Transcript SFN06ARC

DISTINCT MODES OF VISUO-SPATIAL ATTENTION IN MACAQUE V1 AND V4
1,2
1
1
1
Jitendra Sharma* , Hiroki Sugihara , James Schummers , and Mriganka Sur
160.7
1Picower
Institute for Learning and Memory, Department of Brain and Cognitive Sciences, MIT, Cambridge, MA 02139
2Martinos Center for Biomedical Imaging, Massachusetts General Hospital, Charlestown, MA 02129
Introduction
Results
Background
1.
Visuo-spatial attention involves a two way process of enhancing perception of the attended stimulus while inhibiting
non-attended stimuli. Attentional modulations have been likened to contrast gain mechanisms that serve to bias
competitive interactions between multiple stimuli towards the attended stimulus. However, the locus and mechanism
underlying such biased processing remains unresolved. Furthermore, most previous studies on attention have shown
that there occurs a simple increase in neuronal firing rate to the attended stimulus. It however remains unclear if
particular characteristics of stimulus such as size, orientation, contrast or spatial frequency of engage greater or less
attention and whether such a change dynamically modify spatial and temporal dimensions of neuronal responses. In
this study we simultaneously recorded neuronal responses in V1 and V4 in awake behaving monkeys to a dynamically
changing stimulus pattern that overlapped receptive fields (RF) of single neurons in both areas. The monkeys were
trained in a covert spatial attention task where monkeys passively attended towards or away from the RF continuously
for the entire period of the trial. We used a reverse correlation paradigm to study the dynamics of attentional modulation
for different stimulus dimensions.
Attention modulates firing rate response to dynamic grating stimuli
V1 (n = 25)
V4 (n = 22)
Attention Towards
Firing Rate (Hz)
20
Methods:
Behavioral Task: Monkeys were trained to covertly attend to a small attention spot that could appear towards the RF
on the recorded neurons or away from it, in a contralateral location. The task sequence started with monkey holding a
lever that brought a red fixation spot (0.1 deg.). They were required to hold fixation within a small window such that any
jitter in fixation > 0.25 deg. aborted the trial. Once the monkey acquired stable fixation, an attention spot and an
equiluminant distracter were simultaneously presented towards the RF or in the contralateral hemifield. Monkeys were
required to hold fixation release the lever within 350-400ms of the disappearance of the attention spot. The attention
spot could appear towards the RF (attend Towards) or away from the RF (attend Away) in a random order. To prevent
monkeys from anticipating attention spot disappearance, the trials could last in time windows of 600ms, 1200 ms and
1800 ms, and within each window there was a further randomization of up to 500 ms. This made the task attentionally
demanding.
Attention Away
15
4
25
0.01
2
0.005
0
0
-2
V4 early
20
2
500
# Cells
1
3
1000
4
1500
1
2000
-250 0
2
500
Mean = -0.01
Mean = 0.05*
2
1000
3
4
1500
2000
Time (msec)
3
Mean = 0.06*
Mean = -0.01
10
4
4
1
Mean = 0.025**
2
Mean = 0.05**
3
Mean = 0.06**
4
Example V4 cells showing complex temporal dynamics, related to stimulus size. Both cells exhibit simple
dynamics in response to the small stimulus size (thick lines). In response to the large stimulus, however, there is
dominant suppression, which results in “Mexican Hat” shaped tuning curves.
5
2
0
-0.2
0
0.2
Attention cue
off
-0.2
0
0.2
-0.2
0
0.2
-0.2
0
0.2
0
-0.2
0
Modulation Index
t3= t2+500ms
(randomized)
0.2 -0.2
0
0.2 0.2
0
0.2 -0.2
0
0.2
Modulation Index
4.
