency range, but also in the gamma
frequency range and other frequencies; the pattern of changes, however, was not attributable to
one of the two aspects of processing addressed
with this study, semantics or modality. Reasons
that no sensory-specific gamma changes were
observed might be, on one hand, that the gamma
range investigated here was chosen rather low to
avoid contamination with muscle artifacts. On the
other hand, sensory-specific changes in gamma
might occur on a shorter time scale; the window
of our analysis, however, was chosen rather long
Ž2 s. in order to detect late, longer lasting semantical components of processing.
A relation of supramodal, semantical processing to brain structures of temporal and parietal
cortex is in agreement with early neuropsychological studies. The area of posterior parietal cortex,
especially left angular gyrus, is known as a supramodal, integrating area Že.g. Luria, 1973., with
lesions in that region leading to failure in crossmodal matching ŽButters and Brody, 1968. and to
deficits in forming higher integrative concepts
Ž agnosias.. Temporal cortex, on the other hand, is
part of the ventral stream of visual processing and
a higher order visual association cortex Že.g.
Tanaka, 1996., and has also been shown to be
involved in visual imagery Že.g. Roland and
Gulyas, 1994.. An increase of correlated activity
between both structures, temporal cortex and
parietal cortex, during perception of objects might
therefore indicate an interaction between visual
representations Žtemporal cortex. and higher order representations Žparietal cortex. for multimodal semantic integration in humans. Furthermore, parietal cortex is a highly connected, supramodal center in the cortex, connected with different sensory cortices, association cortices, subcortical structures, etc. ŽMesulam, 1990., thus
making it a suitable candidate for the proposed
function of multimodal semantic integration. The
interactions between temporal and parietal cortex
might also be interpreted as indicating an integration of information carried by the dorsal and the
ventral stream. It has often been proposed that
the two have to be integrated at various levels
ŽDeYoe and Van Essen, 1988..
Taken together, our results indicate that similar processes of binding as in the visual domain
are also observable for multi-modal semantic objects; they deviate in the fact that synchronization
during supra-modal processing does not evolve
with a temporal dynamics in the gamma frequency, but in a lower Žbeta1, 12᎐18 Hz. frequency domain.
5. Very long range interactions during working
memory processes evolve with a temporal
dynamics in the theta (4–8 Hz) frequency range
In a third experiment, we investigated synchronization during a process known to involve a
system of largely separated cortical areas
ŽSarnthein et al., 1998.. Working-memory is
known to involve at least two separate structures,
prefrontal cortex and posterior association cortex.
Additionally, during working-memory retention no
external stimuli are presented to the brain, which
allows us to investigate cortical activity during
purely internal mental processes. Working-memory retention might be considered a prototype of
‘top-down’ activity, where the origin of this internal activity is still partly controlled by the previously presented stimulus.
Two subsets of a working memory paradigm
were performed ŽFig. 3a., both requiring the active retention of a stimulus in short term memory.
In Task I Ž character strings., the stimulus was a
string of 6᎐8 standard keyboard characters. First,
the stimulus was presented for 6 s Žperception
interval.. After stimulus offset, subjects actively
retained the stimulus in memory for 4 s Žretention
interval. while fixating the dark screen. Finally,
subjects were cued by a white screen onset to
reproduce the stimulus with paper and pencil.
Although there was no auditory stimulation in
Task I, subjects reported transforming the visu-
307
ally-presented characters into phonological equivalents and rehearsing them subvocally during the
retention interval. In Task II Ž ¨ isuo-spatial condition., stimuli were abstract line drawings activating visuo-spatial working memory. As a control
condition, in the beginning of the session a 2-min
period was chosen during which subjects fixated
the dark computer screen. EEG was recorded
during the 6 s of stimulus presentation, the retention period, and the baseline condition. Power
and coherence were investigated for all four different epochs and changes with respect to a baseline condition were determined. Of major interest
are the changes in powerrcoherence induced
during the retention epoch; changes here reflect
differences between two purely mental states
without stimulus presentation, the baseline state
and a state of mental rehearsal.
