Neural Synchrony in
Attention and
Consciousness
Lawrence M. Ward
University of British Columbia
Collaborators: Sam Doesburg, Kei Kitajo, Alexa Roggeveen,
Tony Herdman, Jessica Green, John J. McDonald
Funded by
Themes
Neural synchrony and its measurement
Neural synchrony in attention: batonpassing in the cerebral cortex
Neural synchrony in consciousness:
binocular rivalry and the thalamic dynamic
core
Implications for LIDA
Gray & Singer’s cats
Neural synchrony
occurs when neural
activity, spiking or
dendritic currents, in
disparate locations
rises and falls in a
fixed relationship
Ward et al’s humans
Varela et al, 2001
Spectral power in a given
frequency range reflects
local synchrony
Theta 6 Hz
Spectral
Power
10 20 30
40 50
Frequency
Gamma 40 Hz
Spectral
Power
10 20 30
40 50
Frequency
Alpha 10 Hz
Spectral
Power
10 20 30
Frequency
40 50
Roles of local neural synchrony
High fidelity neural communication
Perceptual/memorial/motor…. Binding:
Increase 30-70 Hz: amplify post-synaptic effect by increasing spike cooccurrence (Fries et al, 2001)
Decrease 6-15 Hz: increase post-synaptic impact by avoiding spikefrequency adaptation (Fries et al, 2001)
Fries 2005
Roles of global neural synchrony
Integration of neural activity through reciprocal
(re-entrant) interactions between diverse brain
regions (e.g., Varela et al, 2001)
Exchange of data (upward) and hypotheses/templates
(downward); sensory/perceptual processing
Modulation of one region (e.g., hippocampus) by another
(e.g., visual cortex) to store a memory (e.g., of a visual
scene)
Modulation of one region (e.g., visual cortex) by another
(e.g., prefrontal cortex) to enhance processing of
attended information (e.g., a sign for a sushi bar)
Initiating an action in motor regions by computations from
cognitive regions
Consciousness (?)
EEG/MEG synchronization analysis: calculation of phase locking value (PLV)
i.e. surface Laplacian, or MEG field strength,
Step.1 Obtain SCD^of filtered signals f(t) via bandpass filtering at chosen frequencies
(µV)
Fp1
F7 F3
T3
Fp2
●
F4 F8
30
Fz
EEG (C3, O1, Fz)
10Hz
10
20Hz
0
●
C3 Cz
-0.2
C4 T4
-10 0
0.2
0.4
0.6
0.8
1
(sec)
30Hz
-20
T5 P3 Pz P4 T6
O1
●
SCD
20
C3
O1
Fz
-30
stimulus
O2
40Hz
-0.2
0
stimulus
0.2
0.4
0.6
0.8
1
(sec)
Step.2 instantaneous phase and amplitude
3
amplitude
30Hz (C3, O1, Fz)
2
3
30Hz (C3, O1, Fz)
2
1
1
-0.2
0
0
-1 0
0.2
0.4
0.6
0.8
1
(sec)
-0.2
-2
-3
stimulus
C3
O1
Fz
-0.2
0
4
3
2
1
0
-1 0
-2
-3
-4
0.2
0.4
0.6
0.8
1
(sec)
phase 30Hz (C3, O1, Fz)
0.2
0.4
0.6
0.8
1
(sec)
~
 (t )  f (t )  i f t   At expi t 
amplitude
phase
 f ( )
~
~
1
d
where f (t ) is Hilbert transformof f (t ), f (t )   P.V .
t 
Step.3 Calculation of phase locking value (PLV) for each time point
4
3
2
1
0
-1 0
-2
-3
-4
-0.2
PLV t 1, 2
1

N
N
e 
i (t ,n )
n 1
where
 (t , n)  1 (t , n)  2 (t , n)
phase difference (30Hz)
C3-O1
C3-Fz
O1-Fz
(sec)
0.2
0.4
⇒
0.6
0.8
1
complete synchronization:1
random phase difference:0
(phasedifference)
N : the number of trials
t : timepoints
-0.2
1 : the phaseof thesignal fromelectrode1
2 : thephaseof thesignal fromelectrode2
4
3
2
1
0
-1 0
-2
-3
-4
phase difference
(5 trials)
C3-O1 (30Hz)
(sec)
0.2
0.4
High PLV
1
C3-O1 (30Hz)
0.6
0.8
Low PLV
1
PLV (50 trials)
0.8
0.6
0.4
0.2
(sec)
0
-0.2
0
0.2
0.4
0.6
0.8
1
Step.4 standardization of PLV
To reduce the effect of volume conduction of
stable sources and compare between electrode
pairs at different distances
Standardized PLV
6
C3-O1 (30Hz)
4
2
-0.2
0
-2 0
0.2
0.4
0.6
0.8
1
(sec)
-4

PLV  PLVBmean 
PLVz(t ) 
-6
PLVBsd
PLVBmean : themean of PLVin thebaseline period(400ms)
PLVBsd : thestandarddeviationof PLVin thebaseline period(400ms)
Standardized PLV and surrogate PLV
6
PLV (original)
C3-O1 (30Hz)
4
Median
PLVsurrogate
2
Step.5 statistical test using surrogate data
-0.2
0
-2 0
0.2
0.4
0.6
0.8
-4
Note: Local and longrange PLVz must
change together for
spurious
synchronization to be
indicated
(Doesburg,Ward, CC,
2007)
significant PLV increase
-6
(Hz)
C3-O1
60
100
99
98
3-97
2
1
0
50
40
30
20
10
-0.2
1
±95 percentle
(sec)
PLVsurrogate
0
0.2
0.4
0.6
0.8
1(sec)
sync
desync
Alpha, gamma and attention
Local alpha power associated with active
suppression
Local gamma power associated with active
processing
Alpha suppression necessary for gamma binding?
