1990 - c d gilbert - theinfluenceofcontextualstimuliontheorientationsel[retrieved-2016!01!15]
TRANSCRIPT
-
7/25/2019 1990 - C D Gilbert - Theinfluenceofcontextualstimuliontheorientationsel[Retrieved-2016!01!15]
1/13
Vidm Ru. Vol. 30,No. Il. pp.
16894701, 1990
Printed
n Great Britain. All rights rcaerd
0042-6989/903.00 0.00
w Ymt0~~perpmon~pl c
THE INFLUENCE OF CONTEXTUAL STIMULI ON THE
ORIENTATION SELECTIVITY OF CELLS IN PRIMARY
VISUAL CORTEX OF THE CAT
CHARLES . GILBERTand TORSTENN. WIFSEL
The Rockefeller University, 1230 York Ave, New York, NY 10021, U.S.A.
(Received 10 Augwt 1989; in revi sed orm 10 January 1990)
Abatrae-Perception of a visual attribute, such as orientation, is strongly dependent on the context within
which a feature is presented, such as that seen in the tilt illusion. The possibility that the neurophysiological
basis for this phenomenon may be manifest at the level of cells in striate cortex is suggested by anatomical
and physiological observations of orientation dependent long range horizontal connections which relate
disparate points in the visual field. This study explores the dependency of the functional properties of single
cells on visual context. We observed sevtral influences of the visual field area surrounding cells receptive
5elds on the properties of the receptive 5eld ceoterz inhibition or facilitation dependent on the orientation
of the surround, shifts in orientation preference and changes in the bandwidth of orientation tuning. To
relate these changea to perceptual changes in orientation we modeled a neuronal ensemble encoding
orientation. Our results show that the filter characteristics of striate cortical cells are not nazua14y fixed.
but can be dynamic, changing according to context.
visual
cortex
Chientation selectivity Contextual stimuli
Horizontal connect.ions Tilt illusion
Neuronal ensembles
INTRODUCTION
A common theme to many areas of visual psycho-
physics is that perception of a feature in one
part of the visual field is influenced by the visual
context within which that feature is presented.
This has been an area where Gerald Westheimer
has made an important contribution. Together
with collaborators he has found that perception
of position, depth and orientation can be
altered by the presence of neighboring points or
contours (Westheimer, Shimamura 8t McKee,
1976; Westheimer & McKee, 1977; Butler 8c
Westheimer, 1978; Badcock & Westheimer,
1985; Westheimer, 1986). In the domain of
form, where line orientation plays a major part,
the tilt illusion provides dramatic evidence that
estimation of orientation at a given visual field
locus is dependent on information converging
from widely separated points in the visual field
(Gibson & Radner, 1937; for review see
Howard, 1986). In the current study we have
attempted to look for such influences at the
level of single cells in the cat striate cortex. A
dominant feature of the receptive fields of cells
in primary visual cortex is their orientation
selectivity (Hubel & Wiesel, 1959), so it is
natural to ask whether the influences on orien-
tation perception measured psychophysically
may be seen in striate cortex.
An important aspect of cortical circuitry that
may represent the underlying mechanism for
lateral interactions in the cortex and in the
visual field are the long range horizontal
connections (Gilbert & Wiesel, 1979, 1983;
Rockland & Lund, 1983; Martin & Whitteridge,
1984). Evidence from anatomical and cross-
correlation studies indicates that in the
superficial layers the horizontal connections can
mediate communication between cells with
nonoverlapping receptive fields and similar
orientation preference (Gilbert & Wiesel, 1979;
1989; Tso, Gilbert & Wiesel, 1986). Thus an
individual cell integrates information from a
larger part of the visual field than would be
indicated by the receptive field map. One should
keep in mind, however, that the very concept of
receptive field is stimulus dependent, and the
receptive field map obtained by using a simple
stimulus such as a single oriented bar of light
may be very different from that obtained by
using more complex stimuli. The idea that the
horizontal connections may mediate influences
that are orientation dependent is supported by
the finding that they relate cells with similar
orientation specificity.