The major pattern of synchronization that was
seen in both subsets of the mental rehearsal
period was a highly significant interaction between
prefrontal cortex and posterior cortex, in the theta
frequency range ŽFig. 3b.. No such interaction
was found during perception ŽFig. 3c.. Although
significant interactions were found in other frequency ranges, the interactions in the theta range
were the only interactions that were specific for
the retention period ŽTable 1.. Thus, our findings
are evidence that synchronization of activity plays
a role for mediating interactions between prefrontal cortex and posterior association cortex
during working memory processes. The importance of these areas for working memory is well
established by extensive experiments on humans
and non-human primates using a variety of techniques. Single-cell recordings have found prefontal and parietal neurons most active during
delay periods in delayed-response tasks ŽGoldman-Rakic, 1997; Fuster, 1995.. Several imaging
studies on working memory in humans have reported increased activity in both prefrontal cortex
and posterior regions Že.g. Smith et al., 1996;
Tallon-Baudry et al., 1998.. Anatomical studies
indicate high connectivity between frontal and
posterior association cortices ŽGoldman-Rakic,
1988.. A functional interaction was already implied by the fact that cooling parietal cortex reduces the activity of prefrontal neurons in
308
macaque brain ŽFuster, 1995., the interaction itself, however, had remained unclear. Our findings
suggest that synchronized neuronal activity in the
4᎐8 Hz range plays a role in the interaction of
posterior association cortex, where sensory information is thought to be stored, and prefrontal
cortex, where relevant current information is held
and continuously updated.
These findings indicate that very long range
cortico-cortical interactions between prefrontal
and posterior cortex during working memory retention evolve with a temporal dynamics in a low
frequency range Žtheta, 4᎐8 Hz.. Furthermore,
they show that purely mental processes, i.e.
processes without sensory input, may cause specific patterns of interactions and in a particular
frequency range.
6. Relation between the size of a neuronal
assembly and the frequency range of integration
In the set of experiments presented, the tasks
were such that the cortical area involved extends
over different length scales. During local visual
processing, we found evidence for synchronization among members of ensembles reflecting stimulus coherence; in accordance with other reports ŽSinger and Gray, 1995. this synchronization
evolved in the gamma frequency range. During
semantical processing of objects, assemblies are
multimodal and thus should be more extended. A
pattern of synchronization was induced that consists of interactions between temporal cortex and
parietal cortex. In contrast to the synchronization
between columns in visual cortex, this interaction,
however, evolved in a lower, the beta1 Ž12᎐18 Hz.
frequency range. Finally, synchronous activity
could even be found between the largely separated areas of frontal cortex and posterior association cortex during working memory retention.
This very long-range interaction was found in the
theta Ž4᎐8 Hz. frequency range. These results
suggest a relation between the size and distance
of an interaction and the frequency of synchronization: the larger the neuronal assembly involved, the lower the frequency in which activity
in the assembly gets synchronized. Related evidence was found within primary visual cortex: the
Table 1
Similarity between patterns of enhanced coherence. The table
compares the coherence patterns of all task conditions in each
frequency band Žp: perception, r: retention.; delta Ž2᎐4 Hz.,
theta Ž4᎐8 Hz., alpha Ž8᎐12 Hz., beta1 Ž13᎐18 Hz., beta2
Ž18᎐24 Hz. and gamma Ž24᎐35 Hz.. The similarity S between
two patterns is quantified as the ratio of equal connections
with respect to the total number of connections in both maps.
A coherence pattern was only considered as characteristic for
working memory processes if it was Ži. specific for the
retention period Ždistinguished retention period from perception period., and Žii. similar in the working memory subconditions. This condition was only met by the theta coherence
patterns Žhigh similarity between the visual and the verbal
condition, but not between retention and perception., pointing
to the importance of theta activity for working memory
retention. In gamma, the high similarity between the patterns
during retention and those during perception of items might
indicate that changes in the gamma frequency range reflect
primarily coherence induced by sensory processing. In agreement with other studies ŽTallon-Baudry et al., 1998. we
observed an increase in gamma power locally, at temporoparietal and frontal cortical electrodes Žnot shown. Žfrom:
Sarnthein et al., 1998. Copyright Ž1998. National Academy of
Sciences, U.S.A...
