"Andrew Gyles" <syzygium at alphalink.com.au> wrote in message
news:91luip$bgd$1 at nnrp1.deja.com...
> In article <1eltvig.1jzyjdf10kez7kN%harry at dherwin.org>,
>harry at dherwin.org (Harry Erwin, Ph. D.) wrote:
> > Here's my dumb question: why should neurons need a flip-flop memory
> > element? There's no evidence they're digital.
> > --
> > Harry Erwin, PhD, <mailto:harry at dherwin.org>,
> > Senior Lecturer in Computing at the University of Sunderland,
> > Computational Neuroscientist (modeling bat behavior) and
> > Senior SW Analyst and Security Engineer.
>> They are not strictly analogue, either, are they? A neuron either fires
> or it does not fire. There is no difference in the amplitude of
> individual impulses. Information may be encoded in the frequency of
> impulses in a 'volley', or in the length of time for which a volley
> lasts, so in that respect they are analogue.
>> But something has to 'decide' whether a volley will start, what its
> frequency will be and when it will stop. The kind of flip-flop memory
> elements I described could help to make that decision. Whether what
> then happens is analogue or digital I am unable to say.
>> I am aware that my hypothesis is highly speculative.
> Andrew Gyles
The original not-at-all dumb question was almost certainly
posed with the knowledge that many neurons do not fire
action potentials at all. And even those that do use
"microcircuits" including complexes of dendrodendritic
synaptic connections that would seem necessarily to work
entirely on graded (analog) potentials. And even the action
potentials often have the details of their amplitude and
duration modified by presynaptic modulation that
significantly alters their synaptic effect.
So the digital nature of nervous "computation" is really
quite in question. Only when inputs or outputs must be
transmitted any significant distance do things get hard
coded into trains of action potentials.