Transcript Lecture 5

Synchronous firing of a neuronal ensemble
A familiar object
is represented
in the brain by
thousands of neurons
dispersed throughout
the posterior cortex.
When you recall an apple,
the neuronal
ensemble of the apple
is firing in synchrony.
II. Neocortex: The role of the neocortex is
to store and manipulate memories
1. Frontal lobe controls movement: (in
all animals - movement of muscles.
Evolved in primates: purposeful
movement of thoughts. We can look
at conscious purposeful movement of
thoughts as internalization of
movement).
Frontal lobe controls the timing of
neuronal firing: it schedules
movement routines and thoughts in
time.
Frontal
2. Dorsal cortex: encodes sensory
memories. More specifically,
encodes sensory experiences into
ensembles of neurons capable of
self-activation
Parietal
lobe
lobe
Temporal
lobe
Occipital
lobe
1. How is sensory information is encoded by neurons?
• Sensory information is coded by the rate of firing
of action potentials (number of action potentials
per second)
2. How motor information is encoded by a
motor neuron?
• Again, information is coded in frequency of action potentials:
the greater the firing rate  the greater the tension
• In addition CNS can recruit more motor units
• What is the range of firing rate?
• Is every motor neuron action potential followed by a muscle
contraction?
3. How the information about your speed
of movement is encoded in the brain?
• Speed cells in
entorhinal cortex
fire action potentials
at a rate
proportional to
animal speed of
movement.
• This information is
feeding their info to
grid cells.
In the brain…
Action potential
1. Is every presynaptic action potential in the
Postsynaptic
brain followed by a postsynaptic action potential?
potential
2. No. Most postsynaptic potentials are too small
to elicit a postsynaptic action potential. Why?
3. Postsynaptic potentials are local potentials – they are not
usually amplified
4. Postsynaptic potentials are graded potentials: the more APs
arrived  the greater EPSP (as opposed to all-or-none APs)
5. Since postsynaptic potentials are not amplified, postsynaptic
potentials as decreasing in amplitude as one measures
change in Em away from a synapse
Em
10mV
5mV
3mV
1mV
synapse
dendrite
AP in
presynaptic cell
axon
• Reasons for synchronous action potentials: 1. Spatial summation
• It usually requires many presynaptic neurons to fire action
potentials that arrive at the postsynaptic neuron at the same time
to depolarize the postsynaptic membrane potential above
threshold and trigger an action potential in the postsynaptic cell.
Reasons for
synchronous action
potentials: 2. LTP
1
3
Ca++
3
4
5 Influx of
Ca++
Synapse is
modified
by displacing Mg2+
Presynaptic neuron A
6
Presynaptic neuron A
2
postsynaptic
neuron B
postsynaptic
neuron B
Increase of
the number
of synapses
between
presynaptic
axon A and
postsynaptic
neuron B
Reasons for synchronous action
potentials: 2. LTP
• If a presynaptic axon and a
postsynaptic cell are to enhance
their connection  they have to
fire an action potential at the
same time
• Hebbian principle: fire together
 wire together
• This is the only way to develop
memory
The neuronal ensemble binding
mechanism is based on the Hebbian
principle: “neurons that fire together,
wire together” (Hebb, 1949).
Hebb wrote: “… any two cells or
systems of cells that are repeatedly
active at the same time will tend to
become ‘associated,’ so that activity
in one facilitates activity in the other.
… When one cell repeatedly assists in
firing another, the axon of the first
cell develops synaptic knobs (or
enlarges them if they already exist) in
contact with the soma of the second
cell.” (Hebb, 1949).
What is the mechanism of Hebbian learning?
Method to achieve synchronization in a
group of neurons: depolarization wave
• Name examples of pacemaker cells outside of CNS?
• Heart
• Gastrointestinal tract
Examples
• Heart synchronous
beating (outside of CNS)
• Slow wave sleep (1Hz)
• Tremor, such as
essential tremor or
Parkinsonian tremor
• Epilepsy: seizures are
transient
hypersynchronous
neuronal activity in the
Generalized 3 Hz spike and wave
brain
discharges reflecting seizure activity
• In epileptic seizures
a group of neurons
begins firing in an
abnormal, excessive,
and synchronized
manner.
