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

[Andrew Bell <bellring@xxxxxxxxxxxxxxxx>]

Dear Enrique and List:

I appreciate your close interest in this important subject and your positive
feedback. I am glad you have pointed to the paper of Narayan et al, for it
offers instructive insight into how easily data can be misinterpreted.

The key to understanding this paper is to recognise that the tuning tip of
the BM response was _made_ to correspond with the neural tuning tip (see
their footnote 33). That is, the BM tuning curve has an arbitrary position
on the Y-axis relative to the neural curve, and in Fig. 1 the two curves
have been made to correspond at their tips. This relativity can be easily
understood when one recognises that the two Y-axes are not commensurate: the
neural 'threshold' is defined in terms of spike-rate increase on the
spontaneous rate (e.g., a notional 10% increase), whereas the mechanical
threshold relates to the sensitivity (noise floor in units of nm
displacement) of the measuring apparatus and in Narayan et al's case is
measured in terms of dB above the noise floor. Now at neural threshold (13
dB SPL in the case of 'A' and 0.5 dB in 'B'), the BM 'threshold' is made to
match its companion by adjusting down the quoted sensitivity of the
instrument - so in A the BM 'threshold' is set to 2.74 nm, while in B it is
set to 0.26 nm. Therefore a better (less misleading) way of plotting the BM
sensitivity would be relative to the noise floor, which is 0.5 um/s or
about -25 dB SPL (footnote 32). Note also that the neural 'threshold' is
also arbitrary, and will move up and down on the Y-axis depending on the
criterion used for spike-rate increase (e.g. it will move down if a 5%
increase is used as the criterion; up if a 20% increase is decided on).

Once you see that the two curves are arbitrarily placed with respect to each
other, you understand that explanations based on divergent behaviour of the
curves are arbitrary too. [For simplicity, let us confine our discussion to
the BM displacement and neural curves; the BM velocity curve is a
complicating 'adjustment factor' (a high-pass filter of 6 dB/octave) that
can be ignored if we keep to the idea that stereocilia are deflected by
solid structures, not fluid movements.]

IF the two curves are aligned at their tips, then one begins talking of the
tail of the neural curve being "less sensitive" by 15-20 dB than the tail of
the BM curve. But, logically, there is a alternative explanation: IF the
curves are aligned at the tails (say at the 80 dB SPL level), then the tip
of the _BM_ curve is 15-20 dB _less sensitive_ than the neural curve!
Indeed, the latter is the explanation I favour, as I hope the following
makes clear.

At moderate levels (80 dB SPL) the two curves are measuring similar activity
(the whole-scale movement up and down of the partition and its excitation of
the IHCs). At low levels, however, they are measuring different things: the
IHC are efficiently detecting the ripples originating from the OHC SAW
resonator close by, whereas relatively little movement is being communicated
to the BM (what's more, the BM and the relatively large (10-30 um) beads
sitting on it are more or less summing the activity of both OHC2 (in phase)
and OHC1/3 (anti-phase)). It therefore appears, correctly, that the BM is
less sensitive than the IHCs.

We have therefore come to a completely different, but equally valid,
interpretation of the Narayan data. The authors say that differences between
the neural and BM curves are evidence that "certain transformations do
intervene between BM vibration and auditory nerve excitation." Because they
have aligned the tips, they search for differences in the tails, and find it
in terms of high-pass filtering and lack of a high-frequency plateau. My
alternative view sees identity in the tails and looks
for transformations in the tips.

Matching the tails by raising the BM data by 15-20 dB also calls for a
reinterpretation of the observed plateau, which the
authors see in the BM data but not in the neural data. However, the maximum
data-point on the neural high-frequency slope is at 100 dB (in A) or 90 dB
SPL (in B), only some 0 dB (in A) or 10 dB (B) above the BM plateau; if the
BM curve is raised 15-20 dB, it may actually coincide with a similar plateau
in the neural curve, but there is no neural data at high enough levels
to show it. If there were a plateau in the neural curve at about 110 dB SPL,
it would provide confirmation to my alternative explanation (we would want
to align at the plateaus, wouldn't we?).



Andrew.


-----Original Message-----
From: AUDITORY Research in Auditory Perception
[mailto:AUDITORY@LISTS.MCGILL.CA]On Behalf Of Enrique A. Lopez-Poveda
Sent: Tuesday, 20 June 2000 5:30
To: AUDITORY@LISTS.MCGILL.CA
Subject: Re: Wasn't v. Helmholtz right?


Dear Andrew and List,

Like Ben Hornsby, I have been following your discussion very closely.  I
have also read your paper.  I think your model is an excellent piece of
work that leads to many questions that are worth exploring.  There is one
thing, however, that I don't understand.  If BM motion is not the direct
"cause" of IHC excitation, how do you explain, for instance, the
relationship described by Shyamla Narayan, S., Temchin, AN, Recio, A, and
Ruggero, MA [Science 282: 1882-1884] between frequency tuning of BM and
auditory nerve fibres in the same cochleae?  The relationship occurs at
threshold and is almost perfect particularly at the tip of the tuning curve
where, according to your model, BM plays the "least" important of its roles.

-- Enrique
________________________________________________________________
Dr. Enrique A. Lopez-Poveda
Profesor Asociado de Bases Físicas de la Medicina
Facultad de Medicina               Tel. +34-967599200 ext.2749
Universidad de Castilla-La Mancha  Fax. +34-967599272 / 04
Campus Universitario               http://emedica.med-ab.uclm.es
02071 Albacete -- Spain
________________________________________________________________