Transcript Paper

Zicong Zhang
2011.08.15
Authors
• Wendy A. Suzuki
Professor of Neural Science and Psychology,
New York University
Research interest: Organization of memory in the medial temporal lobe and
the effects of exercise on learning, memory and cognition.
Yuji Naya
Post-Doc fellow, associate faculty member in Suzuki’s Lab. He is best known
for the groundbreaking work he did in collaboration with Yasushi
Miyashita on the representation of well-learned associations within the
perirhinal cortex.
2
Backgrounds
• The human medial temporal lobe (MTL) is critical
for episodic memory presumably because of its
role in binding individual stimuli or events to their
temporal and spatial contexts.
• Computation models:
Cortical association areas signal information
about items;
Parahippocampal regions signal information
about items along with their temporal context;
Hippocampus (HPC) supervises these itemcontext associations.
3
Backgrounds
• fMRI studies in humans report both HPC and
parahippocampal activation during tasks of
temporal-order memory.
• Recent neurophysiological studies in the
rodent have highlighted the role of HPC in
signaling either a particular time within a trial
or incremental timing across the entire
recording session.
4
Questions and Ideas
• Little is known about the neurophysiological
basis of how item and timing information is
integrated within MTL.
• In this study, neural activities from MTL areas
and a control visual area were recorded as
nonhuman primates performed a temporalorder memory task that required encoding of
two visual items and their temporal order.
5
Training Design
Cue 1 delay period
“Time cells”: neurons whose responses differentiated between the cue 1 and cue 2
periods, signaling relative timing between cue presentations or temporal order of cue
presentations.
“Item cells”: neurons that showed significant stimulus-selective activity during either
cue 1 or cue 2.
6
Recordings
“Time cells”: neurons
whose responses
differentiated
between the cue 1
and cue 2 periods,
signaling relative
timing between cue
presentations or
temporal order of cue
presentations.
“Item cells”: neurons
that showed
significant stimulusselective activity
during either cue 1 or
cue 2.
7
Recordings
Two-dimensional plots of the population average LFP spectrogram in all areas.
The gamma band activity (>30 Hz) in HPC increased during the cue 1 delay
period and was significantly stronger during cue 2 than cue 1.
8
Question 2
• How did the population of cells signal time
information?
9
Population Vector Analysis: backgrounds
• Original Idea: to explain how populations of motor
cortex neurons encode movement direction.
• Basis: individual neurons tended to discharge more for
movements in particular directions, the so-called
preferred directions for individual neurons.
• In the population vector model, individual neurons
'vote' for their preferred directions using their firing
rate. The final vote is calculated by vectorial
summation of individual preferred directions weighted
by neuronal rates.
• “Population vectors were used to examine information
represented by a population of prefrontal activity and
its temporal change during spatial working memory
processes…” (Takeda and Funahashi, 2004)
• In the present study: “to characterize the temporal
dynamics of time cells observed throughout the MTL.” 10
Population Vector Analysis for time cells
Population vectors were
constructed from the
responses of all individual time
cells for 40-ms time bins
throughout the cue 1 delay
period. Their temporal
dynamics were characterized
relative to two “template” time
periods defined as the cue 1
and the cue 2 states.
Normalized distances (NDs) from the cue 1 states (solid circles) and the cue 2
states (open squares) in each area.
HPC: the distance from the
two templates changed
constantly.
ERC: only the one from cue 1
state gradually increase.
PRC: sudden shifted at the
early period and remained
relatively constant.
11
Population Vector Analysis for time cells
“To determine which area provided the most accurate representation of the
cue 2 state before the presentation of cue 2, we asked which of the three
areas exhibited the shortest distance to the cue 2 state during the last quarter
(240 ms) of the cue 1 delay period.”
12
Summary 1
• As a population, HPC time cells provide an
incremental timing signal that gives an
estimate of the relative time from the last cue
presentation as well as an estimate of the
relative time to the next cue presentation.
13
Question 3
• How did the MTL represent item information?
14
Item information represented in MTL
“PRC and TE represented the
same items across the two
cue periods. A time effect was
revealed in PRC such that the
response to the neuron’s
preferred stimulus differed
across cue 1 and cue 2.”
Mean discharge rates and SEM during cue 1 and cue 2.
“The small correlation
coefficients in ERC can be
explained by the lack of stimulus
selectivity in one of the two cue
periods. This differential level of
stimulus selectivity between cue
periods can serve to integrate
item information with a relative
timing signal.”
Distributions of correlation coefficients between cue 1 and cue 2.
15
Question 4
• Whether the PRC item cells integrate item and
temporal order information by modulating
their stimulus-selective response properties
across the cue periods?
16
Differential time effect in PRC
C: Normalized vector distances between
cue 1 and cue 2
Suggesting “a more
prominent time effect in
PRC”
D: Normalized mean discharge rates
The differential time effect
was not due to general
changes in firing rates
E: Kurtosis measure
Suggesting that “PRC
differentiates between the
cue 1 and cue 2 periods
by changing the
sharpness of its stimulusselective response”
17
HPC provides a robust incremental timing
signal that may serve to anchor the timing
to events within an episode.
ERC neurons can signal incremental
timing information as well as integrate
item and time information at a lower
amplitude.
PRC neurons integrate time and item
information by modulating their stimulusselective response properties across
temporally distinct stimulus presentations.
Hypothesized information flow across MTL regions and TE.
Hypothesis: the incremental timing
signal in HPC is conveyed to PRC via
ERC, where it is integrated with item
information from TE and converted
into a discrete item-based temporal
order signal.
18
References
• Takeda, K. and S. Funahashi, Population vector
analysis of primate prefrontal activity during
spatial working memory. Cereb Cortex, 2004.
14(12): p. 1328-39
• “Neural ensemble”. From Wikipedia.
19