1689
-
7/25/2019 1990 - C D Gilbert - Theinfluenceofcontextualstimuliontheorientationsel[Retrieved-2016!01!15]
2/13
1690
CHARLESD.
GILBERT and TOWN N WIESEL
The current study represents a preliminary
attempt to explore influences outside the recep-
tive field in the domain of orientation. The
results indicate that the functional properties of
individual cells do show contextual sensitivity,
and further suggest that the functional
specificity of a cell is to a degree not necessarily
fixed but can be dynamic, adapting to different
visual environments.
METHODS
We recorded from 65 cells in area 17 of
33 adult cats. The animals were initially
anesthetized with ketamine HCl (10 mg/kg,
i.m.) followed by sodium thiopental (initial dose
20 mg/kg, i.v., supplemented by further injec-
tions as needed). EKG, EEG (through silver
wires implanted between the skull and the dura),
rectal temperature and expired CO2 concen-
tration were constantly monitored. The animal
was intubated with an endotracheal tube, para-
lyzed with succinylcholine (10 mg/kg hr), and
artificially respirated. The stroke rate and the
volume of the respirator was adjusted to yield
4% end-tidal C02. Rectal temperature was
maintained near 38C with a thermostatically
controlled heating pad. During the experiment
sodium thiopental was given i.p. at a rate
suiiicient to produce a slow wave EEG pattern
(l-3 mg/kg hr). The pupils were dilated with 1%
atropine sulfate, and the nictitating membranes
were retracted with 10% phenylephrine. The
refraction of the eye was measured with a
retinoscope and appropriate contact lenses were
used to focus the eyes on a tangent screen 1.5 m
from the animal. The positions of the areae
centrales were back projected onto the screen
with the aid of a fundus camera.
A small hole was drilled in the skull above
the striate cortex, and the dura opened. Single
cell recordings were done with insulated
tungsten microelectrodes (Hubel, 1957) and
were restricted to the superficial layers of the
cortex, in order to sample a more uniform
population of cells.
When a cell was isolated, we mapped the
extent and orientation of the receptive field
using a hand held projector. The orientation
was determined quantitatively with a computer
generated (Adage raster graphics frame buffer
system) bar on a Tektronix 690SR monitor.
Having determined the receptive field character-
istics with a single oriented bar, we then placed
a number of bars surrounding the receptive field
in the arrangement shown in Fig. 1. Using this
arrangement, we could determine the influence
of the surround bars on the response charac-
teristics of the cell. The surround bars did not
activate the cell when presented in isolation, but
they often did modulate the response of the cell
to a bar presented within the receptive field (the
center bar). We determined the dependency
of this modulation on the orientation of the
surround bars and the effect of surround bars
of a given orientation on the tuning of the cell
to the orientation of the center bar. The
cells studied had receptive field located within
5-6 deg of the area centralis, with field size
ranging from 1 to 2 deg in diameter. The length
of the stimulating bars was chosen to approxi-
mate the length of the receptive field of the cell
under study. In most experiments the center and
surround lines were moved in tandem (same
Fig. 1. Stimulus configuration. The dotted line represents the extent of the receptive field center. The lines
in the surround are placed to prevent any one of them from entering the receptive field and activating
the cell. Lines can be placed in the center, surround or both, and moved in tandem to stimulate the cell.
In some instances the center bar is moved and the surround bars are kept stationary.
-
7/25/2019 1990 - C D Gilbert - Theinfluenceofcontextualstimuliontheorientationsel[Retrieved-2016!01!15]
3/13
influence of contextual stimuli
i6 x
velocity and period), though occasionally the
surround lines were kept stationary while the
center line was allowed to move. The separation
between the center line and the innermost sur-
round lines was adjusted so that the surround
lines would not enter the receptive field at any
of the orientations used. The stimulus con-
figuration chosen allowed us to test separately
the effect of simulating outside the receptive
field along the movement axis vs along the
orientation axis.
