Current Biology, Vol. 15, 2142–2148, December 6, 2005, ª2005 Elsevier Ltd All rights reserved. DOI 10.1016/j.cub.2005.10.066

Suppressed Patterns Alter Visionduring Binocular Rivalry

Joel Pearson1,* and Colin W.G. Clifford1

1Colour, Form and Motion LabVisual Perception UnitSchool of PsychologyThe University of SydneySydney, NSW 2006Australia

Summary

Binocular rivalry occurs when incongruent patternsare presented to corresponding regions of the retinas,

leading to fluctuations of awareness between the pat-terns [1]. One attribute of a stimulus may rival whereas

another may combine between the eyes [2–5], but itis typically assumed that the dominant features are

perceived veridically. Here, we show this is not neces-sarily the case and that a suppressed visual feature

can alter dominant perception. The cortical represen-tations of oriented gratings can interact even when

one of them is perceptually suppressed, such that theperceived orientation of the dominant grating is sys-

tematically biased depending on the orientation of thesuppressed grating. A suppressed inducing pattern

has the same qualitative effect as a visible one, butsuppression reduces effective contrast by a factor of

around six. A simple neural model quantifies and helpsexplain these illusions. These results demonstrate that

binocular rivalry suppression operates in a gradedfashion across multiple sites in the visual hierarchy

rather than truncating processing at a single site andthat suppressed visual information can alter dominant

vision in real-time.

Results and Discussion

Despite a multitude of recent developments, a great dealstill remains to be learned about how the primate brainconstructs and maintains an ongoing representation ofthe surrounding environment [6]. The phenomenon ofbinocular rivalry (BR) allows us to study cortical changesin isolation from changes in the environment, thus isolat-ing the neural machinery responsible for perturbations inawareness. BR is induced when dissimilar images areprojected to corresponding regions on the two retinas[1]. BR has been used in an variety of ways as a tool toinvestigate the neural concomitants of visual awareness[6, 7]. It has typically been thought that when one stimu-lus is dominant and the other suppressed, perceptionclosely resembles physically switching off the sup-pressed stimulus. However, in some instances one attri-bute of a stimulus may rival whereas another may becombined between the two eyes [2–5]. Here, we showthat a single attribute, orientation, can simultaneouslybe suppressed due to BR and interact with the dominant

*Correspondence: joelp@psych.usyd.edu.au

pattern, changing how we perceive the orientation of thedominant pattern.

Neuronal orientation selectivity is a property that doesnot emerge until primary visual cortex [8]. For this rea-son, illusions of orientation have provided a valuablemethod of establishing how much information can beprocessed cortically in the absence of awareness [9–13].The perceived orientation of a pattern can be altered bythe orientation of a simultaneously presented super-imposed or surrounding pattern [14–16]. Here, we inves-tigated the simultaneous effect of the orientation of asuppressed pattern on the perceived orientation of thedominant pattern during BR with intermittently pre-sented stimuli (Figures 1A and 1B). We find that a singlevisual feature (orientation) can simultaneously be sup-pressed and modify perception of the dominant stimu-lus. A suppressed inducing pattern has qualitatively thesame effect as a visible one; however, suppression re-duces its effective contrast by a factor of around six.

When rivaling images are presented intermittently, therate of perceptual alternation is drastically lowered suchthat the percept tends to remain the same as in the pre-vious presentation [17]. We made use of this perceptualstability to ensure that the same stimulus was dominantthroughout a block of 60 trials. Each block was sufficientto provide a measure of the magnitude of the illusion ofperceived orientation in the dominant stimulus inducedby the suppressed stimulus. Stimuli were sinusoidalgratings (see Experimental Procedures for details) pre-sented for 500 ms followed by 3000 ms of blank back-ground (Figure 1C). During the blank interruption, sub-jects reported whether the orientation of the dominantgrating was clockwise or counterclockwise from vertical(a single-interval, two-alternative forced-choice task). Ifdominance switched to the other grating, or if the per-cept was piecemeal, observers were instructed to abortthat block of trials. However, this was the case for onlya few blocks.

