www.elsevier.com/locate/visres

Vision Research 45 (2005) 2810–2819

Spatial interactions reveal inhibitory cortical networksin human amblyopia

Erwin H. Wong a,*, Dennis M. Levi a,b, Paul V. McGraw c

a School of Optometry, University of California, Berkeley, CA 94720, USAb Helen Wills Neuroscience Institute, University of California, Berkeley, CA 94720, USA

c Department of Optometry, University of Bradford, Bradford BD7 1DP, UK

Received 24 August 2004; received in revised form 1 June 2005

Abstract

Humans with amblyopia have a well-documented loss of sensitivity for first-order, or luminance defined, visual information.

Recent studies show that they also display a specific loss of sensitivity for second-order, or contrast defined, visual information;

a type of image structure encoded by neurons found predominantly in visual area A18/V2. In the present study, we investigate

whether amblyopia disrupts the normal architecture of spatial interactions in V2 by determining the contrast detection threshold

of a second-order target in the presence of second-order flanking stimuli. Adjacent flanks facilitated second-order detectability in

normal observers. However, in marked contrast, they suppressed detection in each eye of the majority of amblyopic observers. Fur-

thermore, strabismic observers with no loss of visual acuity show a similar pattern of detection suppression. We speculate that

amblyopia results in predominantly inhibitory cortical interactions between second-order neurons.

� 2005 Elsevier Ltd. All rights reserved.

Keywords: Amblyopia; Second-order; Lateral interactions; Visual cortex

1. Introduction

Amblyopia is a disorder of spatial vision, usually

present in one eye, which results from discordant binoc-

ular input to the visual cortex during development.

Amblyopia is typically associated with strabismus (eye

misalignment) or anisometropia (unequal refractive er-

ror). A loss of contrast sensitivity for first-order (lumi-nance defined) spatial information is well documented

in amblyopic eyes, and is widely attributed to neural def-

icits at the level of striate cortex (V1) (Kiorpes &

McKee, 1999). Neurophysiological studies have shown

that the response of V1 neurons to a first-order, near

threshold stimulus placed within its receptive field can

be facilitated (response increased) (Bakin, Nakayama,

0042-6989/$ - see front matter � 2005 Elsevier Ltd. All rights reserved.

doi:10.1016/j.visres.2005.06.008

* Corresponding author. Tel.: +1 314 516 6516; fax: +1 314 516 5150.

E-mail addresses: wonge@umsl.edu (E.H. Wong), dlevi@berkeley.

edu (D.M. Levi), p.v.mcgraw@bradford.ac.uk (P.V. McGraw).

& Gilbert, 2000; Kapadia, Ito, Gilbert, & Westheimer,

1995; Nelson & Frost, 1985; Polat, Mizobe, Pettet,

Kasamatsu, & Norcia, 1998) or suppressed (response re-

duced) (Knierman & Van Essen, 1992; Levitt & Lund,

1997; Walker, Ohzawa, & Freeman, 1999) by flanking

first-order stimuli. The type of interaction, i.e., facilita-

tory or suppressive, depends upon the spatial distance

between target and flanks, the relative orientation differ-ence between the elements that comprise the target and

flanks, and the magnitude of the flank contrast

(Kapadia, Westheimer, & Gilbert, 1999; Polat et al.,

1998). Such cortical interactions are thought to form

the cellular basis to psychophysical demonstrations of

enhanced visibility for first-order stimuli flanked by

facilitatory masks. Psychophysical studies have shown

that target and flank conditions which produce facilita-tion (lower the contrast detection threshold) in normal

eyes (Kapadia et al., 1995; Levi, Hariharan, & Klein,

2002; Polat & Sagi, 1993, 1994; Yu, Klein, & Levi,

E.H. Wong et al. / Vision Research 45 (2005) 2810–2819 2811

2002) can result in suppression (increase the contrast

detection threshold), or reduced facilitation, in amblyo-

pic eyes (Levi et al., 2002; Polat, Sagi, & Norcia, 1997).

However, Polat, Ma-Naim, Belkin, and Sagi (2004,

2005) has reported reduced facilitation in amblyopic

eyes only for high spatial frequency stimuli, and in stra-bismic amblyopes more than anisometropic amblyopes.

Neurophysiological studies have shown that the tran-

sition from facilitatory to suppressive interactions re-

flects the spatial distribution of target and flanks either

within the classic receptive field (CRF) or its inhibitory

surround. The excitatory CRF and larger (P2·) over-lapping inhibitory region form a center-surround mech-

anism (Angelucci et al., 2002; Cavanaugh, Bair, &Movshon, 2002a, 2002b) in which stimulation of the

annular surround suppresses the CRF response through

divisive modulation of the response gain but can not

drive the CRF directly (Cavanaugh et al., 2002a,

2002b). Anatomical evidence indicates that the excitato-

ry spatial limit of the CRF is formed by horizontal con-

nections within V1 (i.e., connections between cortical

columns) and the inhibitory surround is largely formedby feedback connections from V2 to V1 (Angelucci

et al., 2002; Cavanaugh et al., 2002a; but see Stettler,

Das, Bennett, & Gilbert, 2002). Therefore, the abnormal

pattern of spatial interactions for first-order visual stim-

uli reported in amblyopic observers could result from

either abnormal horizontal connections in V1, and/or

feedback connections from V2 to V1.

