Visual perception normally occurs seamlessly and remains stable. However, changes in viewing conditions such as low contrast, clutter, and viewing angle can cause an object to appear ambiguous. For instance, imagine driving down a dark road at night and seeing a light in the distance. The source of the light may be ambiguous and not immediately recognizable: it could be another car, a building, street lamps, etc. Your brain will instinctively try to form a perception of the source of the light. But what neural activity is responsible for this process of generating perception under such visual ambiguity?

In studying this question, many researchers have investigated bistable perception, which occurs when visual ambiguity is extremely high. Ambiguous images, such as the Rubin Face-Vase illusion, are commonly used to elicit bistable perception: when the same stimulus can be perceived as two distinct interpretations that repeatedly switch between each other. Ambiguous images are especially useful for experimentally studying brain responses to perceptual ambiguity because the presented image does not change during viewing but perception does, allowing researchers to dissociate between neural activity related to sensory processing and that related to perceptual experience.

Many studies in recent decades have used fMRI to focus on investigating neural activity around the timing of perceptual switches, often reporting increases in activity in primary visual cortex, extrastriate cortex, and, more recently, regions of the frontoparietal network, most notably the right inferior frontal gyrus. However, a less studied aspect of bistable perception are the brief (typically 2 to 5 seconds) periods of time between perceptual switches, when perception remains relatively stable until the next switch event. Recent studies using electrophysiology have found that oscillatory activity, particularly in the alpha-frequency (8 to 13 Hz) band, is especially important for stabilizing perception. Interestingly, a simple experimental manipulation has also been shown to strengthen perceptual stability: by viewing an ambiguous image intermittently—via inserting brief blank screens in between repeated presentations of the image—perceptual switch rate drastically decreases. This effect is thought to be strongly influenced by perceptual memory, a form of memory that retains information about the most recent percept from before each blank period.

In light of these previous findings, in this study we asked three main questions: Which neural mechanisms promote perceptual stability when viewing ambiguous images? Do neural mechanisms promoting perceptual stability also support perceptual memory? Does the ventral frontal cortex causally contribute to perceptual stability or perceptual memory?

To answer these questions, we used high-definition transcranial direct current stimulation (HD-tDCS) to perturb neural activity over two brain regions: the right inferior frontal gyrus (rIFG) and occipital pole (Occ). After stimulation, subjects then viewed ambiguous images while neural activity was recorded using electroencephalography (EEG).

We found that the amplitudes of alpha and beta oscillations correlate positively with perceptual stability (measured by percept duration) in the continuous-viewing task condition; that amplitudes in these same frequency bands, alpha and beta, also correlate positively with the persistence of perceptual memory in the intermittent-viewing task condition; and that HD-tDCS stimulation did not have a significant effect on behavioral measures in either condition. This null effect of HD-tDCS was surprising, and still needs to be reconciled with the results of another recent causal manipulation study that used transcranial magnetic stimulation (TMS).

On the other hand, the involvement of alpha and beta activity in perceptual stability and perceptual memory is more consistent with previous work and provides further evidence for the importance of oscillatory activity in stabilizing visual perception. In particular, many studies using a wide range of tasks and experiments have found that alpha and beta oscillations are strongly associated with top-down signaling, which is characterized by feedback neural projections from more specialized high-order cortical regions, such as V4 and frontal cortex, to lower-order sensory regions, such as V1. Therefore, we interpret our results as reflections of top-down feedback used by the brain to resolve and stabilize perception when faced with an ambiguous stimulus.

To summarize, we conducted an experiment using a within-subject cross-over design that used HD-tDCS and EEG to study the neural correlates of perceptual stability and perceptual memory.

We report a null result that HD-tDCS applied over the rIFG did not affect perceptual behavior during bistable perception. We also found a unified function of alpha and beta band activity, which correlated with both perceptual stability when viewing a static ambiguous stimulus as well as perceptual memory when viewing an ambiguous stimulus intermittently. Based on past findings, we interpret these neural correlates as signals of top-down feedback that act to stabilize perception.

—Michael Zhu

Michael Zhu is the lead author on this study, which was conducted in NYU Langone’s Perception and Brain Dynamics Laboratory, led by Biyu J. He, PhD, when Michael was an undergraduate student researcher. He received multiple awards for his research, including financial support from NYU’s Dean’s Undergraduate Research Fund, the Simons Foundation Autism Research Initiative, and NYU’s Training Program in Computational Neuroscience. Michael recently completed his master of information and data science at University of California, Berkeley.

Read the paper “Neural oscillations promoting perceptual stability and perceptual memory during bistable perception” in Scientific Reports, published February 17, 2022.

Other recent studies from the Perception and Brain Dynamics Laboratory include “Cortical and subcortical signatures of conscious object recognition,” “Long-term priors influence visual perception through recruitment of long-range feedback,” and “Neural integration underlying naturalistic prediction flexibly adapts to varying sensory input rate” published in Nature Communications, and “Spectral signature and behavioral consequence of spontaneous shifts of pupil-linked arousal in human” published in eLife.