Figure 2A shows data from four subjects for super-posed test and inducing gratings. When the suppressedinducer was oriented at angles between 10º and 30º tothe vertical, the vertical dominant grating was perceivedtilted toward the suppressed inducer (attraction). Whenthe angle of the inducer was larger than 30º the dominantgrating was perceived to be tilted away from the inducer(repulsion). The form of this angular tuning function issimilar to other superposed orientation illusions [14].To obtain a quantitative comparison between the rivalingand nonrivaling illusions, we used a version of the simul-taneous tilt illusion [15] in which the inducing surroundwas suppressed by a colored static noise pattern (Fig-ure 1B). Here, the results can be directly compared tothe dichoptic tilt illusion [10, 18] in which test and sur-round are again simultaneously presented to oppositeeyes but are both visible, thus isolating any effect thatsuppression has on the tilt illusion [19, 20]. The proce-dure in this experiment was otherwise the same as inthe first experiment. Figure 2B shows the magnitude ofthe suppressed surround tilt illusion as a function of theorientation of the suppressed surround grating. The

Suppressed Patterns Alter Vision2143

Figure 1. Illusions of Orientation during Binocular Rivalry

(A) Schematic of the superposed orientation illusion (experiment 1). The images presented to the two eyes are depicted in the left and middle col-

umns, the resulting percept in the right column. The presence of the suppressed grating alters the perceived orientation of the dominant grating.

(B) Schematic of the center-surround orientation illusion (experiment 2) in the same format. Colored static noise suppresses the surround grating,

but the surrounding grating still effects the perceived orientation of the central grating in the other eye.

(C) Illustration of the experimental timeline. The upper two rows show the stimulus timeline for the two eyes, the lower row depicts an observer’s

percepts. On the initial presentation the left eye stimulus comes on 500ms earlier than the right to force a dominance switch: flash suppression (see

methods). Observers signal the orientation of the preceding dominant grating during each blank 3 s interval.

suppressed surrounding orientation had primarily arepulsive effect on the perceived orientation of the dom-inant central patch. The form of the tuning functionhere is similar to that normally observed for the tilt illu-sion [15].

To learn if binocular rivalry suppression weakened thetilt illusion, we compared the magnitude of the sup-pressed surround tilt illusion to that of the dichoptic tiltillusion as a function of the contrast of the inducing sur-round [21]. Figure 2C shows the difference in magnitudebetween the tilt illusion with a suppressed inducer andthe tilt illusion without any suppression (the rival staticcolored noise stimulus removed) when the surround ori-entation was 620º (to the vertical). For low contrast sur-rounds (5%–30%) the magnitude of the tilt illusion gen-erated by a suppressed inducer was markedly smallerthan that of the interocular illusion. However, at highercontrast (50%), the magnitudes of the two tilt illusionswhere similar. Figure 2D shows the average tilt illusionfrom the three subjects as a function of surround con-trast. In the nonrivaling condition, the tilt illusion reacheshalf of its maximum value at 4.6% contrast, whereas inthe suppressed condition the illusion does not reachhalf maximum until 24.4% contrast. Hence, althoughthe magnitude of both versions of the tilt illusion testedhere varies with the contrast of the surround, the illusionwith the suppressed surround requires five to six timesmore contrast to reach an equivalent magnitude.

A simple neural model can help us understand andexplain the illusions of orientation discussed here. A

population of monocular, orientation-selective modelneurons processes the stimulus from each eye. Themagnitude of the response of any given neuron is mod-eled by a difference of Gaussian excitation and inhibitionfunctions (see the Experimental Procedures for details).Signals from these two neural populations interact ata subsequent binocular level. In the superposed stimu-lus condition, the population response at the binocularlevel is modeled as a weighted sum of the two monocu-lar response distributions. The perceived orientation isthen taken to be the preferred orientation of the moststrongly responding neuron in the binocular population:a winner-take-all coding scheme.