Verbal
Visual
Verbal
Visual
Verbal
Visual
Verbal
Visual
Verbal
Visual
Verbal
Visual
Verbal
p
r
p
r
p
r
1
0
0
0
1
0
0
p
r
p
r
1
0
0
0
1
0
0.35
p
r
p
r
1
0
0.31
0
1
0
0.22
p
r
p
r
1
0
0
0
1
0.11
0.21
p
r
p
r
1
0
0.27
0.40
1
0.12
0
p
r
p
r
1
0.54
0.60
0.41
1
0.42
0.42
Visual
p
r
Delta
1
0
1
Theta
1
0
1
Alpha1
1
0.27
1
Alpha2
1
0.60
1
Beta
1
0.14
1
Gamma
1
0.52
1
309
activity in the gamma range was repeatedly shown
to decline with distance, reaching unsynchronized levels at a distance between recording sites
) 8᎐10 mm ŽEckhorn, 1994; Bullock, 1995.. This,
however, raises questions regarding the hypothesis that feature binding is mediated by coherent
neuronal activity, as coherency should not be
restricted to local features but should also be
found between largely separate areas. Our proposal that lower frequencies play a role in mediating long range interactions thus might resolve
that question. In agreement with this proposal,
simulation studies have shown that gamma synchrony is lost over longer distances and that lower
frequency interactions are better suited to sustain
long range synchronization ŽKopell et al., 2000..
Recent evidence obtained with multiple recordings in area 17 and area 7 further confirmed this
larger the spatial extent of a stimulus, the slower
was gamma synchronization ŽEckhorn, 1994.. A
theoretical framework for an inverse relation
between frequency of activity and spatial scale of
a network has been given by Nunez Ž1999.. According to our results, however, the network is
not defined on anatomical grounds but rather is
recruited functionally according to the cognitive
task. Together, we propose that as a manifestation of cooperative activity in larger neuronal
ensemble, inter-areal synchronization during
processes extending over more than one area
evolve with a slower temporal dynamics than local
visual processing.
The finding that gamma synchronization is a
local process is further based on several observations from intracortical recordings. Synchronous
Fig. 3. Ža. Experimental paradigm: Two types of stimuli had to
be retained in short term memory: ŽI. character strings, which
usually were transformed into a phonological equivalent and
rehearsed subvocally Žverbal memory. and ŽII. abstract line
drawings activating visuo-spatial memory. In the sequence of
events, first the stimulus was presented for 6 s Žperception.;
during the ensuing dark-screen interval of 4 s subjects actively
retained the stimulus in memory Žretention.. Finally, subjects
reproduced the stimuli with paper and pencil Žreproduction..
Total number of trials was n s 20. Baseline condition : As a
control condition, in the beginning of the session a 2-min
period was chosen during which subjects fixated the dark
computer screen. This period equals the retention interval
except for the active rehearsal of stimuli in memory. Six
subjects participated in total 10 sessions. Mean memory performance was consistent with normal digit span. EEG recordings and data analysis as in Ža.. Žb. Results: Inter-areal coherence while subjects are mentally rehearsing strings of digits
Žleft. and abstract line drawings Žright.. A significant increase
of coherence Ž P - 0.05 or better. with respect to the baseline
condition is marked by a connection between the two electrodes of a pair. During the retention of character strings
Žleft. coherence between electrodes over prefrontal cortex and
posterior association cortex increased remarkably. A similar
pattern was found during abstract line drawings Žleft.. The
frequency range was theta Ž4᎐8 Hz.. Žc. The fronto-posterior
enhancement of theta coherence was specific for the retention
period. As an example for a different behavioral state, here
theta coherence during perception of stimuli is shown. There
is no similarity between the patterns. Perception does not
seem to rely on low frequent interactions Žfrom: Sarnthein et
al., 1998. Copyright Ž1998. National Academy of Sciences,
U.S.A...
310
hypothesis, showing that intra-areal interactions
evolved in the gamma range, while long range
interactions were mediated by cross-correlations
in lower, alpha and theta frequency ranges in cats
Žvon Stein et al., 2000.. In contradiction to those
results showing gamma synchronization to be locally restricted, there are several reports showing
gamma synchrony over longer distances such as
between the two hemispheres ŽEngel et al., 1991.
or between distant electrodes in EEG recordings
ŽRodriguez et al., 1999.. The study by von Stein et
al. Ž2000. partly solves this conflict, as it gives
evidence that whether two sites show gamma synchronization or not does not depend on the absolute distance between the sites, but rather on the
number of synapses involved in an interaction.
Indeed, a common feature in most examples
showing inter-areal gamma synchronization Že.g.