Neuronal binding by synchronization
• Until recently, the brain was viewed as a collection of neurons that fire in
asynchrony.
• The synchronous firing of cortical neurons was only associated with
epileptic seizures: wavelike electrical activity of a large number of neurons,
often associated with loss of consciousness and involuntary body
movement.
• New evidence indicates that neurons often communicate by firing near
synchronous action potentials.
• In fact, there is evidence that de-synchronization
is associated with disorders such as
schizophrenia and ADHD.
• The solution to the binding problem:
• The synchronous firing is thought to be
involved in binding neurons of a
neuronal ensemble together
(following the principle of
“fire together - wire together”).
• This hypothesis is referred to as
“binding-by-synchronization”
(Singer & Gray, 1995).
binding-by-synchronization
“Neuronal synchronization is a ubiquitous phenomenon in
cortical networks and likely to serve a variety of different
functions in addition to feature binding at early levels of sensory
processing.”
-- PETER J. UHLHAAS et al., Neural synchrony in cortical networks
(2009)
Experiments:
Direct observation of synchronicity of neurons belonging to a
single object
JC- 1_1991_Engel_Direct
physiological evidence for scene
segmentation
• recorded from multiple neurons in the
primary visual cortex of a cat.
• observed that two neurons synchronized
their firing only when their activating
stimulus belonged to a single object.
• In other words, neurons synchronized only
when they were part of the same neuronal
ensemble.
• Stimulation with single light bars of different
orientations (a single object) yielded a
synchronization of oscillatory responses
between all sites activated by the respective
orientation (==a single neuronal ensemble).
• Two different objects
moving separately
result in two
neuronal ensembles
• Cells 1 and 3 belong
to one neuronal
ensemble and cells 2
and 4 belong to
another neuronal
ensemble.
• The ensembles are
asynchronous
between
themselves.
• Hirabayashi & Miyashita, 2005 (JC-Th): Record from pairs of neurons located in
the inferior temporal cortex (responsible for face recognition) in alert monkeys.
• The neurons were selected in such a way that both cells in a pair responded
significantly to facial fragments shown one at a time.
• monkeys were shown pictures from two different sets. In the first set, the nose,
mouth, and eyes were organized into a face-like combination.
• In the second set, the facial fragments - nose, mouth, and eyes - were shown
scrambled up (e.g. mouth could be on top, nose on the left, one eye on the
bottom, and another eye on the right).
• Thus both neurons would significantly increase their activity when presented with
either the face-like pictures or the scrambled face pictures.
• Was there any difference in the neuronal activity between the face-like pictures
and a scrambled collection of facial features?
• The study found that there was no
difference in terms of firing rate,
• but a clear difference was observed
in the timing of the neuronal firing:
• neurons responding to the features
transiently synchronized their
responses when the monkeys
recognized that the
component features of a face
actually formed a face.
• Face perception, i.e. actual
recognition that component
features of a face formed an actual
face was associated with the
synchronization of neurons encoding
the individual facial features
(=perceptual closure =activation of
complete neuronal ensemble.)
• On the local level: groups of neurons within one
cortical area often synchronize in order to “vote
on a particular issue.”
• In this case, local groups of neurons undergo low
amplitude depolarization waves with a frequency
of 30 to 100 Hz.
• These waves are called gamma-frequency waves.
Global synchronization
Neurons tend to fire action potentials on the upswing of the depolarization wave
•
•
•
•
In addition to the local depolarization wave, a global wave with a frequency of 15 to 25 Hz
ensures effective communication between neurons in different cortical areas (analogous to a
company-wide meeting).
This wave is called a beta-frequency wave.
Both beta-frequency and gamma-frequency cortical neuron synchronizations are observed
for consciously perceived stimuli (Meador, 2002; Gross, 2004; Nakatani, 2005; Palva, 2005)
and for conscious perception in binocular rivalry (Fries, 1997; Srinivasan, 1999; Fries, 2002;
Doesburg, 2005).
