ICA decomposes ERP data into a set of components that are distinct and maximally independent time courses but are not necessarily orthogonal scalp projections (Makeig et al. 1997). In other words, ICA spatially and temporally filters data without the assumption of the orthogonality of components to represent the input data as a sum of temporally independent and spatially fixed components that arise from distinct or overlapping source activations. The ICA method has been demonstrated to extract independent components of early and late ERP potentials that can explain functionally distinct brain processes (Makeig et al. 1999a,b), and has been applied to a variety of task paradigms involving perceptual, cognitive and emotional processes (Debener et al. 2005a; Desjardins and Segalowitz 2013; Iidaka et al. 2006; Matsumoto et al. 2005; Sato et al. 2001; Schevernels et al. 2014).