When the orientations of the two stimuli are sufficientlyclose, the peak of the resultant binocular response distri-bution is shifted away from the orientation of the domi-nant stimulus and attracted toward that of suppressedstimulus (Figure 3A). If the orientations of the two stimuliare sufficiently different, the effect of the suppressedstimulus is to reduce the response to the dominant stim-ulus, shifting the peak response away from the orienta-tion of the suppressed stimulus (Figure 3B). The quanti-tative predictions of the model are shown in Figure 3Calong with the data from two of the subjects. The relativeweightings of the two monocular response distributionsused to produce the curves in Figure 3C were 0.14, indi-cating a 7-fold reduction in signal strength as a result ofsuppression, similar to that measured for the tilt illusion.

The same model accounts for the form of the angulartuning function of the center-surround tilt illusion with

Current Biology2144

Figure 2. Illusions Measured as a Function of Inducer Orientation and Inducer Contrast

(A) Illusion magnitude for the superposed condition as a function of the orientation of the suppressed inducer.

(B) Data from the suppressed surround tilt illusion. Error bars show the standard error within subjects.

(C) Data from the suppressed surround (solid lines) and nonrivaling (dashed lines) tilt illusions as a function of the surround grating contrast.

Illusion magnitude is plotted as a function of the contrast of the inducer. Error bars show standard errors within subjects.

(D) For the suppressed surround (circles) and nonrivaling (squares) conditions, the measured tilt illusion from experiment 3 averaged across sub-

jects was fitted with a function of the form:

IðcÞ= Imax

�cn

cn + cn50

�;

where I(c) is the fitted tilt illusion as a function of inducer contrast, Imax is the maximum value of the illusion, c50 is the contrast at which the illusion

reaches half its maximum value, and the exponent n determines the slope of the function. Because the data from the suppressed condition did

not adequately constrain Imax, the following fitting procedure was adopted. First, the data from the nonrivaling condition were fitted. The value of

Imax from this fit was then used to constrain the fit to the suppressed data.

the following auxiliary assumption. The excitatory com-ponent of the population response to the surround grat-ing is assumed to be confined to the cortical represen-tation of the corresponding region of visual space,whereas the inhibitory component is assumed to extendin an attenuated fashion to the cortical representation ofthe central region. Since the effect of the surround on thecenter is purely inhibitory, any resulting tilt illusion is in-evitably repulsive (Figure 3D). The quantitative predic-tions of the model are shown in Figure 3E.

We have shown that a perceptually suppressed pat-tern can alter the contents of visual awareness. The

orientation content of the suppressed pattern can sys-tematically bias the perceived orientation in the domi-nant pattern in qualitatively the same way as a nonsup-pressed pattern. However suppression reduces theeffective contrast of that pattern by a factor of aroundsix. A simple neural model provides a quantitative un-derstanding of these results in terms of crosstalk be-tween the two patterns undergoing BR.

It has been shown that visual cortex can adapt to ori-entation while it is suppressed from awareness due tobinocular rivalry [22]. However, in the current experi-ments the suppressed orientation interacts with the

Suppressed Patterns Alter Vision2145

Figure 3. Neural Model and Quantitative Predictions

(A) Illustration of the behavior of the model for the superposed condition (experiment 1). The top left/right panel depicts the response of a pop-

ulation of monocular model neurons to the dominant/suppressed grating as a function of the preferred orientation of the neuron. When the ori-

entations of the two stimuli are sufficiently close, the peak of the resultant binocular response distribution is shifted away from the orientation of

the dominant stimulus (solid arrow) and attracted toward that of suppressed stimulus (dotted arrow).

(B) When orientations of the two stimuli are sufficiently different, the effect of the suppressed stimulus is to reduce the response to the dominant

stimulus, shifting the peak response away from the orientation of the suppressed stimulus (repulsion).

