Unraveling the Neural Network: Identifying Temporal Labeling of Visual Events through EEG-Based Functional Connectivity Analysis of Brain Regions

The study addresses how the brain processes temporal information and how predictability of event timing modulates neural connectivity underlying time perception. Prior work suggests the brain must constantly predict the timing of future events to optimize sensory and motor responses, and temporal processing engages distributed cortical and subcortical structures. EEG, with millisecond temporal resolution, enables examination of rapidly changing functional connectivity. The authors investigate whether functional connectivity patterns differ between predictable and unpredictable timing conditions, and across delays (83, 150, 400, 800 ms), during a visual labeling task. Building on earlier work [49] that reported predictability effects on alpha power at the longest delay (800 ms), this study tests the hypothesis that predictability modulates inter-regional connectivity across frequency bands and delays, and seeks to identify brain regions, including prefrontal areas, implicated in temporal processing.

The paper reviews perspectives on time as a construct shaped by memory, attention, and prediction, and emphasizes the role of temporal expectations in perception and behavior. Prior neurophysiological evidence implicates cerebellum, thalamus, posterior parietal cortex, prefrontal cortex, and motor cortex in temporal processing across sub-second to multi-second scales. Damage to right frontal cortex and subcortical structures can disrupt timing. EEG/MEG connectivity estimators such as PLI and related measures have been widely used. Behavioral and EEG studies show temporal prediction influences sensory processing, including alpha-band modulation at long delays [49], entrainment mechanisms in attention and temporal expectation, and associations between timing tasks and activity in parietal and prefrontal regions. Studies also relate beta power/coherence to perceived time and report connectivity between frontal-central and posterior regions across frequencies. These works motivate examining frequency-specific connectivity differences under predictable versus unpredictable timing.

Data: EEG data were sourced from a previously published experiment [49]. Participants were 29 healthy right-handed males (mean age ~24 years) with normal vision. EEG was recorded with 64 Ag/Cl electrodes at 1000 Hz, referenced to mastoids, with impedances <6 kΩ. Visual stimuli consisted of clockwise/counterclockwise rotations requiring corresponding button presses. Each trial began with a fixation cross and a 200-ms cue, followed by a variable inter-trial delay sampled from a Gaussian distribution (mean 900 ms, variance 600 ms) to avoid rhythmic entrainment. The critical manipulation was the cue–stimulus interval (delay) under two conditions: predictable blocks (one fixed target delay per block: 83, 150, 400, or 800 ms; 48 trials per block) and unpredictable blocks (delay randomly selected among 83/150/400/800 ms per trial). Monitor frame timing (60 Hz) was considered in delay selection. Preprocessing: Signals were processed in MATLAB/EEGLAB. DC offset was removed; a high-pass filter at 0.5 Hz mitigated low-frequency drifts. Noisy channels were visually identified and replaced via interpolation from neighbors. Common noise was reduced via re-referencing/interpolation procedures, and ICA was applied to remove artifacts (eye blinks, eye/neck movements). From the remaining 62 electrodes, 19 standard 10–20 electrodes (covering the whole scalp) were selected for analysis. Data were segmented from −500 ms to +500 ms around stimulus onset. Connectivity estimation: Functional connectivity was computed using the Phase Lag Index (PLI), which represents the consistency of non-zero phase lags between pairs of signals and is relatively robust to volume conduction. Frequency bands analyzed: delta (1–4 Hz), theta (4–8 Hz), alpha (8–13 Hz), beta (13–30 Hz), and gamma (~30–49 Hz). For each participant, condition, delay, and band, a 19×19 symmetric connectivity matrix was produced; elements reflect pairwise inter-channel coupling (directionality not inferred). Average PLI values were used to summarize connectivity. Statistical analysis: Normality was assessed with Kolmogorov–Smirnov tests (P≥0.05). Paired t-tests compared clockwise vs. counterclockwise stimulation within each delay and band. One-way ANOVAs evaluated differences among the four delays within each condition (predictable, unpredictable), with FDR correction, and post hoc testing to localize effects. Paired t-tests compared predictable vs. unpredictable conditions for each delay. For visualization, edges exceeding mean ± SD thresholds were displayed, with line thickness indicating strength. Significance threshold α=0.05.

