Transcript Gamma Band Oscillation

Gamma Band
Oscillation
The Binding Problem & Binding by
Synchronization
Outline








What are Gamma Oscillations?
Where are they found?
What is the “Binding Problem”?
Hierarchal Model
The Temporal Binding Hypothesis and Binding by
Synchrony
Evidence and the role of Gamma Oscillations
Criticisms or the Temporal Binding Hypothesis
Future Research and Directions
Sources



Rhythms of the Brain – György Buzsáki
Synchrony Unbound: A Critical Evaluation of the Temporal Binding Hypothesis –
Shadlen, M., & Movshon, J.A.
Neural synchrony in cortical networks: history, concept and current status - Uhlhass et
al. (2009)
Gamma Oscillations



The fastest frequency band of neural oscillations
≈ 20 – 100 Hz (typically 40-60 Hz)
“..[Oscillations are] not independent events that
impose timing on neuronal spiking but rather are a
reflection of self-organized interactions of those
same neurons that detect, transfer, and store
information”. (Buzsáki, p 259)
Being fast and having a small amplitude, Gamma
band oscillations were hard to detect in early cell
recording
Where do we find Gamma
Oscillations?



Gamma frequency
oscillations are present
during waking, slowwave and REM sleep.
They are intrinsic to the
neo-cortex, and heavily
rely on GABA-a
receptors, as it mediates
the time constant of the
decay of IPSPs, which
varies from 10 – 25 ms
(40 – 100 Hz)
“Most characteristic field
pattern of the waking,
activated neocortex..”
(Buzsáki, p.259)
The Binding Problem




The problem can be found in both neuroscience and
philosophy, however they are unique in both cases.
In neuroscience, the question is; how higher-order
neural structures are able to segregate and
integrate the proper inputs, both from sensory
organs and internal computations?
In areas such as V1 this is partly accounted for by
the discovery of cortical columns consisting of
simple, complex and hypercomplex cells, which are
attuned to certain stimuli.
However, there remains the question of how we are
able to perceive as unified objects, stimuli in a
robust manner; regardless of out point of view, size
and lighting conditions. (Buzsáki, p 232)

Ex. Our ability to reliably recognize all dogs.
The Binding Problem


Our brains are able to take
different features such as
“…colour, texture, distance,
spatial position and smell
[which are] processed in
separate parts of the cortex
by different sets of neurons
[and] are bound into a
complex representation in a
matter of 200 milliseconds…”
(Buzsáki, p260)
This type of mental
reconstruction has been
largely documented by
Gestaltian psychologists;
showing that human regularly,
and systematically impose
top-down rules on visual
stimuli.
Hierarchical Model



An early solution to the Binding problem; a
feed-forward design where lower level
information is projected to higher level neural
structures, and based on which structures
are activated, .a “Gnostic” cardinal cell
activates. (also called a “grandmother” cell)
This cell represents the object or concept
which fed the sensory input; if it is active then
the mental representation for a particular
stimulus is activated, if it is not activated,
then there is no internal representation.
There has to be a Gnostic cell for every
object, concept or referent in the
environment.


This is problematic because the model
focuses on the role of excitatory cells, while
eliminating the need for inhibitory
interneurons.
Also, there just aren’t enough neurons for
there to be dedicated high-order structures
for each concept or object. “Combinatorial
Explosion”
Hierarchical Model



“Because the number of neurons needed grows exponentially with the
number of unique objects represented by their numerous features, the
brain, so the story goes, quickly runs out of neurons”. (Buzsáki p. 236) –
Combinatorial Explosion
The Hierarchical Model also does not specify the location or spatial
relationship of the Gnostic units.

If they are clustered in particular areas, then they would be easily
susceptible to damage in case of injury or brain damage. However
this does not appear to be the case in patients with cortical column
damage.

If they are widely distributed, then there remains the problem of how
they communicate; and what kind of specialized connectivity they
would require.
Finally, this model lacks a temporal scale, and it would therefore still need
to be explained how it could effectively be used in real-world time and
environments. (Buzsáki p, 237)
The Temporal Binding
Hypothesis





A solution to the Binding problem requires an explanation
of how the various inputs and neural computations are
differentiated so that the right bits of information can be
compared and integrated.
For this, “…it is necessary to “tag” each visual neuron to
signify the object to which its activity relates” (Shadlen &
Movshon)
In the Temporal Binding Hypothesis, this tag is indicated
by synchronous neural spiking.
This offers an endless capacity for coding combinations.
Synchronous spiking also allows for cross modal and long
range communication.
The Temporal Binding
Hypothesis (examples)



In Figure A., focusing on the blue
lines; in order for us to
accurately perceive that there
are two separate lines, and the
proper configuration, we need to
be able to label point y and point
z in the retina and subsequent
retinotopic visual areas so that
higher-order areas properly
segregate the stimulus, and
therefore allowing us to properly
perceive the image.
In Figure B. however, the same
two points in the field of vision
need to be tagged as bound
component in the visual scene,
in order for accurate perception.
It is believed that the manner in
which this labelling occurs is
through neuronal oscillations.
Evidence







