Measurement of suprathreshold binocular interactions in amblyopia
B. Mansouri, B. Thompson, R.F. Hess*
McGill Vision Research, Department of Ophthalmology, McGill University, 687 Pine Ave W (H4-14), Montreal, Que., Canada PQ H3A 1A1
a r t i c l ei n f o
Received 31 March 2008
Received in revised form 13 August 2008
Treatment of amblyopia
Global motion perception
Global form perception
a b s t r a c t
It has been established that in amblyopia, information from the amblyopic eye (AME) is not combined
with that from the fellow fixing eye (FFE) under conditions of binocular viewing. However, recent evi-
dence suggests that mechanisms that combine information between the eyes are intact in amblyopia.
The lack of binocular function is most likely due to the imbalanced inputs from the two eyes under bin-
ocular conditions [Baker, D. H., Meese, T. S., Mansouri, B., & Hess, R. F. (2007b). Binocular summation of
contrast remains intact in strabismic amblyopia. Investigative Ophthalmology & Visual Science, 48(11),
5332–5338]. We have measured the extent to which the information presented to each eye needs to dif-
fer for binocular combination to occur and in doing so we quantify the influence of interocular suppres-
sion. We quantify these suppressive effects for suprathreshold processing of global stimuli for both
motion and spatial tasks. The results confirm the general importance of these suppressive effects in ren-
dering the structurally binocular visual system of a strabismic amblyope, functionally monocular.
? 2008 Published by Elsevier Ltd.
1. General introduction
Our understanding of the detailed architecture of binocular
interaction in normal vision is evolving. Legge’s early work on con-
trast processing provided one of the benchmark models (Legge,
1984), involving purely excitatory pathways, in which each eye’s
contrast information was subjected to a non-linear transduction
stage followed by binocular combination. This is illustrated in
Fig. 1A. In a recent series of experiments an updated model of bin-
ocular interactions in normal observers has been proposed from
experiments using parallel (Baker & Meese, 2007; Baker, Meese,
& Georgeson, 2007a; Baker, Meese, & Summers, 2007c) and
cross-oriented gratings (Baker et al., 2007c) presented to the same
or different eyes. From these studies a two-stage model of contrast
gain control (Meese, Georgeson, & Baker, 2006) has emerged,
involving both excitatory and inhibitory pathways, where the first
stage is placed before binocular summation of signals but receives
suppressive input from the other eye (Baker & Meese, 2007; Baker
et al., 2007c).This model, depicted in Fig. 1B, provides a good ac-
count of a wide variety of phenomena (Baker & Meese, 2007, for
a brief review), including contrast summation, detection and dis-
crimination (Meese et al., 2006) and contrast-matching (Baker
et al., 2007a). The interocular suppressive effects are of a more glo-
bal nature involving non-corresponding regions in the two eye
(Meese & Hess, 2004) and involving global motion (Hess, Hutchin-
son, Ledgeway, & Mansouri, 2007) and form judgments (Mansouri,
Hess, Allen, & Dakin, 2005) as well as contrast detection/discrimi-
nation. While the sites of these two posited stages of processing,
one excitatory and the other inhibitory is not known with cer-
tainty, there is electrophysiological support for the binocular excit-
atory site to be located in layer 4 of area V1 where afferents from
the two eyes are first combined (Hubel & Weisel, 1968) and there is
support for the dichoptic inhibitory stage to be located within the
lateral geniculate nucleus where binocular inhibitory interactions
have been documented (Pape & Eysel, 1986; Sanderson, Bishop,
& Darian-Smith, 1971; Xue, Ramoa, Carney, & Freeman, 1987).
The vast majority of strabismic amblyopes lack binocular func-
tion (McKee, Levi, & Movshon, 2003) and only partially recover it
even after restoration of monocular function by patching and other
therapies (Mitchell, Howell, & Keith, 1983). In a recent study of
monocular, binocular and dichoptic contrast masking in amblyopia
(Baker, Meese, & Hess, 2008) a working model of monocular and
binocular processing in strabismic amblyopia has been proposed.
A schematic is displayed in Fig. 1C. It differs in three important
ways from models dealing with normal vision. Firstly the amblyo-
pic eye is subjected to a stage of signal attenuation. Second, the
amblyopic eye has additional multiplicative noise (Gr) that is ap-
plied to the saturation constant of the contrast gain control located
prior to binocular summation. Finally, the contrast gain control
prior to binocular summation receives suppressive interocular in-
puts with greater weights for the FFE. Two aspects of this model
are noteworthy. First notice that the binocular summation stage
(P) is intact. This is because it has been shown that binocular sum-
ocular inputs to the summation stage are balanced (Baker, Meese,
mation at threshold can occur in strabismic amblyopes if the inter-
0042-6989/$ - see front matter ? 2008 Published by Elsevier Ltd.
* Corresponding author. Fax: +1 514 8431691.
E-mail address: email@example.com (R.F. Hess).
Vision Research 48 (2008) 2775–2784
Contents lists available at ScienceDirect
journal homepage: www.elsevier.com/locate/visres
Mansouri, & Hess, 2007b). Furthermore, recent contrast masking
data (Baker et al., 2008) demonstrates that dichoptic facilitation
is present for several strabismic amblyopes and in some cases is
quite distinct when testing the amblyopic eye. This is pertinent be-
cause dichoptic facilitation is a consequence of excitatory binocu-
lar summation (Baker & Meese, 2007; Meese et al., 2006) and
therefore also suggests that summation remains intact in the
amblyopic visual system. Secondly notice that there are separate
monocular and binocular deficits such that one would not expect
normal binocular interactions to occur automatically by using
stimuli whose performance has been equated monocularly.
