Extended Data Fig. 8: The neural results of the 4-person team, non-competitive cooperative running game.
a, b, Normalized power in the amygdala (a, n = 103 channels) and TPJ (b, n = 162 channels) during cooperation initiation and maintenance states plotted as a function of frequency. We replicated that initiation and maintenance states were differentiated by a significant difference in power in the 30 Hz to 150 Hz high-frequency band. This power difference was further confirmed by a significant power increase in the initiation state but a significant power decrease in the maintenance state within the 30-150 Hz frequency band. Horizontal orange, blue, and black bars indicate the significant frequency bands for the initiation state, maintenance state, and between-state contrast (cluster-based permutation tests, pcorr < 0.05, n = 10000). c, d, Normalized high-frequency power averaged between 30-150 Hz during initiation and maintenance states in the amygdala (c, n = 23 patients) and TPJ (d, n = 18 patients). The state-dependent high-frequency activity pattern was strikingly consistent across sessions in both amygdala (c, initiation vs. maintenance: t22 = 8.160, p = 4.240 × 10-8, Cohen’s d = 2.724, 95% CI: 0.082, 0.137, paired t-test, consistent in 100% teams) and TPJ (d, initiation vs. maintenance: t17 = 8.380, p = 1.929 × 10-7, Cohen’s d = 2.858, 95% CI: 0.056, 0.094, paired t-test, consistent in 94% teams). Specifically, we found significant power increases in the initiation state (amygdala: t22 = 6.310, p = 2.380 × 10-6, Cohen’s d = 1.270, 95% CI: 0.045, 0.089; TPJ: t17 = 6.337, p = 7.444 × 10-6, Cohen’s d = 1.426, 95% CI: 0.033, 0.066, one-sample t-tests) but power decreases in the maintenance state (amygdala: t22 = -10.228, p = 7.995 × 10-10, Cohen’s d = 2.059, 95% CI: -0.051, -0.034; TPJ: t17 = -8.210, p = 2.558 × 10-7, Cohen’s d = 1.848, 95% CI: -0.033, -0.019). Overlaid dots represent session participants. e-j, Time-frequency spectrograms (t-value maps) of epochs from initiation state (e, f), maintenance state (g, h), and between-state contrast (i,j) in the amygdala/TPJ. In the initiation state, we found clusters exhibiting power increases covering a broad frequency range of 30-150 Hz, in both the amygdala (e, significant clusters emerged at initiation onset and lasted up to 1440 ms) and TPJ (f, a significant cluster covering the entire 2000 ms epoch). In the cooperation maintenance state, there were significant power decreases across a broad frequency range of 30-150 Hz in both the amygdala (g, beginning 260 ms after the maintenance onset) and TPJ (h, from 300 ms after the maintenance onset). Direct comparisons of high-frequency power between initiation and maintenance states revealed significant clusters in amygdala (i, one cluster between onset and lasted up to 1690 ms, 30-150 Hz; another covering 1830-2000 ms and 39-142 Hz), and in TPJ (j, covering the entire 2000 ms epoch). For display purposes, power heat maps were smoothed by convolving power time series with a moving Gaussian window of 250 ms for each frequency bin. Black contours delimit significant clusters after two-sided cluster-based permutation tests (pcorr < 0.05, n = 10000). k,l, Inter-brain neural correlation (Pearson correlation coefficients) in 30-150 Hz between two interacting brains across all channel pairs in the amygdala (k, n = 286) and TPJ (l, n = 555), during initiation and maintenance states. The results replicated the findings in the three-legged racing game that the amygdala high-frequency power of the cooperating teammates was significantly correlated in the initiation state (t282 = 8.656, p = 1.904 × 10-16, Cohen’s d = 0.513, lower 95% CI: 0.023, one-sided one-sample t-test) and the maintenance state (t280 = 2.096, p = 0.018, Cohen’s d = 0.125, lower 95% CI: 0.001). The TPJ power in the 30-150 Hz band of two teammates was significantly correlated during both the initiation (t553 = 5.858, p = 4.027 × 10-9, Cohen’s d = 0.249, lower 95% CI: 0.008) and maintenance states (t550 = 5.017, p = 3.554 × 10-7, Cohen’s d = 0.213, lower 95% CI: 0.003). Inter-brain neural correlation also exhibited state-dependency, however, with stronger correlation during initiation than maintenance states in both amygdala (t277 = -7.174, p = 6.655 × 10-12, Cohen’s d = 0.591, 95% CI: -0.032, -0.018, paired t-test) and TPJ (t549 = -3.501, p = 5.005 × 10-4, Cohen’s d = 0.201, 95% CI: -0.011, -0.003). Correlation coefficients were Fisher-transformed. Data are mean ± 95% CI. In k and l, violin plots depict the probability distributions of inter-brain neural correlation across channel pairs, with black dots represent the mean. Prior to statistical tests, any correlation coefficient outliers (mean ± 3 s.d.) were excluded. Statistical tests are two-sided unless otherwise stated. *p < 0.05, ***p < 0.001.
