The role of alpha oscillations for illusory perception.

Alpha oscillations are a prominent electrophysiological signal measured across a wide range of species and cortical and subcortical sites. Alpha oscillations have been viewed for a long time as an "idling" rhythm, purely reflecting inactive sites. Despite earlier evidence from neurophysiology, awareness that alpha oscillations can substantially influence perception and behavior has grown only recently in cognitive neuroscience. Evidence for an active role of alpha for perception comes mainly from several visual, near-threshold experiments. In the current review, we extend this view by summarizing studies showing how alpha-defined brain states relate to illusory perception, i.e. cases of perceptual reports that are not "objectively" verifiable by distinct stimuli or stimulus features. These studies demonstrate that ongoing or prestimulus alpha oscillations substantially influence the perception of auditory, visual or multisensory illusions.

Introduction

Neuronal oscillations in the alpha band (∼8–12 Hz) are a ubiquitous electrophysiological signal in the human and non-human mammalian brain [1]. Alpha oscillations have been found in several cortical areas, including sensory, motor and frontal cortices (see [2,3] for reviews), as well as in subcortical areas such as the thalamus and basal ganglia [4-6]. For sensory cortices, different labels have been assigned to the respective alpha oscillations, e.g. the mu-rhythm in somatosensory [7,8] or the tau-rhythm in auditory cortex [9]. These sublabels refer to the slightly different waveforms of alpha oscillations in different modalities. In addition, the sublabels suggest that the functional role of alpha oscillations in different modalities may not be completely homogenous. Since the aim of this review is to highlight similarities between the oscillations falling in the alpha frequency band in different sensory cortices, we will use the more general term alpha oscillations throughout the manuscript.

The importance of alpha oscillations for neuronal processing has been recognized long ago. For example, early studies have reported an influence of alpha power and/or phase on the perception, evoked response and reaction times (e.g. [10-14]). Only in recent years, neuroscientific research has re-gained increasing awareness that alpha oscillations substantially influence perception, behavior and neuronal processing. It has been known for a long time that alpha oscillations can be modulated intrinsically and extrinsically. For example, the power of alpha oscillations decreases after sensory stimulation, but increases if subjects are not engaged in any task. This latter effect led to the notion of alpha as an “idling” process of brain regions [15]. In contrast to this idling hypothesis, recent studies have argued for an active role of alpha in perception and cognitive processes. A common thread between different proposals on the functional role of alpha is the gating of neuronal processes, such that increased alpha power reflects inhibition of task-irrelevant areas whereas decreased power indicates processing [16-18]. Evidence for the active role of alpha comes from studies investigating working memory and attention. For example, several studies reported a modality specific increase of alpha power in visual, somatosensory, or auditory cortices in the retention period of the respective working memory tasks (e.g. [19-21]). An increase of alpha power in the retention period fits well to the notion of increased alpha power reflecting inhibition. Increased alpha power is suggested to inhibit distracting signals to interfere with the information stored in memory [17]. In addition, studies on spatial attention have demonstrated that the power of alpha oscillations can be modulated by endogenous shifts of attention in the absence of any physical stimulation. When a stimulus is anticipated or attention is directed toward the location of a stimulus, alpha power decreases in visual, auditory or somatosensory cortex, respectively, contralateral to the expected or (covertly) attended site. Conversely, alpha power increases in ipsilateral sites (e.g. [22-27]). Going beyond these well controlled and stimulus-cued alpha modulations, relatively “spontaneous” fluctuations of alpha power prior to stimulus onset can influence detection and discrimination of upcoming near-threshold stimuli [28-33]. Latter studies used stimuli at perceptual threshold and investigated one extreme side of perception where the relevant stimulus is sometimes perceived or goes completely unnoticed. On the other extreme side of perception, there are rare cases where stimuli are perceived which are physically not present. Such misperceptions can appear following stimulation (e.g. illusions) or without an external stimulus (e.g. phantom perceptions like tinnitus). Illusions can be characterized by two perceptual processes: Either subjects demonstrate a distorted perception of the stimulus (-properties) or they perceive a qualitatively and/or quantitatively additive (i.e. illusory) stimulus property which is not present in the physical stimulus.