These results for contrast processing suggest that one impedi-
ment to binocular combination of information in amblyopia, par-
ticularly strabismic amblyopia, is the imbalanced suppressive
drive prior to binocular combination, not a lack of binocular sum-
mation per se. That suppression is an important contributor to the
lack of binocular function in strabismic amblyopia is not new to
the clinician (Travers, 1938) and the more recent finding of normal
binocular combination in strabismic amblyopia (Baker et al.,
2007b, 2008) suggests that it may lie at the heart of the problem
and may be quite separate from the monocular loss of function
(Holopigian, Blake, & Greenwald, 1988) as the contrast model de-
picted inFig. 1C suggests. Since our knowledge of contrast process-
ing is relatively well-developed, this model serves as a useful
starting point for thinking about the combination of information
for other tasks, for example, global motion and global form since
there is evidence that each has a separate binocular and interocular
suppressive stage similar to that already described for contrast
detection (Hess et al., 2007; Mansouri et al., 2005). One important
difference however, is that contrast detection involves the use of
threshold stimuli whereas theses global tasks are suprathreshold
in nature. It should be stressed that the usefulness of these model
descriptions is purely to illustrate certain important principles of
operation that will have general relevance to the later experiments
rather than to develop specific testable predictions.
In the present study we set out to assess the imbalanced sup-
pressive drive that is the defining clinical feature of strabismic vi-
sion and which we believe underlies amblyopes’ inability to
combine information binocularly. We used suprathreshold stimuli
of a more global nature involving both motion and form discrimi-
nations to provide a thorough assessment of dorsal and ventral ex-
tra-striate cortical function. The aim of the study was to assess,
using both motion and form-based tasks, the relative strengths of
the suppressive drive depicted in Fig. 1C and to see if conditions
could be obtained where binocular combination occurs.
dealt with the relative suppressive influences involved in motion
processing and the second with form processing. Despite the use of
cation of interocular suppression using a common approach where-
same ‘coherent’ direction (either up or down) and the noise was a
population of dots that moved in random directions. The task was
to indicate the coherent direction of motion and thresholds could
be measured by varying the relative proportion of signal and noise
dots. For the form task, the signal population was an array of Gabors
each of which was oriented according to a specific distribution, the
ulation was an additional array of Gabors the orientations of which
orientation of the stimulus was to the left or the right of vertical.
Thresholds for this task, for a set signal distribution, could therefore
varying the proportionof signal and noiseGabors. For both tasks we
used dichoptic presentation to present the signal and noise popula-
tions to different eyes. The rationale was that if the amblyopic eye
was presented with the signal, performance on either task should
be at chance if the eye is completely suppressed and unable to con-
tribute to binocular vision. Conversely if the fellow eye received the
be visible and the task would accordingly be trivial. However, if
provide a useful metric of the relative contribution of the different
dot populations presented to either eye and hence quantify the de-
gree of suppression.
The separation of signal and noise between the two eyes had an-
other useful property, namely the fact that the relative contrast
and/or number of samples presented to each eye could be indepen-
dently varied. Therefore, we could test the idea that although under
normal viewing conditions, binocular combination does not occur,
if the inputs to the two eyes are suitably imbalanced and biased to-
wards the AMEs (e.g. more contrast to the amblyopic eye and less to
the fellow eye) then binocular combination may be possible. This is
indeed what we found for both the motion and the form-basedtask.
(or number of samples), performance was largely dependent on
what is presented to the FFEs suggesting a strong suppressive influ-
ence from the fixing eye resulting in little or no binocular combina-
Fig. 1. Schematic illustration of the architecture of a currently proposed models to
account for binocular interactions in normals and strabismic amblyopes (Baker
et al., 2008). In (A), the original model of Legge (1984) involving just summation. In
(B), a more recent two-stage model involving both binocular combination and
balanced interocular suppression (Meese et al., 2006). In (C), Like that of normal
vision (B), the model of strabismic amblyopia (Baker et al., submitted for
publication) has two stages of gain control, one before and one after binocular
combination. The modifications for the amblyopic visual system include a signal
attenuator, the injection of stochastic noise (IGI) and imbalanced interocular
suppressive signals prior to binocular combination Abbreviations: p, q and m are
excitatory exponents; S and Z are semi-saturation constants. Green lines indicate
excitation, red lines suppression and arrows indicate divisive input. Gk indicates a
Gaussian noise generator. (For interpretation of the references to colour in this
figure legend, the reader is referred to the web version of this article.)
B. Mansouri et al./Vision Research 48 (2008) 2775–2784
tion. However,when the relative informationbetween the eyes was
suitably adjusted for contrast (or number of samples) binocular
combination occurred and performance could be measured. Impor-
tantly, the relative strengths of the dichoptic signals that result in
combination (the ‘balance point’) could not be predicted by purely
monocular differences in sensitivity and therefore reflect the rela-
tive weightings of interocular suppressive mechanisms.
2. General methods
Eleven amblyopic and eight normal observers participated in
the two Experiments (seven amblyopic subjects completed each
experiment). Refraction in all observers was tested and corrected
to best visual acuity. The ‘Declaration of Helsinki’ was followed
and informed consent was obtained from all observers before data
collection (see Table 1).
2.2. Eye dominance
Eye dominance for normal subjects was assessed using a sight-
ing test. Subjects were asked to look through a sighting tube at a
distant fixation point and the eye that was aligned was noted
(Rosenbach, 1903). Six normal subjects were right eye dominant,
two were left eye dominant. Stereo acuity was measured using
the Randot test.
3. Experiment 1: Motion
To study the ability of amblyopic observers to binocularly com-
bine motion information we used random dot kinematograms
(RDKs) and a coherence motion discrimination task. These stimuli
are constructed of two populations of moving dots. The ‘signal’
population all move in the same direction termed the ‘coherent’
direction. Conversely, the ‘noise’ population has no common mo-
tion direction as all the dots move in random directions. The ratio
of signal to noise dots required to determine the coherent motion
direction is called the motion coherence threshold. The measure-
ment of motion coherence thresholds is a well-studied paradigm
with regard to global motion integration (Newsome & Pare,
1988). Therefore, by using these stimuli with signal and noise sep-
arated dichoptically, one can assess the degree to which underlying
mechanisms combine information from two eyes.
