It is well established that Granger causality can be sensitive to both uncorrelated and correlated noise that is superimposed on the measurements of the signals of interest. This can lead to a substantial amount of false alarms in the identification of Granger-causality relationships (see refs 64, 65, 66, 67, 68). An effective control for the noise problem can be achieved by using the time reversal of signals67. If the Granger causality from x1 to x2 is stronger than the Granger causality from x2 to x1, then we expect that the reverse holds true when we time reverse the signals67. Thus, we only accept conclusions about the asymmetry of Granger causality when both conditions hold true67. For the analysis of the present dataset, we determined for each channel combination whether the asymmetry in Granger-causality values in the theta-frequency range flipped after time reversing the signals. We only selected those ACC-FEF channel combinations for which this held true. In other words, if an ACC (FEF) channel tended to Granger cause the FEF (ACC) channel in the theta range, then we expected that