Stimuli were presented in blockwise random
fashion. Usually we presented any given stimu-
lus condition 10 times when making a given
tuning curve. When making o~e~tatio~ tuning
curves for a celi, each block of conditions was
used to obtain an estimate of the orientation
optimum by fitting a spline through the points,
and the set of 10 blocks was used to obtain an
estimate of the mean peak position and stan-
dard deviation of the peak position. The statisti-
cal significance of a shift in orientation could
then be determined by using a standard r-test.
For the surround tuning curves, the data were
fit by a polynomial regression, which provided
an estimate of the sizes and positions of the
maxima and minima in the curves.
RESULTS
We first studied single cells in oat striate
cortex for the effect of the surround stimuli on
the responses of cells s~rnula~ with an opti-
mally oriented bar moving within the receptive
field center. Individual cells were only activated
when the line was within the receptive field.
Though the lines surrounding the receptive field
could not by themselves activate the cell, they
altered the response of the ceil when they were
of a certain orientation and position, The effect
of the surround lines on the center response
varied from cell to ce& some cells were inhibited
and some were facilitated by the presence of the
surround lines. The inhibition or fa~li~tion
was dependent on the orientation of the
surround lines, some cells showing maximal
response with surround lines matching the
optimal orientation of the cell, some when the
surround lines were o~hogonal to the cells
optimal orientation and others for orientations
in between. All of the
ells
in our sample were
located in the superficial layers of the cats
primary visual cortex.
Figure 2 illustrates the diversity of orientation
tuning seen for the lines in tbe surround
VR
3X11 K
spi kes/
Sweep
spi kes/
sxeep
70 110 150 10 50
orwntatron
16
o--
100
140 0
40 80
orientahon
Fig.
2. Comparison of center and surround tuning curves for
three superfcial layer complex ceils. The anter tuning
curves are reprcseutcd by the small open circles and the
surrwnd curves by the large so&d triangles. The surround
t~g~~~~~p~
of a center bar at
the optimal urierttatiun, and the level of Wng of the c&l
without the surround is indicated by the horizontal
dotted line. The suxwtmd tmdng curvea are generated by
a polynomial fit, and the symbols indicate the mean value
at each orientation. I$@op part of the figure shows a
call with an optimal oriantation of 8odcg. The surround
tuning curw p#lltJ at 9Ode& and is facilitatory at the
peak, and the minimmu is roughly at the orthogonal
orientation. The oall in the center of the figure has an
optimal orientation of 16Odeg; the psak in the surround
tuning is at 20 deg and the minimum around 60 deg. The ~11
at the buttom has an optimal orientation of fOdeg and the
surrmind tuning curw has a minim- at ~r~~~~ly the
same OtititiQn.
-
7/25/2019 1990 - C D Gilbert - Theinfluenceofcontextualstimuliontheorientationsel[Retrieved-2016!01!15]
4/13
1692
~ZHAIUBD. GILBERTnd TORSTEN . WIESEL
for superficial layer cells in cat area 17. The
response of a complex cell with optimum orien-
tation of 80 deg is shown in the top of Fig. 2.
The orientation tuning curve of this cell is quite
broad, having a half bandwidth (full width at
half height) of 60 deg. When stimulating the
center of the receptive field with a bar of the
optimal orientation, and changing the orien-
tation of the surrounding bars, the response of
the cell was most facilitated when the surround
bars were at orientation 90 deg, 10 deg from the
optimum. The cell was most inhibited by sur-
round bars at 10 deg, 70 deg from the optimum.
The facilitated response was roughly 20%
greater than the response to the center bar alone
(represented by the dotted line), and the most
inhibited response was 65% reduced from the
response to the single center bar. Other cells
were similar in the relationship of the center and
surround orientation tuning curves, but the
degree of inhibition or facilitation differed.
-100
I
0
20
40
60 a0
orientation
An example of a different relationship be-
tween the orientation tuning of the center and
surround is shown in Fig. 2 (middle, bottom).