Engel et al., 1991; Frien et al., 1994; Roelfsema et
al., 1997; Castelo-Branco et al., 1998. are strong
monosynaptic connections between the investigated areas. Similarly, the range of 8 mm where
gamma synchronization was found within an area,
is the range of the horizontal ᎏ monosynaptic ᎏ
connections. Thus, gamma synchronzation seems
a local phenomenon in the sense of being restricted to monosynaptic interactions, either
within one area or between neurons of different
areas that are coupled monosynaptically.
Summarizing, we propose that long range integration during processes extending over more than
one area is mediated by correlated activity in
lower frequency ranges; a relationship might exist
between the extension of the integrating process
and the frequency range of synchronization.
7. Examples of complex multiple integration: the
role of alpha r theta synchronization in top-down
processes
Two of the above described observations
concerning integration over largely extended cortical areas may be related: first, very long range
integration occurred in a low frequency range
Ž4᎐8 Hz., and second, long range interactions
were dominant in a state without actual sensory
input, in the retention period of a working me-
mory paradigm. The mental rehearsal of pictorial
items as during retention is a process where a
‘mental’ sensory experience is produced not by an
external stimulus but by internal processes. Such
a kind of activity is called top-down activity. Is
there a general relation between low frequency
interactions and top-down processes? Indeed, a
similar predominance of low frequency interactions was found during other forms of purely
internal mental activity. In a study on mental
imagery, the predominant feature was an increased inter-areal coherence in the theta Ž4᎐8
Hz. and alpha Ž8᎐12 Hz. frequency range ŽPetsche
et al., 1997., and during mentally playing an instrument we found characteristic coherence patterns in the alpha frequency range Žunpublished
data.. This suggests that low frequency interactions in general might be a characteristic feature
of top-down activities. During top-down processing the input into the sensory areas is highly
convergent from multiple cortical areas. A commonly assumed goal of top-down processing is to
produce a ‘summary’ of all momentary activities
in the cortex to generate an hypothesis about
fruitful future actions that might then guide the
selection of new sensory input. The computation
performed can thus be considered an integration
over activities of most available cortical areas.
Given the proposed relationship between the
complexity of interaction Žthe number of elements taking part in the interaction . and the
frequency range in which the dynamics in a network evolve, an involvement of low frequencies in
the complex interactions of top-down processing
seems a plausible generalization.
The described results imply that theta and in
particular alpha activity are related to top-down
processing. This may seem astonishing since alpha activity is commonly known as an ‘idling
rhythm’ of the brain. We view this not as a
contradiction but rather as supporting evidence.
Our argument is the following: Large amplitude
8᎐12-Hz alpha activity is prevalent above the
visual cortex in states with eyes closed and disappears upon eye opening. Alpha activity is also
found with eyes opened, where it is maximal in
states of quiet restfulness, i.e. in states without
focussed attention, and disappears once attention
is focussed to a specific stimulus ŽAdrian, 1947..
These two observations have led to the notion of
alpha activity as reflecting an inactive cortex. Our
proposal now is to redefine this notion. We propose that states with maximal alpha activity such
as with eyes closed and during rest do not reflect
inacti¨ e brain states, but rather reflect states with
internal mental activity. Rather than reflecting the
absence of cortical processing, they reflect the
absence of bottom up processing and thus can be
classified a pure form of top-down activity. In the
light of this reinterpretation, the hypothesis that
slow frequency interactions reflect top-down processing has gained further support in the classical
alpha rhythm. It is maximal in situations where
cortical processes are not determined by external
stimuli but are driven by free floating associations, mental imagery, planning, etc. There is
evidence that supports the interpretation that alpha is related to top-down processing instead of
idling. Strong alpha activity has been found in cat
intracortical recordings in states of intense expectancy ŽChatila et al., 1992., i.e. in a state that
is classically associated with top-down processing
and generating hypothesis about the environment. Along similar lines, prestimulus background activity ᎏ dominated by activity in the
alpha range ᎏ has a major influence on processing of sensory stimuli ŽArieli et al., 1995, 1996.,
compatible with the top-down process of stimulus
selection. Furthermore, although spindle generation in thalamocortical loops upon unarousal
ŽSteriade et al., 1990. has often been used as a
model for alpha generation, it has been shown
that alpha may also be generated in cortex itself
ŽLopes da Silva and van Leeuwen, 1977.. Finally,
it has been found recently in behaving cats that
top-down contexts exert an effect on alpha and
theta inter-areal synchronization, with phase lags
and a layer specificity compatible with a top-down
directed interaction Žvon Stein et al., 1999b, 2000;
Bernasconi and Konig,
1999.. The plausibility of
¨
these ideas is also supported by a computer simulation on a cellular level ŽSiegel et al., 2000..