We have previously used this approach to study binocular inter-
actions in normals (Hess et al., 2007) and found slight imbalances
in favour of the dominant eye. The effectiveness of the noise was
slightly greater when seen through the dominant eye. We reasoned
that if signal dots were presented to the amblyopic eye (AME) and
noise to the fellow fixing eye (FFE), then the ability to perceive the
coherent motion direction would only be possible if the AME was
able to overcome the suppression of the FFE. In addition we could
ensure that the two eyes were functioning binocularly by measur-
ing motion coherence thresholds, a measurement that reflects the
combination of signal and noise populations.
Experiment 1 contained two different viewing conditions, a
monocular condition where both signal and noise were presented
to one eye at a time and a dichoptic condition where signal and
noise were split between the eyes. The monocular condition was
included to allow for a comparison between the relative contrasts
required to match performance between the eyes on the monocu-
lar task with the relative contrasts between the eyes required to
achieve binocular combination on the dichoptic task. This was
important since the model of binocular combination in strabismic
amblyopia described above would not predict a direct match be-
tween the two conditions since according to the model monocular
performance does not involve interocular suppression whereas the
dichoptic performance measure does. The results of the monocular
presentation are reported in Experiment 1A and the dichoptic pre-
sentation results are presented in Experiment 1B.
Clinical details of the amblyopic observers participating in the experiment
Obs Age/SexType RefractionLA Squint History, stereo
AS 21/F RE
Detected age 4 years, patching at 4 years for 6 months, surgery at 7 years, no stereopsis
Detected age 6 years, no patching, no surgery, no stereopsis
Detected age 4 years, no patching, no surgery, no stereopsis
Detected age 6 years, patching for 1 years, local stereovision 70 s of arc
Detected age 7 years, patching for 1–2 years, No surgery, no stereopsis
Detected at 11 years, no surgery & patching, eye exercise 1–2 years, glasses since 12 years, no
+5.00?2.00 ? 120?
+3.50?1.00 ? 75?
Detected age 5 years, patching for 3 months, no glasses tolerated, 2 strabismus surgery RE age
10–12 years, no stereopsis
Detected age 5 years, patching for 3 years, no surgery, local stereopsis 200 seconds of arc
ML 20/FRE mixed
+1.0?0.75 ? 90?
Detected age 5 years, patching for 2 years, no stereopsis
Detected age 4 years, patching for 6 months, Surgery age 5 years, no stereopsis
Detected age 6 years, glasses since 6 years, no other therapy, near normal local stereopsis
The following abbreviations have been used; strab for strabismus, aniso for anisometrope, RE for right eye, LE for left eye, ET for esotropia, ortho for orthotropic alignment, sph
for dioptre sphere.
B. Mansouri et al./Vision Research 48 (2008) 2775–2784
We used a Macintosh G4 desktop to generate stimuli. The stim-
uli were presented on a gamma-corrected Sony professional Series
P22f monitor. The refresh rate was 75 Hz and the mean luminance
of the display was 50 cd/m2. Dichoptic presentation of stimuli was
achieved by using a Wheatstone Stereoscope with an effective
viewing distance of 114 cm.
Global motion stimuli were translational RDKs presented with-
in two horizontally separated, circular display windows, each equi-
distant from the centre of the screen. Each circular window
subtended 7? and to aid binocular fusion, each display region
was surrounded by a binocular rectangular frame. Dots were pre-
sented on a homogenous mid-grey background (mean luminance
of 50 cd/m2) that filled the entire circular display window. The
luminance modulation (Michelson contrast) and hence the visibil-
ity of the dots could be varied by increasing the luminance of the
dots, with respect to the background, according to the following
Dot luminance modulation ¼ ðLdots?LbackgroundÞ=ðLdotsþLbackgroundÞ;
where Ldotsand Lbackgroundare the dot and background luminance,
respectively. The luminance modulation of the dots ranged from
0.004 to 0.33.
Each RDK was generated anew immediately prior to its presen-
tation and was composed of a sequence of 8 frames, which when
presented consecutively produced continuous apparent motion.
The duration of each frame was 53.3 ms, resulting in total stimulus
duration of 426.7 ms. Each image contained 100 non-overlapping
dots (dot density 0.88 dots/?2) and the diameter of each dot was
0.235?. At the beginning of each motion sequence, the position of
each dot was randomly assigned. On subsequent frames, each dot
was shifted by 0.3?, resulting in a drift speed, if sustained, of 5.9?/
s. On each displacement a dot was randomly assigned to be signal
or noise and at low coherences had a limited directional lifetime.
Importantly, there was no difference in the speed of signal and
noiseelements, onlyintheirdirection.Whena dotreachedthe edge
of the circular display window it was immediately re-plotted in a
random spatial position within the confines of the window.
The global motion coherence level of the stimulus was manipu-
lated by constraining a fixed proportion of ‘signal’ dots on each im-
age update to move coherently along a translational trajectory and
the remaining (‘noise’ dots) to move in random directions. Global
motion thresholds were measured using a single-interval, forced-
choice direction-discrimination procedure. On each trial, observers
were presented with an RDK stimulus in which the signal dots
moved along an upward or downward trajectory. The observers’
task was to identify whether the motion was upwards or down-
wards. Data collection was carried out using an adaptive staircase
procedure (Edwards & Badcock, 1995). The staircase varied the
proportion of signal dots present on each trial, according to the ob-
server’s recent response history. The staircase terminated after
eight reversals and thresholds (79% correct performance) were ta-
ken as the mean of the last six reversals. Each threshold reported
was based on the mean of at least five staircases.
3.2.4. Experiment 1A, monocular presentation condition
In Experiment 1A both signal and noise were presented to one
eye and mean luminance to the other eye (see Fig. 2A). The obser-
ver was not aware of which eye was seeing the stimulus.