- No significant differences between clockwise and counterclockwise stimulation across delays and bands; behavioral accuracy was ~83% for direction identification. - Delay effects within condition: Significant differences across delays were found in both predictable and unpredictable conditions, most prominently in beta, theta, and gamma bands; effects were minimal in delta and slight in alpha. Differences were generally more numerous in the predictable condition. - Detailed band-specific patterns across delays: - Delta: In the predictable condition, differences increased notably when comparing 400 vs. 800 ms and other pairs involving 800 ms, with many frontal connections. In the unpredictable condition, the largest difference was 150 vs. 800 ms with more posterior connections. Across delays, connections often involved central–frontal links; heightened central–prefrontal connectivity appeared particularly at 83 ms. - Theta: Largest differences at 83 vs. 800 ms, 150 vs. 800 ms, and 83 vs. 400 ms in both conditions. Predictable condition differences were widespread initially (83 vs. 800 ms), then prominent in frontal and posterior regions (150 vs. 800 ms). Unpredictable condition showed strongest differences for 83 vs. 800 ms, mainly frontal then posterior. - Alpha: Small but significant differences between 83 and 150 ms (frontal–parietal correlations) in both conditions. Importantly, for 400 ms delay, average alpha connectivity was higher in the unpredictable than predictable condition, especially between occipital and temporal regions. - Beta: Greatest differences between 83 vs. 800 ms and 150 vs. 800 ms in both conditions; predictable condition showed persistent strong effects (83 vs. 800 ms), then 150 vs. 800 ms with stronger parietal then frontal correlations. - Gamma: Predictable condition showed largest differences at 83 vs. 400 ms, then 83 vs. 800 ms, strongest in parietal then posterior regions. Unpredictable condition showed significant differences at 150 vs. 400 ms (posterior-dominant), then 150 vs. 800 ms (parietal-dominant). Unpredictable scenarios exhibited broad gamma connectivity across the brain for all delays; predictable had gamma connectivity mainly at 83 and 150 ms. - Predictable vs. unpredictable comparisons per delay (all bands 1–40 Hz): Significant differences at all delays (α=0.05), with larger differences at 150 and 400 ms. Connectivity generally decreased as delay length increased; 150 ms often showed stronger connections than 83 ms. At 400 ms, strong prefrontal-to-posterior connections were evident; at 800 ms, connectivity in delta, theta, and beta bands declined versus shorter delays in both conditions. - Regional roles: Right prefrontal cortex emerged as pivotal in time perception, with consistent prefrontal involvement across bands and conditions, and interhemispheric connections observed across delays. - Relation to prior work: Unlike [49], which found predictability effects on alpha power only at 800 ms, this study observed significant predictable vs. unpredictable differences at all delays, with specific increases in alpha connectivity in the unpredictable condition at 400 ms and greater anterior differences at 800 ms.

Findings indicate that temporal predictability systematically modulates functional connectivity across distributed cortical networks and frequency bands during visual event labeling. Greater and more widespread differences in beta, theta, and gamma bands, particularly in predictable contexts, suggest enhanced large-scale coordination when event timing is known. The consistent reduction in connectivity at 800 ms (delta/theta/beta) implies changing network demands or reduced preparatory coupling at longer intervals. Elevated alpha-band connectivity in the unpredictable condition at 400 ms, especially between occipital–temporal regions, points to increased sensory processing demands when timing is uncertain. The prominent role of prefrontal regions—especially right prefrontal cortex—and robust prefrontal–posterior interactions across delays align with models positing top-down control and temporal prediction mechanisms. The results extend prior work [49] by demonstrating that predictability affects connectivity not only at the longest delay but across all delays, highlighting frequency- and region-specific dynamics associated with temporal expectation and time perception.

The study demonstrates that temporal predictability and delay length shape EEG-based functional connectivity during visual event labeling. Significant delay-dependent differences were most pronounced in beta, theta, and gamma bands, with broader effects under predictable timing. The unpredictable condition showed higher alpha connectivity at 400 ms (notably occipital–temporal), and connectivity in delta, theta, and beta bands declined at 800 ms in both conditions. Prefrontal—particularly right prefrontal—regions played a key role, with strong interactions to posterior areas. These findings refine understanding of the neural mechanisms of time perception by revealing frequency-specific, regionally distributed connectivity patterns modulated by predictability. Future work should employ additional delays above 500 ms to capture neural dynamics more clearly and consider reducing trials per delay to limit fatigue while preserving experimental integrity.

The authors note that intervals below 500 ms may not elicit markedly different neuronal spiking patterns, suggesting the current delay set constrains interpretability and advocating inclusion of more delays above 500 ms in future studies. They also recommend fewer trials per delay to reduce participant fatigue, indicating potential effects of fatigue on results.