Top-down related changes in gamma frequency
activity or synchronization has been observed in
some experiments Že.g. Tallon-Baudry et al., 1998..
These results provide no contradiction to the
311
presented hypothesis, as for a useful processing of
data, local information has to be integrated with
top-down information, and top-down information
also should be able to instantiate or maintain
local sensory pattern of synchronization, as, e.g. is
supposed to be the case during working memory
retention.
We therefore propose that low frequency Žtheta,
alpha. cortico-cortical interactions reflect ‘topdown’ processing, subserving the function to integrate over multiple simultaneous brain activities.
Low frequency cortico-cortical interactions would
allow the local gamma-synchronized processes to
be integrated, thus forming a unified mental construction such as a mental image, a hypothesis, a
planned action, or a ‘thought’.
8. Conclusions
Our results obtained in several EEG experiments in humans support the hypothesis that
cortico-cortical interactions may occur on different scales, and that the complexity of interaction
influences the temporal dynamics evolving within
the integrating network. It was shown that local
interactions during visual processing involve
gamma frequency dynamics, semantical interactions between temporal and parietal cortex involve beta frequency dynamics, and very long
range interactions during internal mental Žtopdown. processes such as working-memory-retention or visual imagery involve interactions in a
low Žtheta or alpha. frequency range. Thus, an
inverse relation exists between the scale of integration and the frequency of interaction. Additionally, these studies demonstrated the involvement
of alphartheta frequency interactions in processes
reflecting internal mental activity, i.e. top-down
effects. Both results may be related, and top-down
processing may show low-frequency interactions
as a reflection of a complex cortical computation,
where the information of many cortical areas is
massively integrated. In particular, large-scale low
frequency interactions might allow an integration
with the different local, fast gamma processes,
with which sensory information from the periphery seems to be propagated ‘bottom-up’.
312
Acknowledgements
We cordially thank Dr. Peter Konig
¨ for fruitful
discussions on top-down processing. This work
was supported by Swiss National Science Foundation grant. 3100-51 ’059.97 to Peter Konig.
¨
References
Adrian, E.D., 1947. The physical background of perception.
Clarendon Press, Oxford.
Arieli, A., Shoham, D., Hildesheim, R., Grinvald, A., 1995.
Coherent spatiotemporal patterns of ongoing activity revealed by real-time optical imaging coupled with single-unit
recording in the cat visual cortex. J. Neurophysiol. 73,
2072᎐2093.
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.
Bernasconi, C., Konig,
P., 1999. On the direction of interareal
¨
interactions in the cat visual system. Biol. Cybern. 81,
199᎐210.
Bullock, T.H., 1995. Temporal fluctuation in coherence of
brain waves. Proc. Natl. Acad. Sci. 92, 11568᎐11572.
Butters, N., Brody, B.A., 1968. The role of the left parietal
lobe in the mediation of intra- and cross-modal associations. Cortex 4, 328᎐343.
Castelo-Branco, M., Neuenschwander, S., Singer, W., 1998.
Synchronization of visual responses between the cortex,
lateral geniculate nucleus, and retina in the anesthetized
cat. Neuroscience 1998 Ž16., 6395᎐6410.
Chatila, M., Milleret, C., Buser, P., Rougeul, A., 1992. A 10
Hz alpha-like rhythm in the visual-cortex of the waking cat.
Electroencephalogr. Clin. Neurophysiol. 83, 217᎐222.
Cooper, R., Winter, A.L., Crow, H.J., Walter, W., 1965. Comparison of subcortical, cortical and scalp activity using
chronically indwelling electrodes in man. Electroenceph.
Clin. Neurophysiol. 18, 217᎐228.
Creutzfeldt, O., Rosina, A., Ito, M., Probst, W., 1969. Visual
evoked response of single cells and of the EEG in primary
visual area of the cat. J. Neurophysiol. 32 Ž2., 127᎐139.
deCharms, R.C., Merzenich, M.M., 1996. Primary cortical
representation of sounds by the coordination of actionpotential timing. Nature 381, 610᎐613.