Some illusions can be elicited reliably and constant in time (e.g. Kanisza triangle), whereas for other illusions trial-by-trial variability between veridical and illusory perception has been reported (e.g. double-flash illusion [34]). Since physical stimulation remains constant, the variability of perception has to be caused by intrinsic modulations of brain states. Therefore, such illusions offer unique insights into intrinsic brain states modulating veridical and illusory perception.

In this review paper, we will survey studies in which subjects’ perception switches between illusory and veridical perception over time or on a trial-by-trial basis. We will outline that alpha oscillations play an active role for the switching between perceptual states. The modulation of alpha oscillations can be found either in the prestimulus period or in the peristimulus period, i.e. while the illusion is ongoing.

Our review will be structured in three parts. In the first and second part, we will review the role of alpha oscillations on visual and auditory illusions respectively. Finally, we will review multisensory, i.e. audio-visual and visuo-tactile illusions.

Methods to analyze alpha oscillations

Alpha oscillations can be analyzed with several approaches. While the results of these approaches are virtually identical, their applicability and practicality might differ depending on data and/or a priori hypotheses. For example, some of the reviewed studies obtained spectral estimates by Fourier transformation [35-44] others by wavelet analysis [45-47]. As a next step, the alpha-band can be chosen a priori by a defined frequency or frequency-band [35-41,46,47] and statistics are then performed on these alpha-band estimates. Alternatively, spectral estimates can be computed on broader frequency ranges and statistics reveal significant effects confined to the alpha-band [42-45,47]. Finally, some studies analyzed alpha oscillations solely on the sensor level [35-37,39,45,46]. While the topography of these analyzes often gives sufficiently good information about the distribution and sources of the observed effect, it nevertheless limits the interpretation of cortical sources. To overcome this limitation, some studies additionally computed the underlying sources of effects revealed at sensor level using e.g. beamformer techniques [42-44,47]. Other studies used a computationally more demanding approach and performed analyses directly on source level [40,41].

Visual domain

One intriguing example of a visual illusion is the phenomenon of phosphene perception. Phosphenes are flash-like illusory percepts which are typically induced by applying transcranial magnetic stimulation (TMS) over visual cortex, in the absence of retinal input and even in blind subjects [48,49].

In a series of studies, Romei et al. studied the role of alpha oscillations on phosphene perception. The authors applied single pulse TMS over visual cortex and determined individual thresholds of TMS intensity to induce phosphene perception. Simultaneously, they recorded resting state brain activity with EEG and correlated oscillatory brain activity with individual phosphene thresholds. They found that phosphenes could be induced with low TMS intensities in subjects with low resting state alpha power measured over posterior sites, presumably visual cortex. Moreover, TMS intensities required for phosphenes increased with increasing alpha power. In subjects showing the highest alpha power no phosphenes could be evoked at all [35]. In another TMS-EEG study, Romei et al. demonstrated that also individual trial-by-trial variability of phosphene perception was correlated with alpha power at the moment of the TMS pulse. If a TMS pulse coincides with low levels of spontaneously fluctuating alpha power in occipital cortex, the likelihood of perceiving phosphenes was increased. This effect was found in occipital cortex contralateral to the site of the perceived phosphenes [36].

These studies demonstrate that the likelihood of inducing phosphenes is related to the level of alpha power at the moment of application of the TMS pulse. The authors argue that alpha power reflects the state of momentary cortical excitability. If external stimulation by TMS hits visual cortex at states of high excitability (low alpha power), this may induce visual illusory percepts even in the absence of retinal stimulation.

