Parafac and go/no-go: disentangling CNV return from the P3 complex by trilinear component analysis

Abstract

It has been proposed that the fronto-central difference between go-P3 and no-go P3 is due to motor-related negativity overlapping P3 in the go case and positive relaxation of such negativity overlapping P3 in the no-go case, and that this motor-related activation may be identified by its having the same topography as the preceding CNV. Testing these suggestions meets with the problem of how to validly distinguish between P3 and the assumed overlapping continuing or downsweeping CNV. Distinguishing between overlapping components is a major reason why principal component analysis (PCA) has been applied, but the major criterion of conventional rotations (Varimax and Promax) is avoiding overlap. Trilinear component analysis (TCA) defines components independently of overlap, by their having stable topographies across time and observations. By applying TCA, we expected to obtain equal P3 components in go and no-go, and a preceding CNV component that would continue and overlap P3 with negative polarity in the go case and with positive polarity in the no-go case. We analyzed data from warned go/no-go tasks, separately for blocks with hand movements and eye movements. Different from expectation, TCA of the Hand task yielded three relevant components: parietal P3, only slightly affected by go vs. no-go, an overlapping CNV downturn unaffected by go vs. no-go, and an anterior-positive/posterior-negative bipolar N2-P3 that modeled the go/no-go effect. No convincing TCA solution was obtained in the Eye task. Possibilities and limitations of the TCA method are discussed.