These findings have inspired the hypotheses that synchronized oscillations play a role in
consciousness (e.g., Crick and Koch, 1990; Llinas et al., 1998; Buzsaki, 2004).
Experimental evidence
• Pejman Sehatpour et al.
(2008) (JC-Th) used
intracranial electrodes to
record from three human
subjects (who were
undergoing medical tests for
intractable epilepsy) as they
performed a challenging
visual object recognition
task that required them to
identify barely recognizable
fragmented line-drawings of
common objects.
• Neuronal activity recorded
during the “recognition”
moment (also called
perceptual closure) was
compared to the “nonrecognition” event when
scrambled (nonsensical)
versions of the same images
were shown.
only the difference between
scrambles and unscrambled
• Sehatpour and colleagues simultaneously recorded from several areas
involved in visual recognition: the occipitotemporal cortex, the prefrontal
cortex, and the hippocampus.
• The analysis showed a robust coherence in the beta-band between all
three brain areas when participants recognized the fragmented images.
• In contrast, when scrambled versions of the same images were presented, a
significantly lower coherence was observed.
• These results suggest that object-encoding neuronal ensembles are
distributed through several cortical regions (in this case, all the regions
where electrodes were placed) and that neurons of the ensemble do
synchronize when the encoded object is consciously perceived.
Eugenio Rodríguez et al. (1999) (JC-Th)
• Study focused on the
perception of 'Mooney' faces,
high-contrast pictures of a
human faces which are easily
recognized as faces when
presented upright
orientation, but usually seen
as meaningless black and
white shapes when presented
upside-down.
• An electroencephalogram
(EEG) was used to record
from healthy human subjects.
• A consistent pattern of
synchrony between left
occipital, left parietal and
left frontal cortices was
established during face
recognition around 200
ms after presentation of
an upright face.
• When the faces were
presented scrambled up,
no synchrony was
observed.
Synchrony between electrode
pairs is indicated by lines, which
are drawn only if the synchrony
value is beyond the distribution
of shuffled data sets (P < 0.01).
Black lines correspond to a
significant increase in synchrony.
Green lines = a significant
decrease in synchrony. Phase
synchrony is markedly regional
and differs between conditions.
• Singer, 2006: (C)
shows the
topography of phase
synchrony between
20–30 Hz. Synchrony
between electrodes
is indicated by lines,
which are drawn
only if the synchrony
value is beyond a
two-tailed
probability of p <
0.0005.
• Notice the
widespread network
of synchronized
neurons in controls
Hipp et al. (2011) (JC-Th)
• High-density EEG recordings from human subjects
• The subjects’ task was to judge the configuration of an
ambiguous stimulus which consisted of two bars that
approach, briefly overlap and move apart from each
other.
• The perception of this stimulus spontaneously alternates
between two distinct alternatives.
• In half of the trials, the two bars are perceived as two
independent objects passing one another (==two
independent neuronal ensembles = no synchronization).
• In the other half of trials, the two bars are perceived as a
single object with bouncing sides (think of a slinky; a
single neuronal ensemble  neurons synchronize in
time).
• the extrastriate cortex, the posterior parietal, temporal and
prefrontal cortex, showed enhanced beta rhythm synchrony
during stimulus processing.
• the synchronicity between regions in beta frequency
predicted the subjects’ perception of the stimulus even on a
single-trial level!
• when beta frequency synchronization was enhanced, subjects
were more likely to perceive the same sensory stimulus as a
single bouncing object (encoded by a single synchronized
neuronal ensemble) rather than two bars passing each other
(encoded by two independent unsynchronized neuronal
ensembles).
Counter-example
• 2008 - Synchrony and
the binding problem in
macaque visual cortex
• They did not find strong
synchrony
• “A limitation of our
study is also the use of a
behavioral task that
engaged attention at the
fixation target” – were
the monkeys perceiving
two objects? When I
attend to the fixation
point, it is much harder
for me to tell whether
there is one figure or
two figures present
Theta rhythms in
hippocampus
• Neural oscillations, in particular theta (4–7 Hz) activity, are
extensively linked to memory function. Theta rhythms are
very strong in rodent hippocampi and entorhinal cortex
during learning and memory retrieval.
