Asymmetry and non-zero mean of neuronal oscillations: What is essential for the generation of evoked responses?

Abstract

Evoked responses (ERs) from magneto- and electroencephalographic recordings (MEG/EEG) are the primary real-time objective measures of cognitive and perceptual processes in the human brain. For the past fifty years, only two mechanisms for the generation of ERs have been discussed in the literature: „additive“ responses and „phase resetting“ of ongoing oscillations. Recently we introduced a novel mechanism for the generation of ERs (Nikulin et al., 2007), which can be deduced entirely from the properties of the ongoing neuronal oscillations, e.g., alpha oscillations. We have shown analytically and experimentally that neuronal oscillations with a non-zero mean (i.e., the integrated field of one oscillatory cycle is not zero) is a sufficient prerequisite for oscillations to contribute to evoked responses through a mechanism of amplitude modulation. Recently in another study (Mazaheri and Jensen, 2008) it was proposed that an „amplitude fluctuation asymmetry index“ (AFA, the amplitude asymmetry of peaks and troughs around zero) might reveal whether or not an amplitude modulation of oscillations contributes to ERs. Necessarily, oscillations with a non-zero mean will have a non-zero AFA index. Critically, however, the reverse is not true in general, because – as we show here – the AFA index is sensitive also to the non-sinusoidal (e.g., comb-like) shape of neuronal oscillations. We prove analytically and experimentally that the amplitude modulation of non-sinusoidal oscillations with zero mean cannot give rise to evoked responses. Consequently, it is not possible to infer from the AFA index of an oscillation whether its amplitude modulation will contribute to an evoked response. It is therefore crucial to choose methodological approach that is specifically sensitive to the non-zero-mean properties of neuronal oscillations in order to be able to make conclusions about the contribution of the novel ER mechanism to the measured data.