3.2.5. Experiment 1B, dichoptic presentation
Experiment 1B was identical to Experiment 1A with the excep-
tion that presentation was now dichoptic whereby both eyes
viewed a part of the stimulus, either the signal dots or the noise
dots, i.e. signal was presented to one eye and noise to the other
(see Fig. 2B). Since we varied the contrast of the signal and noise
independently, we were able to present stimuli with high contrast
to the AME and low contrast to the FFE.
3.3. Results and discussion
Fig. 3 represents the average coherence threshold data for mon-
ocular (Fig. 2A) and dichoptic (Fig. 2B) conditions. In the monocular
condition (A) AMEs showed higher thresholds than all other eyes
tested. FFEs also had in some cases slightly higher thresholds than
controls, but this effect diminished at medium suprathreshold con-
trasts (e.g. 5–8%). In the dichoptic condition (B) AMEs again
showed significantly higher thresholds at all contrasts tested, an
effect that was more pronounced than for the monocular testing.
The normal eye average thresholds fall between those of the
AME and FFE at the higher contrasts, suggesting that the AME
not only suffers from FFE suppression, but also that the FFE bene-
fits from this phenomenon.
Dichoptic results for one representative amblyope (ML) are dis-
played in Fig. 4. In A, the Y-axis represents the ratio of the AME to
FFE coherence thresholds and the X-axis represents the contrasts
of the stimuli that were presented to the AME. The corresponding
contrast of the stimuli presented to the FFE is presented as different
curves (filled circle for 2.34%, open circle for 3.13% filled square for
gle for 6.25%). The dotted line represents a ratio of 1 where the
thresholds in both eyes are the same. The threshold ratio is seen to
Fig. 2. Schematic presentation of the random dot kinematogram is shown for
monocular (A) and dichoptic (B) conditions. Black arrows indicate the motion
direction of the signal dots which moved in the same ‘coherent’ direction (up vs.
down) within a trial. White arrows represent the motion directions of the noise dots
which were moved in random directions. In the monocular condition, signal and
noise dots were presented to one eye at a time (A). In the dichoptic condition, signal
and noise dots were presented to different eyes within each trial.
B. Mansouri et al./Vision Research 48 (2008) 2775–2784
reduce as the contrast in the AME is increased. The AME contrast at
which this occurs is a function of the contrast shown to the FFE; the
lower the contrast shown to the FFE, the lowerthe contrast needsto
be in the AME before thresholds are comparable in the two eyes un-
der dichoptic viewing. The numbers attached to different points on
these curves in Fig. 4A give the ratio of the contrasts shown to the
FFE and AME. This is summarized in Fig. 4B where the dichoptic
threshold ratio is seen to simply depend on the interocular contrast
ratio; as the interocular contrast of stimuli is increased, dichoptic
performance for the AME and FFE comes together, suggesting that
efficiency irrespective of what eye sees the signal and what eye sees
the noise, similar to that of a normal binocular observer.
Fig. 5 shows representative dichoptic results for six amblyopes
plotted in the same way as described for Fig. 4B. There is in all cases
a relative contrast of signal and noise where dichoptic performance
for AME and FFE is matched (i.e. similar effectiveness of signal and
noise independent of which eye is viewing the signal). The contrast
across subjects. These results suggests that the interocular contrast
can be adjusted (i.e. reduced in the fixing eye) to the point where
the information, be it signal or noise, seen by the two eyes is com-
to accomplish this global motion direction task. Importantly, the
interocular contrast ratio to achieve this balance in dichoptic perfor-
manceisnotpredictable(correlationcoefficient = 0.65:p = .13;n = 6)
from the monocular contrast threshold difference to Gabors whose
spatial frequency (5c/d) matches that of the fundamental frequency
(diameter 0.235?, approx 4.2c/d) of the elements in the motion
4. Experiment 2: Form
In Experiment 2 we applied the same principle of (a) signal/
noise binocular integration and (b) manipulating the relative infor-
mation content of the two eye’s images (i.e. number of samples and
contrast) for dichoptically presented form (e.g. orientation) stimuli.
We used a global mean orientation discrimination task where an
array of oriented Gabors was presented to observers and they were
asked to judge whether the mean orientation was tilted to the left
or right of vertical (Dakin, 2001). The orientations of the signal Ga-
bors were randomly selected from a predetermined population
with a specific mean and variance. The orientations of the noise
Gabors were selected from a flat distribution. Similar to Experi-
ment 1, we reasoned that by separating the signal and noise pop-
ulations and presenting them to separate eyes, we could measure
the contribution of each eye to the final percept, i.e. all signal, all
noise or a combination of each. We could therefore objectively
Monocular global motion direction
Coherence threshold (%)
Dichoptic global motion direction
Coherence threshold (%)
Fig. 3. Average contrast sensitivity threshold data for motion direction (Exp. 1) for 7 amblyopic and 8 normal subjects is shown for amblyopic subjects’ AME (solid line and
filled squares) and FFEs (dashed line and open squares) and for normal subjects’ non-dominant (solid line and filled circles) and dominant (dashed line and open circles) eyes
for monocular (A) and binocular (B) conditions. The Y-axis represents the coherence threshold (%) in linear scale. The X-axis represents the contrast in logarithmic scale.
Fig. 4. In (A), dichoptic coherence threshold data for different combinations of contrasts to amblyopic and FFEs is represented for one amblyopic subject (ML). The Y-axis
represents the ratio of the AME to FFE coherence threshold. The X-axis represents the contrast of the stimuli which were presented to the AME. The corresponding contrast of
the stimuli presented to the FFE is presented as different curves (open circle for 2.34%, filled square for 3.13% open square for 3.91%, filled triangle (with line) for 4.69%, open
diamond for 5.47% and filled triangle (without line) for 6.25%. The dotted line represents a ratio of 1 where the thresholds in both eyes are the same. The numbers attached to
different points on these curves in (A) give the ratio of the contrasts shown to the FFE and AME. In (B), the data in (A) is re-plotted in terms of the ratio of the dichoptic
contrast to show that as this varies so too does the balance of dichoptic sensitivity.
