*2.2. Visual Stimuli*

We used an experimental paradigm with an ambiguous bistable visual stimulus in the form of the Necker cube, which allows two possible interpretations [4,16,18,19]. The non-perceptually impaired volunteer interpreted this two-dimensional (2D) image as a three-dimensional (3D) object which is oriented either left or right. The balance between the brightness of three inner lines (1,2,3) located in the left bottom corner and three inner lines (4,5,6) in the right upper corner determines the ambiguity and orientation of the 3D cube (Figure 1A). The contrast parameter *a* ∈ [0, 1] is the normalized brightness of the inner lines (1,2,3) in the grayscale palette. In turn, the normalized brightness of the other inner lines (4,5,6) is defined as 1 − *a*. Thus, the limiting cases *a* = 0 and *a* = 1 correspond to unambiguous 2D projections of the cube oriented to the left or to the right, respectively, whereas *a* = 0.5 implies a completely ambiguous spatial orientation of the 3D cube.

In the experiment, we used a set of the Necker cube images with *a* = {0.15, 0.25, 0.4, 0.45, 0.55, 0.6, 0.75, 0.85} (Figure 1B), which we divided by subsets of cubes oriented to the left *a* = {0.15, 0.25, 0.4, 0.45} and to the right *a* = {0.55, 0.6, 0.75, 0.85}. At this set, stimuli with low ambiguity (LA) *a* = {0.15, 0.25, 0.75, 0.85}, are easily interpreted, whereas the interpretation of stimuli with high ambiguity (HA) *a* = {0.40, 0.45, 0.55, 0.60} requires more effort [4]. We also assume that HA processing engages more top-down control [20].

**Figure 1.** Experimental paradigm: (**A**) An example of the Necker cube image with the labeled inner edges. (**B**) Visual stimuli (Necker cubes) with different values of the contrast parameter *a*, which determines orientation and ambiguity. (**C**) Experimental protocol including 150-s resting state recordings and presentation of 400 stimuli alternating with the pauses. Colored horizontal stripes indicate 575-s time intervals. These intervals equally divide the stimuli presentation session. Each interval includes 100 stimuli. (**D**) Detailed illustration of a single stimulus presentation and abstract image. The cube presentation starts at the presentation time PT and lasts *τ<sup>i</sup>* ∈ [1, 1.5] s. The decision time (DT) is determined by the interval between PT and button pressing. The pause time *γ<sup>i</sup>* varies from 3 to 5 s. The time interval of interest (TOI1) is the 1.5-s pre-stimulus segment time-locked to the PT.

## *2.3. Experimental Protocol*

Necker cubes (22.55 × 22.55 cm) were shown on a white background using a 24 monitor (52.1 × 29.3 cm) with a 1920 × 1080 pixels resolution and a 60 Hz refresh rate. The distance between the participant and the monitor was 0.79—0.8 m, and the visual angle was ∼0.39 rad.

The duration of the entire experiment was about 40 min for each participant, and included EEG recordings of the eyes-open resting state (≈150 s) before and after the main part of the experiment. Cubes with predefined *a* values (selected from the set in Figure 1B) were randomly presented 400 times during the experimental session. Each cube with a particular ambiguity appeared about 50 times.

Each *i*-th stimulus presentation lasted for time interval *τi*, which ranged from *τmin* = 1 s to *τmax* = 1.5 s. The pauses between subsequent presentations of the Necker cube images, *γi*, ranged from *γmin* = 3 s to *γmax* = 5 s (Figure 1D) and contained an abstract image demo. The abstract image was the white noise picture (Figure 1D).