Other studies reported that in addition to power, the phase of the alpha oscillation in occipital areas seems relevant for phosphene perception [45]. The authors found that applying a TMS pulse to occipital cortex at a certain preferred alpha phase increases the likelihood of perceiving phosphenes by 15% relative to the opposite phase. This correlation of alpha phase and phosphene perception was found in occipital and frontal areas within 400 ms before application of the phosphene-inducing TMS pulse. It is well known that the phase of oscillations can also be modulated, e.g. by crossmodal stimuli in which an event in one sensory modality can reset the phase of oscillations in another modality [50]. In this study, Lakatos et al. measured the phase of oscillations in auditory cortex of macaques using intracranial multi unit recordings. They found that somatosensory stimulation lead to a phase reset in auditory cortex across various frequency bands. By applying auditory stimuli, Romei et al. used such a crossmodal paradigm to entrain the phase of occipital alpha oscillations [51]. At variable times after auditory stimuli, TMS pulses were applied over occipital cortex to elicit phosphenes. The rate of the induced phosphene perception revealed a periodic pattern of ∼10 Hz. This pattern was phase-locked to the auditory stimulus: at specific time points after auditory stimulation (e.g. ∼100 ms and ∼200 ms before the TMS pulse) the likelihood of phosphene perception was increased relative to other time points. In contrast, the likelihood of phosphene perception was lowest a ∼50 ms, ∼150 ms and ∼250 ms, i.e. the pattern of phosphene perception followed a periodic pattern of ∼10 Hz (Fig. 1).

Fig. 1

The right panel illustrates an experiment where TMS is applied over visual cortex to induce phosphenes. TMS is applied at varying time points relative to an auditory stimulus. The left panel shows percentage of trials with TMS-induced phosphene as a function of delay of TMS pulse relative to the auditory stimulus. The shaded areas (75–120 ms and 180–225 ms) represent windows of significantly increased visual cortex excitability by auditory input.

In addition to these studies on phosphene perception, studies on other visual illusions demonstrate that alpha oscillations can also play a role when stimulation of visual cortex is provided by retinal input. As we will review below, in such cases alpha oscillations can correlate with a modulation of perception.

One example for a modulatory role of alpha oscillations is given by the “wagon wheel illusion”. In this illusion, a wheel continuously rotating in one direction spontaneously appears to reverse the direction of rotation. This phenomenon is frequently observed on movies when the sampling rate of the camera is too slow compared with the temporal frequency of, for example, the rotating wheel of a car. In rare cases this illusion has been reported to occur under real conditions, i.e. when the sampling rate of a movie cannot account for the illusion. VanRullen et al. studied this illusion using EEG [37]. Subjects viewed a continuously rotating wheel and reported the perceived direction of rotation by continuously pressing corresponding buttons. VanRullen et al. found that alpha power decreased before a subjective reversal of motion direction and increased before transition back to real motion. This effect was evident in centro-parietal sensors and was independent of the speed of the rotation. VanRullen et al. argued that the human visual system might act similar to a video camera and sample the world in snapshots. The duration of a snapshot is provided by an alpha oscillation, thus lasting ∼100 ms.

Spontaneous subjective reversals of motion direction can also appear in other paradigms. Strüber and Herrmann [46] used an alternating dot pattern that induces a bistable perception: similar to the wagon wheel illusion, the dots appear to rotate clockwise or anti-clockwise and the perceived direction of the rotation reverses spontaneously. In this study, oscillatory brain activity was measured with EEG while subjects indicated their subjective time point of motion reversal. The authors found that alpha power in posterior sensors monotonically decreased before the subjective motion reversal. Strüber and Herrmann argued that physical motion direction is ambiguous and thus the subjectively perceived motion direction is generated by (unknown) internal processes. The authors argue that high alpha power might stabilize the current subjectively perceived motion direction, i.e. make the subjective perception more robust against internal or external influences or noise. Consequently, a decrease of alpha power might reflect a destabilisation of the current subjective perception. If alpha power finally decreases below a threshold, perception is unstable and the alternative visual perception can develop.

