Available online at www.sciencedirect.com

www.elsevier.com/locate/visres

Vision Research 48 (2008) 63–69

What does the illusory-flash look like?

David McCormick *, Pascal Mamassian

Laboratoire Psychologie de la Perception, CNRS & Universite Paris Descartes, 45 rue des Saints-Peres, 75270 Paris Cedex 06, France

Received 21 February 2007; received in revised form 8 October 2007

Abstract

In the illusory-flash effect (Shams, L., Kamitani, Y., & Shimojo, S. (2000). Illusions. What you see is what you hear. Nature, 408, 788),one flash presented with two tones has a tendency to be seen as two flashes. Previous studies of this effect have been ill-equipped to estab-lish whether this illusory-flash is the result of a genuine percept, or that of a shift in criterion. We addressed this issue by using a stimuluscomprising two locations. This enabled contrast-threshold measurement by means of a location detection task. High-contrast white orblack flashes were presented simultaneously to both locations, followed by threshold contrast flashes of the same contrast polarity at thetwo locations in half of the trials; observers reported whether or not the low-contrast flashes had been present. Irrelevant to the task, halfof the trials contained one tone, the other half contained two tones. In this way, we were able to compute the change in sensitivity andshift in criterion between illusory and non-illusory trials. We observe both a decrease in visual sensitivity and a criterion shift in the illu-sory-flash conditions. In a second experiment, we were interested in determining whether this change in visual sensitivity gave rise tomeasurable visual attributes of the illusory-flash. If it has a contrast, it should interact with a spatio-temporally concurrent real flash.Using a similar two-location stimulus presentation, we found that under certain conditions, we were able to infer the polarity of the per-ceived illusory-flash. We conclude that the illusory-flash is indeed a perceptual effect with psychophysically assessable characteristics.� 2007 Elsevier Ltd. All rights reserved.

Keywords: Multisensory perception; Audio–visual; Illusory-flash

1. Introduction

It is well established that what we perceive visually isreadily influenced by our auditory perception, and viceversa. If cues from both modalities are temporally synchro-nous, the nervous system has been shown to categorise thecues as emanating from the same external event at bothneural (Meredith, Nemitz, & Stein, 1987) and behavioural(Bertelson & Radeau, 1981; Sekuler, Sekuler A. B., & Lau,1997; Slutsky & Recanzone, 2001; Watanabe & Shimojo,2001) levels. Low level audio–visual integration has alsobeen shown to be influenced by the spatial location of thecues from the two modalities (Alais & Burr, 2004; Meyer,Wuerger, Rohrbein, & Zetzsche, 2005). Multisensory‘binding’ has been demonstrated in the presence of a tem-poral disparity between visual and auditory signals

0042-6989/$ - see front matter � 2007 Elsevier Ltd. All rights reserved.doi:10.1016/j.visres.2007.10.010

* Corresponding author. Fax: +33 142863322.E-mail address: david.mccormick@univ-paris5.fr (D. McCormick).

(McDonald, Teder-Salejarvi, & Ward, 2001; Meredithet al., 1987). The ‘multisensory temporal integration win-dow’ defines the potential extent of such a disparity thatcan still facilitate such cross-modal perception. A robustand highly persuasive demonstration of how audition andvision influence each other can be seen in the McGurkeffect (McGurk & MacDonald, 1976). Here, lip and tonguemovements presented with incongruous auditory informa-tion can bring about a misperception of the auditory signal.This interaction between hearing and vision in speech per-ception indicates that it involves information from morethan one sensory modality. Perhaps the most striking evi-dence for sensory trading in healthy individuals comesfrom synaesthesia. In this condition, involuntary conscioussensations from one sensory modality can be induced bystimuli from another modality. For example, in sound–col-our synaesthesia, individuals experience colours inresponse to some aspect of a musical stimulus, such aspitch or timbre. Furthermore, Ward, Huckstep, and Tsak-

64 D. McCormick, P. Mamassian / Vision Research 48 (2008) 63–69

anikos (2006) showed that there are consistent trendsamong such synaesthetes to experience higher pitchednotes as more brightly coloured. In light of this and thepresence of similar patterns of pitch-brightness matchingin non-synaesthetic subjects, they posit that this form ofsynaesthesia shares mechanisms with non-synaesthetes.