– From: Gray and Singer (1989); as shown in Buzsáki (2006)
Singer and Gray recorded both multi-unit activity and local field potentials
(LFP) from single electrodes placed in the Primary visual cortex of
anesthetised and paralyzed cats.
Using a correlational analysis of cell activity and Fourier analysis, they
noticed that a significant proportion of the recordings showed Gamma
frequency oscillations. (30-60 Hz)
This oscillatory response was induced by visual stimulus, consisting of
moving bars.
The cell activity was phase-locked to the trough of the field oscillations.
“These findings provided conclusive evidence that the oscillatory ensemble
events emerged locally… [and] were not directly related to the stimulus but
were added on by the brain.” (Buzsáki, p. 240)
Synchrony between various locations occurred only when neurons at
those locations responded to related visual features of the object.
Furthermore, the determining factor of the “vigour of synchrony” was the
response features of the neurons.
During stimulus-induced transient oscillations, neurons several millimetres
apart, and even contralateral to each other synchronized.
Evidence and the Role of
Gamma Oscillations





In experiments recording multiple neurons in two separate
recording sites in the ‘motion-sensitive’ MT area of waking
monkeys, experimenters stimulated both sites simultaneously,
using 2 bars moving in the preferred directions of the neurons.
 In these trials, there was rarely oscillatory coupling!
Experimenters then substituted the stimuli with a single bar,
which activated both neuronal sites.
 In these trials, there was robust synchrony!
Therefore the oscillatory synchrony was produced not by the
simultaneous excitations of both recording sites, but induced by
the coherence of the stimulus.
Gamma-frequency power has been shown in motor areas during,
and more typically prior to voluntary movement.
Gamma oscillations are commonly induced between 150-300 ms
after stimulus onset, “approximately at the time when stimulus
acquire meaning” (Buzsáki, p.244)
Binding by Synchrony Gamma Oscillations

If Gamma oscillations are to tag certain signals, then it should
follow that they are only found in selective brain areas, and are
not entirely identical; in fact, Intracranial and Subdural
recordings in human corroborate this prediction.



“Recording sites as close as 3-4 millimetres from each other in the
visual cortex yielded quite different amplitudes of gamma
oscillations”. (Buzsáki, p.245)
Experiments in patients with many subdural electrodes showed
that gamma power increased linearly with memory load (when
memorizing strings of syllables), especially above the prefrontal
cortex; with power levels remaining high until retention was lost
and working memory was relaxed.
There is increasing evidence such as this, suggesting that;
“…gamma oscillations are used in the brain for temporally
segmenting representations of different items”. (Buzsáki)
Criticism of the Temporal
Binding Hypothesis
Michael N. Shadlen and J. Anthony Movshon, in their 1999 Review
entitled “Synchrony Unbound: A Critical Evaluation of the Temporal
Binding Hypothesis” brought up roughly a dozen critical concerns about
the reality of Oscillatory Binding.

1.
2.
3.
4.
5.
6.
7.
The hypothesis is not a theory about how binding is computed; it is a theory
only of how binding is signalled.
How does the visual system decide which elements are part of single objects
and which belong to different objects?
Full image segmentation (and recognition) probably requires even higherlevel analyses, including the explicit inclusion of information from memory
about the nature and structure of previously viewed objects and scenes.
If binding is not computed in the primary visual cortex [as this level of
computation focuses on extremely particular features (i.e. edge/contrast
detection and orientation) at a micro level] why is synchrony to be expected
there?
Proponents have stated that “synchronized signals would be particularly
effective in activating post-synaptic neurons that operate as coincidence
detectors. But, how would these coincidence detectors differ from Gnostic
cells?
Oscillations are observed in the cortex which have nothing to do with
perceptual binding, as well, there will always be asynchronous “renegade”
spiking; How do the postsynaptic neurons distinguish which is “special”
synchrony that is suppose to convey additional information?
How is the brain supposed to distinguish the temporal modulation due to
visual input from the temporal modulation produced intrinsically?
Criticism of the Temporal
Binding Hypothesis
8.
9.
10.
11.

Temporally precise visual activity is sufficient for binding, but it is
not necessary for binding and its disruption does not affect binding
elicited by other cues. (In the segmentation of visual stimuli.)
The prevalence of gamma oscillatory responses varies widely
from laboratory to laboratory, for unknown reason.
1.
While Singer, Eckhorn and Livingstone find oscillatory
responses in about half their recordings, most others find
their prevalence to be in about 2%-5% of recording sites.
Initial experiments, from which the theory was developed were
conducted on anesthetised animals. How much did this affect the
results?
Since no perceptual judgements were made during the
experiments, evidence that the chosen stimulus configurations
actually promoted perceptual binding was circumstantial. The
experiments typically used stimuli that promoted binding-like
effects in human observers, but did not establish that experimental
animals perceived the stimuli in the same way.
So what does this mean for the Temporal Binding Hypothesis?
Future Research





Research of the Temporal Binding Hypothesis has suggested that there
may well be a solution to the Binding Problem.
However, experimental results are fairly heterogeneous, and many
researchers fail to observe the type of gamma band oscillations which are
implicated in perceptual, and conceptual binding, as well as those which
may be implicated in memory and consciousness.
Shadlen and Movshon bring up many questions which still need to be
address in order for Binding by Synchrony to become a complete theory.
More experimentation needs to be done in order to answer some of these
questions; however finer grained analysis of local and long range
oscillations are hard to record. Higher spatial and even temporal resolution
is required in order to give researchers a better picture of the behaviour of
gamma oscillations, and its role in Mental Binding.
Regardless of whether gamma band oscillations are relevant in perceptual
binding, the question still needs to be addressed; Whether gamma
oscillations play a role in our brain function, or whether they are merely an
epiphenomenon, a by-product of the actual causal agents of our central
nervous system.