Transcript Fredericks

Primary Cortical
Representation of Sounds
by the Coordination of
Action-Potential Timing
R. Christopher deCharms & Michael M.
Merzenich
Take-home: “Population coding based on relative spike timing
can systematically signal stimulus features, it is topographically
mapped, and it follows the stimulus time course even where mean
firing rate does not.”
Rationale
Cortical population coding is still not understood.
Firing rates versus relative timing of firing
Changes in firing rates of most neurons do not
reflect duration of a stimulus.
Sensorimotor neurons in frontal cortex change
the relative timing of firing even when firing rate
does not change.
Purpose: Investigate the role of relative action
potential timing in population coding of stimuli in
the primary auditory cortex.
Methods
3 marmoset monkeys
Extracellular recordings from pairs of locations of
neurons in the supragranular primary auditory cortex 751,000 µm apart
Computer sampling at 20 kHz and spike sorting
Thresholded signals gave 54 pairs of single units & 369
pairs of multiunit groups
Auditory stimuli produced by a DSP chip and presented
binaurally
Statistical significance est. between cross-correlograms
Mean, s.e. & confidence limits derived from all trials
(~100)
Question 1
Do cortical neurons that fire at low rates
provide information about stimuli?
Figure 1
Conclusion:
“Cortical neurons can thus maintain signals
a–d: average crosscorrelations
about
ongoing stimuli by temporally coordinating the few
between 2 locations
action
potentials present
at low
firingtorates.”
(100 reps.)
f & g:even
firing-rate
response
a long-lasting
stimulus similar to a brief 50 ms stimulus
e: amplitude
envelope; 70 dB at
4 kHz; purestimulus tone
f & g: mean firing
rates at the 2
locations
d:&Control
b
c: the change
for remaining
in coordinated
relative spike
rate was
timing or
-5 for
-4 for
highly significant
short-term
synaptic
(P<1x10
plasticity
after
b, P>3x10
the
c, permutation
introduction
of atest)
stimulus: unchanged
Question 2
Is neuronal coordination effective at the
level of individual pairs of neurons or at
the population level?
Figure 2
a: single pair of neurons: increased
cross-correlation during the
stimulus, but variability too high
b: single pair of neurons with a
significant cross-correlation, but
little or no change in tonic firing
rate
Conclusion: “Neuronal
c:coordination
comparison of two
is agroups of
well-isolated units with similar
population effect.”
effect to b, but more robust
d: two thresholded multiunit
groups; similar effect as in c
e: single unit and the local field
potential (adjacent electrode):
increase in correlation with sound
Question 3
Is neuronal coordination stimulus specific,
and if so, how is this specificity
represented on the cortex?
Figure 3
a-h: frequency tuning of
spike time coordination
between two groups of
neurons
when
mean
3e: coordination rate increases even
Conclusion:
“ Where firing rate and relative timing
firing
decreases
both change, the absolute number of coordinated
Coordinated
rateboth
of action
was increased
events reflects
effectspotentials
.”
significantly by a 4 kHz stimulus & significantly
decreased by a 6.35 kHz stimulus
Figure 3
Conclusion: Topographic map of coordinated spike
tuning reflects the map for traditional onset burst tuning.
4 kHz
4 kHz
70 dB
40 dB
i-k: spatial locations
on auditory cortex
Red: increased
coordinated rate
Blue: decreased
coordinated rate
Center frequency of tonic phase coordinated spike rate tuning
2.52 kHz
40 dB
Question 4
Does neuronal coordination mimic the
duration of the stimulus at locations where
the firing rate does not?
Figure 4
e: mean
plus s.e.across
of all the a
Conclusion:
“…a target
which
samples
a: coordinated discharge
rate atlocation
two
coordinated spike-rate time
locations& non-homogeneous population
broad
of neurons that are
courses
firing
onsignificant
average
nearinbackground
can nonetheless
b & c: no
change
firing rate at f:rate
average of all individual firing
extract
the
time
course
of
an
ongoing
using as in
the same two locations as in a (not even
ratesstimulus
from sameby
population
transients)
temporal coincidence information from
e
that population.”
Conclusions
At stimulus transients, population firing rate &
coordinated firing rate both reflect the change in
signal
Results reveal that relative spike timing of action
potentials in the primary auditory cortex may be
vital for signaling features of a stimulus once
population firing rate has returned to baseline.
More basic than the concept of neuronal
grouping
Evolving temporal pattern of activity