The study sample consisted of 41 individuals (34 females, mean age 26.7, SD = 7.18 and 7 males, mean age 26.7, SD = 5.71). The age range of the whole sample was from 18 to 42 years. According to the power analysis conducted using the G*power software (Faul et al., 2007), the sample size exceeds the minimally required sample size (N = 29) to detect a medium effect in correlational analysis (with R of 0.5, a two-tailed test, alpha error probability of 0.05 and 80% power).

Participants in the study were screened for any mental, neurological, or serious somatic diseases, recent head injuries, or taking psychotropic drugs. Additionally, they were asked to refrain from smoking or drinking coffee for a few hours prior to the study. All applicable subject protection guidelines and regulations were followed in conducting the research in accordance with the Declaration of Helsinki. All participants signed written informed consent. The study protocol was approved by the Ethical Commission of the Institute of Higher Nervous Activity and Neurophysiology of the Russian Academy of Sciences (Ethics protocol: No. 2, 30 April 2021).

We employed the Subjective Minute Test, a variation of the duration production task designed to assess time perception by requiring participants to estimate an interval of 1 min. This test was programmed using Presentation software (Neurobehavioral Systems, Inc., Berkeley, CA, USA, www.neurobs.com), and was conducted alongside EEG recordings. Participants were instructed to produce a minute-long interval four times, alternating between open and closed eyes. Specifically, they completed two trials with their eyes open and two with their eyes closed. To initiate the task, participants pressed the spacebar, signaling the start of the minute, followed by the appearance of a cross on the screen. Then they pressed the spacebar a second time when they believed a minute had elapsed (Figure 1). Participants were explicitly instructed to refrain from using any external aids, such as tapping or counting, relying solely on their internal sense of time.

When selecting the number of trials, we adhered to the fact that a duration starting from 20 s is appropriate for analyzing resting-state activity (Möcks and Gasser, 1984). The alternation between open and closed eyes aimed to mitigate drowsiness during the closed-eye intervals (Maltez et al., 2004). This approach yielded EEG recordings consisting of two trials under the open-eye condition and two under the closed-eye condition, presented in an alternating sequence. The length of the EEG segment for each condition varied according to the individual participant’s responses.

During the experiment, 64-channel EEG was acquired. The electrodes were positioned according to the 10-10 international system. This number is generally sufficient to source reconstruction as can be referred from previous studies (Michel et al., 2004). BrainProduct amplifier with a 0.1–100-Hz analog bandpass filter was used for signal amplification. The sampling rate was 500 Hz. Electrode impedances were kept at or below 15 kΩ. Due to the active electrode technology, the recording system we use can record high quality EEG data with higher impedances (up to 25–50 kΩ). The AFz electrode was used as a ground and the Cz as a reference. A finite impulse response (FIR) filter was used with a frequency range of 1–30 Hz. Bad channels were manually selected and interpolated. The data were recomputed to the average reference. Artifact correction was performed using the independent component analysis (ICA) method (FastICA algorithm, 30 components). Artifacts associated with eye blinks, and movements were removed. The selection of components adhered to the recommendations provided by the ALICE toolbox (Soghoyan et al., 2021). EEG data were preprocessed and analyzed using MNE-python (Gramfort, 2013).

The EEG segments were cropped according to the shortest subjective minute production in the sample (∼30 s). Preprocessed EEG segments corresponding to open and closed-eye conditions (60 s each) were divided into 2.5-s epochs with a 50% overlap. Epochs containing artifacts were automatically discarded if their peak-to-peak amplitude surpassed a threshold of 180 μV. The mean number and the standard deviation of epochs were 45.2 ± 1.6 per condition per participant.

We investigated cortical sources activity within the following frequency ranges: theta (4–7 Hz), alpha (8–12 Hz), and beta (13–30 Hz). The current study focuses on identifying the locations and characteristics of brain activity through measurements obtained from electrodes positioned on the scalp. This undertaking is linked to the EEG “inverse problem,” which emerges from the fact that multiple source configurations can produce identical EEG signals, complicating the accurate identification of source locations without additional information or assumptions regarding the sources. As a result, various algorithms have introduced different assumptions and constraints to tackle this issue. Among these approaches, sLORETA (Pascual-Marqui, 2002) is particularly notable as a robust method. It is a non-parametric technique that estimates the distribution of electrical activity in the brain while implementing constraints that enhance spatial resolution and increase stability against noise. sLORETA has demonstrated efficacy in numerous cognitive studies (e.g., Dyson et al., 2010; Ocklenburg et al., 2012; Nguyen et al., 2014). Furthermore, sLORETA is recognized for having the minimal localization error for test sources (Pascual-Marqui, 2002).

A forward solution model was calculated based on the 10-10 montage and averaged brain MRI data (MNE “fsaverage” dataset). The boundary element model (BEM) (see, e.g., Mosher et al., 1999 was constructed using the linear collocation method with the isolated skull approach (Hamalainen and Sarvas, 1989). The default MNE electrical conductivities were applied (0.3 S/m for the brain and the scalp, and 0.006 S/m for the skull) (Gramfort et al., 2014). The source space was computed with an “oct6” resolution, which leads to 4,098 vertices per hemisphere with an average distance of ∼5 mm between them. The resulting CSD distribution was separated into time series for 68 cortical structures (regions of interest, ROI) according to Destrieux Atlas (Freesurfer “aparc” parcellation). The CSD estimations were averaged over epochs, as the temporal dynamics of the signal were not of interest. As a result, mean sLORETA CSD value was obtained for each of 68 ROI.