Roles of long-range alpha and gamma synchrony?
Coordination of local and long-range
synchronization?
Theta-modulated gamma synchrony
e.g., Klimesch, Jensen & Colgin, Palva & Palva, Ward
Alpha/gamma local synchrony (power)
indexes spatial attention orienting
•Arrow cued box to attend to
•Press button only if + in attended box,
not if x nor if in unattended box
•Cue-target SOA 1000-1200 ms
•Cue onset at 0 ms in figures
15
+
20
P7
10
5
P8
10
0
0
Left Cue 11 Hz
-5
Left Cue 11 Hz
-10
Right Cue 11 Hz
Right Cue 11 Hz
Left Cue 39 Hz
-10
Right Cue 39 Hz
Left Cue 39 Hz
-20
Right Cue 39 Hz
-15
-30
-20
-40
-25
P7
-30
P8
-50
0
250
500
750
1000ms
0
250
500
Doesburg, Roggeveen, Kitajo & Ward, 2007, Cerebral Cortex
750
1000ms
Long-distance gamma synchrony establishes focused attention network
QuickTime™ and a
decompressor
are needed to see this picture.
•At 250 ms (same as local power max gamma/min alpha) increase in global
phase locking at diverse frequencies
•Lateralized in gamma band: P7 (left) for right target, P8 (right) for left
target (orienting?)
•Increased synch in beta band also but not lateralized (readying?)
•Desynch at 100 ms in gamma: erasing old network?
MEG replication
Increased
lateralized
synchronization in
high alpha band
from ~400 ms
post cue onset
until end of epoch
Decreased
synchronization
side ipsilateral to
cue
Synchronization in
alpha band
associated with
maintenance of
attentional focus
at cued location
Left cue
14 Hz
Doesburg, Ward, 2007, Proceedings BIOMAG2006
Right cue
What is SAM beamformer?
Synthetic Aperture Magnetometry (Vrba & Robinson 2001,
Methods)
Based on time/space covariance matrix of sensors
Achieves estimate of source power for each voxel in brain
region from weighted linear combination of all measurements
(where weights are selected to attenuate signals from all
other voxels):
t
ˆ
S (t )  W m(t )
Optimal coefficients found by minimizing total power over
time (computational tricks used in practice).
MEG
Replication
MEG filtered at 14 Hz
SAM beamformer sources in
parietal (SPL?) and visual
cortices (n=5)
PLV analysis applied to
broadband source activity
filtered from 6 Hz to 60 Hz
14 Hz PLV shows lateralized
increased synchrony similar
to sensor analysis from 4001000 ms post cue onset
(right; n=2)
40 Hz (gamma-band) PLV
shows burst of lateralized
increased synchrony at
~200-250 ms post cue onset
Theta rate gamma synch
bursts at least for right
parietal-occipital when
attending left
Doesburg,Herdman, Ward, 2007, Cognitive Neuroscience Society
MEG replication
Beamformer sources
projected to cortical
surface (star=source in
a sulcus so projection
done by hand)
Lateralized increases
and decreases in
synchrony between
parietal and occipital
sources in alpha band
from 400-800 ms post
cue onset (here 800 ms
and 14 Hz)
Occipital 14 Hz power
replicates EEG data
(blue bars, 800 ms)
Doesburg, Herdman, Ward, 2007, Cognitive Neuroscience Society
MEG Replication
Endogenous orienting:
dorsal fronto- parietal
(FEF, IPS, V1/V2)
We also found FEF
sources but not
consistent enough for
PLV analysis
Not enough to activate
relevant areas - must be
synchronized to be
functionally effective?