DeYoe, E.A., Van Essen, D.C., 1988. Concurrent processing
streams in monkey visual cortex. Trends Neurosci. 11,
219᎐226.
Eckhorn, R., 1994. Oscillatory and non-oscillatory synchronizations in the visual cortex and their possible roles in the
associations of visual features. In: van Pelt, J., Corner,
M.A., Lopes da Silva, F.H. ŽEds.., Progress in Brain Research, Elsevier.
Elul, R., 1972. The genesis of the EEG. Int. Rev. Neurobiol.
15, 227᎐272.
Engel, A.K., Konig,
P., Kreiter, A.K., Singer, W., 1991. Inter¨
hemispheric synchronization of oscillatory neuronal responses in cat visual cortex. Science 252, 1177᎐1179.
Frien, A., Eckhorn, R., Bauer, R., Woelbern, T., Kehr, H.,
1994. Stimulus-specific fast oscillations at zero phase
between visual areas V1 and V2 of awake monkey. Neuroreport 5 Ž17., 2273᎐2277.
Fuster, J.M., 1995. Temporal processing. Ann. N.Y. Acad. Sci.
769, 173᎐181.
Goldman-Rakic, P., 1997. Neurobiology. Space and time in the
mental universe. Nature 386 Ž6625., 559᎐560.
Goldman-Rakic, P.S., 1988. Topography of cognition: parallel
distributed networks in primate association cortex. Annu.
Rev. Neurosci. 11, 137᎐156.
Homan, R.W., Herman, J., Purdy, P., 1987. Cerebral location
of International 10᎐20 System electrode placement. Electroencephr. Clin. Neurophysiol. 66, 376᎐382.
Kimura, D., 1973. The asymmetry of the human brain. Sci.
Am. 228, 70᎐78.
Kopell, N., Ermentrout, G.B., Whittington, M.A., Traub, R.D.,
2000. Gamma rhythms and beta rhythms have different
synchronization properties. Proc. Natl. Acad. Sci. 97 Ž4.,
1867᎐1872.
Konig,
P., Engel, A.K., 1995. Correlated firing in sensory-mo¨
tor systems. Curr. Opin. Neurobiol. 5, 511᎐519.
Lopes da Silva, F.H., Storm van Leeuwen, W., 1977. The
cortical source of alpha rhythm. Neurosci. Lett. 6, 237᎐241.
Lopes da Silva, F., 1991. Neural mechanisms underlying brain
waves: from neural membranes to networks. Electroencephalogr. Clin. Neurophysiol. 79, 81᎐93.
Luria, A.R., 1973. The Working Brain. New York:Penguin.
Lutzenberger, W., Pulvermueller, F., Elbert, T., Birbaumer,
N., 1995. Visual stimulation alters local 40 Hz responses in
humans: an EEG study. Neurosci. Lett. 183 Ž1-2., 39᎐42.
Makeig, S., 1993. Auditory event-related dynamics of the EEG
spectrum and effects of exposure to tones. Electroencephalogr. Clin. Neurophysiol. 86 Ž4., 283᎐293.
Mesulam, M.M., 1990. Large scale neurocognitive networks
and distributed processing for attention, language and memory. Ann. Neurol. 28, 597᎐613.
Muller,
M.M., Bosch, J., Elbert, T. et al., 1996. Visually
¨
induced gamma-band responses in human electroencephalographic activity ᎏ a link to animal studies. Exp.
Brain Res. 112 Ž1., 96᎐102.
Murthy, V.N., Fetz, E.E., 1992. Coherent 25-Hz to 35-Hz
oscillations in the sensorimotor cortex of awake behaving
monkeys. Proc. Natl. Acad. Sci. USA 89, 5670᎐5674.
Nunez, P.L., 1995. Neocortical Dynamics and EEG Rhythms.
New York: Oxford University Press.
Nunez P.L., 2000. Toward a quantitative description of large
scale neocortical dynamic function and EEG. Behav. Brain
Sci. 23, 371᎐398.
Petsche, H., Kaplan, S., von Stein, A., Filz, O., 1997. The
possible meaning of the upper and lower alpha frequency
for cognitive and creative tasks: a probability mapping
study. Int. J. Psychophysiol. 26, 77᎐97.
Rappelsberger, P., Petsche, H., 1988. Probability Mapping:
power and coherence analysis of cognitive processes. Brain
Topogr. 1 Ž1., 46᎐54.