It is clear that different sensory modalities do not neces-sarily process perceptual information separately and it is ofinterest to establish how, and under what circumstances,such multi-modal interactions take place.

A useful strategy for disambiguating the contributionsof individual modalities in multimodal perception is toput sensory modalities in conflict with one another in situ-ations where they would normally concur (as with theMcGurk effect). Shams et al. (2000) demonstrated that asingle flashed disc, presented with two brief tones, has atendency to be perceived as two flashes. This findingillustrated that vision does not always win when visionand audition are in conflict. They went on to investigatethis effect using the brain imaging techniques EEG (Bhat-tacharya, Shams, & Shimojo, 2002; Shams, Kamitani,Thompson, & Shimojo, 2001) and MEG (Shams, Iwaki,Chawla, & Bhattacharya, 2005). The MEG study showedthat activity in the visual cortical areas could be modulatedby sound, with modification of activity in occipital, parie-tal and anterior areas. The EEG studies indicated signifi-cantly higher oscillatory and induced gamma bandresponses in illusory than non-illusory trials, as well assignificant supra-additive audio–visual interactions inillusory trials. Furthermore, the brain potentials for theillusory-flashes were similar to those for a genuine flash.These studies connote neurophysiological correlates tothe perception of the illusion; the processing of informa-tion in visual cortical regions can be modulated by sound.It is not clear, however, that a visual cortical area responseto an auditory stimulus implies a perceptual event havingbeen ‘seen’.

Other research in this area has demonstrated that, undercertain conditions, an equivalent but opposite ‘fusion’effect is also possible; one tone presented with two flashescan give rise to the perception of only one flash (Andersen,Tiippana, & Sams, 2004). This ‘illusory-flash effect’, plusanalogous visuo-haptic (Ernst, Banks, & Bulthoff, 2000)and audio-haptic (Violentyev, Shimojo, & Shams, 2005)effects, have recently received considerable attention inthe literature, both from experimental and modelling(Andersen, Tiippana, & Sams, 2005; Ernst & Banks,2002; Shams, Iwaki, Chawla, & Bhattacharya, 2005a;Shams, Ma, & Beierholm, 2005b) perspectives. Much ofthe research into this effect has purported to show thatthe illusion is the result of a genuine perceptual effect, sug-gesting that the illusory-flash has characteristics of appear-ance akin to those of real flashes. As yet, however, nopsychophysical study has sought to answer questions per-taining to the specifics of what one might perceive at thetime of the illusion. Here, we are interested in establishingwhether the sound merely creates confusion, or whether it

evokes illusory-flashes with visual characteristics similar tothose resulting from genuine visual stimulation—a percep-tual event.

We created a new experimental set-up in which wewanted to present two simultaneous illusory-flashes attwo separate screen locations. In our first experiment,we were interested in two aspects of the illusory-flash asthey pertained to our design. First, to measure the efficacyof the method in generating illusory-flashes. Second, andmore importantly, to ascertain whether the effect’s inci-dence was the product of more than a shift in observers’criteria alone. That is, did the presence of two tones pre-sented with one flash bring about a change in visual sen-sitivity? We found that not only was our methodproficient in generating illusory-flashes, but that trials inwhich illusory-flashes were reported were associated withboth a shift in observers’ criterion and a decrease in visualsensitivity.

If the illusory-flash has a measurable contrast, it shouldinteract with a real flash presented at the same time andlocation. If this is the case, and if real and illusory flasheswere to occur simultaneously, we would expect to see anenhanced detectability of a real flash with the same contrastpolarity as that of an illusory-flash. Conversely, a flash ofopposite contrast to the illusory-flash would diminish indetectability. With this premise in mind and the confirmedeffectiveness of our presentation method in experiment 1,we devised a second experiment. In this second experiment,one location contained only one high-contrast flash, whilethe other contained a high-contrast flash followed by alow-contrast flash. Observers were then required to indi-cate which of the two locations contained two flashes, thatis, which of the two locations contained a low-contrastflash. We were then able to infer what, if any, influencethe presence of an illusory-flash had on the perceived con-trast of a real flash by determining the difference in observ-ers’ thresholds in the one-tone (non-illusory) and two-tone(illusory) conditions.

2. Experiment 1

2.1. Methods

2.1.1. Subjects

Eight observers (5 male, 3 female, aged 23–35) with nor-mal or corrected-to-normal vision participated withoutpayment in experiment 1.