Spearman’s test was employed to calculate the correlation between the subjective minute duration and the CSD during the task in both close and open eye states (p < 0.05). In addition, the accuracy of subjective minute production was assessed as a modulus of the difference between a subjective estimation and the objective minute. The Benjamini–Hochberg correction was used to control the multiple comparisons problem (Benjamini and Hochberg, 1995). This method is designed to control the false discovery rate (FDR), which is the expected proportion of false discoveries among the rejected null hypotheses. All calculations were performed by means of the free R software packages.

The involvement of numerous brain regions and differences between eye-open and eye-closed conditions may indicate that beta activity facilitates the integration of sensory information, attention, and memory, thereby shaping our subjective experience of time. The obtained results are of interest in light of extensive recent research on the role of beta oscillations in time perception. Initially, alpha oscillations were identified as a potential contributor to time perception (Anliker, 1963; Treisman, 1984; Treisman et al., 1990). Subsequently, the role of other brain rhythms in this process was explored, yet the findings were not always consistent as it soon became apparent that the experimental design shapes the outcomes of such studies. This issue is thoroughly explained by Wiener and Kanai (2016). They examined the role of different frequency bands and regions in time perception and highlighted that temporal processing in the brain is an adaptable mechanism, which induces dynamic interactions among brain structures and networks (Wiener et al., 2011; Wiener and Kanai, 2016). Following this, Wiener et al. (2018) determined that the beta-rhythm power might reflect a shift in the value of the memory standard over time. Using transcranial alternating current stimulation (tACS), they observed that beta-rhythm stimulation of the fronto-central cortex leads to perception of visual stimuli as lasting longer. As a result, the researchers inferred the critical role of beta oscillations in encoding and retaining memories of temporal intervals (Wiener et al., 2018).

In studying the function of beta oscillations, the problem of distinguishing between motor and perceptual aspects of timing arises. Time perception tasks often imply conducting a movement to indicate a task-relevant response. Frank (2011) emphasizes the difficulty of separating temporal and motor components given the interaction of cognitive and motor circuits in the corticostriatal loop. Tasks that necessitated motor system involvement were labeled as “motor” timing, whereas those that did not were referred to as “perceptual” timing. The use of motor timing paradigms in non-human primates has revealed that putamen-originating beta power is larger for longer durations (Bartolo et al., 2014; Bartolo and Merchant, 2015). However, Kononowicz and Rijn (2015) hypothesized that if beta power dynamics were solely linked to motor sequence generation, fluctuations in beta power would not correlate with behavior on a time production task. At the same time, if beta power were linked to the internal sense of time, it should correlate with the length of the produced interval. The authors conclude that the initial motor inhibition at the onset of an interval affects the duration of produced intervals. This suggests that beta oscillations play a role in interval timing beyond just motor preparation (Kononowicz and Rijn, 2015). The problem of the influence of motor response and the impact of stimulus predictability on the outcomes is also discussed in detail in Wiener et al. (2018).

Thus, there is more and more evidence in favor of the existence of a division into cognitively controlled timing and motor timing, which are probably characterized by the involvement of different brain regions. In our study, we used the cognitively controlled timing paradigm, which presupposes the use of supra-second intervals (the Subjective Minute Test), which assumes a production of an interval of about 60 s. The data analysis was carried out by cutting the EEG into epochs, which then were averaged; therefore, we can claim a general absence of a motor component in our results. Nevertheless, the only frequency range in which we observed correlations with the duration of the subjective minute—within our analysis of the alpha, beta and theta bands—was the beta range. A MEG study conducted by Kulashekhar et al. (2016) employing a non-motor paradigm, revealed that comparing durations led to notably increased beta power amplitudes in the parietal-frontal network, which includes the SMA, while encoding stimulus durations. Moreover, elevated beta amplitudes were linked to both correct and incorrect selections, highlighting their substantial behavioral importance (Kulashekhar et al., 2016). These findings, when considered alongside our own, indicate that the beta range is likely to be associated not only with motor aspects but also with the cognitive components of time perception.

During eyes-closed sessions, beta activity of the left precuneus correlated positively with subjective minute duration, meaning higher beta power was linked to longer subjective minutes and deceleration (dilation) of subjective time. Beta activity in the right SFG and left superior parietal lobule showed negative correlations, indicating higher beta power association with shorter subjective minutes and acceleration (compression) of subjective time. The right SFG exhibited a tendency for relation of another direction with subjective minutes during the open-eye condition.

