significance. This may be due to the relatively short number of trials used in the experiment, the low signal-to-noise ratio of MEG recordings (Goldenholz et al., 2009), variations in the peak alpha and gamma oscillatory frequencies (Muthukumaraswamy et al., 2009), combined with the fact that the KL-MI-Tort approach is relatively conservative (van Driel et al., 2015). More generally, it is important to emphasize that all four PAC metrics are highly sensitive to a range of factors (Dvorak and Fenton, 2014; Aru et al., 2015), which are often hard to control (Berman et al., 2012), resulting in both type I and type II statistical errors.