In addition to correlating with excitability and perceptual changes of bistable stimuli, it was recently shown that alpha power can be related to illusory motion perception in a static visual stimulus [38]. In this illusion, subjects perceived a physically static “wheel” as flickering. Sokoliuk and VanRullen measured EEG activity while subjects indicated when they perceived the flickering illusion by pressing a button at flicker onset and releasing the button at flicker offset. The illusion was most pronounced during epochs of enhanced alpha power in occipital EEG sensors. Furthermore, the frequency of the subjectively perceived flickering was ∼9 Hz and correlated with individual alpha peak frequencies of the EEG.

In summary, TMS studies suggest that power and phase of prestimulus alpha oscillations reflect momentary excitability of visual cortex. At states of optimal excitability, when alpha power is low and/or alpha phase is at a preferred phase, TMS is capable of inducing a visual perception even in the absence of retinal input. The excitability shows a periodic pattern which is determined by the phase of the alpha oscillation. Other studies using continuous stimulation with bistable stimuli argue that a change of alpha power indicates a switch of perception.

Auditory domain

Although far less in number compared to visual illusions, there are several reports of auditory illusions. The probably most familiar illusion in the auditory domain is tinnitus. Tinnitus is characterized by sensation of a tone in the absence of auditory stimuli or any other external stimulation of auditory cortex.

In an MEG study, Weisz et al. [39] studied the role of alpha oscillations for tinnitus perception by comparing ongoing alpha oscillations in individuals with and without tinnitus across all MEG sensors. The authors found a significant reduction of ongoing alpha power for the tinnitus group compared to normal hearing group, predominantly in bilateral temporal regions. These results suggest that decreased alpha power indicates reduced inhibition/increased excitability of auditory cortex. Such abnormally reduced alpha power might induce perception of tones even in the absence of stimulation.

To test this hypothesis, two studies exploited methods of neuromodulation to entrain auditory alpha oscillations. Müller et al. [40] used external stimulation by applying repetitive TMS (rTMS), a method which has been frequently explored as potential clinical tool, however with mixed outcomes [52]. By measuring MEG, the authors revealed that the tinnitus relief after applying rTMS was associated with an increase of alpha power in auditory cortex. In another study, Hartmann et al. [41] used neurofeedback to modulate alpha oscillations. Subjects were trained to intrinsically increase auditory alpha power in response to an auditory tone. After four weeks of neurofeedback training, MEG recordings revealed an increase of ongoing alpha power in right auditory regions. In addition, the authors reported that neurofeedback training significantly decreased tinnitus symptoms. These results imply that alpha oscillations in auditory cortex play a fundamental role for phantom perception. Conforming with notions in the visual modality, the function of alpha power might be to balance excitation and inhibition. If loss of inhibition is abnormally strong, the increased internal excitation might lead to phantom perception [53].

Additional evidence for a role of alpha oscillations for auditory illusions comes from a study investigating illusory music perception in noise. Müller et al. [42] presented subjects with familiar or unfamiliar songs. The songs were interrupted by short segments containing only pink noise. Although no music was played during the noise segments, subjects reported an illusory music perception which was stronger for familiar songs compared to unfamiliar songs. Müller et al. simultaneously recorded brain activity using MEG or ECoG and analyzed alpha power during the noise segments inducing illusory music perception. A stronger alpha power decrease in auditory cortices, mostly pronounced in right primary auditory cortex, was found if subjects perceived familiar compared to unfamiliar music. This finding was interpreted as an increased excitability of auditory cortex leading to an increased probability of noise to induce an illusory perception of music.

Similar findings are reported in a recent study on the Zwicker Tone illusion [43]. In this paradigm, an “auditory afterimage” is induced by stimulation with a notch-filtered auditory noise stimulus [54]. White noise stimuli with different notch widths were presented and subjects reported a linear increase in the auditory illusion with increasing notch width. Alpha power in the Heschl Gyrus exhibited a pronounced negative trend with increasing notch width. This study shows that the change in the strength of the “auditory afterimage” is inversely related to alpha power. It supports the notion that alpha power is inversely related to cortical excitability.