The SFG includes subregions which are often reported to be engaged in temporal processing, such as the SMA and dorsolateral part of the prefrontal cortex (PFC). It is well-known that the SMA plays a crucial role in temporal processing (Casini and Vidal, 2011; Schwartze et al., 2012; Naghibi et al., 2024). According to a recent meta-analysis of fMRI studies, the SMA is one of the structures which is consistently involved in time perception across diverse experimental paradigms and presentation modalities (Naghibi et al., 2024). The SMA’s role in time perception aligns with embodied cognition theories, which propose that bodily states generated through action can serve as anchors for cognitive processing. This suggests that motor activity, including that in the SMA, may represent an active form of perception, structuring the perception of time through sensorimotor integration and top-down predictions (Fernandes and Garcia-Marques, 2019). While the SMA may represent motor derivatives in time perception, not specifically sensitive to the time scale, the PFC is involved in cognitively controlled timing mechanisms for longer durations, which are tightly linked with working memory and attention (Coull et al., 2011). It has been demonstrated that the disruption of the right PFC activity with transcranial magnetic stimulation (TMS) leads to impairment of supra-second timing (Jones et al., 2004; Koch et al., 2007), and there is causal evidence, obtained with transcranial direct current stimulation, of the involvement of the right dlPFC in time perception (Li et al., 2022; Johari et al., 2023).

The joint activity of the right SFG and parietal regions (superior parietal left – lateral part and precuneus as a medial part) represents “cognitively controlled” timing system, which is systematically observed in neuroimaging time perception studies (Lewis and Miall, 2003; Üstün et al., 2017; Naghibi et al., 2024). In addition, parietal regions are implicated in multisensory integration for body positioning and movement in space. There is evidence that within these areas the convergence between the space and time representations occurs, and the underlying mechanism could be cross-modal magnitude encoding (Walsh, 2003; Buzsáki and Llinás, 2017). Another point is that the sensory integration occurring in these regions is key to represent the bodily self. With intracranial stimulation, it was shown that disruption of anterior precuneus leads to loss of the bodily physical self accompanied with feeling of falling, dropping, and dizziness, and the superior parietal lobule showed functional connections with this “hot zone” (Lyu et al., 2023).

Together the left SPG, right SFG, and left precuneus comprised the core “closed-eyes” time estimation network although the left precuneus and right SFG had a tendency for an effect of different direction during the eyes-open condition. Their coherent involvement specifically during the eyes-closed condition could be interpreted as an inward appeal to internal sources of information to judge about time passage. In line with the known functions of these structures, these internal sources of information could be related to spatial-time representations of a bodily self according to an embodied framework (Naghibi et al., 2024).

During eyes-open sessions, beta activity exhibited negative correlations with subjective minute duration in the left caudal ACC, bilateral cuneus, and bilateral PCC. This suggests that higher beta power in these regions was related with shorter subjective minutes and acceleration (compression) of subjective time. Positive correlations between the beta activity and subjective minute duration were found in the right lingual gyrus and bilateral parahippocampal gyrus. Therefore, higher beta-rhythm power in these areas was associated with longer subjective minutes and deceleration (dilation) of subjective time. In addition, the PCC and cuneus exhibited tendencies toward inverse relations with subjective minutes in the eyes-closed condition.

In contrast to the eyes-closed condition, the set of structures which showed relationships with a subjective minute during the eyes-open condition might indicate attentional focus to external sources of information to judge about time passage. For instance, a combination of the caudal ACC, bilateral cuneus, and right lingual gyrus involvement could point to activation of the mechanisms of visually guided attention (Menon, 2015). The bilateral cuneus and lingual gyrus are components of a visual network, thereby indicating activity of modality-dependent temporal mechanisms (Bueti, 2011). The caudal ACC is relatively rarely reported to be engaged in time perception processes but its role in supra-second timing is highlighted in several fMRI studies, which is usually interpreted as attentional control or decision-making derivative (Pouthas et al., 2005; Üstün et al., 2017).

The PCC and parahippocampal gyrus are the nodes of the default mode network (DMN) (Zhao et al., 2022). Being one of the most metabolically active cortical regions both at rest and during cognitive tasks, the PCC possesses various functions including pain processing, episodic memory retrieval, and conscious awareness (Raichle et al., 2001; Leech and Sharp, 2014). It was found that in the context of time perception, the PCC stands for the aspects related to working memory (Üstün et al., 2017). The parahippocampal gyrus is also involved in working memory processes. However, while the PCC is responsible for retrieval, the parahippocampal gyrus may underlie the integrative and maintenance functions (Luck et al., 2010; Ward et al., 2014). Taken together, the activity of these structures points to involvement of working memory to encode, maintain, and retrieve perceptual moments, which constitute subjective time.

Among the limitations of the study, it is important to mention that EEG possesses a relatively low spatial resolution; thus, we analyzed the activity of cortical sources only. However, the subcortical structures, such as the basal ganglia, also play a crucial role in time perception. Another limitation of this study is its correlational design, which prevents establishing causal links between neural activity and time perception. Future research using causal methods, such as brain stimulation, is needed to confirm the functional roles of the identified brain regions. In addition, we could not completely rule out the influence of the motor component in our study, although we believe that it was largely eliminated by the peculiarities of the paradigm, which involved pressing only once at the beginning and end of the subjective minute, and by the division of the data into epochs with subsequent averaging.