B. Mansouri et al./Vision Research 48 (2008) 2775–2784
measure the contribution of either eye to performance, based on
which eye viewed signal and which eye viewed noise.
The first stage of the experiment was to equate monocular
thresholds for both the FFE and the AME for discriminating the ori-
entation of a single element by reducing the contrast of the ele-
ment when viewed by the FFE. We then used a dichoptic
presentation of stimuli of multiple elements with the contrast to
each eye matched for performance based on the monocular single
element threshold measurements. We found that although both
eyes could perform similarly when the contrast of the stimuli
was manipulated under monocular conditions, when the same
monocular performance-equated contrast stimuli were presented
dichoptically, the AME was suppressed. However, when weaker
stimuli (i.e. less samples or less contrast) were presented to the
FFE under dichoptic viewing conditions, the AME started to con-
A Power Macintosh G3 computer was used to generate and dis-
play the stimuli. Stimulus presentation was controlled by the Mat-
lab environment (MathWorks Ltd) and Psychophysics ToolBox
(Brainard, 1997). All stimuli were displayed on a 20-in. Sony Trin-
itron GDM-F520 monitor. The monitor was calibrated and linear-
ized using a Graseby S370 photometer and the Video Toolbox
(Pelli, 1997) package. Pseudo 12 bit contrast accuracy was
achieved by using a video attenuator that combined the RBG out-
puts of the graphic card (ATI Rage 128) into the green (G) gun.
The refresh rate was 75 Hz. The mean luminance of the screens
was 28 cd/m2. The resolution was 1152 ? 870 pixels. One pixel
on the screen was 0.32 mm, which was 2.12 arcmin of the observ-
ers’ visual angle from the viewing distance of 52 cm.
mirror stereoscope. The images were 6? ? 6? wide and arranged on
the screen centrally and adjacent to each other. Stimuli were arrays
of Gabor micro-patterns presented on a mean luminance back-
ground. The envelope of each Gabor had a standard deviation of
0.4?. The spatial frequency of sinusoidal modulation within the Ga-
bors was 0.52 cycles per degree (cpd). Typically, 16 Gabors were
cular area inside the box outline, centred on the centre of the box.
When the patches overlapped (as would occasionally occur), their
grey levels were added, if this led to brightness levels outside the
possible luminance range (only occurred rarely), they were clipped
appropriately at the maximum or minimum contrast values.
The orientation of each Gabor was controlled by the standard
deviation of its parent distribution. Two types of parent distribu-
tions were used, producing two different Gabor populations:
‘noise’ and ‘signal’. The orientation of each Gabor micro-pattern
in the signal population was selected from a Gaussian distribution
with a mean equal to the orientation cue (i.e. 90±? the cue gener-
ated by APE, an adaptive method of constant stimuli (Watt & An-
drews, 1981) and a variable bandwidth. The distribution’s
standard deviation, rext, was varied from 0± (all elements aligned)
to 28± (high orientation variability). The orientations of Gabors in
the noise population were selected from a Gaussian distribution
with a standard deviation of 90±. We used the same method to
generate the parent distribution of the noise Gabors as we used
to generate the parent distribution of the signal array. This meant
that the noise population distributions had a randomly selected
(on each trial) mean orientation, however, given the breadth of
the distribution this was not discernable. Note also that since ori-
entation is a circular variable (i.e. 0? and 180? wrap around), our
noise populations were equivalent to uniform distributions be-
tween 0? and 180?.
A single temporal interval two alternative forced choice para-
digm was used. The observers’ task was to judge whether the mean
orientation of the array of Gabors was rotated clockwise or coun-
ter-clockwise (tilted to right or left of vertical—seeFig. 6). The stim-
ulus was presented for 500 ms. On each trial, observers indicated
their decision with a button press. The mean orientation of the sig-
nal population was controlled by APE, an adaptive method of con-
Fig. 5. Summary results for six strabismic amblyopes in which AME/FFE coherence motion threshold ratios (Y-axis) are plotted against the AME/FFE contrast ratios (X-axis).
At a contrast ratio of 1, AMEs show higher thresholds compared to those of the FFEs. However, presenting stimuli with higher contrast improves the coherence thresholds for
AME relative to those of the FFE where at high AME/FFE contrast ratios, AMEs show better performance than FFEs. The arrows in (B) show where a linear fitted line to the data
meets the unity line and hypothetically where the two eyes show similar performances.
B. Mansouri et al./Vision Research 48 (2008) 2775–2784
stant stimuli (Watt & Andrews, 1981) which sampled a range of
orientations around vertical.
4.3.1. Single element threshold measurements
We equated the monocular performance of the FFE and AME in
terms of the contrast required to obtain criterion levels of orienta-
tion discrimination for a single Gabor element. The contrast was
nate forcedchoice paradigmusingthe APE adaptive procedure until
the FFE exhibited comparable levels of performance to that of the
AME (Mansouri, Allen, Hess, Dakin, & Ehrt, 2004, for more details).
Two different combinations of signal and noise were tested.
Depending on which condition was tested, each eye’s image could
contain a signal population, a noise population or both. To prevent
any bias, the observers were not informed which population (e.g.
signal or noise) was being presented at any time and if different
Gabor populations were presented to different eyes, the process
was randomized within a run so that observers were unaware of
which stimulus was presented to which eye. Observers did not re-
The two combinations of signal and noise were:
(A) Signal population presented to FFE and mean luminance to
the AME, and vice versa (Fig. 6A).
(B) Signal population presented to FFE and noise population to
the AME, and vice versa (Fig. 6B).
All subjects started the experiment with the signal and noise
populations both comprised of 16 Gabors and continued with dif-
ferent proportions of signal and noise elements and different ratios
of contrast for stimuli to either eye.
