Under continuous lighting, moving stimuli such as ceiling fans and car wheels can sporadically appear to move in the reverse direction—this phenomenon is known as illusory motion reversal (IMR). We have previously suggested that IMR results from the spurious activation of motion detectors tuned for the opposite direction of motion, leading to a rivalry between two possible motion percepts. To determine if this hypothesis is supported by evidence from electrophysiology, we used EEG to directly compare neural signatures in IMR and binocular rivalry (BR), a well-studied form of rivalry. We find that both IMR and BR show large changes in power in the beta range (14–30 Hz) at the time of a perceptual switch. More importantly, during a stable perception, beta power correlates with the probability of a perception. Specifically, beta power associated with veridical motion perception (experienced the majority of the time) is higher than the power during illusory motion perception (experienced a minority of the time). The BR percepts, each 50% probable, are associated with an intermediate beta amplitude. We propose that the amplitude of synchronized beta activity reflects the size of currently active neural coalitions, with less likely percepts associated with smaller coalitions.

*α*= (

*s*)

^{2}and

*β*=

*s*

^{2}/

*s*is the standard deviation). Significance was tested using a Kolmogorov–Smirnov goodness-of-fit test (

*α*= 0.05).

*f*

_{0}/

*σ*

_{ f }equal to 5 (Tallon-Baudry, Bertrand, Delpuech, & Permier, 1997). This parameter defines a wavelet duration of 398 ms at 4 Hz and 29 ms at 55 Hz as extreme values. The power of the oscillatory activity was then averaged across the 124 channels covering the whole scalp as there was no

*a priori*hypothesis about the location of the activity. Power was computed for the classic frequency bands (theta, 4–7 Hz; alpha, 8–13 Hz; beta, 14–30 Hz; and gamma, 31–55 Hz, see Nunez & Srinivasan, 2006).

*t*-statistic for each time point in the data, the algorithm clusters together those significant data points (

*p*< 0.05, two-tailed) that are temporally adjacent. In this way, it creates clusters characterized by a test statistic defined as the sum of the

*t*-values of its component data points. For the cluster with the largest test statistic, the algorithm computes a surrogate Monte Carlo distribution of the test statistic by randomly reassigning the data points to the two conditions of interest. This reference distribution is then used to estimate the

*p*-value of the test statistics of all the clusters. Data points belonging to clusters with

*p*-values <0.05 are thus considered significant after controlling for multiple comparisons.

*t*-test over the 17 blocks with an expected mean of 50% (in the case all the predictions were random).

*p*> 0.05) for the veridical motion perception (9 participants out of 10) and for the illusory motion perception (7 participants), in line with our hypothesis that IMR is a form of rivalry (Kline et al., 2004; Kline & Eagleman, 2008). We also replicated the finding that dominance durations in binocular rivalry are usually modeled by a gamma distribution (7 participants, see Leopold & Logothetis, 1999). The motor control condition was well fit only in 4 cases. Participants reported seeing the veridical motion direction more often than the illusory one (81.76% and 18.24% of total viewing time, respectively;

*p*< 10

^{−8}with a paired

*t*-test), while the two patterns of BR were equiprobable (50.30% and 49.70%,

*p*= 0.82).

*F*(3,21) = 0.819,

*p*= 0.50). In the motor condition, by contrast, the amount of beta power was significantly higher than both IMR and BR conditions for the same interval (Figure 2E, black line; paired

*t*-test between the motor condition and the mean of the rivalrous conditions,

*t*(8) = 2.85,

*p*< 0.025).

*p*< 0.005, and −800 to −700 ms,

*p*< 0.025, and +900 to +1250 ms,

*p*< 0.005). As can be seen clearly in Figure 2, during the switch from veridical into illusory motion perception, beta power decreases (Figure 2A), while the opposite effect is observed during the switch from illusory to veridical motion perception (Figure 2B). Importantly, in the BR condition, the amplitude remains indistinguishable before and after the fluctuation temporarily caused by the switch (Figures 2C and 2D). A paired

*t*-test to confirm this observation between the beta power before and after the switch yielded no significant difference for the BR condition (

*t*(8) = 0.15,

*p*= 0.89 for Figure 2C and

*t*(8) = −0.36,

*p*= 0.73 for Figure 2D) while the paired

*t*-test for the IMR condition was significant in both cases (

*t*(8) = −2.89,

*p*< 0.025 for Figure 2A and

*t*(8) = 2.75,

*p*< 0.025 for Figure 2B). Below, we will explore the speculation that the probability of a percept correlates with the synchronized activity of underlying neural circuits—that is, more probable percepts are associated with larger neural populations, which, when they dominate, yield a stronger EEG signal.

*t*(16) = 5.79,

*p*< 10

^{−4}), and in all the blocks except one, the correct classification was above 50%. Although its accuracy is still rather low, this method has a higher success rate than previous presented methods (VanRullen et al., 2006). Generally, this type of analysis is substantially hindered by the poor signal-to-noise ratio of single-trial EEG (Jung et al., 2001), yet in the future more sophisticated algorithms might identify the conscious percept with greater precision.

*p*< 0.05 for each time window, Figure 3). Preceding a perceptual switch, veridical motion perception is correlated with larger alpha power than the illusory perception. This result is consistent with VanRullen et al. (2006), in which they report a 13-Hz component associated with veridical motion perception.

*t*(16) = 4.55,

*p*< 10

^{−3}).

*amplitude*of these oscillations reflects the effectiveness of the neural coalitions representing the competing percepts—that is, which of two rivaling percepts dominates at any moment (Figure 4D). In other words, the size of the neuronal ensemble that drives the widespread oscillatory activity determines the power of the EEG signal. In the case of motion viewing, more neurons encode the veridical direction than the illusory one, and when the perception is driven by the former, the amount of EEG activity measured at the scalp will be larger. In contrast, equally sized coalitions are responsible for the two equiprobable percepts during BR, resulting in an almost equal level of beta power during the two possible percepts.