In comparison with the striate cortex, much less isknown about the effects of amblyopia on extra-striate

cortical structure and function. Visual processing in

the extra-striate cortex (V2) can be investigated using

second-order spatial stimuli, e.g., a visual stimulus de-

fined by contrast modulations. Contrast modulation fre-

quencies are not represented in the Fourier spectrum of

an image, and therefore demodulation is required for

stimulus detection—this has been extensively modeledas a filter-rectify-filter processing cascade (Chubb &

Sperling, 1988). Briefly, luminance modulations of high

spatial frequencies undergo linear filtering in V1, the

output is rectified (the demodulation step), and this en-

ables contrast modulations of low spatial frequencies to

be detected by a second-stage of linear filtering. There is

compelling psychophysical evidence that first-order and

second-order spatial information can be processed inde-pendently in the visual cortex (Schofield & Georgeson,

1999, 2003; Willis, Smallman, & Harris, 2000). Further-

more, physiological studies in cat (Mareschal & Baker,

1998; Zhou & Baker, 1994) and monkey (Leventhal,

Wang, Schmolesky, & Zhou, 1998; von der Heydt &

Peterhans, 1984, 1989) place the locus of the second fil-

tering stage predominantly in area 18/V2.

In a previous study, we demonstrated a specific loss ofsecond-order sensitivity in individuals with amblyopia

(Wong, Levi, & McGraw, 2001). However, it is presently

unknown whether the pattern of spatial interactions

which occur in the visual cortex of normal, or amblyopic

observers, are qualitatively or quantitatively similar for

first- and second-order stimuli. We examine this issue

by psychophysically determining contrast detection

threshold for a second-order target in the presence of col-linear or orthogonal second-order flanks (equated for

visibility) in normal observers (control), amblyopic

observers, and observers with strabismus but no loss of

visual acuity. We found the flanking effect to be facilita-

tive in normals but suppressive in each eye of most

amblyopic and strabismic observers (subsequently re-

ferred to as non-control observers). We speculate that

human amblyopia results in predominantly inhibitoryhorizontal interactions between second-order neurons.

2. Methods

2.1. Observers

Six amblyopic observers, two observers with strabis-mus but no loss of visual acuity, and five normal (con-

trol) observers participated in the experiment. All

observers were adults and the visual characteristics of

the non-control observers are presented Tables 1A and

B. Control observers had normal or corrected-to-normal

vision. All observers were highly practiced at making

psychophysical judgements, wore refractive correction

as required, and all but the author (EW) were naı̈ve tothe task. Informed consent following the guidelines of

either the University of Houston or the University of

California was obtained from all observers prior to data

collection.

2.2. Apparatus

Stimuli were generated using the macro capabilitiesof NIH Image 1.62f (available from http://rsb.

info.nih.gov/nih-image/). The host computer was an

Apple Power Macintosh 6500/225 and stimuli were pre-

sented on a Dell monitor (21-inch screen, resolution

1024 · 768 pixels, frame refresh rate 75 Hz, and mean

luminance 15 cd/m2). The monitor output was made lin-

ear over the entire range used in the experiment via cal-

ibration with a photometer (Minolta LS-110 digitalluminance meter). To obtain accurate control of lumi-

nance contrast we increased the number of intensity lev-

els from 8 to 12 bits by combining the outputs of the red,

green, and blue guns via a video summation device (Pelli

& Zhang, 1991).

2.3. Stimuli

We used stationary, contrast modulations of random

static noise as second-order stimuli (Fig. 1). Stimuli were

Table 1

Visual characteristics of non-control observers

Subject Type Refractive error Acuitya Fixation Strabismus Stereoacuityb Treatment/age

(A) Visual characteristics of amblyopic observers

JF Aniso R �0.50–0.25 · 177 20/20 R/L central None None No surgery or patch;

25, M L +3.50 DS 20/50 glasses at 11

AH Strab R +0.50–1.00 · 93 20/50 R/L central R esotrope 12D None No surgery; patch at 7;

19, F L plano-0.75 · 100 20/25 glasses at 16

RH Strab R �1.00–0.50 · 170 20/15 R/L central, L esotrope 2D None No surgery or patch;

35, M L �1.50–1.50 · 010 20/40 L unsteady glasses at 12

AM Aniso R +2.50–1.00 · 010 20/50 R/L central None 320 s arc No surgery or patch;

22, F L �0.25 DS 20/20 contact lens at 7

DM Strab & R �0.50–0.25 · 92 20/20 R central, L exotrope 3D None No surgery or patch;

40, F Aniso L +2.50–1.00 · 160 20/80 L 0.5� nasal glasses at 12

DS Strab & R +2.25 DS 20/40 R 2� nasal L esotrope 8D None No surgery;

26, M Aniso L +0.50 DS 20/20 L central patch and glasses at 5

(B) Visual characteristics of observers with strabismus and no loss of visual acuity

RC Strab R plano-0.25 · 167 20/15 R/L central alt exotrope 8D, None No surgery, patch, or

33, M L +0.25–0.50 · 170 20/15 L hypertrope 6D glasses

WS Strab & R +2.75 DS 20/15 R/L central R exotrope 10D, None Surgery at 0.5, 3, 7;

50, F Aniso L +0.50 DS 20/20 R hypertrope 5D patch and glasses at 2

a Snellen chart.b Borish card.