• Theta rhythms are believed to be vital to the induction of
long-term potentiation, a cellular mechanism for learning
and memory.
• When people are finding their way or looking at something
novel (when they are learning), the theta rhythm is
particularly strong .
• The stronger theta oscillations are, the better the person
will remember the new material.
Theta in human
hippocampus
• EXPERIMENT: Rutishauser team used the implanted microelectrodes to
track electrical activity in the hippocampus and the amygdala as well
as local field potential in nine epileptics.
• Patients viewed 100 slides, each of which showed an image of a person, an
animal, or an everyday object such as a car or a tool. The patients had 1s to
remember each picture as best they could before the next one appeared.
• prominent theta activity when the patients were memorizing the images.
• The team later tested the patients’ recall by showing them a second set of
100 photographs, half of which were novel and half of which were repeats.
• On average, participants recognized 60% of the initial pictures.
• What predicted successful recall?
• The firing rate while a subject viewed an image during learning phase did
not predict whether or not the patient would later recall it.
• However, if a picture flashed on the screen at a moment when neuronal
spikes in the hippocampus lined up with the local theta rhythm, patients
were more likely to remember the image!
• Common knowledge: memory formation is
influenced by
– attention
– novelty
– emotional impact
• TIMING!
• Neurons always spike in response to new images
and experiences.
• But when the spikes happen to coincide with the
theta rhythm, this coordinated electrical activity
alters the brain’s synapses, enabling memories to
form.
•
How widespread can neuronal
ensemble be?
In the past, fMRI studies have depicted the cortex as a patchwork of function-specific regions.
How widespread
can neuronal
ensemble be?
• Siegel & Buschman & Miller, 2015 - found significant
encoding for all information across all regions 
multiple cortical regions work together
simultaneously to process sensorimotor information
• 108 electrodes that measured neural spikes in 2,694
sites across six cortical regions: [middle temporal area
(MT/V5), visual area four (V4), inferior temporal
cortex (IT), lateral intraparietal area (LIP), prefrontal
cortex (PFC), and frontal eye fields (FEF)]
• Neural activity, near simultaneously: Sensory information
— for cue, and color or motion — started in the MT and
V4, but flowed to the LIP, IT, FEF, and PFC.
• Task information started in V4 and IT, but flowed forward
to PFC and LIP, and onward to the FEF and back to the V4.
• Choice signals built up in PFC and LIP, before flowing
forward and backward to FEF and the V4.
• In short, despite neural spikes in specific areas, all
information was shared widely.
• Conclusion by Miller: Previously, it was thought that
decisions rise solely in specific cortical areas. “But you see
the decision percolating up all over many parts of the
cortex simultaneously, so even decision-making is more
of an emerging property of many cortical areas” .
Summary of different rythms
• Delta wave – (0.1 – 3 Hz) - associated with the deep non-REM slow
wave sleep (training of neuronal ensembles by hippocampus).
Pacemaker cells are in subcortical structures thalamus, reticular
formation, suprachiasmatic nuclei in the hypothalamus. Right
hemisphere dominance during sleep.
• Theta wave – (4 – 7 Hz) – associated with learning and memory
(training of neuronal ensembles by hippocampus). Pacemakers in
the CA1 layer of hippocampus.
• Alpha wave – (8 – 15 Hz) – e.g. originates from the occipital lobe
during wakeful relaxation with closed eyes, hence its use in
biofeedback during meditation. Pacemaker cells are in the
thalamus.
• Beta wave – (16 – 31 Hz) - associated with active thinking and
active concentration, formation of global neuronal ensembles.
• Gamma wave – (32 – 100 Hz) - associated with formation of local
neuronal ensembles. Pacemakers are inhibitory interneurons
releasing GABA.
Normally we have
meditation
•
•
•
•
•
•
10 novice vs. 8 long-term meditators – energy in gamma-band
activity (relative to more slowly changing brain waves).
