question: job of a neuron

Matt Jones jonesmat at
Tue Sep 11 10:39:56 EST 2001

"John H" <John at> wrote in message news:<C5Lm7.2370$iH4.174324 at>...
> Matt,

> Inhibition.
> Thanks for the evolution lesson. Reading today a basic neuro primer (layman,
> reads journal articles while reading primers) the author stated that in
> humans interneurons occur in very high numbers relative to other animals. I
> believe Matt that you have done research into GABA(still are?) and wonder if
> you can provide any insight into the idea of this high density of
> interneurons. If true, why, what is so special about human brains that we
> would require so many interneurons?
> No, I cannot make any sense of inhibition being first. What a mystery ... .
> John.

I think Richard makes interesting points elsewhere in this thread
about why humans would have more interneurons, and I haven't really
thought much about -why- we should have a lot, and in fact wasn't even
aware that we did. I think Richard is referring to "interneurons" in
the broadest sense, i.e., any neuron that doesn't project outside of
aits local circuit, but interconnects cells within the circuit. These
can be either excitatory or inhibitory. Below, I am referring to the
narrower meaning of GABAergic inhibitory interneurons.

Mostly, I wonder why we have so few! Here's what I mean: In cortex and
hippocampus (and many other places), the projection neurons that send
signals to other brain regions tend to be excitatory, often arranged
in a regular sheet, and are very numerous, outnumbering inhibitory
interneurons by about 10-to-1. Meanwhile, the inhibitory neurons are
fewer, and scattered around without nearly as much obvious
orientational regularity. Also, there are often just one or two kinds
of excitatory neurons, but lots and lots of distinctly different kinds
of interneurons. In hippocampus, there are about 3 main types of
excitatory cells (granule cells and CA1 and CA3 pyramidal cells), but
there are about 20 kinds of identifiable interneuron types (depending
on who you read). These interneurons can be distinguished by the
histochemical markers they express (i.e., calcium binding proteins,
neurotransmitter receptors), by their location and shape, and by
-where- onto the principle cells they form their inhibitory synapses
(i.e., dendrites vs soma vs axon). Also, they tend to make thousands
of synapses through axons that branch widely and contact many many
principal cells at once.


Well, it's probably -not- just to stop principal cells from firing.
You don't need anything like this much complexity to do such a
pedestrian task. One of the more well-studied aspects of  interneuron
function appears to be that they control the periodicity of
oscillations, for example theta and gamma oscillations. These
oscillations have been proposed to act something like a clocking
signal, that allows different circuits to synchronize the activity of
many principal cells, thereby marking those cells as conveying
information about a common object or percept. The evidence for this
role of interneurons in dictating oscillation properties is now quite
strong, though the "purpose" of the oscillations themselves is still
very much in debate.

It turns out you can get fairly interesting behavior out of circuits
that contain -nothing- but inhibitory neurons, because each cell has
its own native excitability even without excitatory synaptic drive.
Thus, two interneurons coupled to each other can produce either
synchronous or alternating oscillations. Many interneurons wired up
together can create a "pacemaker" circuit, even though none of the
cells themselves are inherently periodic. Interneuron-interneuron
connections are widespread in cortex, and this may be part of their

Aside from oscillations, interneurons probably contribute to
information processing by deciding -exactly- when a target neuron is
allowed to spike and when not. This is clearly important at the gross
level, because if you block inhibition you get seizures, and if you
enhance it, you get amnesia, anxiolysis, analgesia and unconsciousness
(e.g., almost -all- general anesthetics potentiate GABA-A receptor
function). However, on a finer level, it makes sense for information
processing to be able to regulate spiking in a complex way. If you
want to send a high amount of information over a channel, you need to
be able to send an awful lot of different patterns, not just an
oscillation with a fixed frequency. For example, a perfectly periodic
pulse train contains literally only 1 bit of information (i.e.,
whether the train is on or off). That is, it can be compressed into a
single binary number (on or off). However, a string of pulses
separated by random times cannot be compressed at all. Therefore it
takes as many bits to describe it as there are time intervals in the
pulse train (i.e., t contains a lot more information). Highly
random-looking spiketrains are much more likely to be high in
information content than regular ones (note that it can therefore be
quite difficult to tell the difference between efficient signalling
and noise).

Inhibition helps neurons to produce highly random-looking spiketrains
(e.g., see papers by Dieter Jaeger), and thus increases the
information capacity. It is not clear whether this capacity is actualy
used, but it is there. And largely because of  the complexity of
inhibitory connectivity.



More information about the Neur-sci mailing list