Perceptual Coherence : Hearing and Seeing

Transformation of Sensory Information Into Perceptual Information 71
modulation that characterizes speech, animal communication, and music.
Neurons in the cochlear nucleus can amplify the depth of the amplitude
modulation over a wide range of sound levels. Moreover, the effect of
background noise is small, and there are neurons in which the response to
the amplitude modulation actually is enhanced in background noise. The
rate of amplitude modulation is coded by the synchronous responses (i.e.,
the phase-locked responses) to the modulation.
The inferior colliculus integrates almost all the ascending acoustic information
and determines the form in which information is conveyed to the
auditory cortex. The receptive fields of many neurons match those from
lower nuclei in the auditory pathways and also match the results from psychophysical
experiments using the same paradigms. The frequency bandwidths
are similar and the receptive fields are relatively constant across
changes in intensity.
This differs from the receptive fields found at the auditory brainstem.
There, the bandwidth greatly increases at higher intensities. This suggests
that the ability to resolve frequencies should decrease at higher intensities
because the bandwidths of the tuning curves measure the ability to resolve
differences between frequencies, but that is not the case. We can measure
the behavioral ability to resolve and integrate frequencies by determining
the frequency range of a noise that masks the detection of a tone at the
middle frequency of the noise. This range has been termed the critical band.
Noise energy outside this frequency range does not affect the detectability
of the tone. This experimental procedure yields bandwidths that are roughly
one third of an octave wide and are basically constant across a wide range of
intensities. Frequencies that fall within one critical band are not resolved
(into distinguishable frequencies) and are summed. It is critically important
to understand that there are many critical bands, and there is a great deal of
Figure 2.24. Panel (A) is a schematic drawing of place × timing models. The
cochlea performs an initial frequency analysis and then the outputs at the different
frequencies are analyzed separately, the place component. The timings between the
spikes of each such output are autocorrelated (equivalent to measuring the frequency
distribution of the intervals), and the excitation is assumed to be proportional
to the size of the autocorrelation at each delay interval. The summary
autocorrelation histogram is calculated by adding the autocorrelations across receptors,
as illustrated at the right. Figure courtesy of Dr. Peter Cariani. Panel (B) is an
example of a summary correlogram based on the first ten harmonics of a 100 Hz
sound. The evenly spaced peaks at each frequency are due to the autocorrelations
at multiples of the period. The summary correlogram isolates the 10 ms period of
the 100 Hz fundamental. From “Virtual Pitch and Phase Sensitivity of a Computer
Model of the Auditory Periphery. I. Pitch Identification,” by R. Meddis and
M. J. Hewitt, 1991, Journal of the Acoustical Society of America, 89, 2866-2882.
Copyright 1991 by the American Institute of Physics. Reprinted with permission.