Subject: Re: Control of jitter From: Peter Cariani <peter(at)epl.meei.harvard.edu> Date: Tue, 3 Jun 1997 14:44:15 +0000Charles S. Watson wrote: > I certainly do not disagree in principle, nor did Lloyd who was well > aware of the potential power of the law of large numbers, averaging > across lots of neural elements, etc. But still, 10 microsec thresholds > (for auditory localization/lateralization) are pretty darned small, and > the combined jitter of, say, 4 to 8 synapse crossings (not to mention the > temporal dispersion created by a wide variety of diameters).... but most > of all, there DOES seem to be a delay-to-place mapping at the level of > the olivary nucleus, as LAJ predicted. I wasn't taking issue with the notion that, if possible, one wants to do fine timing analysis as early in the pathway as possible -- that, all other things being equal, fewer steps in the transmission chain are better. I was taking issue with the common (entropic) assertion that timing must <inevitably> deteriorate as one ascends the auditory pathway. The counterexamples that I mentioned suggest that in at least some systems, timing can be preserved at higher stations -- but note this is not to imply that timing must be preserved in all pathways...... (By the way, there could also be other additional reasons why binaural cross-correlators are situated low in the auditory pathway that have to do with getting rapid motor response times for orienting one's ears and one's body to moving sounds. One may not want to wait 100 or 200 msec for these decisions to be made at higher stations.) The rest are more general ruminations................ Of course, we all think that the auditory system is special and is exquisitely adapted for the processing of temporal information. However, when I look into the literature on other sensory systems there are phenomena that look very similar to what we see in audition. My understanding is that in somatoception, one can distinguish by apparent location on the body time differences between two electrical impulses at different loci down to < 1 msec apart (Bekesy's experiments). I do think we need to look more seriously at exactly how much timing information might be available in higher auditory centers. There is clearly a huge overabundance of timing information related to frequency in the auditory nerve, so that even if a small fraction of the information were preserved (maybe 5-10%) in the primary auditory cortex, this could well be enough to account for frequency discrimination. (It continually astonishes me that the argument is made that timing information is "too good" to account for the frequency resolution of the human auditory system, since it's extremely easy to throw away one kind of information in the pathway in order to include other kinds.) If there is enough timing information available for frequency/periodicity discrimination, then some of the terrible complexities of saturating and nonmonotonic rate-level functions and the multiplicity of factors that can drive cortical discharge rates can be avoided by using timing instead. The dominant impression one gets from the literature is that there is barely any timing information above 200 Hz in the IC and barely anything above 25 Hz in the auditory cortex. Most of the time these assertions are about the lack of obvious, synchronized responses that hit you over the head (like they do in the auditory nerve). On the other hand, if you look more carefully, you see more. At ARO, Shamma's group presented pooled population interval data from their ferret ICs that showed robust encoding of 1200 Hz periodicities (they thought it might go up to about 2 kHz), and this cutoff is consistent with the conclusions that were reached for IC by Steven Greenberg in his analysis of his FFR data. So it's entirely possible that the periodicity information doesn't need to be stuffed into modulation-tuned rate-based periodicity detectors as early as possible. The temporal representations may still be better, even at the IC. This remains to be seen experimentally. Maybe timing information is distributed onto more and more neurons -- sparse temporal coding -- as one goes up the pathway.......and it becomes less and less obvious. We need to look with this in mind. At primary auditory cortex averaged potentials can show periodicities up to 200 Hz or more, which is only the synchronized component of the response, and there are accounts of single unit activity showing locking from several hundred Hz to 1 kHz (de Ribaupierre). Whatever the upper limit, this is a far cry from 25 Hz. What it means is that the modulations that we care about in speech, from voice pitch to VOT down to rhythm and still larger patterns could all be directly temporally coded at the level of the cortex in time patterns of onset and offset responses. Maybe, as in the IC with periodicity pitch, rate profiles of modulation-tuned units aren't the (only) way that these modulations can be represented...... Peter Cariani Here are Joris and Phillips references that were requested: ----------------------------------------------------------------------- Joris, P.X., Carney, L.H., Smith, P.H., and Yin, T.C.T. Enhancement of neural synchronization in the anteroventral cochlear nucleus. I. Responses to tones at the characteristic frequency. J. Neurophysiol. 71: 1022-1036, 1994. Phillips, D.P., and Hall, S.E. Response timing constraints on the cortical representation of sound tIme structure. J. Acoust. Soc. Am. 88: 1403-1411, 1990. (Variability of cortical first-spike latencies.)