2812 E.H. Wong et al. / Vision Research 45 (2005) 2810–2819

constructed by multiplying a random static noise back-

ground by a 2-D Gabor, and are mathematically de-

scribed by:

Lðx; yÞ ¼ ðLmean þ ðrand� 0.5ÞÞ� ðLmean þ Lmean � C sinð2pðFxþ /ÞÞ� expð�ðx2 þ y2Þ=r2ÞÞ; ð1Þ

where Lmean is the mean luminance of the background,

rand is a uniformly distributed random variable between

0 and 1, C is the contrast of the modulation, F is the spa-

A

B

Fig. 1. Examples of our stimuli: horizontal target with (A) collinear

flanks and (B) orthogonal flanks, separated (center-to-center) by 4k(where k (wavelength) = r, the standard deviation of the Gaussian

window). Target and flanks are sinusoidal modulations (1.0 or

2.0 c/deg) of the background contrast (static, random noise).

tial frequency of the modulation, / is the spatial phase,

r is the standard deviation of the Gaussian envelope,

and x and y are the respective horizontal and vertical

distances from the peak of the Gaussian envelope.

Eachpixel subtended 0.93 min of arc at amean viewing

distance of 1.4 m and noise patches were 4 · 4 square pix-els. The luminance increment or decrement of each noise

patch was taken randomly from a uniform distribution.

The mean contrast of the noise background was 50%.

The contrast of each noise patch depended on the value

of the convolving sinusoid at that position. The stimulus

area (768 · 256 pixels) was centered within the screen

and contained a horizontal target and flanks (collinear or

orthogonal pair) of equal spatial frequency (1.0 or2.0 c/deg). The standard deviation (r) of the 2-D Gabor

for all stimuli was 1k (where k = 1 carrier cycle). This pro-

duced targets and flanks of equal bandwidth, spatial

isolation at the smallest separation (4k), and decreasing

center-to-center separationas spatial frequency increased.

Flankswere equidistant from the target andwe tested sep-

arations (center-to-center) of 4, 5, 6, and 8k. Separationsfrom1k to 3kwerenot tested to avoidoverlapbetween tar-get and flanks. Three sets of stimuli were created and used

randomly during the experiment to offset any luminance

artifacts produced from the clumping of random static

noise.

2.4. Experiment

We used a lateral masking design (Polat & Sagi, 1993)and incorporated normalization of flank contrast

A

B

C

E.H. Wong et al. / Vision Research 45 (2005) 2810–2819 2813

between all observers. We measured second-order con-

trast detection for (a) flanks, (b) horizontal target, and

(c) horizontal target in the presence of collinear (hori-

zontal) or orthogonal (vertical) flanks (at 2· and in some

cases 4· the contrast detection threshold measured in

(a)). For all measurements, the observer sat in a dark-ened room, head positioned on a chin rest, non-tested

eye occluded with a black patch, and fixating the center

of the screen. We tested each eye of the non-control

observers, and the dominant eye of the control

observers.

For measurement (a) we used a method of limits par-

adigm consisting of a self-paced, 7-step staircase (con-

trast changes in .075 log steps). Stimulus duration was500 ms and the inter-stimulus interval (500 ms) con-

tained an un-modulated random noise field of the same

size and average Michaelson contrast (50%) as that con-

tained in the stimulus intervals. Each run contained 25

trials, and threshold was calculated from pooling at least

10 consecutive runs taken over 2 or more days.

For measurement (b) and (c) we used a self-paced,

temporal two-alternative forced-choice paradigm withthe method of constant stimuli. Flanks were presented

in each of two 500-ms stimulus intervals, separated by

a 500-ms interval as in measurement (a), and, respective-

ly, signaled by simultaneous single or dual tones. The

subject�s task was to detect the target presented random-

ly in one of the stimulus intervals. A keyboard press sig-

naled the response and no feedback was given. Each

trial contained the target at one of seven contrast levels(in .075 log steps), chosen to span the psychometric

function, and presented in random order. Each run con-

sisted of 145 trials, with the first 5 being discarded to al-

low for task adaptation, and tested one combination of

spatial frequency and flank to target separation. We col-

lected at least 5 consecutive runs over 3 or more days

and calculated contrast detection threshold (75% correct

response) by Weibull function fit to the data.We note that in pilot studies amblyopic observers

were unable to perform the experimental task adequate-

ly when stimuli were presented for 200 ms. We therefore

used 500 ms stimulus presentations in all experimental

measures.

Fig. 2. Normal observers show facilitation of contrast detection by

flanks. Contrast of the collinear flanks (column 1) and orthogonal

flanks (column 2) is in multiples of their respective contrast detection

threshold: (A) 2· threshold and (B) 4· threshold. Spatial frequency is

1 c/deg. Lines connect the weighted average at each separation and

error bars represent ± SEM. The weighted averages of (A) and (B) are

shown in (C).

3. Results

All observers viewed second-order stimuli consisting

of 1 c/deg (modulation frequency), non-overlapping tar-

get and flanks (center-to-center separations of 4, 5, 6,

and 8k) (where k (wavelength) = r, the standard devia-

tion of the Gaussian window) (Fig. 1). We calculated a

threshold modulation ratio: (contrast detection thresh-

old with flanks)/(contrast detection threshold withoutflanks), for the dominant eye of normal observers,

amblyopic eye (AE) and fellow non-amblyopic eye

(NAE) of amblyopic observers, and the right eye (RE)

and left eye (LE) of non-amblyopic observers with stra-

bismus. Threshold modulation ratios greater than 1

indicate suppressive interactions, whilst ratios less than

1 indicate facilitative interactions. We plot threshold

modulation ratio as a function of target to flankseparation.

We found that normal eyes showed facilitation in the

presence of both collinear and orthogonal flanks at 4, 5,

and 6k separation with a reduced effect at 8k (Fig. 2).