2.1.2. Apparatus

Stimuli were presented via an Apple PowerMac G4computer on a 2100 Sony FD Trinitron CRT monitor at arefresh rate of 75 Hz. The spatial resolution of the displaywas 1024 · 768 pixels. Observers were positioned at a dis-tance of 57 cm from the monitor, maintained by a chin-rest. The experiment was conducted in a dark room. Stim-uli were generated using MATLAB and the Psychophysics

D. McCormick, P. Mamassian / Vision Research 48 (2008) 63–69 65

Toolbox (Brainard, 1997; Pelli, 1997). Tones were pre-sented through Sennheiser HD280 headphones.

1 We report the result of an exact binomial test because there is noguarantee that data are normally distributed. t-Test results for the samedata are: white first flash, t(14) = 3.7, p < .05; black first flash, t(14) = 3.3,p < .05.

2.1.3. Design

Experiment 1 comprised two parts: first, two prelimin-ary sessions were conducted for each observer to obtaintheir contrast detection thresholds. Second, each observerran 10 blocks of trials in the main experiment. Two squarelocations comprising mid-level grey, each measuring fourdegrees of visual angle across, were embedded in a 25%contrast grey-level checkerboard pattern. The inner edgesof the locations were separated horizontally by 10� visualangle. Observers were required to maintain fixation on across midway between the two locations. In the two preli-minary sessions, observers’ thresholds for determining thelocation (left or right) of a low-contrast flash presentedafter a high-contrast flash were measured. These sessionscomprised two-interleaved staircases of stochastic approx-imation (Robbins & Monro, 1951, as cited in Treutwein,1995) converging on 75% correct location detection. Whiteand black first flashes were separated into these two ses-sions and the low-contrast flashes had the same contrastpolarity as the high-contrast flashes. First, a high-contrastflash of 2 frames (26 ms) was presented simultaneously toeach location. Second, an inter-stimulus interval (ISI) of4 frames (52 ms) was presented (mid-level grey at bothlocations). Third, a low-contrast flash of 2 frames (26 ms)was presented at one of the two locations (left or right)while the other location remained mid-level grey. The loca-tion of the low-contrast flash was determined pseudo-ran-domly. The background outside of the checkerboardpattern was mid-level grey. In these sessions, one tone(3750 Hz, 26 ms) was always presented simultaneous withthe low-contrast flash. The task in these preliminary ses-sions was to determine at which of the two locations thelow-contrast flash occurred; observers were required topress the left cursor key if it occurred at the left locationor the right cursor key if it occurred at the right location.No feedback was given.

These 75% correct location detection threshold valueswere then used as the low-contrast flashes in the mainexperiment. Observers ran 10 blocks of trials and alter-nate blocks contained white and black high-contrastflashes. High-contrast flashes were presented at bothscreen locations simultaneously. These were followed bylow-contrast flashes (of the same contrast polarity) in halfof the trials. When they occurred, the low-contrast flasheswere present in both screen locations. Half of the trialscontained one tone, half contained two tones (3750 Hz,26 ms). When only one tone was present, it occurredsimultaneous with the low-contrast flashes. When twotones were presented, they occurred simultaneous withthe two flashes. Fig. 1 depicts a diagrammatic representa-tion of the stimulus set-up. Observers reported whetherlow-contrast flashes had been present on a trial usingthe computer keyboard. d’ and criterion (c) values were

computed for the four combinations of high-contrast flashcolour and number of tones.

2.2. Results

In signal detection terms, a false alarm is an instancewhere an observer reports having seen something whichwas not physically present. In this experiment, an illu-sory-flash is evidenced by the presence of a false alarm.That is, when observers reported having seen two flasheswhen only one was present. Our data contain two sets ofsuch false alarms: the first when only one tone was present(FA1); the second when two tones were present (FA2). Theillusory-flash as described in the literature is the result of anaudio–visual interaction when tones outnumber flashes.Therefore, an accurate measurement of the proportion ofillusory-flashes for any given observer is the differencebetween these two sets of false alarm rates:FA2 � FA1 = D.

