Discriminating valid from spurious indices of phase-amplitude coupling

Abstract

Recently there has been a strong interest in cross-frequency coupling, the interaction between neuronal oscillations in different frequency bands. In particular, measures quantifying the coupling between the phase of slow oscillations and the amplitude of fast oscillations have been applied to a wide range of data recorded from animals and humans. Some of the measures applied to detect phase-amplitude coupling have been criticized for being sensitive to non-sinusoidal properties of the oscillations and thus spuriously indicate the presence of coupling. While such instances of spurious identification of coupling have been observed, in this commentary we give concrete examples illustrating cases when the identification of cross-frequency coupling can be trusted. These examples are based on control analyses and empirical observations rather than signal processing tools. Finally, we provide concrete advice on how to determine when measures of phase-amplitude coupling can be considered trustworthy.

Significance Statement Neuronal oscillations at different frequencies are thought to reflect processing within and across brain networks. To fully understand how these oscillations support neuronal computation, it is essential to understand how they interact. It is however not straightforward to quantify cross-frequency interactions. We here discuss the problems associated with quantifying cross-frequency coupling and put forward examples in which indices of cross-frequency interactions can be considered reliable.