Note that the target–flank separation is specified in both

degrees and wavelengths (k)—which are identical at a

stimulus spatial frequency of 1 c/deg. The amount of

facilitation was essentially equal for flanks of 2· or 4·their flank contrast detection threshold (flank detection

data not shown). Therefore, facilitative interactions in

normal observers appear to be independent of flank ori-

entation and contrast for second-order stimuli. At 2·flank contrast detection threshold, the mean facilitation

at 4k (the separation producing the greatest effect) was

13% ± 4% for collinear flanks and 11% ± 4% for orthog-

onal flanks. We subsequently tested all non-controlobservers at 4k separation, and most observers at 5, 6,

2814 E.H. Wong et al. / Vision Research 45 (2005) 2810–2819

and 8k separations, with flanks set at twice the flank

contrast detection threshold.

Inmarked contrast to normal observers, we found that

flanks suppressed target detection thresholds in each eye

of the majority of non-control observers. Fig. 3 presents

data from each eye of all non-control observers, testedat 4k separation. Collinear flanks produced suppression

in both eyes of five amblyopic observers, and in 3 of 4 eyes

of strabismic, non-amblyopic observers WS and RC

(non-dominant eye shown as AE and dominant eye as

NAE) (Fig. 3A). Amongst amblyopes, observer DS

showed the largest amount of suppression (38% ± 12%).

All amblyopic observers generally showed very similar

amounts of suppression in each eye.Orthogonal flanks at a separation of 4k produced

suppression in all amblyopic and non-dominant eyes

(Fig. 3B). For each of these eyes the amount of suppres-

sion was very similar to that produced by collinear

flanks (Fig. 3A). In contrast, orthogonal flanks pro-

duced suppression in only 3 of 6 non-amblyopic eyes

(DS, JF, and AM) and 1 of 2 dominant eyes (RC)

(Fig. 3B). Interestingly, the non-amblyopic eye ofobserver RH showed the largest amount of suppression

for orthogonal flanks (39% ± 10%), which was very sim-

ilar to the largest amount produced by collinear flanks

(36% ± 12%, amblyopic eye of observer DS).

Fig. 3. Collinear (A) and orthogonal (B) flanks produce suppression of

contrast detection in each eye of most non-control observers. Data are

for 1 c/deg stimuli at 4k separation (i.e., the condition of maximum

flank effect) and error bars represent ± SEM. Suppression is shown by

most amblyopic eyes (filled symbols) and fellow preferred eyes (open

symbols). The facilitation by normal observers (weighted average) is

shown by the solid line. For observers RC and WS (both strabismic

without loss of visual acuity), their non-dominant eye is represented as

AE and the dominant eye as NAE.

Overall, our findings indicate that target detection

thresholds are clearly influenced by neural activity asso-

ciated with both collinear and orthogonal flanks. In

visually normal observers, the presence of these flanking

stimuli enhance detection, whilst in non-control observ-

ers they impair detection, in most cases regardless ofthe eye investigated. This strongly suggests that the

mechanism that mediates facilitation, or suppression in

the case of non-control observers, has a cortical locus

at or beyond the site of binocular combination. In sup-

port, the amblyopic observers who showed the least

amount of suppression (Fig. 3—JF and AM) were not

strabismic and amblyope AM had demonstrated some

residual binocular function in the form of gross stereop-sis (32000).

The amount of suppression shown by non-control

observers decreased as the separation between target

and flank increased, and dissipated almost entirely at a

separation of 8k (Fig. 4). This is demonstrated by data

from four amblyopic observers who showed suppression

by both collinear and orthogonal flanks. Taken together

with the results from normal observers (Fig. 2), it islikely that a separation of 8k reflects the spatial limits

of the inhibitory surround on the CRF (see Section 4).

The non-control observers had higher contrast detec-

tion threshold for flanks (data not shown) which limited

the contrast level at which flanks could be presented at

twice their respective detection threshold—the exception

being amblyope DS. Whilst the normal observers

showed similar interactions for different levels of flankcontrast, when tested with flanks at 4· detection thresh-

old the amblyopic eye of DS showed reduced suppres-

sion for both collinear (0% ± 9% vs 38% ± 12%) and

orthogonal flank conditions (10% ± 10% vs 36% ±

12%) (Fig. 4). The reduction in target suppression may

reflect the unmasking of excitatory connections by high-

er levels of flank contrast (see Section 4).

For our sample of non-control observers, we foundno association between flank-induced suppression and

absolute second-order sensitivity (isolated target detec-

tion threshold in the absence of flanks, relative to nor-

mal observers). This was the case for amblyopic (or

non-dominant) eyes (Fig. 5A) and fellow non-amblyopic

(or dominant) eyes (Fig. 5B). These results also demon-

strate that flank-induced suppression is not associated

with visual acuity (Tables 1A and B).In control experiments we found that flank-induced

suppression in non-control observers was relatively im-

mune to changes in scale and size (Fig. 6). Amblyopic

observers showed flank-induced suppression for 2 c/deg

stimuli at similar target to flank separations (4k and

5k) as the 1 c/deg stimuli. The overlap of the 2 c/deg data

(representing 2 and 2.5� separations) and 1 c/deg data

(representing 4 and 5� separations) suggests that the sup-pressive effects are invariant to spatial frequency (scale)

and size. Physiologically, this suggests the presence of

A

B

Fig. 4. Suppression by collinear and orthogonal flanks occurs at 4k–6k separations. Amblyopic eye (A) and preferred eye (B) data from four

amblyopic observers. Data are for 1 c/deg stimuli at 4k–8k separations and error bars represent ± SEM. Only amblyopic observer DS could be tested

with flank contrast at 4· flank contrast detection threshold. The gray line represents the facilitation shown by control eyes (weighted average).