Meditators were asked to attain a state of “unconditional loving-kindness and compassion”
Experienced meditators (monks) produce increased gamma waves in the brain (25-42Hz)
synchronized across the frontal and parietal cortices
Such activity is thought to be the hallmark of focusing attention that involves
synchronization of spatially dispersed groups of neurons.
gamma activity in monks is the largest seen in nonpathological conditions and 30 times
greater than in the novices. The more years the monks had been practicing meditation, the
stronger the gamma activity.
Normally: all different departments activate spontaneously (most of them subconsciously)
 EEG waves cancelling each other. From time to time a neuronal ensemble achieves
greater synchronization and gets conscious attention to itself  this spontaneous activation
of neuronal ensembles produces thought clutter.
Experienced meditators can synchronize all departments  significant increase in EEG
amplitude  no thought clutter
Conclusions
• There is significant experimental evidence that
long-range synchronization of neurons encoding
an object plays an important role in the binding
of multiple features into one integrated percept.
• This integration is involuntary. It is driven by the resonant
activation of the object’s neuronal ensemble stored in memory
as enhanced connections between neurons of the ensemble.
• The complete neuronal ensemble of an object is automatically
activated and the object is perceived whether the recollection
is triggered by seeing a partially visible object, hearing about
the object, or just thinking of the object.
• The complete neuronal ensemble of the object is activated
because the neurons of the ensemble have preexisting
enhanced connections (greater number of synapses).
II. Neocortex: The role of the neocortex is
to store and manipulate memories
1. Frontal lobe controls movement: (in
all animals - movement of muscles.
Evolved in primates: purposeful
movement of thoughts. We can look
at conscious purposeful movement of
thoughts as internalization of
movement).
Frontal lobe controls the timing of
neuronal firing: it schedules
movement routines and thoughts in
time.
Frontal
2. Dorsal cortex: encodes sensory
memories. More specifically,
encodes sensory experiences into
ensembles of neurons capable of
self-activation
Parietal
lobe
lobe
Temporal
lobe
Occipital
lobe
• Wolf Singer (1h, 2013; start on 27min to avoid
philosophy):
https://www.youtube.com/watch?v=WgEWMdV1
Q4w
• stop here
•
synchronicity has to be understood in terms of synchronicity of the arrival of action potentials to a
target neuron rather than absolute equality of action potential conduction times over different
paths. Consider the following example: suppose neuron A is receiving excitatory input from neurons
B and C via two different pathways (neuron A is the target neuron for both neurons B and C).
Suppose that the action potential conduction time is 2ms from neuron B to neuron A and 22ms
from neuron C to neuron A (i.e. the axonal pathway B-A has a significantly shorter conduction time
than the axonal pathway C-A). Does it mean that the connections B-A and C-A are always
asynchronous? The answer depends on the predominant neural activity rhythm in this network. At
the firing rate of 50Hz (inter-spike interval of 20ms that correspond to Gamma rhythm), neurons B
and C can actually be considered isochronous in relationship to neuron A: consider a train of action
potentials synchronously fired by neurons B and C. The first action potential from neuron B will
reach neuron A in 2ms and the first action potential from neuron C will reach neuron A in 22ms.
Obviously, there would be no coincidence in the arrival times of the 1st action potentials from
neurons B and C. However the second action potential from neuron B will arrive to neuron A in
22ms, concurrently with the 1st action potential from neuron C. Thus, starting with the second
action potential, neuron A will receive isochronous activation from neurons B and C. The
synchronous activation has a significantly greater probability of enhancing synaptic connections
between neurons A and B, and A and C (Hebbian learning: ‘neurons that fire together, wire
together’ (Hebb, 1949)). Thus, isochronicity does not need to imply absolute equality in the
conduction time over different pathways. Rather isochronicity implies near-zero phase-shift
between the two firing trains of action potentials at the postsynaptic cells. The phase-shift, of
course, depends on both conduction times over each pathway and the dominant firing frequency in
the neural network.
Long-term potentiation (LTP)