Date: Mon, December 10, 15:00-
Place: Room Dw601, D Block, IIS, The University of Tokyo

Invited Speaker: Prof. James A. Simmons (Brown University)

Title: Bat sonar as a computational process in neuroscience

Abstract:
In FM echolocation of big brown bats (Eptesicus fuscus), the minimum
"object" in perception is a point-reflector (a single glint) at a
specific target range.  The echo has a spectrum identical to the
broadcast (~20-100 kHz), and the bat's image of the target's distance is
based on echo delay.  Two or more glints comprise a typical real object.
While the image of the target's distance is based on echo delay, the
image also represents the distances to the glints.  This part of the
image is based on the interference spectrum for closely spaced delay
separations.  The bat's delay accuracy is 10-15 nanoseconds, and the
limit for delay resolution is about 2 microseconds.  Nevertheless, big
brown bats perceive targets in terms of the actual distances to the
individual glints, even for separations of only a few microseconds.
Often, more than one object is present, so that the content of an
auditory scene in echolocation consists of the distances to the various
objects, including the distances to their individual glints.  Neurons in
the bats ascending auditory pathway register the timing of reflections
by single spikes at stable latencies, with a precision of ~10-40
microseconds or less in the brainstem and ~300 microseconds in the
midbrain and cortex.  The sharpness of delay-tuning for cortical neurons
is much less precise than spike latency.  Delay-tuning as a feature is
~50% of the best delay for each cell, which is ~1-10 milliseconds at
different best delays.  In contrast, spike latency variability is only
about 300 microseconds.  Because the sharpness of delay-tuned neurons is
too coarse to account for perception of closely-spaced delays, and
because different subpopulations of delay-tuned neurons respond to
different glint interference patterns, even though the bat perceives all
interference patterns as corresponding delays, the origin of the bat's
perception of targets cannot simply be which population coding from
which neurons are active and which neurons are inactive.  Instead,
perception of each "object" must be based on the timing of spikes
circulating in recurrent neuronal circuits between the midbrain and
cortex.  Neuronal responses reveal such reverberatory signaling.  An
auditory model is presented that postulates how these timing signals are
generated and how they explain what the bat perceives.