Schematic of latency analysis.

We presented a 7-minute auditory narrative twice to each participant. (A) Moving time windows were applied to the pair of chosen channels. (B) Within each time window, we measured the cross correlation between the pair. The latency τ was defined as the time lag yielding the maximum cross-correlation between the channels. (C) Cross correlation (blue to yellow colors) and latency τ (black line) of this channel pair are illustrated for all time windows. The latency τ fluctuates throughout the time course.

Latency and power changes for an electrode pair in the middle STG.

Time windows are sorted by value of Latency τ between electrodes X and Y. For each channel pair, we arranged the time windows in the increasing order of the latency τ, and measured mean alpha-band (10 Hz) power and mean broadband (65+ Hz) power across channels for each corresponding time window. (A) Cross correlation between two chosen channels is shown for each time window, sorted by the value of latency τ. Black line denotes the inter-electrode latency τ. (B) Alpha-band (7-14Hz) and broadband power in each channel for each 2s time window. Blue asterisks denote alpha peaks and red asterisks denote broadband power peaks (top 10% of values). The latency τ is correlated with alpha power and anti-correlated with broadband power.

Latencies and coupling strengths increase with low-frequency power.

(A) Global latency-flow analysis. Latency-flow patterns for high alpha power time windows and low alpha power time windows are shown. Top 10% of the windows yielding highest alpha power and bottom 10% of the windows yielding lowest alpha power are chosen, and latency flows are computed across the chosen windows respectively. Yellower color of arrows denotes higher cross correlation. Larger size of arrows denotes longer latencies. The size of arrows on each flow map is scaled for readability: the grey arrow next to each participant’s brain map denotes scale of latency flow arrows. (B) Time delay vs. alpha power. Channel pair analysis: Correlation between latencies of pairs and global alpha power for each pair across time windows are computed and the distribution is shown in histograms. Red histogram represents nearest neighbor pairs and blue histogram represents next-nearest neighbor pairs. The distribution is positively skewed. Time window analysis: Mean latencies across pairs and mean global alpha power for each time window are computed and shown as scatterplots. The distributions yield positive correlation. (C) Maximum cross correlation vs alpha power. Channel pair analysis: Correlation between maximal cross correlation of pairs and global alpha power for each pair across time windows are computed and the distribution is shown in histograms. The distribution is positively skewed. Time window analysis: Mean maximal cross correlations across pairs and mean global alpha power for each time window are computed and shown as scatterplots. The distributions yield positive correlation.

Reliability of latency patterns between run 1 and run 2.

(A) Two runs of the same auditory stimulus were applied to a participant., and channels with high reliability between two runs were chosen. (B) The latency pattern for run 1 is shown in the order of increasing latency. (C) The latency pattern for run 2 is shown in the same order used in run 1. The latency reliability between two runs was 0.25 for the chosen channel pair. (D) Latency reliability vs. LF power reliability for all channel pairs from 10 participants.

Coupled oscillator model on brain network.

(A) Stuart-Landau coupled oscillator model is applied to diffusion tensor imaging (DTI) structural brain network consisting of 78 nodes. (B) Mean delay across the nodes, mean of maximal cross correlations across the nodes, mean of power of the nodes are shown. Below a certain critical coupling strength S (∼1.0), all three measures are positively correlated with each other as the coupling strength S increases.