(C) Quantitative predictions of the model (solid line) with the following parameters (sE = 24º, sI = 36º, a = 1.09, w = 0.14) and psychophysical data

(circles and squares represent data from observers JP and CC, respectively) from experiment 1. The value of w is important because it quantifies

the relative strength of the inputs from the representations of the dominant and suppressed stimuli. The value of 0.14 used to model the exper-

imental data indicates a 7-fold reduction in effective signal strength as a result of suppression.

(D). In the center-surround condition, the effect of the surround is purely inhibitory so any resulting tilt illusion is repulsive.

(E) Model predictions (solid line) and psychophysical data (circles and squares) from experiment 2. For the center-surround configuration, w0 was

0.09, somewhat smaller than the value of 0.14 used to model interactions between superposed stimuli.

dominant in real time, the effect is not driven by adapta-tion. This demonstrates not only that the suppressedstimulus is processed at a cortical level, as is the caseduring adaptation, but that the suppressed stimulus isconcurrently altering the perception of the dominantstimulus.

Previous work has shown some attributes of a stimuluscan combine between the eyes while at the same timeothers rival [2–5]. However, the attraction effect ob-served in the current study cannot be explained by fusionof the two gratings’ orientations while their colors rival. Atno time did subjects report any plaid-like percepts [23] ora grating slanted in depth. Instead, the percept was of asingle grating oriented in the fronto-parallel plane, dem-onstrating that not only was the color undergoing binoc-ular rivalry but also the orientation information. Further-more, we found that a significant attraction effectremained when the spatial frequency of one grating wasraised to twice that of the other to avoid any possibility ofstereoscopic fusion (data not shown). Indeed, when dif-ferent oriented patterns are superimposed such thatboth are clearly discernable, perceptual attraction

results without any stereoscopic depth perception [14].This demonstrates that the reported perceptual attrac-tion from superposed orientations is not a result of ste-reoscopic fusion.

The results of the current study are diagnostic regard-ing the site and mechanisms of binocular rivalry sup-pression in the visual processing hierarchy. Four hypo-thetical functional architectures are illustrated in Figure4. If the effect of suppression were to truncate the pro-cessing of the suppressed stimulus prior to any interac-tion with the representation of the dominant stimulusthen there would be no illusion of orientation with thesuppressed superposed or surround stimuli (Figure 4A).Conversely, if the effect of suppression were simply totruncate processing of the suppressed stimulus subse-quent to the site of interaction with the representation ofthe dominant stimulus then suppression would not at-tenuate the tilt illusion at all (Figure 4B). Thus, we canconclude that binocular rivalry is not mediated at a singlesite early or late in the visual processing hierarchy.

Previous evidence suggests that the tilt illusion maybe the product of interactions at multiple sites along

Current Biology2146

Figure 4. Four Hypothetical Functional Archi-

tectures

(A) If the effect of suppression were to trun-

cate the processing of the suppressed stimu-

lus (S) prior to any interaction with the domi-

nant stimulus (D), there would be no illusion

of orientation.

(B). If the effect of suppression were simply to

truncate processing of the suppressed stim-

ulus subsequent to the site of interaction (I)

with the representation of the dominant stim-

ulus, the tilt illusion would be unaffected by

suppression.

(C) Binocular rivalry suppression could trun-

cate processing of the suppressed stimulus at

a stage of processing intermediate between

the interactions mediating the tilt illusion.

We would then expect the effect of suppres-

sion to compress the illusion versus contrast

function vertically, as illustrated in (D).

(E) Alternatively, suppression could be medi-

ated in a graded fashion across two or more

levels of processing, as depicted here.