Corbetta and Shulman (2002) Nat Rev
Neurosci; Wright & Ward, 2008,
Orienting of Attention
EEG Replication: BESA beamformer
sources from theta band signals
Green & McDonald, 2008, PLoS Biology
EEG analyses
BESA beamformer sources in theta
band identified
Broadband activity of dipoles at peak
voxels computed based on EEG
recordings
Broadband signals filtered and
analytic signal, PLV etc., computed
between sources
Lateralized increased
synchronization in
the alpha band
Right brain
Left-neutral
Right-neutral
Left brain
350 ms post cue onset
until target onset:
maintains attention at
cued location
Doesburg, Green, Ward &
McDonald, in prep.
Baton-passing in the cerebral
cortex
Auditory cues and targets
Cue types: up or down glide
for orient left or right (or
vice versa), both for do not
orient; each on 1/3 of trials
White noise targets (respond
to all targets) for gap
discrimination presented left
(1/3 of trials) or right (1/3 of
trials) at random regardless
of cue type; probes (respond
only if at cued location)
presented on 1/3 of trials at
random
BESA beamformer source
RT
analyses for theta-band
signal
Green, Doesburg, Ward & McDonald, submitted
% Corr
Valid
Neutral Invalid
694 ms
(25.5)
718 ms
(26.6)
716 ms
(26.7)
90.9
(.016)
90.7
(.014)
90.6
(.018)
Theta-band BESA
beamformer source
activations
Baton-passing:
•1 Cue activates STG
which in turn activates IPL
and SPL
•2 IPL/SPL activate IFG
•3. IFG interprets cue
•4. IFG tells IPL/SPL
where to orient
•5. IPL/SPL activate
relevant STG and maintain
activation until target
Green, Doesburg, Ward & McDonald, submitted
Interim Summary
Brain configured to attend to a specific location by
Confluence of burst of decreased local alpha power,
increased local gamma power and increased long-range
gamma synchrony at 250 ms post cue
Attention maintained at specific location by increased
long-range alpha-band synchrony (at least parietal-visual)
that coincides with decreased local alpha power
Information/control signals passed between brain regions
by establishing and breaking synchronization in gamma
band between those regions
Still not understood: mechanism of synchrony used
(thalamic control?); nature of signals exchanged
(control/compliance signals?); how reactivity of
sensory cortex is increased by control signal; ……
Binocular rivalry: window to
awareness
Stimuli
Apparent
locus of
fused
object
Constant
stimulation,
involuntarily
alternating
experience
Prisms
Eyes
Corresponding retinal areas
Neural synchrony and consciousness
Cosmelli et al, (2004) Waves of consciousness: Ongoing cortical patterns
during binocular rivalry. NeuroImage, 23, 128-140
Widespread 5 Hz synchrony associated
with perception of the 5 Hz stimulus
Face
Rings
Face Rings Face
Rings
Face
Rings
Face
Synchrony of brain
areas at 5 Hz for
different subjects;
note individual
differences in both
locations of active brain
areas and in amount of
synchrony between
them.
Binocular rivalry: window to
awareness
Constant
stimulation,
Stimuli
involuntarily
alternating
experience;
subjects press
one of two
buttons to
indicate which
pattern they
see, neither to
indicate parts of
both
Apparent
locus of
fused
object
Prisms
Eyes
Corresponding retinal areas
Long-range theta-rate bursts of gamma-band
synchrony in binocular rivalry of striped patterns
Frequency
(Hz)
Button
press at 0
Time (ms)
45Hz -450
7Hz -450
-400
-400
-350
-300
-250
-200
-150
-100
-50
0
50
100ms
-350
-300
-250
-200
-150
-100
-50
0
50
100ms
Doesburg, Kitajo & Ward, 2005, NeuroReport
Frequency (Hz)
BESA beamformer (280-220 ms prebutton-press, 36-46 Hz) applied to
replication
R
L
Time (ms – 0 indicates button press)
Average PLVz over 8 active sources: R/L V1/V2, R sup
occipital, R parietal, L temporal pole, L DLPF, R/L
prefrontal
Bursts of gamma-band synchronization occur at theta
rate (arrows) around button press in presence of
persisting synchronization in theta band; also bursts of
synchronization in beta band
Theta-modulated synchronization
Left V1/V2 – Left DLPFC
Right
brain
Left V1/V2 – Right parietal
Left
brain
40-50 Hz alt
synch/desynch
40-50 Hz synch
30-40 Hz synch
20-30 Hz synch
Right parietal – Right occipital
40-50 Hz desynch
30-40 Hz desynch
surrogate level 0.025/0.025
Dynamic core hypothesis
Neural correlate of conscious awareness is
what Tononi & Edelman called the "dynamic
core"
Large-scale (brain-wide, 200-msec time
scale)
Coherent (statistically synchronous)
activity
Millions of neurons involved
Dynamic core hypothesis
DC simultaneously integrates activities of many
brain areas (not all of them, a constantly changing
subset) …
And also differentiates current conscious state
from many other, possible conscious states.
