Re: Wasn't v. Helmholtz right? (BM/neural tuning)

[Jont Allen <jba@xxxxxxxxxxxxxxxx>]

Dear Andrew, and List,

Andrew Bell wrote:
>
> Dear Enrique and List:
>
> It is clear that this data (like similar data over previous years)
> demonstrates that the BM and neural curves are equally sharply tuned.

Actually the majority of the data do NOT show such agreement. Hand selected
basilar membrane (BM) data, usually taken from the very base of the cochlea, sometimes
agree with select neural tuning, and then only near the tip of the tuning curve.
Never do BM data show "flat tails," which are common in neural data.
Thus, in my opinion (IMO), the situation is more complex than is being presented in
these many BM tuning papers.  After all, the cochlea is nonlinear.

Also the variance is huge in such mechanical data.
If you look at ALL the data, you will find almost any tuning response you can imagine,
except cat neural tuning curves, which have frequency independent tails.
The cat is important here because it is the animal for which we have the most
and best neural tuning data.

> The puzzling question then naturally arises - and it is a question that has been
> vexing hearing science for quite some time (a century or more) - how could
> the BM (or any passive cochlea structure) be so sharply tuned when it is
> essentially a slack membrane, or conjoined fibres, immersed in fluid? Such a
> structure would have a very low Q.

Actually the cochlea is very low loss. Gold failed to do a simple "viscous
boundary layer" calculation that would fix his error. If you check this calculation,
you will find that the viscous boundary layer in water is about 10 microns. This is
much less than the dimensions of the scalae. Thus the cochlear damping is actually
very small. The  major source of damping is the fluid drag in the small region where
the hair cell cilia reside.
From an efficiency point of view this is ideal. T
[[portion lost - dpwe]]
middle ear, and is deposited with no loss at the cilia, which are in a resistive
luid space. The cilia are about 6 microns high, which defines this fluid filled
region. It is also close to a viscous boundary layer depth

@article{Allen80,
   author = {Allen, J. B.},
   title = {Cochlear micromechanics: {A} physical model of transduction},
   journal = JASA,
   volume = {68},
   number = {6},
   pages = {1660-1670},
   year = {1980}
}

>
> So if we assume that 'intermediate' structures merely amplify the BM
> response, we end up with something with a similarly low Q. As Gold
> suggested, some positive feedback is needed to sharpen the response.

There are many paper written on this subject, and not one of them has
provided a physically motivated model which can provide the necessary cochlear
amplifier gain (typically assumed to be between 40 and 60 dB). One of the main
problems is the wrong sign for the OHC feedback. This is typically dealt with
by assuming a place dependent delay, which is not physically motivated (IMO).

@article{Allen92,
   author = {Allen, J. B. and Neely, S. T.},
   title = {Micromechanical models of the cochlea},
   journal = {Physics Today},
   pages = {40-47},
   volume = {45},
   number = {7},
   month = jul,
   year = {1992}
}

IMO your reasoning is based on a faults assumption (that the cochlear damping is
large).

> I have suggested a mechanism involving the OHC and tectorial membrane that
> provides high Q in a fluid environment. Incidentally, the BM picks up this
> vibration. However, the BM is not the primary tuned element of the ear; it
> is an accessory structure designed to absorb excess energy.

This is your assumption, not a proven fact. Please, let's distinguish our
observations from our assumptions.

>
> Andrew.
>

Jont

> -----Original Message-----
> From: AUDITORY Research in Auditory Perception
> [mailto:AUDITORY@LISTS.MCGILL.CA]On Behalf Of Enrique A. Lopez-Poveda
> Sent: Tuesday, 27 June 2000 12:38
> To: AUDITORY@LISTS.MCGILL.CA
> Subject: Re: Wasn't v. Helmholtz right? (BM/neural tuning)
>
> Andrew,
>
> I take your point and I agree with your comment that Narayan et al. could
> have made the curves to match at their base.  In that case, neural
> thresholds would be lower than the BM threshold.  However, even if you did
> that, their data show that BM response is as sharply tuned as AN response.
> The bandwidths of their tuning curves are comparable at the tip, even
> though their thresholds may be different.  Assuming that the BM and AN
> tuning curves are made to match at their tails, one could then argue that
> the role of intermediate structures (e.g., OHC, IHC, etc) could be only to
> "amplify" the BM response, but NOT to sharpen the tuning of the system.
> Any comments on this?
>
> -- Enrique

--
Jont B. Allen
AT&T Labs-Research, Shannon Laboratory, E161
180 Park Ave., Florham Park NJ, 07932-0971
973/360-8545voice, x7111fax, http://www.research.att.com/~jba