#### *2.4. Behavioral Estimates*

The participants were instructed to press either the left or right key in response to the left or right stimulus orientation, respectively. For each stimulus, we registered presentation time (PT)—the time between the beginning of the experiment and the moment when the current stimulus appeared on the screen. The behavioral response to each stimulus was assessed by measuring the decision time (DT), which corresponded to the time passed from the stimulus presentation to the button pressing (Figure 1C). We also monitored the correctness using error rate (ER) by comparing the actual stimulus orientation with the subject's response. The actual orientation of the Necker cube was defined by the contrast of the inner edges. Thus, *a* = {0.15, 0.25, 0.4, 0.45} defined the left-oriented cubes, while *a* = {0.55, 0.6, 0.75, 0.85} stood for the right-oriented ones. To define the correctness, we checked whether the subject pressed the left button for *a* = {0.15, 0.25, 0.4, 0.45}, or the right button for *a* = {0.55, 0.6, 0.75, 0.85}. Otherwise, their response was considered as incorrect. We excluded two subjects with *ER* > 20%, as they exceeded the 90th percentile of ER distribution in the group.

## *2.5. EEG Recording*

For registration of EEG signals, a monopolar method and a classical extended 10–10 electrode scheme were used. We recorded signals from 31 channels using an electrode cap, with two reference electrodes on the earlobes (*A*1 and *A*2) and a ground electrode *N* above the forehead. Ag/AgCl cup adhesive electrodes placed on the "Tien–20" paste (Weaver and Company, Aurora, CO, USA) were used for signal acquisition. Immediately before the experiments, a special abrasive "NuPrep" gel (Weaver and Company, Aurora CO, USA) was applied to the electrode attachment areas to increase skin conductivity. We maintained the impedance values in the range of 2–5 kΩ. For registration, amplification, and analogto-digital conversion of the EEG signals, we used a multichannel electroencephalograph "Encephalan-EEG-19/26" (Medicom MTD company, Taganrog, Russian Federation) with a two-button input device (keypad). This device holds the registration certificate from the Federal Service for Supervision in Health Care No. FCP 2007/00124 of 07.11.2014 and European Certificate CE 538571 from the British Standards Institute (BSI).

The raw EEG signals were filtered by a fourth-order Butterworth (1–100)-Hz bandpass filter and a 50-Hz notch filter with built-in acquisition hardware and software. In addition, we performed an independent component analysis (ICA) to remove eye blinking and heartbeat artifacts. To determine components with artifacts, we examined their scalp map projections, waveforms, and spectra. The components containing eye-blinking artifacts usually had leading positions in the component array due to high amplitude. They demonstrated a smoothly decreasing spectrum, and their scalp map showed a strong far-frontal projection. Finally, eye-blinking artifacts had the typical waveform; therefore, those segments of EEG signals were marked by the experienced neurophysiologist, and used for determining the corresponding independent components.

We then segmented the EEG signals into 4-s trials, where each trial was associated with a single presentation of the Necker cube, and included a 2-s interval before and 2-s interval after the moment of the stimulus demonstration. After the EEG pre-processing procedure, we excluded some trials due to the remaining large-amplitude artifacts. To exclude trials containing large amplitude artifacts, we used the *z*-value threshold *z* < 1. The rejection procedure was performed using FieldTrip toolbox in Matlab [21].

After all preprocessing procedures, we had 52 ± 11 SD trials for the interval 1, 47 ± 11 SD trials for the interval 2, 47 ± 11 SD trials for the interval 3, and 55 ± 11 SD trials for the interval 4. We calculated the wavelet power for each trial in the (4–40)-Hz frequency range using the Morlet wavelet, and the number of cycles *n* was defined as *n* = *f* , where *f* is the signal frequency. Finally, we computed the event-related spectral perturbation (ERSP) by normalizing the wavelet power estimates *W* to the wavelet power of 40-s resting-state EEG as *ERSP* = (*W* − *Wrest*)/*Wrest*. All processing procedures were performed offline.

Our goal was to study how the participant's state changed in the course of the experiment, regardless of the type of stimulus. Therefore, we measured brain activity before the start of the stimulus presentation (1.5-s prestimulus interval, TOI1 in Figure 1D).