In summary, in line with studies from the visual domain, these studies on auditory illusions support the view that alpha power reflects excitability of sensory cortex. During states of low alpha power auditory stimulation with noise can induce an illusory perception of music. In addition, tinnitus patients revealed an abnormally low level of alpha power compared to healthy subjects. This suggests that for abnormally low alpha auditory perception can also be induced in the absence of any stimulation.

Multisensory illusions

In this section, we will review two recent studies investigating the so-called double-flash illusion (DFI [34]). In this multisensory illusion, one briefly presented visual stimulus is accompanied by two rapidly presented either auditory (audio-visual DFI) or tactile stimuli (visuo-tactile DFI). Both paradigms frequently induce the perception of a second, illusory visual stimulus. The DFI is optimally perceived if the visual stimulus is presented between the two auditory or tactile stimuli and if all stimuli are presented within ∼100 ms [55].

In a recent MEG study, Lange et al. [44] investigated how fluctuations of prestimulus neuronal activity influence perception of the visuo-tactile DFI. The authors found that the illusory perception of a second visual stimulus was preceded by a reduction of prestimulus alpha power in visual cortex relative to the perception of one visual stimulus. In a second task, subjects were presented with two unisensory visual stimuli. Here, subjects frequently missed one stimulus. Comparable to the DFI task, perception of two visual stimuli correlated with decreased prestimulus alpha power relative to the perception of one visual stimulus. Hence, in both tasks, decreased prestimulus alpha power correlated with the perception of two stimuli, irrespective of whether this perception reflected veridical or illusory perception (Fig. 2). The authors concluded that prestimulus alpha power indexes excitability of visual cortex. If alpha power in visual cortex is low, i.e. excitability of visual cortex is high, two visual stimuli are more likely to be processed distinctly and thus perceived correctly. Conversely, low alpha power, i.e. high excitability, makes visual cortex also more susceptible for heteromodal input, e.g. from somatosensory cortex. Heteromodal input is thus capable of eliciting additional activity in visual cortex which can induce an illusory perception in the DFI.

Fig. 2

Analysis of the visuo-tactile double-flash illusion (DFI) and the visual fusion effect. (A) Left panel shows time–frequency representation of the contrast between two perceived visual stimuli (i.e. DFI) vs. one perceived stimulus for MEG sensors over occipital cortex. The significant decrease of power in the alpha-band is highlighted. Middle and right panel show source reconstructions of the significant alpha effect. (B) Same as A, except that these figures show the contrast between two perceived visual stimuli (i.e. the “fusion effect”) when subjects had to discriminate two visual stimuli.

In another MEG study, Keil et al. [47] examined the audio-visual DFI. While most studies reviewed in this paper investigated power of alpha oscillations as a marker of illusions, Keil et al. examined whether the perception of the DFI correlated with functional connectivity between cortical areas. Subjects had to indicate by button press whether they perceived two (i.e. the DFI) or one (i.e. non-DFI) visual stimuli. Functional connectivity was measured by phase synchrony between cortical areas in an interval of 500–100 ms prior to stimulus onset. For the alpha band, the authors found that right primary auditory cortex (A1) was more strongly connected to visual areas (BA18) prior to DFI trials compared to non-DFI trials in the prestimulus period. In addition, primary visual cortex (V1) showed stronger connections to medial frontal (MFG) and parietal (BA4) areas and weaker connections to the inferior frontal cortex (BA44) prior to illusion trials (Fig. 3). The results imply that phase synchrony in the alpha band might index a functional network which promotes illusory perception by integration or segregation of auditory and visual stimuli. MFG has been reported frequently to be involved in top-down attentional control [56,57]. A stronger connection between MFG and V1 might suggest a stronger impact of top-down control in the period before DFI perception.