Using this method, we find a higher false alarm rate inthe two-tone than one-tone trials, but that the colour ofthe initial flash (white or black) has no bearing on the num-ber of illusory-flashes reported: D = 0.15 in white-first-flashtrials and 0.14 in black-first-flash trials. We used an exactbinomial test (Hollander & Wolfe, 1973) to establish thatFA2 was significantly greater than FA1 in the white firstand black first flash conditions. In both instances, exactbinomial p (one-tailed) < .05.1 This illustrates that ourmethod was indeed effective in generating illusory-flashesin approximately 15% of trials.

Signal detection analysis reveals no differences in sensi-tivity (d 0) or criterion (c) between black- and white-first-flash trials. However, in both of these conditions, d 0 wasreduced—from 1.6 to 0.8 across observers—when a flashwas accompanied by two tones relative to when only onetone was presented: F(1, 30) = 18.8, p < .05. This indicatesthat the presence of illusory-flash inducing tones reducesvisual sensitivity. We also see an accompanied shift in cri-terion—from 1.1 to 0.5 across observers—between thesetwo conditions indicating the illusion pertains to more thanjust visual sensitivity: F(1, 30) = 45.1, p < .05. Fig. 2 is areceiver operating characteristic (ROC) diagram with hitrates plotted against false alarm rates for black/white firstflashes with one/two tones. This figure indicates that d 0 isreduced by approximately 1 in the trials containing twotones. It also shows that there is very little difference inthe proportions of hits and false alarms between the white-and black-first-flash trials. ROC diagrams of the individualobservers’ data are available as supplementary material,online.

We thus established the efficacy of our method in gener-ating illusory-flashes. We also witnessed the second tones’

Visual Auditory

1000ms

26ms

52ms

26ms

>300ms

26ms

26ms

Fig. 1. Stimulus set-up and timeline for the main part of experiment 1. One tone was always presented simultaneous with the low-contrast flashes. Half ofthe trials also contained a tone simultaneous with the high-contrast flashes. Low-contrast flashes had the same contrast polarity as the preceding high-contrast flashes: white high-contrast flashes with lighter than mid-level grey low-contrast flashes are shown.

0

0.2

0.4

0.6

0.8

1

0 0.2 0.4 0.6 0.8 1

N = 8

False Alarm Rate

Hit

Rat

e

2 Tone, Black 1st Flash2 Tone, White 1st Flash

1 Tone, Black 1st Flash1 Tone, White 1st Flash

d' = 2

d' = 1

d' = 0

Fig. 2. A receiver operating characteristic (ROC) diagram with hit ratesplotted against false alarm rates, averaged across observers with standarderrors. Filled symbols represent black-first-flash trials, open symbolsrepresent white-first-flash trials with one (triangles)/two (squares) tones.Dashed lines represent iso-sensitivity curves for d 0 values of 0, 1 and 2.

66 D. McCormick, P. Mamassian / Vision Research 48 (2008) 63–69

effect on observers’ visual sensitivities and criteria. Itshould be noted that signal detection theory does not makeany claims on a responder’s conscious state and hence isnot qualified to permit an interpretation of the data interms of what a responder has ‘perceived’. Subsequentlywe were interested to know if this lowered visual sensitivitywould be manifested in a perceptual event. That is, is thereany extent to which we can measure what observers see atthe time of the illusory-flash?

3. Experiment 2

3.1. Methods

3.1.1. Subjects

Twelve graduate student observers (7 male, 5 female,aged 23–35) with normal or corrected-to-normal visionparticipated without payment in the two conditions ofexperiment 2.

3.1.2. Apparatus

Apparatus for experiment 2 was the same as experiment 1.

3.1.3. Design

Stimulus presentation (potential locations and timing)remained similar to the preliminary sessions in experiment1. A low-contrast flash always occurred after two initialsimultaneous high-contrast flashes, and was presented toone location only (varying from trial to trial, determinedpseudo-randomly) while the opposite location remainedat mid-level grey. The low-contrast flash could have eithercontrast polarity. White and black first flashes were sepa-rated into two conditions—condition 1: white; condition2: black. Observers ran 10 blocks of trials per condition,each of which comprised two-interleaved staircases con-verging on 75% (threshold) correct. Alternate blocks con-tained one or two tones with the same timing andfrequency as those of experiment 1. Threshold detectionvalues were averaged over the 10 blocks for the four com-binations of low-contrast flash polarity and number oftones. In each condition (white first flash/black first flash)there were four types of trial; the number of beeps (1/2)was blocked, while the contrast polarity (above/belowmid-level grey) of the low-contrast flash varied pseudo-ran-domly from trial to trial.