Given that thresholds are estimates of response variance, the
non-ideal behaviour of observers with noiseless stimuli can be ex-
pressed as an additive internal noise. The level of internal noise is
measured by increasing the amount of external noise in the stim-
ulus and determining the point at which observers’ performance
begins to deteriorate. If the task requires integration, then observ-
ers’ robustness to increasing amounts of external noise will depend
decreasingly on internal noise and increasingly on how many sam-
ples are averaged. Thus, the form of the equivalent noise model is:
where robsis the observed threshold, rextis the external noise, rint
is the estimated equivalent intrinsic or internal noise and n is the
estimated number of samples being employed. In terms of the ori-
entation discrimination task, robscorresponds to the threshold for
orientation discrimination, rextto the standard deviation of the dis-
tribution from which the samples are derived;rintto the noise asso-
ciated with the measurement of each orientation sample and their
combination and n corresponds to the estimated number of orien-
tation samples being combined by the visual system. It is important
to note that this is an equivalent noise model and that the model
supplies equivalent estimated parameters. The usefulness of such
a model is in providing a more detailed description of performance
than provided by a single threshold estimate. We don’t believe that
these estimates necessarily have direct biological correlates. Orien-
tation discrimination thresholds were derived from between 192
and 340 presentations for each of a number of standard deviations
of the parent distribution, i.e. external noise (8 levels typically be-
tween 0? and 28?). The orientation threshold for each level of vari-
ance of the parent distribution was estimated as the slope of the
best fitting cumulative Gaussian function using a maximum likeli-
hood procedure in which the threshold was equal to 82% correct.
1000 bootstrap replications of the fitted function were carried out
and used to generate 95% confidence intervals (CIs) for the thresh-
old estimates (Foster & Bischof, 1997). The orientation discrimina-
tion thresholds at each level of external noise were fitted by the
equivalent noise model to derive the measures of internal noise
and number of samples.
Fig. 6. Arrangement of the stimuli used in the mean orientation task are shown for
monocular (A) and dichoptic (B) conditions. In (A), only eight signal Gabor are
presented to one eye and mean luminance plus fixation point to the other. The
accuracy of discriminating the mean orientation of the array is measured for arrays
whose Gabor orientations are samples from parent distributions of different
standard deviation. The example here is of a moderate parent standard deviation. In
(B), eight signal Gabors (samples from an infinitely narrow parent distribution) are
presented to one eye (right image in this presentation) and eight noise Gabors
(samples from an infinitely wide parent distribution) to the other eye (right image
in this presentation). Measurements were made for Gabor signals whose orienta-
tions were sampled from parent distributions having different standard deviations.
In the experiments reported here we always used 16 signal element and 16 noise
elements when ER = 1).
Fig. 7. Monocular mean orientation discrimination thresholds are presented for FFE
(circles and dashed line) and AME (stars and solid line) for one amblyopic subject
(ML). X-axis represents orientation standard deviation (?). Y-axis represents
threshold orientation offset (?). Internal noise (IN) and sampling efficiency (NS)
parameters, which were derived from fitting the equivalent noise model to data, are
presented in inset. MLG and MLA represent the FFE and the AME for subject ML,
respectively. The contrast of the stimuli to the FFE is 50% and to AME is 75%. At this
combination of contrasts, the two eyes of this subject showed similar local
orientation discrimination thresholds.
B. Mansouri et al./Vision Research 48 (2008) 2775–2784
4.5. Results and discussion
Fig. 7 shows a condition where signal is presented to one eye
and at the same time mean luminance is presented to the other
eye (see Fig. 6A). The contrasts of the stimuli to FFE and AME are
set at the level that produced similar monocular performance for
the two eyes for local orientation discrimination (e.g. 50% contrast
to FFE and 75% contrast to AME for this example subject). This fig-
ure shows that if the monocular performance of the AME is equa-
ted in terms of contrast for local orientation discrimination
performance (i.e. a single Gabor), its performance on a global mean
orientation task (i.e. an array of Gabors) is similar to that of the FFE.
Similar results for a sample of amblyopes have been presented pre-
viously (Mansouri et al., 2004).
In Fig. 8 we show representative dichoptic results for the same
amblyope (ML) where orientation discrimination thresholds are
plotted against signal variance under the condition where signal
is presented to AME and noise to the FFE (filled symbols) and vice
versa (unfilled symbols). Each panel (Fig. 8A–C) represents a differ-
ent number of elements presented to the FFE and AME (e.g. in A, 16
to FFE and 16 elements to AME, ER = 1; in B, 8 elements to FFE and
32 elements to AME, ER = 4; in C, 4 elements to FFE and 64 ele-
ments to AME, ER = 16). As the number of elements (be they signal
or noise) presented to the FFE is reduced, dichoptic performance of
the FFE and AME come together, suggesting balanced dichoptic
performance. This is summarized in panel (D). Since the individual
orientations of the signal elements are samples from a distribution
whose width can be varied, to derive a single estimate, we aver-
aged performance across different signal variances. The unfilled
bar on the left represents the relative dichoptic performance
(AME/FFE) under conditions where monocular performance is
matched (Fig. 7) using the single element task (see Methods) and
it is clear that dichoptic performance is imbalanced (i.e. above
unity). The filled bar on the right shows balanced dichoptic perfor-
mance of the FFE and AME when the number of elements is re-
duced to the FFE (in this case the interocular element ratio, ER is
16 but the interocular contrast is unity).
