Re: paper on human cochlear tuning

[Andrew Bell <bellring@xxxxxxxxxxxxxxxx>]

Christopher:

Thankyou for bringing this paper to our attention. Using SFOAEs at
40 dB SPL, you and your colleagues show that the human ear can
display Qs of 15-30, increasing with frequency from 1 to 10 kHz --
sharper than the standard experiments indicate. It well demonstrates
the simplicity and power of SFOAEs.

One thing your paper omits to do is acknowledge psychophysical
studies which, more than 50 years ago, came to about the same
conclusion as you do. In 1948, Gold and Pumphrey concluded that "the
selectivity of the resonant elements... of the human ear is in fact
roughly proportional to frequency and is very much higher than has
generally been supposed" [Proc. Roy. Soc. B, 135, 462]. This sounds
similar to your statement on p.3322 that "human filter bandwidths
are considerably narrower than previously believed throughout the
cochlea." The only difference is that Pumphrey and Gold's figures,
derived from working at lower intensities than you did, show even
higher Qs.

In fact, your paper can be seen as really underlining how far ahead
of their time Pumphrey and Gold were in the late 1940s. I think it
would have been appropriate to make reference to their pioneering
and revolutionary work.

In Nature 160 (1947), 124 [Transient Reception and the Degree of
Resonance of the Human Ear], Pumphrey and Gold give a description of
a revealing experiment in which they measured the Q of the human ear
psychophysically and found that, at SPLs near threshold, the Q is in
the range 200-350 from 2.5-10 kHz, falling to 50 at 1 kHz. This
fundamental experiment involved an observer comparing the
detectability of a continuous tone (at threshold) with a similar
wave-form that comprised only a finite number of cycles. Clearly,
the latter will need to have a higher intensity to be audible, and
this difference will depend on the Q of the system. The higher the
Q, the more cycles will be needed to get the resonant elements
ringing at audible amplitude; at the same time, for a fixed number
of cycles, a higher intensity will be needed. From the trade-off
between number of cycles and intensity increase, Pumphrey and Gold
calculate the equivalent Q. A very nice experiment.

An even more elegant piece of work is reported soon after in Nature
161 (1948), 640 [Phase Memory of the Ear: a Proof of the Resonance
Hypothesis], in which Pumphrey and Gold use a different paradigm
based on a series of wavetrains separated by an integer number of
cycles of silence. If an observer can distinguish that stimulus from
a similar one in which every second wavetrain is of inverted
polarity, then the ear must be able to 'remember' the inversion
across the silent gaps. Again, the experiments were done at low
intensity -- about 20 dB above threshold -- and the resulting Q
values were similar to before: the ear could remember up to 30
cycles, translating to a Q of at least 100 and as high as 250 at 5
kHz. At this point it is worth emphasising your statement that human
auditory frequency selectivity is determined at the level of the
periphery (that is, frequency discrimination is all there physically
in the cochlea, and neural sharpening is a concept to be discarded).

It would be interesting if you were to repeat your experiments at 20
dB SPL, and see whether you again need to raise the Q values
attributed to the human ear. This is particularly so in the light of
your comment [p.3321] that "previous behavioural measures of
frequency selectivity [meaning Fletcher's paradigm, actually, not
Pumphrey and Gold's] greatly overestimate the bandwidths of
peripheral cochlear filters..." due to suppression by the masker or
self-suppression. One can only expect Qs to increase as SPLs
decrease. For example, it is not unusual for low-level SOAEs between
1-2 kHz to have bandwidths of less than 1 Hz, implying an equivalent
Q of more than 1000.

This example is relevant to your comment on p.3322: "The development
of an objective, noninvasive measure of cochlear tuning solves a
long-standing problem in the hearing sciences." Granted, your work
has given us another tool on the objective front, but haven't simple
measurements of SOAE bandwidths already given us the same sort of
information?

And given that SFOAEs are entraining the same oscillators as give
rise to SOAEs, I would venture to say that SFOAE-derived Q values
will undoubtedly rise as SPLs diminish. The Pumphrey and Gold
experiments are of course measuring the very same resonators. I am
surprised that Pumphrey and Gold's findings were not taken up at the
time, and their work still seems to be overlooked. The title of
their second paper remains a challenge to today's auditory
theorists. It seems clear that SOAEs represent the continuous
ringing of the ear's resonant elements for which Pumphrey and Gold
were searching. They have been proved right about their seminal idea
of a "regenerative receiver" mechanism in the cochlea; why then,
with detailed knowledge of SOAEs, does the resonance theory, for
which they offered solid evidence, continue to be ignored?

Can I suggest that list members again take time to read Pumphrey and
Gold, particularly the more detailed companion papers in Proc. Roy.
Soc. B, 135 (1948), 462-498. The quote on the first page throws down
the gauntlet: "Previous theories of hearing are considered, and it
is shown that only the resonance hypothesis of Helmholtz... is
consistent with observation."

In this context, your own recent JASA paper adds an important
perspective to the issue. You say that "These results [the
phase-constancy of BM responses to clicks -- the intensity of which
range over almost the entire dynamic range of hearing] contradict
many, if not most, cochlear models in which OHC forces produce
significant changes in the reactance and resonant frequencies of the
partition" [JASA 110 (2001), 332]. Perhaps you could offer us a list
of which models remain in the running? You point out that this
result implies, physically, that the local resonant frequencies of
the cochlear partition are virtually independent of intensity. I
take this to mean that the resonance theory would be on your list?

Thankyou again for raising this important issue.


Andrew.

________________________________

Andrew Bell
PO Box A348
Australian National University
Canberra, ACT 2601
Australia
Phone {61 2} 6258 7276
Fax {61 2} 6258 0014
Email bellring@smartchat.net.au
________________________________

Endnote:
For the record, here are other specific places in your paper where
due recognition of Pumphrey and Gold calls for changes.

p.3318. "Filter bandwidths in humans... have relied on
psychophysical (i.e., behavioral) measures of filter bandwidth based
on the phenomenon of masking". You have overlooked the Pumphrey and
Gold threshold-detection approaches.

p.3321. "To obtain the most accurate behavioral measures of cochlear
tuning that psychophysics can currently provide..."; again Pumphrey
and Gold measured these parameters at lower intensities and arrived
at superior Q values. They commented that the scatter of their
results was small and that agreement with the theoretical slope was
good.



 |>-----Original Message-----
 |>From: cochlea-admin@auditorymodels.org
 |>[mailto:cochlea-admin@auditorymodels.org]On Behalf Of
 |>Christopher Shera
 |>Sent: Thursday, 21 March 2002 1:04
 |>To: auditory@lists.mcgill.ca; cochlea@auditorymodels.org
 |>Subject: [Cochlea] paper on human cochlear tuning
 |>
 |>
 |>Dear list members,
 |>
 |>If you can forgive the shameless advertisement,*
 |>some of you may be interested in the following paper
 |>recently buried in an out-of-the-way location:
 |>
 |>"Revised estimates of human cochlear tuning
 |>from otoacoustic and behavioral measurements"
 |>C.A. Shera, J.J. Guinan, Jr, and A.J. Oxenham
 |>Proc. Natl. Acad. Sci. USA 99:3318-3323.
 |>
 |>http://www.pnas.org/content/vol99/issue5/#PHYSIOLOGY
 |>