The cell in the middle part of the figure was
optimally oriented at 160 deg. The surround
tuning curve was most inhibitory at 60-70 deg
(80-90 deg away from the optimal orientation)
and least inhibitory at 20 deg, 40 deg from the
optimum. For the orientations tested, the cell in
Fig. 2 (bottom) was strongly inhibited by a
surround oriented within 10 deg of the optimum
center orientation of the cell, the reduction in
response being 40% of the center only response.
Fig. 3. Plot of the maxima and minima of the surround
tuning of a sample of 13 cells. The maxima (12) are
rqrcscnted by the open circles and the minima (13) by the
solid triangles. The abscissa shows the absolute vale of the
orientation difference between the optimum orientation of
the center and the peak or valley in the surround tuning
curve. The amount of facilitation or inhibition is indicated
on the ordinate. Most of the &ects were inhibitory. Though
the peaks and valleys covered a range of relative orien-
tations, the maxima tended to be found near the orientation
prefmncc of the cell (0 deg orientation difference) and the
minima congregated towards the orthogonal orientation.
Other cells showed a broadly tuned surround, or
were uniformly inhibitory at all orientations.
We also observed, however, that the signs and
magnitudes of the surround effects were not
hxed for a given cell but could be influenced
by changing the brightness or contrast of the
surround lines.
The relationship between the orientation pro-
To explore other effects of the surround that
ducing maximal excitation or inhibition in the
may pertain to the tilt illusion, we focused on
surround and the orientation optimum of the surround effects for surround orientations near
center is shown for 13 cells in Fig. 3. We
the optimum of the cell (f40 deg), since in
included in the sample only cells for which we humans the most effective orientations for pro-
determined orientation tuning to the surround
ducing tilt illusions are those with an orien-
across the full range of orientations, and which tation contrast*
of 20-30 deg (Westheimer,
had a statistically significant peak and/or valley 1990). Surprisingly, the effect of the surround
in the surround tuning curve. Usually, the was not restricted to simple inhibition or facili-
surround inhibition was not complete, and tation, in that surround lines of the appropriate
the orientation tuning curve of the surround
orientation could in fact change the orientation
was broader than the orientation tuning of preference of a cell. In these experiments we
the center bar response. The positions of the fixed the orientation of the surround and deter-
minima and maxima of the surround tuning mined the orientation tuning to the center bar
curves varied from cell to cell, but there was a when displayed concurrently with the surround
tendency for the maxima to aggregate towards and with all lines moved in tandem. This was
the optimal orientation of the cell, and for the repeated for surround lines of different orien-
minima to aggregate towards the orthogonal tations. For the cell illustrated in Fig. 4 the
orientation. Furthermore, surround effects
optimum orientation without surround lines
that were largely inhibitory were much more
was 30deg. However, when the surround lines
commonly seen than facilitatory surrounds. were oriented at 0 deg (30 deg away from the
-
7/25/2019 1990 - C D Gilbert - Theinfluenceofcontextualstimuliontheorientationsel[Retrieved-2016!01!15]
5/13
Influence of contextual stimuli
1693
central Gtr angle O)
Fig. 4. Orientation tuning curves for a superficial layer
complex cell without surround (bold line) and with sur-
rounds of different orientations. The peak orientation is at
30 deg, but in the presence of a surround of 0 deg the peak
shifts away from the surround orientation, to 40deg. In
addition to the peak shift seen for a 0 deg surround, there
is varying degrees of inhibition for surrounds of other
orientations. Note also the broadening of the tuning for the
Odcg surround condition.
optimal orientation of the cell, which is near
the most inhibitory surround orientations), the
tuning curve peaked at 40 deg, representing a
shift in the peak position of 10 deg. Note that in
this instance the effect of the surround was
repulsive, with the tuning curve shifting in
a direction away from the orientation of the
surround lines. The shift was reversible and
repeatable, such that surround lines of any other
orientation produced tuning curves peaking
again at 30 deg, and each time the surround
orientation was put at Odeg, the peak shifted
back to 40 deg (as seen by the three curves on
the right in Fig. 4). The effect was statistically
significant, with
P