E.H. Wong et al. / Vision Research 45 (2005) 2810–2819 2815

different size CRF-surround mechanisms in V2, operat-

ing over a large range of cortical distances.

4. Discussion

We report novel visual deficits in amblyopic observ-ers, and strabismic observers with no loss of visual acu-

ity, which result from anomalous interactions between

spatially isolated, second-order visual stimuli. Specifical-

ly, under conditions where normal (control) observers

show enhanced visual performance, non-control observ-

ers show a very different pattern of spatial interaction: in

virtually every case, visual sensitivity is compromised by

the presence of nearby stimuli. Taken together, the

results of this study, when considered in light of known

physiology of second-order processing, represent an

amblyopic deficit at an early stage of extra-striate visual

processing (V2).

Almost all amblyopic observers showed suppressionin each eye, both amblyopic and preferred, often of al-

most equal magnitude. Furthermore, strabismic observ-

ers RC and WS showed suppression in both eyes (Fig.

3), without any loss of visual acuity or second-order sen-

sitivity in either eye (Fig. 5 and Table 1B). Our finding

of bilateral second-order deficits should perhaps have

A

B

Fig. 5. Flank effect is not associated with absolute second-order

contrast sensitivity. Data are from (A) amblyopic eyes (AE) and (B)

fellow non-amblyopic eyes (NAE). For each eye, threshold modulation

ratio for collinear flanks (filled symbols) and orthogonal flanks (open

symbols) is plotted against relative sensitivity for the isolated target.

Relative sensitivity = (contrast detection threshold)/(average contrast

detection threshold for five normal observers). For observers RC and

WS (both strabismic without loss of visual acuity), their non-dominant

eye is represented as AE and the dominant eye as NAE.

Fig. 6. Suppressive flank effect is size and scale invariant. Data are

from two amblyopic observers. Stimuli are 1 c/deg (filled symbols) and

2 c/deg (open symbols) and error bars represent ± SEM.

2816 E.H. Wong et al. / Vision Research 45 (2005) 2810–2819

been expected, given the predominantly binocular nat-

ure of V2 neurons (Hubel & Livingstone, 1987).In normal observers, collinear and orthogonal flanks

facilitated second-order contrast detection thresholds by

approximately 12% (Fig. 2). This value is somewhat

lower than that found for first-order stimuli (up to

50%) (Levi et al., 2002; Polat & Sagi, 1993; Yu et al.,

2002). A similar relationship has been shown for con-

trast discrimination: the reduction of apparent (supra-

threshold) contrast by flanking stimuli is significantlyless for second-order cues than for first-order cues

(Ellemberg, Allen, & Hess, 2004). Previous studies

examining the degree and nature of flank-induced facil-

itation for first-order visual stimuli present a somewhat

contradictory picture. Some studies suggest that facilita-

tion is orientation dependent (and is greatest for collin-

ear flanks and targets) (Polat & Sagi, 1993), whilst

others posit orientation independent mechanisms (Leviet al., 2002; Yu et al., 2002). The results of the present

study show qualitative agreement with the latter class

of first-order studies where the magnitude of facilitation

is unrelated to the internal image structure of isolated

patches. Therefore, facilitation of contrast detection

thresholds, by both first-order and second-order flank-

ing stimuli, appears to involve a process that pools

information across a broad range of oriented filters.Physiologically, this suggests that horizontal connec-

tions exist between neurons with a range of different ori-

entation preferences, rather than being restricted to

connecting common orientations. Our results suggest

orientation independence: both collinear and orthogo-

nal flanks produced nearly equal facilitation in control

eyes (Fig. 2), and nearly equal suppression in the major-

E.H. Wong et al. / Vision Research 45 (2005) 2810–2819 2817

ity of amblyopic eyes as well as the fellow preferred eyes

(Fig. 3). Furthermore, Ellemberg et al. (2004) reported

that in normal observers, flank orientation effects on

contrast discrimination were more broadly tuned for

second-order stimuli than for first-order stimuli. For

the large flank to target separations used in this study,we hypothesize that our results indicate a CRF-sur-

round mechanism in which the surround input to the

CRF is broadly tuned to orientation.

We used spatially isolated stimuli and thus did not

measure the previously reported suppression effects pro-

duced at small target–flank separations (1k–2k) when

using first-order stimuli (Polat & Sagi, 1994). Suppression

results from the overlap of supra-threshold flanks thateffectively transform the task from one of contrast detec-

tion, to contrast discrimination. The latter task produces

contrast increment thresholds that are greater (poorer)

than the corresponding contrast detection thresholdmea-

sured at wider flank–target separations. Our finding that

spatially isolated stimuli up to 4�–6� produced suppres-

sion in non-control observers and facilitation in normal

observers possibly suggests unequal CRF sizes betweenthe two groups. However, we have recently conducted a

spatial summation experiment using identical stimuli

(Wong & Levi, in press) and found that amblyopic and

normal observers show similar performance, i.e., a simi-

lar improvement in contrast detection threshold as target

size increases (up to 8� at 4r). This suggests that the spa-tial extent of the CRF is similar in non-control and nor-

mal observers. In total, it is more probable that thefindings in the present study reflect the influence of the

flanking stimuli on the balance between the CRF-sur-

round gains, rather than an elaborate recruitment of hor-

izontal connections over large cortical distances (but see

Stettler et al., 2002). Furthermore, the inhibitory sur-

round in non-control observers appears to exert a greater

influence than that seen in normal observers.