Fig. 9 shows summary dichoptic results similar to those already
described in Fig. 8B for six amblyopes. Sixteen signal and 16 noise
Gabors were dichoptically presented. The unfilled bars represent
the discrimination threshold ratio (AME views signal/FFE views
signal) under dichoptic conditions. An interocular performance
asymmetry (i.e. AME/FFE discrimination ratio greater than unity)
is evident in all cases, suggesting a strong suppressive influence
for a task whose monocular performance has been equated for
FFE and AME (Fig 7). The filled bars represent the dichoptic perfor-
mance ratio when the relative (i.e. interocular) number of elements
(i.e. element ratio, ER) and/or relative contrasts (i.e. contrast ratio,
CR) has been adjusted to produce balanced performance (same
performance irrespective of whether the signal is presented to
the AME or the noise is presented to the AME). In all cases it is pos-
sible to balance performance between the eyes by manipulating
the number and/or contrast of the elements comprising the signal
seen by one eye and the noise seen by the other. Under these arti-
ficial conditions, information coming from the AME was combined
with information coming through the FFE, similar to that expected
Fig. 8. In (A–C), comparison of dichoptic performance of FFE and AME for the orientation discrimination task where one eye sees the signal and the other, the noise.
Orientation thresholds are plotted against the standard deviation of the parent population of which the signals are samples. The solid line is the best fitting line according to
the summation-variance model. As the number of elements comprising either the signal or noise seen by the FFE is reduced (e.g. in (A), 16 to FFE and 16 elements to AME,
ER = 1; in (B), 8 elements to FFE and 32 elements to AME, ER = 4; in (C), 4 elements to FFE and 64 elements to AME, ER = 16), dichoptic performance of the two eyes comes
together. In (D), relative dichoptic performance (averaged across signal standard deviation) is compared under two conditions; equated monocularly on the left and equated
dichoptically, on the right. CR and ER refer to the interocular contrast and element ratio for the dichoptic stimuli.
B. Mansouri et al./Vision Research 48 (2008) 2775–2784
of a normal binocular observer (in whom it does not matter which
eye sees signal and which eye sees noise) (Mansouri et al., 2005).
5. General discussion
The results from the motion and orientation tasks demon-
strate that under natural viewing, information from the two eyes
of an amblyope interact anomalously, but that this can be
changed by altering the relative information seen by each eye.
Under these special conditions, the amblyopic visual system
can exhibit balanced interocular performance. This is explicable
in terms of a model of binocular combination similar to that de-
scribed in the introduction for contrast detection (Fig. 1C) and
(a) The presence of an intact binocular summation stage (Baker
et al., 2007b, 2008), and
(b) Different strengths of suppressive drive to the gain control
stage prior to binocular summation.
It follows that matching monocular performance of fixing and
amblyopic eyes will not result in a matched dichoptic performance
for the amblyopes because this does not take into account the main
reason for imbalanced dichoptic performance, namely the role
played by the stronger suppressive drive coming from the FFE un-
der dichoptic conditions. In this aspect the models for contrast
detection and global processing might differ because the former in-
volves threshold stimuli whereas the latter, suprathreshold stim-
uli. We conclude that the weights of the interocular suppression
are different in amblyopic compared with normal observers. By
reducing either the contrast or the number of elements seen binoc-
ularly by the FFE we can counteract the higher intrinsic gain asso-
ciated with the suppressive drive coming from the FFE and thereby
more adequately balance the left and right inputs to the final stage
of binocular summation. The magnitude of this suppressive drive
varied from individual to individual and in some cases was very
strong (e.g. GN, Fig. 9). However, even in this case a suitable bal-
ance point could be obtained such that signal and noise were being
combined with equal efficiency to accomplish the signal/noise
task. This was a general finding for both types of global signal/noise
tasks used here.
Since this is true for both motion and form, we take it to rep-
resent a general principle of operation of both the dorsal and
ventral cortical processing streams. The available neurophysio-
logical evidence suggests that this interocular suppression is
present in area V1 (Harrad, Sengpiel & Blakemore, 1996; Sengpiel
& Blakemore, 1996; Sengpiel, Blakemore, Kind, & Harrad, 1994;
Sengpiel, Jirmann, Vorobyov, & Eysel, 2006) and so it may not
be surprising that comparable effects are seen for stimuli de-
signed to activate dorsal as well as ventral extra-striate pathways
if the problem arises in an early common site (Ungerleider &
Mishkin, 1982). The present study not only provides further sup-
port for intact binocular mechanisms in amblyopia but also at-
tests to the importance of suppression in rendering what is
essentially a structurally binocular visual system, functionally
monocular. The finding that suppression provides a fundamental
limit to the extent to which signals from the two eyes of amblyo-
pes can be combined provides the basis upon which to develop
more effective binocularly based treatments, avoiding the need
for patching therapy with its inherent low compliance and psy-
cho-social problems (Koklanis, Abel, & Aroni, 2006). Such a bin-
ocularly based treatment, based on a manipulation of the
relative information content of left/right images, would be direc-
ted at shifting the dichoptic balance point over time towards that
found in normal observers where comparable dichoptic perfor-
mance is achieved with the same physical stimuli presented to
each eye. If the balance point can be changed over time then
conditions would be established whereby binocular combination
would occur without the need for adjusting the relative informa-
tion content to each eye.
This study is supported by a CIHR Grant (# MOP 53346) to
R.F.H. We are grateful to Drs. Tim Ledgeway and Steven Dakin for
their technical support.
Baker, D. H., & Meese, T. S. (2007). Binocular contrast interactions: Dichoptic
masking is not a single process. Vision Research, 47(24), 3096–3107.
Baker, D. H., Meese, T. S., & Georgeson, M. A. (2007a). Binocular interaction: Contrast
matching and contrast discrimination are predicted by the same model. Spatial
Vision, 20(5), 397–413.
Baker, D. H., Meese, T. S., & Hess, R. F. (2008). Contrast masking in strabismic
amblyopia: Attenuation,noise, interocular
summation. Vision Research, 48(915), 1625–1640.
Baker, D. H., Meese, T. S., Mansouri, B., & Hess, R. F. (2007b). Binocular summation of
contrast remains intact in strabismic amblyopia. Investigative Ophthalmology &
Visual Science, 48(11), 5332–5338.