0.2

0.25

0.3

0.35

Low-Contrast Flash PolarityDark Light

TA

Con

trast

Thr

esho

ld

2Tones1 Tone

White 1st Flash

-0.1

-0.05

0

0.05

0.1

-0.1 -0.05 0 0.05 0.1Difference in Dark Flash Contrast Thresholds

Diff

eren

ce in

Lig

ht F

lash

Con

trast

Thr

esho

lds White 1st Flash

a b

Fig. 3. (a) Mean absolute threshold contrast detection values for a typical observer with white first flashes—split into light/dark low-contrast flashes withstandard errors across blocks. Filled discs represent two-tone thresholds, open discs represent one-tone thresholds. (b) Summary plot of the 12 observers’data with white first flashes. Each observer’s mean thresholds for the one-tone trials are subtracted from those of the two-tone trials for light/dark low-contrast flashes. The diagonal (bottom left to top right) therefore represents a dimension along which there is no difference between the change in light/dark low-contrast flashes when in the presence or absence of two tones. Means/standard errors across observers shown.

D. McCormick, P. Mamassian / Vision Research 48 (2008) 63–69 67

3.2. Results

In condition 1, the high-contrast flashes were white.Fig. 3(a) shows absolute threshold contrast detection val-ues for a typical observer. These data are split intoinstances where the low-contrast flash was either lighteror darker than mid-level grey.

Fig. 3(b) is a summary plot of the 12 observers’ data.This was generated by subtracting the threshold valuesfor the one-tone trials from those of the two-tone trialsfor both the light and dark low-contrast flashes. The diag-onal (bottom left to top right) therefore represents a dimen-sion along which there is no difference between the changein light and dark second flashes when in the presence orabsence of two tones. The observers’ data tend to lie belowthis diagonal, indicating that the presence of two tones(and thus an illusory-flash) facilitates the detection of a reallight flash, but not that of a dark one. We conducted anexact binomial test of the probability of this configurationof points around the diagonal.2 The test indicates that wecan discount the null hypothesis that this arrangement hap-pened by chance and that a significantly greater proportionof points lie below the diagonal (exact binomial p (two-tailed) < .05). From this result, we infer that when illu-sory-flashes were preceded by high-contrast white flashes,they had a tendency to be perceived as light flashes.

2 t-Test results for the same data are: condition 1, t(22) = 2.7, p < .05;condition 2, t(22) = 0.1, p = .89.

In condition two, the high-contrast flashes were black.Fig. 4(a) again shows data for a typical observer, and4(b) shows a summary of all observers’ data. In conditionone, where initial flashes were white, the results point to theperception of light illusory-flashes. In condition two, wherethe initial flashes were black, we might expect to see theopposite effect—dark illusory-flashes. In this condition,however, there is little deviation of the data from the diag-onal and the exact binomial test indicates an insignificantdeviation in either direction (p = .77).

4. Discussion

The experiments reported here lend further support tothe assertion that the illusory-flash is at least partlyexplained in terms of a genuine perceptual effect and notsimply the product of artefacts (Shams, Kamitani, & Shim-ojo, 2002). It should be noted, however, that a significantproportion of the effect, classically described, is almost cer-tainly attributable to an increased willingness (in a non-visual decision) to report having perceived a greater numberof flashes than were presented when they occurred in thepresence of a greater number of concomitant tones (McCor-mick & Mamassian, 2005). This can be explained in terms ofthe multi-sensory integration window—illusory-flashes willtend to occur when flashes and tones are not separated bymore than approximately 100 ms (Shams et al., 2000)—cou-pled with the fact that the auditory system has a higher tem-poral resolution than that of the visual system. We have

0.2

0.25

0.3

0.35

Low-Contrast Flash PolarityDark Light

TA

Con

trast

Thr

esho

ld

2 Tones1 Tone

Black 1st Flash

-0.1

-0.05

0

0.05

0.1

-0.1 -0.05 0 0.05 0.1

Black 1st Flash

Difference in Dark Flash Contrast ThresholdsD

iffer

ence

in L

ight

Fla

sh C

ontra

st T

hres

hold

s

a b

Fig. 4. (a) Mean absolute threshold contrast detection values for a typical observer with black first flashes—split into light/dark low-contrast flashes withstandard errors across blocks. Filled discs represent two-tone thresholds, open discs represent one-tone thresholds. (b) Summary plot of the 12 observers’data with black first flashes; details as per Fig. 3(b).