(F) The effect of suppression in such an archi-

tecture would be to scale the effective con-

trast of the inducing stimulus, shifting the illu-

sion versus log-contrast curve rightwards.

the form processing pathway [10, 24, 25]. Thus it is con-ceivable that binocular rivalry suppression could trun-cate processing of the suppressed stimulus at a stageof processing intermediate between the stages of inter-action mediating the tilt illusion (Figure 4C). We wouldthen expect the suppression to reduce, but not abolish,the tilt illusion, as we observed experimentally. However,in this case we might reasonably expect the effect of thesurround at each stage of interaction to be a monotoni-cally increasing function of contrast, at least for sur-round contrasts less than or equal to that of the test(as used here). If the effect of suppression were to elim-inate interactions between the representation of thedominant and suppressed stimulus beyond some pointin the processing hierarchy then this would be evidentas a vertical compression of the illusion versus contrastfunction (Figure 4D). Just such a pattern of results wasobserved previously when backward masking of the sur-round was used to truncate processing of the tilt illusion[10]. However, in the current experiments the effect ofsuppression was a lateral rather than a downward shiftof the illusion versus contrast function (Figure 2D).

Having established that a single site of binocular rivalrysuppression either early (Figure 4A), late (Figure 4B),or intermediate (Figures 4C and 4D) in the processinghierarchy cannot account for the lateral shift of the func-tion relating the strength of the tilt illusion to surroundcontrast found experimentally, we must consider thepossibility that suppression is mediated in a gradedfashion across two or more levels of processing (Figure4E). In this case, the initial stage(s) of suppression is as-sumed to weaken the effective strength of the signal rep-resenting the suppressed stimulus, but not to truncateits processing altogether. Such a functional architecturewould result in a reduction of the effective contrast ofthe signal from the surround that interacts with the

representation of the dominant stimulus, shifting the illu-sion versus log-contrast function laterally as illustratedin Figure 4F.

Thus, we can conclude that binocular rivalry is medi-ated at multiple sites along the processing hierarchy [7,26]. This supports findings from single-cell electrophys-iology showing that the proportion of neurons whosepattern of firing correlates with modulations in visualawareness during BR increases as the visual processinghierarchy is ascended [27–29]. However, it was only re-cently that theories regarding binocular rivalry proposedthat suppression took place at a single location either atan early sensory level or a higher level [7]. Although sub-sequent studies have suggested that different stimulusparadigms can induce rivalry at different levels of the vi-sual processing hierarchy [7, 26], we have demonstratedhere that interactions between rivaling stimuli can occurat multiple levels simultaneously.

The multiple level model of binocular rivalry proposedhere (Figure 4E) can account for the dissociation be-tween measures of rivalry suppression showing rela-tively modest increases in increment thresholds [33, 34]and the striking phenomenological impression of com-plete disappearance of the suppressed stimulus. It is as-sumed that detection of increment thresholds can becarried out by mechanisms at a level in the visual hierar-chy intermediate to the initial stage of attenuation of therepresentation of the suppressed stimulus and the siteat which processing of the stimulus is truncated. If theresponses of mechanisms at this intermediate stage ofprocessing are not accessible to visual awareness [10]then the subsequent truncation of processing will pre-vent conscious perception of the suppressed stimulus.

If, as indicated, rivalry is a fundamentally distributedprocess along the visual hierarchy, this prompts thequestion of how suppression across multiple sites is

Suppressed Patterns Alter Vision2147

coordinated. If competition between the representationsof the rivaling stimuli early in the processing hierarchy at-tenuates the effective signal strength of one more thanthe other, this could bias competition at subsequentlevels [30, 31]. In addition, it is possible that feedbackof information from higher levels of processing servesto boost the effective signal strength of the dominantstimulus and attenuate that of the suppressed stimulus[31, 32].

Our use of intermittent rivalry to investigate interac-tions between dominant and suppressed vision shouldprove useful in future studies of the neural correlatesof visual awareness. Our results show that even if a visualpattern is suppressed from vision, the information con-tained in that pattern can simultaneously infiltrate visualawareness via another stimulus. In particular, sup-pressed form information can simultaneously alter thedominant visual percept. It follows that the classical un-derstanding of BR under which the dominant stimulusattributes are perceived veridically warrants revision.