Through psychophysical measures we can only infer,based on contemporary neurophysiology (see Section 1)

that the amblyopic deficits found in this study reflect

neural deficits in V2. An alternative or contributory

mechanism could be deficient reentrant connections to

V2 from higher-order visual areas; connections analo-

gous to those found to V1 (for a review see, Angelucci

& Bullier, 2003). A feedback mechanism is especially rel-

evant to this study in light of evidence that second-ordercues stimulate neurons in primate MT (O�Keefe &

Movshon, 1998) and multiple extra-striate areas in hu-

mans via fMRI (Mendola, Dale, Fischl, Liu, & Tootell,

1999; Smith, Greenlee, Singh, Kraemer, & Hennig,

1998; but see Nishida, Sasaki, Murakami, Watanabe,

& Tootell, 2003). Furthermore, greater influence from

feedback mechanisms may have occurred due to our

use of 500 ms stimulus presentations.Another possible explanation for our results is that

the imbalance between excitation and inhibition reflects

an imbalance in the excitatory and inhibitory inputs to

second-order neurons from first-order (V1) neurons

(Morgan & Baldassi, 1997). However, we note that

our amblyopic observers show these abnormalities at

low spatial frequencies (1 and 2 c/deg) where amblyopic

observers show normal facilitation with first-order stim-uli of similar spatial frequency (Polat et al., 2004, 2005).

Moreover, our use of static, random noise as the carrier

in the second-order stimuli greatly reduced the likeli-

hood that side-band spatial frequencies (first-order

structures produced by the modulation of the carrier)

contributed to the interactions between the second-order

stimuli. That is, prominent side-bands could elicit first-

order spatial interactions (abnormal in amblyopicobservers) that could then be passed forward to the sec-

ond-order mechanism (i.e., filter-rectify-filter model).

However, past studies show that side-bands, and their

adjacent spatial frequencies, are more likely to be signif-

icant in second-order stimuli that contain a grating car-

rier of high contrast rather than a noise carrier (of any

contrast) (Dakin & Mareschal, 2000; Jamar, Campagne,

& Koenderink, 1982).Our hypothesis, of altered CRF-surround gain mech-

anisms in second-order neurons, predicts that in normal

observers, flanks disinhibit the (tonic) surround suppres-

sion. Disinhibition is plausible because the flanks were

present in each trial (500 ms each) of the two-alternative

design, thereby acting as an almost constant mask. In

this situation, the contrast detection threshold reflects

the state of contrast adaptation of the inhibitory sur-round. A reduction in inhibition produces an increase

in the CRF gain and a net facilitation of visual sensitiv-

ity. Psychophysically, the increase in CRF gain is dem-

onstrated by an improvement in contrast detection

threshold. In the first-order domain, improved contrast

sensitivity following adaptation to flanks has been

shown by Ejima and Takahashi (1985). We speculate

that an analogous mechanism could occur for second-order stimuli.

Our hypothesis also predicts that for non-control

observers that showed suppression, the flanks did not

completely disinhibit the (dominant) surround. The net

effect of surround stimulation on the CRF would be

inhibition, or less excitation, relative to normal observ-

ers. Psychophysically, the inhibition of the CRF is dem-

onstrated by an increase (deterioration) in contrastdetection threshold. This finding appears to be specific

to second-order stimuli in light of evidence that first-

order stimuli of low spatial frequency, like that used in

our study, produce nearly equal amounts of flank-in-

duced facilitation in amblyopic and normal observers

(Polat et al., 2004, 2005). A stronger inhibitory surround

could represent a shift in the balance of horizontal excit-

atory and inhibitory connection strength betweensecond-order neurons, i.e., a functional increase in inhi-

bition between second-order neurons.

2818 E.H. Wong et al. / Vision Research 45 (2005) 2810–2819

Exactly this type of inhibition-dominated network is

supported by the results of amblyope DS who showed

substantially less suppression in both eyes when flank

contrast was doubled (the other amblyopic observers

could not be tested at four times flank detection con-

trast). The visual performance of normal observers, onthe other hand, remained unchanged under identical

conditions (Figs. 2 and 4). The reduced suppression ef-

fect shown by amblyope DS could be explained by the

mechanism of short-term plasticity—in this case further

disinhibition resulting in an unmasking of subthreshold

excitatory connections to supra-threshold activity. The

limited contrast range for second-order flanks prevented

other amblyopic observers from being tested at 4· detec-tion contrast. The presentation of stimuli for greater

than 500 ms may have also produced further disinhibi-

tion, i.e., greater contrast adaptation of the inhibitory

surround; however, we did not test this condition.

Although the horizontal connections and CRF-sur-

round mechanism in V2 are largely unexplored, an inhi-

bition-dominated network in amblyopia appears likely

based on evidence from V1. Synaptic weighting has itsbasis in the reciprocal connections between columns of

similar orientation in striate cortex (Weliky, Kandler,

Fitzpatrick, & Katz, 1995) and the balance of neural in-

put (push-pull arrangement) is modeled to determine the

contrast gain of single neurons (Carandini, Heeger, &

Movshon, 1997). In strabismus, anomalous horizontal

connections are found following abnormal visual experi-

ence during development in cat (Lowell & Singer, 1992)and monkey (Tyschen & Burkhalter, 1995), and may ex-

plain the less robust synchronization between neurons

responsive to the amblyopic eye in cat (Roelfsema,

Konig, Engel, Sireteanu, & Singer, 1994). Moreover,

inhibition dominated horizontal connections between

both ocular dominance columns and orientation col-

umns are thought to underlie the interocular suppres-

sion seen in strabismus (Sengpiel & Blakemore, 1996).These evidences lead us to speculate the existence of

analogous mechanisms in V2.