My (radical) proposal: the thalamic dynamic core
is the neural correlate of phenomenal awareness
Cortex computes, thalamus experiences
Human cortex, with more neurons and more corticocortical fibers per thalamic fiber computes much more
complex contents than do, e.g., rat, dog, or chimp
cortices; pace Paul Nunez
Cortical DC arises from synchronous activity in thalamus
Thalamic dynamic core
Dynamic core and supporting binocular
rivalry data (Tononi & Edelman, Varela
group); neural synchrony is key
Lesions, neurosurgery, and anesthetic
action point to thalamic “relay” nuclei as
critical (Penfield, Alkire et al)
Anatomy and function of thalamus and
cortex (Mumford)
We experience results (products) of
computations, not the computations
(processes) themselves (Lashley,
Kinsbourne, Prinz, Rees, Koch, Baars)
Lesions: Karen Ann Quinlan
Karen Ann Quinlan’s Brain at Autopsy (see Kinney et al 1994)
Drug/alcohol reaction;
permanent vegetative
state for 14 years
Thalamus-massive loss
Cortex-little loss
Penfield’s neurosurgery and
stimulation mapping
M.M.’s cerebral cortex mapped via
electrical stimulation by Penfield
Patient
M.M.
treated for
intractable
epilepsy
Neurosurgery and stimulation
mapping
Penfield (The Mystery of the Mind, 1975):
The mechanisms of epilepsy and electrical stimulation
mapping imply that “…there are two brain mechanisms
that have strategically placed gray matter in the
diencephalon …, viz. (a) the mind’s mechanism (or highest
brain mechanism); and (b) the computer (or automatic
sensory-motor mechanism).” (p. 40).
Penfield’s “mind mechanism”
Merker (2006, BBS):
argued SC is locus of
conscious analog
simulation of world
General anesthesia
Alkire et al (2000) Consciousness &
Cognition:
Common brain loci and mode of action of
different general anesthetics imply that
the critical mechanism of general
anesthesia is a hyperpolarization block
of the thalamic relay nuclei neurons
Common brain areas where halothane and isoflurane
anesthesia significantly depresses activity; a.
thalamus, b. midbrain reticular formation
Alkire et al (2000) Consciousness & Cognition
The thalamus
Synchronizes cortical oscillations
“Gateway” to cortex for major sensory
systems (except smell)
Evolved along with the cerebral cortex; a
“seventh layer” of cortex (but with
different neuron type)
Each cortical area has an associated
subnucleus of thalamus with massive
cortico-thalamic projections and smaller
thalamo-cortical projections; most thalamic
subnuclei have no other inputs!
Where is the thalamus?
Thalamus
Pineal body
Gross Anatomy of some cortico-thalamic circuits
Roles of the thalamus
Relay station and gateway (attentional
engagement) to cortex for sensory systems
Synchronizes neural activity in remote cortical
areas
Active blackboard that echoes back to cortex
results of latest computations (Mumford)
Site of dynamic core of neural activity that gives
rise to phenomenal experience(?): thalamic
dynamic core
We experience products…
Crovitz: maximum rate of consciously
following strobe light = 4 to 5 Hz (250 to
200 msec per cycle)
Sternberg STM scanning: no awareness of
process; 40 Hz (25 ms/item) scan?
LTM search & Retrieval: no awareness of
memory search codes, only memories
Speech: not aware of composing
utterences, phonemes, (co-)articulation,
etc.
Conclusions
Synchronous neural oscillations occur at several
scales and in particular frequency bands
Local and global synchrony coordinate to establish
an attentional network that enhances processing of
attended stimuli
Neural synchrony in the thalamo-cortical circuits,
particularly in the thalamus, establishes a dynamic
core of brain activity that supports (is?) conscious
awareness
Implications for GWT/LIDA
Dynamic core is the GW
GW (broadcasting) established by synchronization
between various active processing modules
mediated by thalamic dynamic core
Transient networks of brain regions (processing
coalitions) established by low frequency
synchronization (carrier?), with high frequency
synchronization establishing transient information
exchange between processing areas in necessary
temporal sequences (baton passing)
Perception/action cycle governed by several
temporal interactions based on processing speeds
and communication requirements vis a vis available
oscillation frequencies.