Riehle, A., Grun, S., Diesmann, M., Aertsen, A., 1997. Spike
synchronization and rate modulation differentially involved
in motor cortical function. Science 12 278 Ž5345.,
1950᎐1953.
Rodriguez, E., George, N., Lachaux, J.P., Martinerie, J., Renault, B., Varela, F.J., 1999. Perception’s shadow: long-distance synchronization of human brain activity. Nature 397
Ž6718., 430᎐433.
Roelfsema, P.R., Engel, A.K., Konig, P., Singer, W., 1997.
Visuomotor integration is associated with zero time-lag
synchronization among cortical areas. Nature 385 Ž6612.,
157᎐161.
Roland, P.E., Gulyas, B., 1994. Visual imagery and visual
representation. Trends Neurosci. 17, 281᎐287.
Sarnthein, J., Rappelsberger, P., Shaw, G.L., von Stein, A.,
1998. Synchronization between prefrontal and posterior
association cortex during working memory tasks in humans’.
Proc. Natl. Acad. Sci. USA 95, 7092᎐7096.
Srinivasan R., Tucker D.M., Murias M. Žin press. Estimating
the spatial nyquist of the human EEG. Behav. Res. Instrum. Comput.
Shadlen, M.N., Newsome, W.T., 1996. Motion perception:
seeing and deciding. Proc. Natl. Acad. Sci. USA 93 Ž2.,
628᎐633.
Siegel, M., Kording,
K.P., Konig,
P., 2000. Integrating top-down
¨
¨
and bottom-up sensory processing by somato-dendritic interactions. J. Comput. Neurosci. 8, 161᎐173.
Singer, W., Gray, C.M., 1995. Visual feature integration and
the temporal correlation hypothesis. Annu. Rev. Neurosci.
18, 555᎐586.
Smith, E.E., Jonides, J., Koeppe, R.A., 1996. Dissociating
313
verbal and spatial working memory using PET. Cereb.
Cortex 6 Ž1., 11᎐20.
Steriade, M., Gloor, P., Llinas, R.R., Lopes da Silva, F.H.,
Mesulam, M.M., 1990. Basic mechanisms of cerebral
rhythmic activities. Electroencephalogr. Clin. Neurophysiol.
76, 481᎐508.
Tallon, C., Bertrand, O., Bouchet, P., Pernier, J., 1995.
Gamma-range activity evoked by coherent visual stimuli in
humans. Eur. J. Neurosc. 7 Ž6., 1285᎐1291.
Tallon-Baudry, C., Bertrand, O., Peronnet, F., Pernier, J.,
1998. Induced gamma-band activity during the delay of a
visual short-term memory task in humans. J. Neurosci. 18
Ž11., 4244᎐4254.
Tanaka, K., 1996. Inferotemporal cortex and object vision.
Annu. Rev. Neurosci. 19, 109᎐139.
von Stein, A., Rappelsberger, P., Filz, O., Petsche, H., 1993.
EEG correlates of mental imagery: an amplitude and
coherence study. Z. EEG-EMG 24, 1᎐8.
von Stein A., Sarnthein J., Petsche H., 1995. EEG synchronization seems to reflect functional properties of the underlying cortical networks in the human brain. Eur. J. Neurosci., supp.8
von Stein, A., Rappelsberger, P., Sarnthein, J., Petsche, H.,
1999a. Synchronization between temporal and parietal cortex during multimodal object processing in man. Cerebral
Cortex 9, 137᎐150.
von Stein, A., Chiang, C., Konig,
P., 1999b. The role of alpha
¨
and gamma frequency interactions in top-down processing
in the cat, Soc. Neurosci. Abstr.
von Stein, A., Chiang, C., Konig,
P., 2000. Top-down process¨
ing mediated by inter-areal synchronization. Proc. Natl.
Acad. Sci. USA, in press.

Different frequencies for different scales of cortical integration: from

Transcription

Documents pareils

The Frequency of Visually Induced Gamma

MS Project Synchronization with TeamDirection® Project

Exercise Equipment for Your Mind

Jeu Roulette Casino Jouer Gratuit Jouer Roulette Casino 0 01

burgundy neuromeeting 2011

Vorfeld

Raymond Depardon

How cortical neurons help us see: visual recognition

Synchronization in Distributed Systems