In summary, we found that spatially isolated, second-

order stimuli produced facilitatory interactions in nor-

mal observers but suppressive interactions in each eye

of amblyopic observers and strabismic observers with

no loss of visual acuity. Based on contemporary neuro-

physiology, our results suggest an early, higher-orderprocessing deficit in amblyopia. We further speculate

that early abnormal visual experience may result in pre-

dominantly inhibitory cortical networks.

Acknowledgments

Portions of the results of this study were presented inabstract form at the 2001 Association for Research in Vi-

sion and Ophthalmology Meeting. This study was sup-

ported by NIH NEI grants K23EY14261 (E.H.W.) and

RO1EY01728 (D.M.L.), and a Wellcome Trust (UK)

Research Career Development Fellowship to P.V.M.

References

Angelucci, A., Levitt, J. B., Walton, E. J. S., Hupe, J. M., Bullier, J., &

Lund, J. S. (2002). Circuits for local and global signal integration in

primary visual cortex. Journal of Neuroscience, 22, 8633–8646.

Angelucci, A., & Bullier, L. (2003). Reaching beyond the classical

receptive field of V1 neurons: Horizontal or feedback axons?

Journal of Physiology—Paris, 97, 141–154.

Bakin, J. S., Nakayama, K., & Gilbert, C. G. (2000). Visual responses

in monkey areas V1 and V2 to three-dimensional surface config-

urations. Journal of Neuroscience, 20, 8188–8198.

Carandini, M., Heeger, D. J., & Movshon, J. A. (1997). Linearity and

normalization of simple cells of the macaque primary visual cortex.

Journal of Neuroscience, 17, 8621–8644.

Cavanaugh, J. R., Bair, W., & Movshon, J. A. (2002a). Nature and

interaction of signals from the receptive field center and surround

in macaque V1 neurons. Journal of Neurophysiology, 88,

2530–2546.

Cavanaugh, J. R., Bair, W., & Movshon, J. A. (2002b). Selectivity and

spatial distribution of signals from the receptive field surround in

macaque V1 neurons. Journal of Neurophysiology, 88, 2547–2556.

Chubb, C., & Sperling, G. (1988). Drift-balanced random stimuli: A

general basis for studying non-Fourier motion perception. Journal

of the Optical Society of America A, 5, 1986–2007.

Dakin, S. C., & Mareschal, I. (2000). Sensitivity to contrast modula-

tion depends on carrier spatial frequency and orientation. Vision

Research, 40, 311–329.

Ejima, Y., & Takahashi, S. (1985). Effect of localized grating

adaptation as a function of separation along the length axis

between test and adaptation areas. Vision Research, 25, 1701–1707.

Ellemberg, D., Allen, H. A., & Hess, R. F. (2004). Investigating local

network interactions underlying first- and second-order processing.

Vision Research, 44, 1787–1797.

Hubel, D. H., & Livingstone, M. S. (1987). Segregation of form, color,

and stereopsis in primate area 18. Journal of Neuroscience, 7,

3378–3415.

Jamar, J. H. T., Campagne, J. C., & Koenderink, J. J. (1982).

Detectability of amplitude- and frequency-modulation of supra-

threshold sine-wave gratings. Vision Research, 25, 407–416.

Kapadia, M. K., Ito, M., Gilbert, C. D., & Westheimer, G. (1995).

Improvement in visual sensitivity by changes in local context:

Parallel studies in human observers and in V1 of alert monkeys.

Neuron, 15, 843–856.

Kapadia, M. K., Westheimer, G., & Gilbert, C. D. (1999). Dynamics

of spatial summation in primary visual cortex of alert monkeys.

Proceedings of the National Academy of Sciences USA, 96,

12073–12078.

Kiorpes, L., & McKee, S. P. (1999). Neural mechanisms underlying

amblyopia. Current Opinion in Neurobiology, 9, 480–486.

Knierman, J. K., & Van Essen, D. C. (1992). Neuronal response to

static texture patterns in area V1 of the alert macaque monkey.

Journal of Neurophysiology, 67, 961–980.

Leventhal, A. G., Wang, Y., Schmolesky, M. T., & Zhou, Y. (1998).

Neural correlates of boundary perception. Visual Neuroscience, 15,

1107–1118.

Levi, D. M., Hariharan, S., & Klein, S. A. (2002). Suppressive and

facilitatory spatial interactions in amblyopic vision. Vision

Research, 42, 1379–1394.

Levitt, J. B., & Lund, J. S. (1997). Contrast dependence of contextual

effects in primate visual cortex. Nature, 387, 73–76.

E.H. Wong et al. / Vision Research 45 (2005) 2810–2819 2819

Lowell, S., & Singer, W. (1992). Selection of intrinsic horizontal

connections in the visual cortex by correlated neuronal activity.

Science, 255, 209–212.

Mareschal, I., & Baker, C. L. Jr., (1998). Temporal and spatial

response to second-order stimuli in cat area 18. Journal of

Neurophysiology, 80, 2811–2823.

Mendola, J. D., Dale, A. M., Fischl, B., Liu, A. K., & Tootell, R. B.