Experimental Procedures

Stimuli

Stimuli were sinusoidal gratings (Figures 1A and 1B) generated us-

ing Matlab software to drive a VSG 2/5 Graphics Card (Cambridge

Research Systems), displayed on a g-corrected 21’’ Sony Trinitron

GM 520 monitor (1024 3 768 resolution; 120 Hz refresh rate) and

viewed through a mirror stereoscope adjusted for each observer.

This ensured that there was no optical cross talk between the two

rivaling stimuli. All gratings had a spatial frequency of 0.5 cycles

per degree. The gratings in the superposed condition and the central

patch in the surround condition had a diameter of 3º. In the center

surround condition the annulus had an external diameter of 6.2º

and an internal diameter of 3º. In all experiments, a bulls eye fixation

point that subtended 0.5º of visual angle was used to aid in conver-

gence in addition to circular fusion locks surrounding the stimulus.

The average luminance of all stimuli was equal to that of the back-

ground (10.7 Cd m22). One grating was green while the other was

red. The color coordinates for red were CIE, x = 0.585, y = 0.376;

and green, CIE, x = 0.413, y = 0.512. The gratings were defined by

a sinusoidal change in luminance across space. The contrast of all

gratings was 50%, except the surround annulus in experiment 3

whose contrast ranged from 2.5%–50%. The background color sur-

rounding the stimulus was an average luminance and chromaticity

of both colors in the stimulus. To suppress the oriented annulus in

experiment 2 (see Figure 1B), static colored noise was presented

to the corresponding location in the opposing eye at 100% contrast.

The static colored noise consisted of square elements subtending

0.075º x 0.075º (2 pixels 3 2 pixels).

Subjects and Procedure

Four subjects took part in experiments 1 and 2 (two subjects were

naive to the purpose of the study) and three subjects (two were naive

to the purpose of the study) in experiment 3. Subjects were seated in

a darkened room and used a chin rest to stabilize their heads. To

control the dominant pattern at the outset of each block of trials,

we used flash suppression. This consists of the asynchronous onset

of the patterns presented one to each eye [35]. One grating is pre-

sented to one eye, joined 500 ms later by the opposing grating in the

other eye. As the second pattern is presented perception switches

to it. Flash suppression thus enabled us to choose one of the gra-

tings to be dominant at the outset of each block of trials. No re-

sponse was made on this first trial. Once one pattern was dominant,

due to the intermittent presentation, that pattern almost invariably

remained dominant for the subsequent presentations. Even chang-

ing the suppressed pattern across trials did not disrupt perceptual

stability due to the intermittent presentation, and subjects were

clearly instructed to abort that block of trials if rivalry ever became

piecemeal (incomplete) or if dominance switched to the inducer

grating. In different blocks of trials the orientation of the suppressed

inducer was varied from 10º–80º. Subjects signaled the orientation

of the dominant grating using a button box. The task was single-in-

terval, two-alternative forced-choice clockwise or counterclockwise

from vertical. Two interleaved adaptive psychophysical procedures

[36] of 30 trials each (25 in experiment 3) were used to estimate the

orientation of the dominant grating at which it was perceived to be

vertical. Each procedure displayed an inducer of opposite orienta-

tion 6Qº from vertical in a randomized order to avoid the build up

of orientation-selective adaptation over time. For each block, the

psychometric estimation of perceived tilt was based on half the dif-

ference of the outcome of the two procedures. Data collection took

place in blocks of 60 trials (50 in experiment 3), and estimates of the

illusion of perceived orientation were the average from at least 4

blocks.

Model

A population of monocular, orientation-selective model neurons

processes the stimulus from each eye. The magnitude of the re-

sponse, R, of any given neuron in the population is modeled by a dif-

ference of Gaussian excitation and inhibition functions:

R(qstim; qpref ) = a exp

�2

�qstim 2 qpref

sE

�2�2 exp

�2

�qstim 2 qpref

sI

�2�

where qstim is the orientation of the stimulus, qpref is the preferred ori-

entation of the neuron, sE and sI control the spread of excitation and

inhibition, respectively, and a determines their relative weighting.