Fig. 3

Analysis of the audio-visual double-flash illusion (DFI). Alpha band functional connectivity contrast between two perceived stimuli (i.e. double-flash illusion) vs. one perceived stimulus. Positive values indicate stronger functional connectivity prior to illusions, whereas negative values indicate stronger functional connectivity prior to the perception of one visual stimulus.

In summary, studies on multisensory illusions support the view that alpha power indicates excitability of visual cortex. Low alpha power reflects states of heightened excitability in which multiple visual stimuli can be processed distinctively, yet crossmodal influence may elicit visual illusions. Moreover, this crossmodal influence can be influenced by changes in functional networks. Prestimulus alpha oscillations seem to shape functional networks underlying illusory perception and change the relative effect of bottom-up crossmodal vs. top-down influence.

Summary and conclusions

In this review we have surveyed studies which provide cumulative evidence for a critical role of alpha oscillations for perception of visual, auditory, or multisensory illusions. These studies indicate that alpha oscillations influence perception by regulating excitability of sensory cortex and/or by regulating neuronal information flow within and between cortical areas.

While some studies report the critical modulation of alpha oscillations before the perception of the illusion, other studies report an ongoing modulation of alpha oscillations which appear while the illusion is ongoing. Note that ongoing modulations of alpha oscillations are found in the absence of any stimulation (tinnitus) or when the sensory stimulation is ongoing and illusory perception switches spontaneously [37-39,42,43,46]. Prestimulus modulations of alpha oscillations are found when sensory or TMS stimuli are abruptly presented for only a short period [35,36,44,45,47,51]. A potential explanation for differences in duration and time periods of alpha modulations might thus be found in the type of stimulation. Abrupt presentation of stimuli might interrupt or overshadow otherwise ongoing modulations of alpha oscillations. We will thus treat both phenomena (prestimulus and peristimulus modulations of ongoing activity) as one common effect.

Alpha oscillations can be modulated intrinsically or extrinsically for example by modulations of attention. In addition, alpha oscillations are also modulated by spontaneous fluctuations. When alpha power is not explicitly controlled, but fluctuates spontaneously, low prestimulus alpha power has been shown to improve neuronal processing of sensory input and thus improve perception, especially for weak, near-threshold stimuli (e.g. [28-33]). A potential concern with these studies is that alpha power might simply reflect arousal or attention. If alpha power is high, stimuli might be missed simply because high alpha power reflects inattention. In this paper, we have reviewed several studies which complement this view. The reviewed studies on illusions reveal that low alpha power does not lead to more veridical perception per se. Instead low alpha power sometimes promotes illusory perception. A unifying explanation that combines these seemingly contradictory findings in studies using near-threshold and illusory stimuli is that alpha power reflects excitability of sensory cortices. At states of high excitability stimuli are more precisely processed in sensory cortex. This means that at states of high excitability (low alpha power) the threshold of activation of the underlying neural population is lowered. Therefore the neuronal population is more likely to be activated, either by direct stimulation of the specific population (e.g. sensory stimulation or TMS) or by heteromodal input to the neuronal population. The reviewed studies on illusory perception suggest that optimal states of excitability correlate with low alpha power or optimal phase. Studies on the perception of single, near-threshold stimuli suggest, however, that too low alpha power/too high excitability might be detrimental for neuronal processing (e.g. [28,32,33]). Note, that such a detrimental relationship between alpha power and perception has not been reported in the reviewed illusions. In addition, at states of high excitability neuronal activity in sensory cortex is more likely to be elicited by other sources than modality-specific sensory input. In such cases the induced subjective perception does not match the sensory input, which eventually leads to illusions or other misperceptions, e.g. external stimulation by TMS leads to phosphene perception [35,36,51], input from other modalities induces the DFI [44], noise induces illusory music perception [42] or tones are perceived even in the absence of any stimulation [39].

