I. Introduction
There are three main challenges to improving the accuracy of EEG source imaging: 1. Better Head Modeling: Determining and dealing with error and uncertainty in the electrical head model of how potentials flow from the brain source space to the scalp sensors. 2. Better Source Modeling: Understanding and incorporating spatial source uncertainty and temporal source non-stationarity in solving the EEG inverse problem. 3. Better Spatial Modeling: Allowing the cortical effective sources found by ICA decomposition as projecting coherently to the scalp need to have a smooth projection strength distribution over a compact set of cortical surface voxels. Another important feature of EEG source dynamics is non-stationarity of the effective sources. Here, we consider the stability of effective source scalp projections, and using a single case example illustrate how correctly modeling spatial uncertainty in ICA-decomposed EEG effective source processes may improve the accuracy of their localization.