Fig. 6: The approach to establish shared contributions of eye movements and neural activity to speech tracking.

a With a mediation analysis approach, we investigated the shared contribution of eye movements and neural responses to the speech envelope encoding. For this, we compared the plain effect (c) of the speech envelope on neural activity to its direct (residual) effect (c’) by including an indirect effect via eye movements. The two models described above were calculated to predict neural responses of all 102 magnetometer channels separately. b Exemplary temporal response functions (TRF) for the plain effect (c) and direct (residual) effect (c’) and respective differences (c – c’) to visualize the mediation effect over time. For illustration, we chose the channel that showed the highest prediction accuracy for the plain effect (c) in the single-speaker target condition (r_s = 0.12). TRFs were resampled to 500 Hz for visualization. Center lines represent the mean, shaded areas represent 95% confidence intervals c We used a cluster-based permutation-dependent t-test to compare TRFs (using absolute values) of both models at each sensor and report effect size Cohen’s d averaged over sensors within significant clusters. Cluster-based permutation tests (one-sided) for the contrasts (c’ < c) revealed a small mediation effect by eye movements for the relationship between the speech envelope and neural responses over left parietal sensors in the single speaker condition (bright purple; t(29) = −4.40, p < 0.001, Cohen’s d = −0.80), in the multi-speaker condition for a target (t(29) = −4.90, p < 0.001, Cohen’s d = −0.90) as well as a distractor speaker(yellow; t(29) = −4.68, p < 0.001, Cohen’s d = −0.85). Marked sensors in topographies belong to sensor clusters on the basis of which the null hypothesis of no difference was rejected. N = 30.