"George Bajszar" <gyuri at usa.net> wrote in message
news:91b9be$bas$1 at nnrp1.deja.com...
> I am trying to understand the basics behind neural spike trains and I
> only have guesses at this point. Can neural spike trains be viewed as
> an information stream where the timing between each spike represents
> unique data bits of the specific information stream, i.e. visual
> explicit memory data?
>> Does timing between spikes carry such significant information? If so,
> can neurons be trained to record data streams through their input
> synapses and then be somehow triggered for replaying them through their
> output synapses?
>> Thanks for any info.
The classical view of spike trains is that information is primarily
encoded simply in the average frequency of potentials. There are
cases, as in the control of crustacean muscle, where the synapse
(actually neuromuscular junction) is highly facilitating. In that case,
patterned activity, as in two or three pulses per burst in a series of
bursts can produce significantly different results from a uniformly
distributed spike train. Even a single action potential interpolated
into a uniform spike train can make a significant difference. And
the medial superior olivary nucleus in the mammalian auditory
system is a "coincidence detector" identifying simultaneity of
inputs from the two ears. It can detect time differences as fine
as 10 microseconds. So timing does matter.
However, I don't know of any case where neurons can actually
identify a binary coded pulse train where each interpulse
interval codes for an independent data value. Even in these
cases, the relatively crude pattern has to be repeated and
averaged over a number of instances.