mats_trash at hotmail.com (mat) wrote in message news:<43525ce3.0207240235.5989ab74 at posting.google.com>...
> Im reading around the literature on nerual oscillations given the
> current vogue for explaining many aspects of function (binding
> problem, consciousness) through them. However, I'm a little confused
> as to what is actually referred to by 'oscillation' (i.e. what is
> oscillating?). Is it the fluctuation of resting membrane potentials
> or is it more about the sequential firing of spatially distributed at
> certain frequencies. i.e. given neurons A, B Is the oscillation
> A-B-A-B-A-B
>> Cheers for any explanations or refs to that effect.
Hi mats,
When people talk about oscillations in the brain, they essentially
mean rhythmic activity that can be detected at the EEG level. The EEG
(electroencephalogram) works by recording synchronous activity across
fairly large populations of neurons.
As you know, neural electrical activity is comprised of changes in the
membrane conductance of individual neurons. When ion channels open
(e.g., during synaptic transmission or during an action potential),
current flows across the membrane. This has two related effects: 1) it
changes the cells membrane potential, and 2) it causes a small change
in the distribution of charges inside and outside the cell. If you
perform a "single unit" recording in a living animal, what you are
doing is placing a small wire -outside- a neuron. Since you're
outside, you can't see small changes in membrane potential the way you
could if you were making an intracellular recording. But you -can-
detect the small change in extracellular charge distribution that
occurs when the neuron fires a spike. This is because during a spike,
a truly huge number of ion channels are doing the same thing all at
the same time. The tiny local effects of each channel all sum up
together to give a just-barely-detectable change in the extracellular
charge. This change in charge results in a small, quick change in the
local voltage, and if your electrode is near enough (i.e., a few
microns), you can measure it.
However, the small local voltage change from a single cell, even
during a spike, is too small to be seen with the EEG, which is usually
an array of electrodes farther away in the tissue, or most often, at
the surface of the scalp. But if you can get -many- neurons to fire
spikes at the same time, then the local extracellular potentials all
add up together, and you can detect it. This change in potential from
many synchronous individual neurons goes by many names: local field
potential (LFP), event-related potential (ERP), or
electroencephalogram (EEG). The core idea is that you are recording
simultaneous (but not necesserily rhythmic) activity from many cells,
usually hundreds, thousands, or millions in the case of the scalp EEG.
Now, oscillations: During many different behaviors, it turns out that
the EEG does in fact display some rhythmic behavior. That is, if you
look at it, you can actually see repeating wiggles at certain
frequencies. In practice, you would take the wiggly signal and pass it
through a spectrum analyzer or FFT, and note where the peaks occurred.
For example, during exploration in rats, the EEG recorded near the
hippocampus develops a prominent peak around 4-14 Hz (so called theta
rhythm). During certain other behaviors (usually involving that
"binding" thingy), the prominent rhythm in various parts of cortex is
around 20-80 Hz (so called gamma rhythm). There's a bunch of other
greek letter rhythms too, but I forget when and where they occur.
What these rhythms signify is that lots and lots of neurons are doing
whatever they're doing in a roughly synchronous concerted manner. THIS
DOES NOT MEAN THAT EVERY NEURON IS FIRING AT 20 Hz !!!! In fact, I'm
pretty sure that what EEG, ERP and LFP are most often measuring is
-not- spike firing, but rather the slower subthreshold potentials
associated with synaptic potentials. In a "single-unit" recording, you
can also see these synaptic potentials from large numbers of cells,
but typically one is interested in the spikes, so one high-pass
filters out anything slower than spikes. It is possible to get both
the spike infrmation and the LFP information by passing the same
signal throu two different filters.
OK, so that's what people -really- mean when they talk about
oscillations. But nowadays, people are always talking about
synchronous spiking in the same way that they talk about oscillations,
so what's up with that? Well it turns out that you can try and
determine whether two neurons (possibly in different regions of the
brain) are participating in oscillations with the same frequency and
phase, by analyzing the crosscorrelation between their spiketrains.
The idea is that if the two spike trains are crosscorrelated with each
other, one way of getting such a crosscorrelation (xcorr) would be if
they were both firing spikes at approximately the same times. This
shows up as a peak near zero milliseconds in the xcorr. If you see
such a peak, you might be tempted to shout from the rooftops that
these two cells were spiking together, and since one is in the
auditory system and the other one is in the visual system (for
example), you might conclude that they were both encoding part of the
same complex "percept" (e.g., the sound of screeching tires and the
image of a rapidly approaching truck). This is the idea behind the
whole business of synchrony, oscillations, and the "binding" thingy.
And maybe this is indeed what's going on.
However, there are also other possibilities. For example, there are
other ways of getting two spiketrains to be correlated besides
synchronizing the spikes between two neurons. One way is to simply
have both neurons -start- firing at the same time, but fire each spike
randomly (i.e., correlated latency, without synchrony). Another way is
to have both neurons fire at random times but have their mean spike
rates be correlated. So crosscorrelataion does not imply synchrony
(which is in fact the title of a very nice paper on the subject by
Carlos Brody - do a medline search on him for more info on the these
issues). On the other hand, I believe synchrony -does- strongly imply
crosscorrelation.
Cheers,
Matt