68 D. McCormick, P. Mamassian / Vision Research 48 (2008) 63–69

illustrated here, however, that this criterion shift cannot beexclusively responsible for the effect; a change in visualsensitivity must also be taken into account.

In experiment 1, we have demonstrated that the illusory-flash effect is attributable to both a decrease in visual sen-sitivity when tones outnumber flashes and a simultaneousshift in observers’ criterion as to what constitutes a flash.This is in line with results from the haptic illusory-flashstudy of Violentyev et al. (2005).

Previous studies of the illusory-flash have tended to besusceptible to response bias. Because we are interested ingenuine visual attributes of this effect, by using a locationdetection task, experiment 2 measures along a differentdimension than ‘present/absent’ and overcomes biasesinherent in ‘yes/no’ tasks. Using this paradigm, we didnot measure the prevalence of the illusory-flash directly,but were able to infer its properties nevertheless.

In the two-tone trials in both experiments, we envisagedthere being an effect of the first tone cueing the timing ofthe second tone, and hence the onset of the low-contrastflash. This was a necessary asymmetry between the twotypes of trials (one tone/two tone). The absence of a sensi-tivity change between one- and two-tone trials in experi-ment 2, condition 2, however, renders it unlikely thatcueing affected the results in either experiment. In experi-ment 2, our results elucidate a trend suggesting that theillusory-flash can have a measurable contrast: lighter-than-grey low-contrast flashes have a tendency to increasein salience in the presence of simultaneous tones when theyare preceded by high-contrast white flashes. This indicatesthat, in this instance, the illusory-flash is of the same

contrast polarity of contrast as the flash which precedesit. However, this effect is evident only when the low-con-trast flash was preceded by a high-contrast white flash (dar-ker-than-grey flashes are not better detected following ahigh-contrast black flash).

We witnessed a relatively small proportion of illusory-flashes in the data from experiment 1 compared to previousstudies of this effect. This is most likely attributable to twofactors. First, the inherent assumption that illusory-flasheswill be created at both locations. While we have no reasonto believe that this is not the case, it may have increaseddoubt for observers and hence caused them to choose amore conservative estimate of whether or not two flasheswere indeed present. Second, observers had perceptualaccess to a genuine (albeit threshold) real flash in half ofthe trials. It is likely that this point of reference would alsoreduce willingness to report having seen two flashes in one-flash trials.

Our signal detection data from experiment 1 indicatethat illusory-flashes preceded by real white or black flashesinvolve the same perceptual processes. The asymmetry inthe data from experiment 2 could be attributable to differ-ences between the physical symmetry (on a calibratedscreen) and the perceptual symmetry of what constitutesthe mid-point between black and white; this seems unlikely,however, given the lack of asymmetry witnessed betweenblack first flash and white first flash trials in exper-iment 1. Furthermore, experiment 2 was not attuned toestablishing which two-tone trials were perceived as indeedcontaining illusory-flashes (experiment 1 indicates thatapproximately only 15% did). Thus, a genuine effect in

D. McCormick, P. Mamassian / Vision Research 48 (2008) 63–69 69

the data could well have been contaminated by the neces-sary inclusion of those trials which were not perceived asillusory.

We have illustrated that the sound induced illusory-flashis a product of both a shift in observers’ criteria and achange in visual sensitivity. Furthermore, we have shownthat under certain conditions, inferences regarding the per-ceptual nature of the illusory-flash can be made. Our datareject a model whereby the illusory-flash is the results of alower threshold for after-image detection (because thiswould predict perceived illusory-flashes of the oppositecontrast polarity to the real high-contrast flash). They alsoappear to reject a model in which the illusory-flash isalways the same contrast polarity as the initial inducingflash. Our paradigm is attuned to measuring the perceivedcontrast of the illusory-flash; further study using a similartechnique could establish whether other visual attributes(for example shape and orientation) are equally susceptibleto influence via the illusory-flash.

Acknowledgements

We would like to thank Tobias Andersen, Simon Bart-helme and Andrei Gorea for their help. This research wasfunded by a chaire d 0excellence from the French Ministryof Research.

Appendix A. Supplementary data

Supplementary data associated with this article can befound, in the online version, at doi:10.1016/j.visres.2007.10.010.