Time course of suppression in V1 and V4
V4 (n = 33)
Stimulus on
Top: Population average PSTH representation of responses to the dynamic grating stimulus for V4 neurons (left) and V1 neurons
(right), under two behavioral conditions: attention towards the RF, or away from the RF. In both visual areas, attention to the RF
increases the firing rate, compared to the attention away condition. Bottom: Histograms of the Modulation Index (MI) of the firing
rate ([FRtow-FRawy]/[FRtow+FRawy]) for four different epochs during the stimulus presentation (250-750, 750-1350, 1350-1950
and 1950-1550 msec). During the later epochs, there is a strong trend for cells to shift to higher MI values, indicating an increase
in response for attend towards condition. (* = p<0.1; ** = p<0.05)
CRF
Attention Cue
Attention
spot
Distractor
t1= t+300ms
Fixation
Fix-spot
-3
8
8
V4 awy
x 10
-3
12
7
6
-3
V1 tow
x 10
12
10
10
8
8
6
6
4
4
2
2
0
0
6
V1 awy
x 10
5
5
4
4
3
3
2
2
1
1
0
Time = t
V1 (n = 37)
-3
V4 tow
x 10
7
Normalized
Probability
Stimulus
0.025
0.015
Mean = 0.06*
Stimulus
0.03
10
Time (msec)
t2= t1+ 600/1200/1800ms
(randomized)
0.035
0.02
Behavioral Control & Recording: The behavioral task and stimulus presentation was controlled by CORTEX (NIMH
freeware). Extracellular unit recordings were done in V1 and V4 from neurons having receptive fields centers within 6°
from the center of gaze. Before commencing the recording, RF size, location and other characteristics were carefully
mapped to optimize stimulus location. Spike sorting and post processing were done using custom software written in
Matlab.
400 ms
Attention Away
12
6
-250 0
Response
(Bar Release)
Attention Towards
V1 awy sm late
8
1
Reward
-3
x 10
30
10
6
3.
-1
0
20
40
60
80
100
120
140
Time (msec.)
160
180
200
0
0
20
40
60
80
100
120
140
160
180
200
-2
Time (msec.)
0
20
40
60
80
100
120
140
160
Time (msec.)
180
200
-2
0
20
40
60
80
100
120
140
160
180
200
Time (msec.)
Bar Press
2.
Population average dynamic tuning curves
RF
Attention Towards
Small
Attention Away
Large
Time
Stimuli: The stimulus consisted of patches of gray
sinusoidal gratings consisting of 8 orientations, two
different sizes (2° and 5° for V1 neurons and 4°
and 8° for V4 neurons) and two contrasts of 20%
and 60%. Spatial frequencies were optimized for V1
and V4 neurons (typically 3 cycles /deg and 0.5
cycles / deg. respectively). Each stimulus frame was
presented in rapid sequence at the monitor refresh
rate (75 Hz) and lasted from 1100-2300 ms. There
were a total of 360 conditions that were presented in
a pseudo-random order. Identical patterns of stimuli
were presented covering the RF and in the
contralateral location. The attention spot and
distracter locations were randomly chosen at two
points just outside the RF so as to obviate any
anticipation bias for a particular location around the
RF.
Data Analysis: Responses to individual frames in
the stimulus were assessed with the reversecorrelation technique. For each time delay (tau)
from 0-200msec, the probability distribution of
stimulus frames presented “tau” msec before a
spike was calculated. The response to the blank
stimulus frames was subtracted from the response
to each stimulus. Orientation tuning curves were
then calculated for each “tau” under four conditions
consisting of two attention states: towards or away
from the RF, and stimulus size: small or large.
V4 (n = 33)
Spike Train
V1 (n = 37)
Population average response dynamics. Each plot shows the average time course of response across cells, and
across stimulus orientation. V4 neurons show positive-going responses for small stimulus size, but the response
is almost entirely suppressive for large stimuli. In V1 neurons, the large stimuli result in larger, tuned responses.
Towards
t
Towards
Stimulus
-3
x 10
7
6
5
4
3
2
1
0
-1
-2
~13 msec
Summary
-3
x 10
10
8
6
4
2
1) Spatial attention to the receptive field location leads to increased coarse firing
rate in response to dynamic grating stimulus movies
0
-2
Away
Away
2) Time course of this effect includes both early and late components in V4
neurons, but only late component in V1 neurons
…
…
…
Small
Large
Small
Large
Population average tuning curve vs. time plots for V4 neurons (left) and V1 neurons (right). Each of the four
stimulus/behavioral conditions is plotted independently. The top row contains tuning curves for the attend toward condition, and
the bottom row for attend away. The stimulus size is indicated at the bottom of each column. V1 neurons are more sharply
tuned for the large stimulus where excitation is followed by a prolonged suppression, V4 neurons on the other hand show an
early short excitatory response followed by a sharp suppression by the large stimulus.
3) Temporal dynamics in V4 neurons depend on stimulus configuration;
surround stimuli largely suppress responses, and in some cases lead to
“mexican hat” tuning
4) Surround suppression in V1 is preceded by facilitatory responses.
5) This is the first report of reverse-correlation analysis of temporal dynamics
underlying behavioral effects