In the superposed stimulus condition, the population response at

the binocular level, Rbinoc, is modeled as a weighted sum of the two

monocular response distributions:

Rbinoc(qpref ) = Rdom(qdom; qpref ) + wRsupp(qsupp; qpref );

where qdom and qsupp are the orientations of the dominant and sup-

pressed stimuli, Rdom and Rsupp are the response distributions of

the populations representing them at the monocular level, and w is

the relative strength of the input from the representation of the sup-

pressed stimulus. The perceived orientation is modeled as the pre-

ferred orientation of the most strongly responding neuron in the bin-

ocular population: a winner-take-all coding scheme.

In the center-surround stimulus condition, the excitatory compo-

nent of the population response to the surround grating is assumed

to be confined to the cortical representation of the corresponding re-

gion of visual space, while the inhibitory component is assumed to

extend in an attenuated fashion to the cortical representation of

the central region. Thus, in the center-surround condition, the re-

sponse of the population representing the central test region is given

by:

Rbinoc(qpref ) = Rdom(qdom; qpref ) 2 w0 exp

�2

�qsupp 2 qpref

sI

�2�;

where w0 is weight of the inhibitory effect of the surround on the cen-

ter. Although the population response at the binocular level was

modeled as a linear combination of the monocular population re-

sponses, we observed qualitatively similar behavior when multipli-

cative interactions were implemented [37, 38].

Acknowledgments

This work was supported by a Queen Elizabeth II Fellowship and

Discovery Project Grant awarded to CC by the Australian Research

Council and an Australian Postgraduate Award held by J.P.

Received: August 8, 2005

Revised: September 17, 2005

Accepted: October 17, 2005

Published: December 5, 2005

References

1. Wheatstone, C. (1838). On some remarkable, and hitherto unob-

served, phenomena of binocular vision. Philos. Trans. R. Soc.

Lond. B Biol. Sci. 128, 371–394.

Current Biology2148

2. Andrews, T.J., and Blakemore, C. (2001). Form and motion have

independent access to consciousness. Nat. Neurosci. 2, 405–

406.

3. Carlson, T., and He, S. (2000). Visible binocular beats from invis-

ible monocular stimuli during binocular rivalry. Curr. Biol. 10,

1055–1058.

4. Treisman, A. (1962). Binocular rivalry and stereoscopic depth

perception. Quart. J. Exp. Psych. 14, 23–37.

5. Creed, R.S. (1935). Observations on binocular fusion and rivalry.

J. Physiol. 84, 381–392.

6. Koch, C. (2004). The Quest for Consciousness: A Neurobiologi-

cal Approach (Englewood, Colorado: Roberts).

7. Blake, R., and Logothetis, N.K. (2002). Visual competition. Nat.

Rev. Neurosci. 3, 13–21.

8. Hubel, D.H., and Wiesel, T.N. (1968). Receptive fields and func-

tional architecture of monkey striate cortex. J. Physiol. 1, 215–

243.

9. Blake, R., and Fox, R. (1974). Adaptation to ‘invisible’ gratings

and the site of binocular rivalry suppression. Nature 249, 488–

490.

10. Clifford, C.W.G., and Harris, J.A. (2005). Contextual modulation

outside of awareness. Curr. Biol. 15, 574–578.

11. He, S., Cavanagh, P., and Intriligator, J. (1996). Attentional reso-

lution and the locus of visual awareness. Nature 263, 34–37.

12. He, S., and MacLeod, D. (2001). Orientation-selective adaptation

and tilt aftereffect from invisible patterns. Nature 411, 473–476.

13. Rajimehr, R. (2004). Unconscious orientation processing. Neu-

ron 41, 663–673.

14. Blake, R., Holopigian, K., and Jauch, M. (1985). Another visual il-

lusion involving orientation. Vision Res. 25, 1469–1476.

15. Gibson, J.J., and Radner, M. (1937). Adaptation, after-effect and

contrast in the perception of tilted lines. J. Exp. Psychol. 20,

453–467.

