Control of jitter (Peter Cariani )

Subject: Control of jitter
From:    Peter Cariani  <peter(at)epl.meei.harvard.edu>
Date:    Fri, 30 May 1997 14:32:01 +0000

Charles S. Watson wrote:

> And,
> as Lloyd Jeffress used to point out, if (1) it takes a millisecond or so
> to cross a synapse, (2) there is some reasonably proportional variance in
> that crossing time, and (3) there are several synapses between cochlea and
> cortex....then you'd darned well better extract any microsec-level
> temporal information early on in the system, because it won't be in very
> good shape further on...if it can be found at all.  (Lloyd probably said
> it more simply)  That was the point in putting his "coincidence detector"
> in the medial geniculate, and then later moving it down to the superior
> olive.

I think this is not necessarily true. First, jitters can be improved if
there is convergence of many small inputs, such that time to threshold
is really what one is looking at. I'm thinking of both the Joris et al
results that the outputs of VCN neurons can have less jitter than their
inputs and the results of Phillips that the jitters of first spikes in
primary auditory cortex are on the same order as those in the auditory
nerve.
Presumably this the result of the Central Limit Theorem at work.
Other sensory systems, such as electroception in weakly electric fish,
use massive convergence to reduce jitters.

There is the question of whether timing precision can be maintained
even in the cortex with an appropriate degree of fan-out and fan-in,
even despite jitters introduced by synaptic transmission
(and even these need to be examined more closely --
maybe they are different for periodic vs.nonperiodic inputs).

Lestienne, Abeles, and others have found precisely replicating triplet
patterns
in cortical spike trains that suggest that precise timings could be
maintained even in visual cortex.

As for timing in the visual system, we only need to look at the work of
Reichardt in the fly visual system and the information theoretic
analysis of fly vision data by Bialek & coworkers. At CNS*96 last
year Strong, who works with Bialek, reported that there was
information at the submillisecond scale available in the timing
of spikes. A great deal of the problem with vision research is
that almost nobody looks at timing precisions below 5 msec. They
are also manifestly NOT looking for stimulus-driven time patterns.
The issue is not the latency of the visual response (which is slow),
but how variable is it (how tight are those threshold crossings
when an edge is drifted across the receptive field -- what are
the jitters like?). As in audition, the interval information
might be a means of explaining apparent hyperacuities of
visual discriminations, provided that the jitters are of the
order of a millisecond or less. This may not be implausible.

I need to run, but I'd be happy to post references to the list
if there is interest.

Peter Cariani


Peter Cariani, Ph.D.
Eaton Peabody Laboratory
Massachusetts Eye & Ear Infirmary
243 Charles St, Boston MA 02114

tel       (617) 573-4243
FAX      (617) 720-4408
email    peter(at)epl.meei.harvard.edu