(1999). The representation of illusory and real contours in human

visual areas revealed by functional magnetic resonance imaging.

Journal of Neuroscience, 19, 8560–8572.

Morgan, M. J., & Baldassi, S. (1997). How the human visual system

encodes the orientation of a texture, and why it makes mistakes.

Current Biology, 7, 999–1002.

Nelson, J. I., & Frost, B. J. (1985). Intracortical facilitation among co-

oriented, co-axially aligned simple cells in cat striate cortex.

Experimental Brain Research, 61, 54–61.

Nishida, S., Sasaki, Y., Murakami, I., Watanabe, T., & Tootell, R. B.

(2003). Neuroimaging of direction-selective mechanisms for sec-

ond-order motion. Journal of Neurophysiology, 90, 3242–3254.

O�Keefe, L. P., & Movshon, J. A. (1998). Processing of first- and

second-order motion signals by neurons in area MT of the

macaque monkey. Visual Neuroscience, 15, 305–317.

Pelli, D. G., & Zhang, L. (1991). Accurate control of contrast on

microcomputer displays. Vision Research, 31, 1337–1350.

Polat, U., & Sagi, D. (1993). Lateral interactions between spatial

channels: Suppression and facilitation revealed by lateral masking

experiments. Vision Research, 33, 993–999.

Polat, U., & Sagi, D. (1994). The architecture of perceptual spatial

interactions. Vision Research, 34, 73–78.

Polat, U., Sagi, D., & Norcia, A. M. (1997). Abnormal long-range

spatial interactions in amblyopia. Vision Research, 37, 737–744.

Polat, U., Mizobe, K., Pettet, M. W., Kasamatsu, T., & Norcia, A. M.

(1998). Collinear stimuli regulate visual responses depending on

cell�s contrast threshold. Nature, 391, 580–583.

Polat, U., Ma-Naim, T., Belkin, M., & Sagi, D. (2004). Improving

vision in adult amblyopia by perceptual learning. Proceedings of the

National Academy of Sciences USA, 101, 6692–6697.

Polat, U., Bonneh, Y., Ma-Naim, T., Belkin, M., & Sagi, D. (2005).

Spatial interactions in amblyopia: Effects of stimulus parameters

and amblyopia type. Vision Research, 45, 1471–1479.

Roelfsema, P. R., Konig, P., Engel, A. K., Sireteanu, R., & Singer, W.

(1994). Reduced synchronization in the visual cortex of cats with

strabismic amblyopia. European Journal of Neuroscience, 6,

1645–1655.

Schofield, A. J., & Georgeson, M. A. (1999). Sensitivity to modu-

lations of luminance and contrast in visual white noise: Separate

mechanisms with similar behavior. Vision Research, 39,

2697–2716.

Schofield, A. J., & Georgeson, M. A. (2003). Sensitivity to contrast

modulation: the spatial frequency dependence of second-order

vision. Vision Research, 43, 243–259.

Sengpiel, F., & Blakemore, C. (1996). The neural basis of suppression

and amblyopia in strabismus. Eye, 10, 250–258.

Smith, A. T., Greenlee, M. V., Singh, K. D., Kraemer, F. M., &

Hennig, J. (1998). The processing of first- and second-order

motion in human visual cortex assessed by functional magnetic

resonance imaging (fMRI). Journal of Neuroscience, 18,

3816–3830.

Stettler, D. D., Das, A., Bennett, J., & Gilbert, C. D. (2002). Lateral

connectivity and contextual interactions in macaque primary visual

cortex. Neuron, 36, 739–750.

Tyschen, L., & Burkhalter, A. (1995). Neuroanatomic abnormalities of

primary visual cortex in macaque monkeys with infantile esotropia:

Preliminary results. Journal of Pediatric Ophthalmology and Stra-

bismus, 32, 323–328.

von der Heydt, R., & Peterhans, E. (1984). Illusory contours and

cortical neuron responses. Science, 224, 1260–1262.

von der Heydt, R., & Peterhans, E. (1989). Mechanisms of contour

perception in monkey visual cortex. I. Lines of pattern disconti-

nuity. Journal of Neuroscience, 9, 1731–1748.

Walker, G. A., Ohzawa, I., & Freeman, R. D. (1999). Asymmetric

suppression outside the classical receptive field of the visual cortex.

Journal of Neuroscience, 19, 10536–10553.

Weliky, M., Kandler, K., Fitzpatrick, D., & Katz, L. C. (1995).

Patterns of excitation and inhibition evoked by horizontal

connections in visual cortex share a common relationship to

orientation columns. Neuron, 15, 541–552.

Willis, A., Smallman, H. S., & Harris, J. M. (2000). Comparing

contrast-modulated and luminance modulated masking: Effects of

spatial frequency and phase. Perception, 29, 81–100.

Wong, E. H., & Levi, D. M. Second-order spatial summation in

amblyopia. Vision Research, in press, doi:10.1016/j.visres.

2005.05.020.

Wong, E. H., Levi, D. M., & McGraw, P. V. (2001). Is second-order

spatial loss in amblyopia explained by the loss of first-order spatial

input? Vision Research, 41, 2951–2960.

Yu, C., Klein, S. A., & Levi, D. M. (2002). Facilitation of contrast

detection by cross-oriented surround stimuli and its psychophysical

mechanisms. Journal of Vision, 2, 243–255.

Zhou, Y. X., & Baker, C. L. Jr., (1994). Envelope-responsive neurons

in areas 17 and 18 of cat. Journal of Neurophysiology, 72,

2134–2150.