Frontiers in Behavioral Neuroscience,
Год журнала:
2018,
Номер
12
Опубликована: Март 6, 2018
The
gradual
accumulation
of
sensory
evidence
is
a
crucial
component
perceptual
decision
making,
but
its
neural
mechanisms
are
still
poorly
understood.
Given
the
wide
availability
genetic
and
optical
tools
for
mice,
they
can
be
useful
model
organisms
study
these
phenomena;
however,
behavioral
largely
lacking.
Here,
we
describe
new
evidence-accumulation
task
head-fixed
mice
navigating
in
virtual
reality
(VR)
environment.
As
navigate
down
stem
T-maze,
see
brief
pulses
visual
on
either
side,
retrieve
reward
arm
with
highest
number
pulses.
occur
randomly
Poisson
statistics,
yielding
diverse
yet
well-controlled
stimulus
set,
making
data
conducive
to
variety
computational
approaches.
A
large
different
genotypes
were
able
learn
consistently
perform
task,
at
levels
similar
rats
analogous
tasks.
They
sensitive
side
differences
single
pulse,
their
memory
cues
stable
over
time.
Moreover,
using
non-parametric
as
well
modeling
approaches,
show
that
indeed
accumulate
evidence:
use
multiple
from
throughout
cue
region
maze
make
decision,
albeit
small
overweighting
earlier
cues,
performance
affected
by
magnitude
not
duration
evidence.
Additionally,
analysis
mice's
running
patterns
revealed
trajectories
fairly
stereotyped
modulated
amount
evidence,
suggesting
navigational
this
may
provide
continuous
readout
correlated
underlying
cognitive
variables.
Our
which
readily
integrated
state-of-the-art
techniques,
thus
valuable
tool
circuit
dynamics
particularly
under
more
complex
contexts.
Choice
confidence,
an
individual's
internal
estimate
of
judgment
accuracy,
plays
a
critical
role
in
adaptive
behaviour,
yet
its
neural
representations
during
decision
formation
remain
underexplored.
Here,
we
recorded
simultaneous
EEG-fMRI
while
participants
performed
direction
discrimination
task
and
rated
their
confidence
on
each
trial.
Using
multivariate
single-trial
discriminant
analysis
the
EEG,
identified
stimulus-independent
component
encoding
which
appeared
prior
to
subjects'
explicit
choice
report,
was
consistent
with
measure
predicted
by
accumulation-to-bound
model
decision-making.
Importantly,
trial-to-trial
variability
this
electrophysiologically-derived
signal
uniquely
associated
fMRI
responses
ventromedial
prefrontal
cortex
(VMPFC),
region
not
typically
for
perceptual
decisions.
Furthermore,
activity
VMPFC
functionally
coupled
regions
frontal
linked
decision-making
metacognition.
Our
results
suggest
that
holds
early
representation
arising
from
dynamics,
preceding
potentially
informing
metacognitive
evaluation.
Decisions
are
often
made
by
accumulating
ambiguous
evidence
over
time.
The
brain’s
arousal
systems
activated
during
such
decisions.
In
previous
work
in
humans,
we
found
that
evoked
responses
of
decisions
reported
rapid
dilations
the
pupil
and
track
a
suppression
biases
accumulation
decision-relevant
(de
Gee
et
al.,
2017).
Here,
show
this
arousal-related
decision
bias
acts
on
both
conservative
liberal
biases,
generalizes
from
humans
to
mice,
perceptual
memory-based
challenging
sound-detection
tasks,
impact
spontaneous
or
experimentally
induced
choice
was
reduced
under
high
phasic
arousal.
Similar
occurred
when
drawn
memory.
All
these
behavioral
effects
were
explained
biases.
Our
results
point
general
principle
interplay
between
decision-making.
Learning
from
successes
and
failures
often
improves
the
quality
of
subsequent
decisions.
Past
outcomes,
however,
should
not
influence
purely
perceptual
decisions
after
task
acquisition
is
complete
since
these
are
designed
so
that
only
sensory
evidence
determines
correct
choice.
Yet,
numerous
studies
report
outcomes
can
bias
decisions,
causing
spurious
changes
in
choice
behavior
without
improving
accuracy.
Here
we
show
effects
reward
on
principled:
past
rewards
future
choices
specifically
when
previous
was
difficult
hence
decision
confidence
low.
We
identified
this
phenomenon
six
datasets
four
laboratories,
across
mice,
rats,
humans,
modalities
olfaction
audition
to
vision.
choice-updating
strategy
be
explained
by
reinforcement
learning
models
incorporating
statistical
into
their
teaching
signals.
Thus,
mechanisms
continually
engaged
produce
systematic
adjustments
even
well-learned
order
optimize
an
uncertain
world.
Learning
should
be
adjusted
according
to
the
surprise
associated
with
observed
outcomes
but
calibrated
statistical
context.
For
example,
when
occasional
changepoints
are
expected,
surprising
weighted
heavily
speed
learning.
In
contrast,
uninformative
outliers
expected
occur
occasionally,
less
influential.
Here
we
dissociate
from
degree
which
they
demand
learning
using
a
predictive
inference
task
and
computational
modeling.
We
show
that
P300,
stimulus-locked
electrophysiological
response
previously
adjustments
in
behavior,
does
so
conditionally
on
source
of
surprise.
Larger
P300
signals
predicted
greater
changing
context,
context
where
was
indicative
one-off
outlier
(oddball).
Our
results
suggest
provides
signal
is
interpreted
by
downstream
processes
differentially
order
appropriately
calibrate
across
complex
environments.
Frontiers in Behavioral Neuroscience,
Год журнала:
2018,
Номер
12
Опубликована: Март 6, 2018
The
gradual
accumulation
of
sensory
evidence
is
a
crucial
component
perceptual
decision
making,
but
its
neural
mechanisms
are
still
poorly
understood.
Given
the
wide
availability
genetic
and
optical
tools
for
mice,
they
can
be
useful
model
organisms
study
these
phenomena;
however,
behavioral
largely
lacking.
Here,
we
describe
new
evidence-accumulation
task
head-fixed
mice
navigating
in
virtual
reality
(VR)
environment.
As
navigate
down
stem
T-maze,
see
brief
pulses
visual
on
either
side,
retrieve
reward
arm
with
highest
number
pulses.
occur
randomly
Poisson
statistics,
yielding
diverse
yet
well-controlled
stimulus
set,
making
data
conducive
to
variety
computational
approaches.
A
large
different
genotypes
were
able
learn
consistently
perform
task,
at
levels
similar
rats
analogous
tasks.
They
sensitive
side
differences
single
pulse,
their
memory
cues
stable
over
time.
Moreover,
using
non-parametric
as
well
modeling
approaches,
show
that
indeed
accumulate
evidence:
use
multiple
from
throughout
cue
region
maze
make
decision,
albeit
small
overweighting
earlier
cues,
performance
affected
by
magnitude
not
duration
evidence.
Additionally,
analysis
mice's
running
patterns
revealed
trajectories
fairly
stereotyped
modulated
amount
evidence,
suggesting
navigational
this
may
provide
continuous
readout
correlated
underlying
cognitive
variables.
Our
which
readily
integrated
state-of-the-art
techniques,
thus
valuable
tool
circuit
dynamics
particularly
under
more
complex
contexts.