Extended Data Fig. 9: Behavioral and neural profiles between win and loss trials.
To examine whether the behavioral and neural profiles differed between win and loss trials, we performed ANOVAs on the behavioral and neural data, including the trial outcome as a within-subject independent variable. For the behavioral analysis, we conducted a three-way ANOVA of repeated measurement including Velocity, State, and Outcome as within-dyads variables. For the intra- and inter-brain analysis, we averaged high-frequency power (30-150 Hz), and calculated inter-brain correlation across win and loss trials respectively. We conducted two-way ANOVAs of repeated measurement for power and inter-brain correlation including State and Outcome as within-dyads variables. a, b, Behavioral results revealed a significant Velocity × State × Outcome three-way interaction, indicating stronger Velocity × State interaction for win trials than loss trials (F1,22 = 13.897, p = 0.001, ηp2 = 0.387, 90% CI: 0.120, 0.564, ANOVA of repeated-measurement). Further decomposing the three-way interaction, we conducted ANOVA separately for each cooperation state and found significant Velocity × Outcome interaction in both initiation (F1,22 = 10.526, p = 0.004, ηp2 = 0.324, 90% CI: 0.074, 0.514) and maintenance (F1,22 = 5.117, p = 0.034, ηp2 = 0.189, 90% CI: 0.008, 0.397) states. Specifically, for initiation state (a), we found wining trials are featured by increased \({v}_{C}\) (t22 = 2.711, p = 0.013, Cohen’s d = 0.657, 95% CI: 0.117, 0.882) and decreased \({v}_{G}\) (t22 = -3.128, p = 0.005, Cohen’s d = 0.536, 95% CI: -0.345, -0.070), suggesting that during cooperation initiation, more focus on coordination and stronger suppression of collective goal pursuit contributed to wining the trial. For maintenance state (b), lower level of \({v}_{C}\) (t22 = -4.509, p = 1.738×10-4, Cohen’s d = 0.536, 95% CI: -0.345, -0.070) and a trend of higher \({v}_{G}\) (t22 = 1.848, p = 0.078, Cohen’s d = 0.313, 95% CI: -0.022, 0.383) were found in win trials compared to loss trials, indicating less concentration on teammate coordination and more effort exerted toward collective goal during cooperation maintenance facilitates winning the trial. These behavioral findings once again highlight the important role of teammate coordination in the success of cooperation. More engagement in interpersonal coordination during cooperation establishment, and less during cooperation maintenance enable cooperators to achieve cooperation more efficiently. c, Contrast of power between win and lose trials showed a significant State × Outcome interaction only in amygdala (F1,156 = 15.069, p = 1.526×10-4, ηp2 = 0.088, 90% CI: 0.030, 0.164), but not in TPJ (F1,229 = 0.348, p = 0.556, ηp2 = 0.002, 90% CI: 0.000, 0.021). We found that high-frequency band neural activity increased in initiation state (t156 = 2.495, p = 0.014, Cohen’s d = 0.296, 95% CI: 0.004, 0.036) and decreased in maintenance state (t156 = -2.470, p = 0.015, Cohen’s d = 0.249, 95% CI: -0.018, -0.002) for win trials compared with loss trials. d, For inter-brain correlation, we found a significant State × Outcome interaction only in TPJ (F1,457 = 9.381, p = 0.002, ηp2 = 0.020, 90% CI: 0.004, 0.046), but not in amygdala (F1,246 = 1.106, p = 0.294, ηp2 = 0.004, 90% CI: 0.000, 0.028). This interaction was mainly attributed to a significant larger inter-brain correlation at high-frequency band in maintenance state in win trials than loss trials (t458 = 3.506, p = 5.002×10-4, Cohen’s d = 0.228, 95% CI: 0.005, 0.019). The result indicates that more synchronized TPJ activities between teammates during well-maintained cooperation contributed to higher efficiency and a satisfactory outcome. Together, these results suggested that the significant behavioral and neural patterns might facilitate efficient cooperation since these patterns were more prominent in winning trials than losing trials. In a and b, overlaid dots represent session dyads with both win and loss trials (n = 23). In c and d, violin plots depict the probability distributions of power or inter-brain neural correlation across channel (c, amygdala: n = 157) or channel pairs (d, TPJ, n = 479). Correlation coefficients were Fisher-transformed, and coefficient outliers (mean ± 3 s.d.) were excluded prior to statistical tests. ANOVA of repeated-measurement are used for multi-factor analyses. Paired t-tests are used for pair-wise comparisons. Data are mean ± 95% CI. Statistical tests are two-sided unless otherwise stated. †p < 0.05, *p < 0.05, **p < 0.01, ***p < 0.001, n.s. not significant.