16. Virsu, V., and Taskinen, H. (1975). Central inhibitory interactions

in human vision. Exp. Brain Res. 23, 65–74.

17. Leopold, D.A., Wilke, M., Maier, A., and Logothetis, N.K. (2002).

Stable perception of visually ambiguous patterns. Nat. Neuro-

sci. 5, 605–609.

18. Walker, P. (1978). Orientation-selective inhibition and binocular

rivalry. Perception 7, 207–214.

19. Rao, M. (1977). Tilt illusion during binocular rivalry. Vision Res.

17, 327–328.

20. Wade, N.J. (1980). The influence of colour and contour rivalry on

the magnitude of the tilt illusion. Vision Res. 20, 229–233.

21. Blake, R., and He, S. (2005). Adaptation as a tool for probing the

neural correlates of visual awareness: Progress and precau-

tions. In Fitting the Mind to the World: Aftereffects in High-Level

Vision, C.W.G. Clifford, and G. Rhodes, eds. (Oxford, UK: Oxford

University Press), p. 281.

22. Wade, N.J., and Wenderoth, P. (1978). The influence of colour

and contour rivalry on the magnitude of the tilt aftereffect. Vision

Res. 18, 827–835.

23. Wolfe, J. (1983). Influence of spatial frequency, luminance, and

duration on binocular rivalry and abnormal fusion of briefly pre-

sented dichoptic stimuli. Perception 12, 447–456.

24. Clifford, C.W.G., Spehar, B., Solomon, S.G., Martin, P.R., and

Zaidi, Q. (2003). Interactions between colour and luminance in

the perception of orientation. J. Vis. 3, 106–115.

25. Wenderoth, P., and Johnstone, S. (1987). Possible neural sub-

strates for orientation analysis and perception. Perception 16,

693–709.

26. Wilson, H.R. (2003). Computational evidence for a rivalry hierar-

chy in vision. Proc. Natl. Acad. Sci. USA 100, 14499–14503.

27. Leopold, D.A., and Logthetis, N.K. (1996). Activity changes in

early visual cortex reflect monkeys’ percepts during binocular

rivalry. Nature 379, 549–553.

28. Logothetis, N.K., and Schall, J.D. (1989). Neuronal correlates of

subjective visual perception. Science 245, 761–763.

29. Sheinberg, D.L., and Logothetis, N.K. (1997). The role of tempo-

ral cortical areas in perceptual organization. Proc. Natl. Acad.

Sci. USA 94, 3408–3413.

30. Tong, F., and Engel, S.A. (2001). Interocular rivalry revealed in

the human cortical blind-spot representation. Nature 411, 195–

199.

31. Polonsky, A., Blake, R., Braun, J., and Heeger, D.J. (2000). Neu-

ronal activity in human primary visual cortex correlates with per-

ception during binocular rivalry. Nat. Neurosci. 11, 1153–1159.

32. Watson, T.L., Pearson, J., and Clifford, C.W.G. (2004). Percep-

tual grouping of biological motion promotes binocular rivalry.

Curr. Biol. 14, 1670–1674.

33. Fox, R., and Check, R. (1972). Independence between binocular

rivalry suppression duration and magnitude of suppression.

J. Exp. Psychol. 93, 283–289.

34. Wales, R., and Fox, R. (1970). Increment detection thresholds

during binocular rivalry suppression. Percept. Psychophys. 8,

90–94.

35. Wolfe, J. (1984). Reversing ocular dominance and suppression

in a single flash. Vision Res. 24, 471–478.

36. Kontsevich, L.L., and Tyler, C.W. (1999). Bayesian adaptive esti-

mation of psychometric slope and threshold. Vision Res. 39,

2729–2737.

37. Carandini, M., and Heeger, D.J. (1994). Summation and division

by neurons in primate visual cortex. Science 264, 1333–1336.

38. Wilson, H.R., and Humanski, R. (1993). Spatial frequency adap-

tation and contrast gain control. Vision Res. 33, 1133–1149.