Power Law Scaling in Human and Empty Room MEG Recordings

In 2009, we published a paper in PLOS Computational Biology [1] that described using a new, wavelet-based metric of phase synchronization in human MEG data. Specifically, we showed that this metric of phase synchronization, that we called the phase lock index (PLI), demonstrated power law scaling across all frequency intervals or wavelet scales from the low frequency delta band (1–2 Hz) to the high frequency gamma band (35–70 Hz). Based on these experimental results, and additional confirmatory data obtained from PLI measurements on time series generated by computational models of critical systems, we offered the interpretation that power law scaling of phase synchronization in human MEG recordings was compatible with the prior theory that human brain dynamics demonstrate self-organized criticality. More recently, in collaboration with a group at the National Institutes of Health (NIH), we published a paper in the Journal of Neuroscience [2] that described using a similar phase synchronization metric to explore scaling behaviour in MEG data recorded from normal human subjects and, crucially, in MEG data recorded with no human subject present, so-called “empty room” data. In Figure 9 of [2], we showed data indicating that phase synchronization appeared to demonstrate power law scaling even in empty room data; as reproduced here in Fig 1. Fig 1 First two panels of Figure 9 from Shriki et al. [2]. We originally judged this issue to be of minor concern, because, as shown in Figure 1 of [2], NIH empty scanner amplitude variance is about 1–2 orders of magnitude less than equivalent brain scans. This is most likely an underestimation given that data are Z-normalized and absolute amplitudes of empty scanner data should be lower than brain scans. Thus, we reasoned the contribution from “empty scanner” effects to PLI scaling in brain recordings should be insignificant. Nonetheless, the issue was addressed in the Discussion of [2], where we stated: “because of the ambiguity of PLI for brain scans and empty scanner, additional steps such as amplitude comparisons need to be taken into account” (page 7089 of [2]).