Peripheral nonlinearities

[Walt Jesteadt <JESTEADT@xxxxxxxxxxxx>]

        I have a general question about the order of events in the peripheral
auditory system and it occurred to me that it would be worthwhile to ask those
of you on the auditory list.  If you send responses directly to me rather than
to the list as a whole, I will put together a summary and send it out.

        I will couch this discussion in terms of a 1995 paper in Hearing
Research by Oxenham and Moore, which I think I have read and have remembered
correctly.  My apologies to them if I have not.  They present a black box model
of peripheral processing consisting of auditory filters, followed by a
nonlinearity, followed by a temporal window, then a decision device.  They cite
other examples of similar models.  Other examples exist where the filter or
excitation pattern segment of peripheral models is assumed to have linear
input, but later processing is not defined. There are a few others (Zwicker's
specific loudness model, for example), where the filter or excitation pattern
process is assumed to have nonlinear input, or to be nonlinear itself.  I am
curious as to what the current consensus is regarding the location of the
nonlinearity and implications of that for models at the psychoacoustic level.

        Oxenham and Moore note two sources of assumed nonlinearity, basilar
membrane mechanics and hair cell transduction.  For their purposes, they lump
these sources together.  Only the latter source, however, clearly comes after
the filter.  They chose their conditions to avoid any role of the filter, so the
possibility of a nonlinearity before the filter would have no impact on their
particular study, but it would in many other cases.

        Given nonlinearity as well as frequency selectivity in the basilar
membrane response, and stimuli of unequal level presented within the frequency
range of a given auditory filter, should I just sum the power across frequency,
as the roex program does, and assume a nonlinear transform afterwards, or should
I incorporate a nonlinear transform ahead of the auditory filter, or should I do
both?  I would appreciate people's opinions and supporting data if such exist.
If there appears to be any interest in this question, I will get a summary out
to people as soon as possible.

Walt Jesteadt               INTERNET: jesteadt@BoysTown.ORG
Boys Town National Research Hospital
555 North 30th Street          PHONE:  (402) 498-6704
Omaha, NE 68131                  FAX:  (402) 498-6351