Fig. 9. Summary results for six strabismic amblyopes in which AME/FFE mean orientation discrimination offsets averaged over eight different orientation distribution
variances (Y-axis) are presented for monocularly matched (open bars) and dichoptically matched (filled bars) conditions. ER stands for AME/FFE stimulus elements ratio and
CR shows the AME/FFE contrast ratio. At monocularly matched conditions, AMEs show higher thresholds compared to those of the FFEs. However, presenting more stimulus
elements and with higher contrast to AME improves the relative mean thresholds for AME to equal those of the FFE.
B. Mansouri et al./Vision Research 48 (2008) 2775–2784
Baker, D. H., Meese, T. S., & Summers, R. J. (2007c). Psychophysical evidence for two Download full-text
routes to suppression before binocular summation of signals in human vision.
Neuroscience, 146(1), 435–448.
Brainard, D. H. (1997). The psychophysics toolbox. Spatial Vision, 10, 433–446.
Dakin, S. C. (2001). An information limit on the spatial integration of local
orientation signals. Journal of the Optical Society of America. A, Optics and Image
Science, 18, 1016–1026.
Edwards, M., & Badcock, D. R. (1995). Global motion perception: No interaction
between thefirst-and second-order
Foster, D. H., & Bischof, W. F. (1997). Bootstrap estimates of the statistical
accuracy of thresholds obtained from psychometric functions. Spatial Vision,
Harrad, R., Sengpiel, F., & Blakemore, C. (1996). Physiology of suppression in
strabismic amblyopia. The British Journal of Ophthalmology, 80(4), 373–377.
Hess, R. F., Hutchinson, C. V., Ledgeway, T., & Mansouri, B. (2007). Binocular
influences on global motion processing in the human visual system. Vision
Research, 47(12), 1682–1692.
Holopigian, K., Blake, R., & Greenwald, M. J. (1988). Clinical suppression and
amblyopia. Investigative Ophthalmology & Visual Science, 29(3), 444–451.
Hubel, D. H., & Weisel, T. N. (1968). Receptive fields and functional architecture of
monkey striate cortex. The Journal of Physiology (London), 195, 215–243.
Koklanis, K., Abel, L., & Aroni, R. (2006). Psychosocial impact of amblyopia and its
treatment: A multidisciplinary study. Clinical and Experimental Ophthalmology,
Legge, G. E. (1984). Binocular contrast summation—II. Quadratic summation. Vision
Research, 24(4), 385–394.
Mansouri, B., Allen, H. A., Hess, R. E., Dakin, S. C., & Ehrt, O. (2004). Integration of
orientation information in amblyopia. Vision Research, 44(25), 2955–2969.
Mansouri, B., Hess, R. F., Allen, H. A., & Dakin, S. C. (2005). Integration, segregation,
and binocular combination. Journal of the Optical Society of America. A, Optics and
Image Science, 22(1), 38–48.
McKee, S. P., Levi, D. M., & Movshon, J. A. (2003). The pattern of visual deficits in
amblyopia. Journal of Vision, 3(5), 380–405.
Meese, T. S., Georgeson, M. A., & Baker, D. H. (2006). Binocular contrast vision at and
above threshold. Journal of Vision, 6(11), 1224–1243.
Meese, T. S., & Hess, R. F. (2004). Low spatial frequencies are suppressively masked
across spatial scale, orientation, field position, and eye of origin. Journal of
Vision, 4, 843–856.
Mitchell, D. E., Howell, E. R., & Keith, C. G. (1983). The effect of minimal occlusion
therapy on binocular visual functions in amblyopia. Investigative Ophthalmology
& Visual Science, 24(6), 778–781.
Newsome, W. T., & Pare, E. B. (1988). A selective impairment of motion perception
following lesions of the middle temporal visual area (MT). The Journal of
Neuroscience, 8, 2201–2211.
Pape, H. C., & Eysel, U. T. (1986). Binocular interactions in the lateral geniculate
nucleus of the cat: GABAergic inhibition reduced by dominant afferent activity.
Experimental Brain Research, 61(2), 265–271.
Pelli, D. G. (1997). The VideoToolbox software for visual psychophysics:
Transforming numbers into movies. Spatial Vision, 10(4), 437–442.
Rosenbach, O. (1903). Ueber monokulare Vorherrschaft beim binikularen Sehen.
Munchener Medizinische Wochenschriff, 30, 1290–1292.
Sanderson, K. J., Bishop, P. O., & Darian-Smith, I. (1971). The properties of the
binocular receptive fields of lateral geniculate neurons. Experimental Brain
Research, 13(2), 178–207.
Sengpiel, F., & Blakemore, C. (1996). The neural basis of suppression and amblyopia
in strabismus. Eye, 10(Pt 2), 250–258.
Sengpiel, F., Blakemore, C., Kind, P. C., & Harrad, R. (1994). Interocular suppression
in the visual cortex of strabismic cats. The Journal of Neuroscience, 14(11 Pt 2),
Sengpiel, F., Jirmann, K. U., Vorobyov, V., & Eysel, U. T. (2006). Strabismic
suppression is mediated by inhibitory interactions in the primary visual
cortex. Cerebral Cortex, 16(12), 1750–1758.
Travers, T. (1938). Suppression of vision in squint and its association with retinal
correspondence and amblyopia. British Journal of Ophthalmology, 22, 577–604.
Ungerleider, L. G., & Mishkin, M. (1982). Two cortical visual systems. In D. J. Ingle
(Ed.), Analysis of visual behaviour (pp. 549–586). Cambridge, MA: MIT Press.
Watt, R. J., & Andrews, D. (1981). APE: Adaptive probit estimation of psychometric
functions. Current Psychological Review, 1, 205–214.
Xue, J. T., Ramoa, A. S., Carney, T., & Freeman, R. D. (1987). Binocular interaction in
the dorsal lateral geniculate nucleus of the cat. Experimental Brain Research,
B. Mansouri et al./Vision Research 48 (2008) 2775–2784