References

Alais, D., & Burr, D. (2004). The ventriloquist effect results from nearoptimal crossmodal integration. Current Biology, 14, 257–262.

Andersen, T., Tiippana, K., & Sams, M. (2004). Factors influencingaudiovisual fission and fusion illusions. Cognitive Brain Research, 21,301–308.

Andersen, T., Tiippana, K., & Sams, M. (2005). Maximum likelihoodintegration of rapid flashes and beeps. Neuroscience Letters, 380,155–160.

Bertelson, P., & Radeau, M. (1981). Cross-modal bias and perceptualfusion with auditory–visual spatial discordance. Perception & Psycho-

physics, 29, 578–584.

Bhattacharya, J., Shams, L., & Shimojo, S. (2002). Sound-induced illusoryflash perception: Role of gamma band responses. NeuroReport, 13,1727–1730.

Brainard, D. H. (1997). The psychophysics toolbox. Spatial vision, 10,433–436.

Ernst, M. O., & Banks, M. S. (2002). Humans integrate visual and hapticinformation in a statistically optimal fashion. Nature, 415, 429–433.

Hollander, M., & Wolfe, D. A. (1973). Nonparametric statistical inference.New York: John Wiley and Sons, pp. 15–22.

McCormick, D., & Mamassian, P. (2005). Response biases in the illusory-flash effect (Abstract). Journal of Vision, 5(8), 878, 878a.

Ernst, M. O., Banks, M. S., & Bulthoff, H. H. (2000). Touch can changevisual slant perception. Nature Neuroscience, 3, 69–73.

McGurk, H., & MacDonald, J. W. (1976). Hearing lips and seeing voices.Nature, 264, 746–748.

McDonald, J. J., Teder-Salejarvi, W. A., & Ward, L. M. (2001).Multisensory integration and crossmodal attention effects in thehuman brain. Science, 292, 1791.

Meredith, M. A., Nemitz, J. W., & Stein, B. E. (1987). Determinants ofmultisensory integration in superior colliculus neurons. I. Temporalfactors. Journal of Neuroscience, 7, 3215–3229.

Meyer, G. F., Wuerger, S. M., Rohrbein, F., & Zetzsche, C. (2005). Low-level integration of auditory and visual motion signals requires spatialco-localisation. Experimental Brain Research, 166, 538–547.

Pelli, D. G. (1997) The VideoToolbox software for visual psychophysics:Transforming numbers into movies, Spatial Vision 10, 437–442.

Robbins, H., & Monro, S. (1951). A stochastic estimation method. Annals

of Mathematical Statistics, 22, 400–407.Treutwein, B. (1995). Adaptive psychophysical procedures. Vision

Research, 35, 2503–2522.Sekuler, R., Sekuler, A. B., & Lau, R. (1997). Sound alters visual motion

perception. Nature, 385, 308.Shams, L., Iwaki, S., Chawla, A., & Bhattacharya, J. (2005a). Early

modulation of visual cortex by sound: An MEG study. Neuroscience

Letters, 378, 76–81.Shams, L., Kamitani, Y., & Shimojo, S. (2000). Illusions. What you see is

what you hear. Nature, 408, 788.Shams, L., Kamitani, Y., & Shimojo, S. (2002). Visual illusion induced by

sound. Cognitive Brain Research, 14, 147–152.Shams, L., Kamitani, Y., Thompson, S., & Shimojo, S. (2001). Sound alters

visual evoked potentials in humans. NeuroReport, 12, 3849–3852.Shams, L., Ma, W. J., & Beierholm, U. (2005b). Sound-induced flash

illusion as an optimal percept. NeuroReport, 16, 1923–1927.Slutsky, D. A., & Recanzone, G. H. (2001). Temporal and spatial

dependency of the ventriloquism effect. NeuroReport, 12, 7–10.Violentyev, A., Shimojo, S., & Shams, L. (2005). Touch-induced visual

illusion. NeuroReport, 13, 1107–1110.Ward, J., Huckstep, B., & Tsakanikos, E. (2006). Sound–colour synaes-

thesia: To what extent does it use cross-modal mechanisms common tous all? Cortex, 42, 264–280.

Watanabe, K., & Shimojo, S. (2001). When sound affects vision: Effects ofauditory grouping on visual motion perception. Psychological Science,

12, 109–116.