In addition to alpha power, alpha phase has been shown to influence illusory perception [37,45,51]. These studies argue that alpha power and phase define temporal windows in which perception is improved [37,45,51]. This view complements the hypothesis of excitability by arguing for periodic patterns of excitability [17].

The hypothesis that ongoing alpha oscillations define periodic cycles of excitability has been proposed already a long time ago (e.g. [10]). For example, Bishop [58] found in visual cortex of rabbits that evoked responses could be evoked only at certain intervals corresponding to the spontaneous alpha cycle [58]. Similar results have been reported by Chang [59] for auditory cortex in cats [59]. In the present review, we extend this view to the perception of illusions.

The concept of alpha oscillations defining excitability of sensory cortex is intriguing and supported by several studies. However, alpha oscillations might subserve more than one functional role. In addition to determining excitability, one potential role of alpha oscillations might be integration of cortical areas to a common network. Pre-established neuronal networks might determine information flow within and between cortical areas and thus determine how an upcoming stimulus is processed and perceived [60]. In line with this view, global cortical networks defined by alpha phase determine whether subjects perceive the DFI [47].

An important concern is whether alpha oscillations shape perception, i.e. play a causal role for illusions or whether alpha oscillations represent neuronal processing without a causal role. Such scenarios are difficult to differentiate if the illusory perception is ongoing or modulations of alpha oscillations are observed parallel to the illusion [39,42]. However, other studies studying similar illusions or phantom perceptions demonstrate that a modulation of alpha oscillations by neurofeedback or TMS has direct consequences for illusory perception [40,41]. This direct consequence argues for a causal role of alpha oscillations. Further evidence for a causal role comes from studies reporting that the level of alpha oscillations in the prestimulus period has consequences for illusory perception [35-37,44-47]. Since modulations of alpha oscillations appear before the stimulation, these studies suggest that prestimulus alpha oscillations are likely to be involved in shaping of brain networks for upcoming stimuli [17,60].

While the reviewed papers comprehensively and convincingly demonstrate that alpha oscillations play a crucial role in perception of illusions, other frequency-bands might also contribute to the complex pattern and dynamics of illusory perception. For example, in addition to alpha power, Lange et al. [44] found prestimulus beta- and gamma-power to correlate with the perception of the visuo-tactile DFI, while Keil et al. [47] reported beta-power to correlate with the perception of the audio-visual DFI. Other studies imply that low alpha power is not a prerequisite for illusory perception. For example, an MEG study found prestimulus beta-band power to correlate with the perception of the McGurk illusion, but did not find an influence of prestimulus alpha power [61]. We can only speculate why no influence of alpha oscillations has been found for the McGurk illusion. One reason might be that the McGurk illusion is more complex than the illusions surveyed in this review. Thus the perception of the McGurk illusion might involve other mechanisms and/or cortical areas. In line with this interpretation, beta-band effects for the McGurk illusion have been reported mainly for the superior temporal gyrus and frontal and temporal regions [61]. In contrast, the alpha effects surveyed in this review have been reported mainly in (early) sensory cortices. Another reason for the lack of effects in the alpha band might be that the McGurk illusion involves ongoing visual stimulation prior to the onset of the auditory stimulus (i.e. prior to the illusion). Therefore, pre-illusion activity in the McGurk illusion is actually intermixing with visual stimulation and thus might be overshadowed. Additional studies are needed to elucidate whether alpha oscillations might play a role for the perception of the McGurk illusion.

In summary, alpha power and phase define a momentary state of sensory cortices and cortical networks subserving perception. This momentary state defines how sensory input is effectively processed. A state of low alpha power – indicating increased excitability – leads to increased crossmodal interaction. In case of the double-flash illusion the increased crossmodal interaction leads to erroneously perception of an additive stimulus. In addition, a state of low alpha power might also open up the window for crossmodal influence and switches between bistable perceptions.